4-th ICEGE Workshops - Tessaloniki, Greece PDF

ARISTOTLE UNIVERSITY OF THESSALONIKI
LABORATORY OF SOIL MECHANICS, FOUNDATION & GEOTECHNICAL EARTHQUAKE ENGINEEERING
4th International Conference

on Earthquake Geotechnical
Engineering
25-28 June Thessaloniki Greece
Proceedings of:
Workshops 1, 2, 4
Invited Lectures
Delayed Papers
editor: Kyriazis Pitilakis
Table of Contents
Invited Lectures Presentations
TITLE Authors Page
STABILIZATION OF THE LEANING TOWER OF PISA. BEHAVIOUR OF M. JAMIOLKOWSKY, C. VIGGIANI 1
THE TOWER AFTER STABILIZATION WORKS 2001-2004
NEW ORLEANS LEVEE PERFORMANCE IN HURRICANE KATRINA: R. B. SEED 29

LESSONS FOR CALIFORNIAS LEVEE SITUATION
Keynote Lecture Presentation

TITLE Authors Page
PILE RESPONSE TO LATERAL SPREADING: FIELD OBSERVATIONS R. DOBRY, C. MEDINA, T. ABDOUN, 61
AND CURRENT RESEARCH S. THEVANAYAGAM
Delayed papers
ID TITLE Authors Page
1700 SEISMIC DESIGN LOADS FOR METROPOLITAN SUBWAY TUNNELS: Kiriazis PITILAKIS, Anastasios ANASTASIADIS, 131
THE CASE OF THESSALONIKI METRO Dimitrios RAPTAKIS, Nikolaos BOUSOULAS, Elena
PAPAGEORGIOU
WORKSHOP 1: LARGE SCALE FACILITIES, GEOTECHNICAL STRONG MOTION ARRAYS

AND EXPERIMENTAL SITES
W1-1005 REFLECTIONS ON THE IMPORTANCE OF THE QUALITY OF THE Jean-Louis CHAZELAS, Gopal SP MADABHUSHI 146
INPUT MOTION IN SEISMIC CENTRIFUGE TESTS
W1-1004 STUDY OF PILE RESPONSE TO LATERAL SPREADING USING Ricardo DOBRY, Sabanayagam THEVANAYAGAM, 161
PHYSICAL TESTING AND COMPUTATIONAL MODELING Tarek ABDOUN, Ahmed ELGAMAL,
Usama EL SHAMY, Mourad ZEGHAL, and Claudia
MEDINA
W1-1010 TRENDS AND OPPORTUNITIES IN THE FURTHER USE AND Michel GERADIN and Fabio TAUCER 174
DEVELOPMENT OF THE EU LARGE RESEARCH INFRASTRUCTURES
FOR EARTHQUAKE ENGINEERING
W1-1006 VISUALIZATION OF LARGE-SCALE SEISMIC DATA RECORDS Falko KUESTER , Tara C. HUTCHINSON, Tung-Ju 175
HSIEH
W1-1007 NONLINEAR WAVE PROPAGATION AND TRENDS AT A LARGESCALE Bruce KUTTER and Dan WILSON 187
CENTRIFUGE FACILITY
W1-1011 A 3-D VISUALIZATION SYSTEM FOR LARGE-SCALE Jorge MENESES, Masayoshi SATO and Akio ABE 199
EXPERIMENTAL GEOTECHNICAL EARTHQUAKE DATABASES
W1-1012 15 YEARS OF EUROSEISTEST Kyriazis PITILAKIS, Dimitris RAPTAKIS, Konstantia 211

MAKRA, Francisco CHAVEZ GARCIA, Maria
MANAKOU, Pashalis APOSTOLIDIS, George MANOS
W1-1008 LARGE-SCALE GEOTECHNICAL SIMULATIONS TO ADVANCE Glenn J. RIX, Ellen M. RATHJE, Patricia M. 225
SEISMIC RISK MANAGEMENT FOR PORTS GALLAGHER, and Ross W. BOULANGER
W1-1009 INSTRUMENTED GEOTECHNICAL SITES: CURRENT AND FUTURE Jamison H. STEIDL 234
TRENDS
W1-1001 DENSE SEISMIC INSTRUMENTATION OF SMALL SOFT BASINS Bill (W.R.) STEPHENSON 246
W1-1003 RECENT DEVELOPMENTS OF GEOSYNTHETIC-REINFORCED Fumio TATSUOKA, Junichi KOSEKI, Masaru 256
SOIL STRUCTURES TO SURVIVE STRONG EARTHQUAKES TATEYAMA, Daiki HIRAKAWA
W1-1002 LARGE-SCALE SHAKE TABLE TESTS FOR EARTHQUAKE Tzou-Shin UENG, Meei-Ling LIN, Wen-Jong CHANG, 274
GEOTECHNICAL ENGINEERING AT NCREE, TAIWAN Chia-Han CHEN, and Kuo-Lung
WANG
WORKSHOP 2: GEOTECHNICAL EARTHQUAKE ENGINEERING RELATED TO MONUMENTS
AND HISTORICAL CENTRES
W2-1012 SEISMIC RESPONSE ANALYSIS OF ANCIENT COLUMNS Nikolaos ARGYRIOU, Olga-Joan KTENIDOU, Maria 284
MANAKOU, Pashalis APOSTOLIDIS, Francisco
CHAVEZ GARCIA, Kyriazis PITILAKIS
W2-1015 DESIGN AND IMPLEMENTATION OF ENGINEERING MEASURES FOR Stavros BANDIS, Christos SCHINAS, Elias BAKASIS 302
THE PROTECTION OF A HISTORICAL MONUMENT AT THE SEISMIC
AREA OF MOUNT ATHOS PENINSULA GREECE
W2-1007 SEISMIC RESPONSE OF HISTORICAL CENTERS IN ITALY: SELECTED Antonio COSTANZO, Anna D ONOFRIO, Giuseppe 319
CASE STUDIES LANZO, Alessandro PAGLIAROLI, Augusto PENNA,
Rodolfo PUGLIA, Filippo SANTUCCI DE MAGISTRIS,
Stefania SICA, Francesco SILVESTRI, Paolo
TOMMASI
W2-1001 A RESEARCH ON THE PERFORMANCE OF THE CONCRETE Roozbeh ETTEHAD, Hamed JAHANGIRI 343
STRUCTURES AND THE REASONS OF THEIR FAILURE IN BAM
EARTHQUAKE AND DESIGN SUGGESTIONS
W2-1013 UNDERGROUND MONUMENTS (CATACOMBS) IN ALEXANDRIA, Sayed HEMEDA, Kyriazis PITILAKIS, Ioanna 348
EGYPT PAPAYIANNI, Stavros BANDIS, Mohamed GAMAL
W2-1016 EFFECT OF STRONG WIND TO THE CENTRAL TOWER, BAYON, Yoshimori IWASAKI 370
ANGKOR THOM, CAMBODIA
W2-1011 DAMAGES OF EARTHEN STRUCTURES AT ARG-E-BAM CAUSED BY Yoshinori IWASAKI, Mahmoud NEJATI 378
THE EARTHQUAKE OF DEC. 26, 2003, THE CITADEL AT BAM, IRAN
W2-1002 THE RECONSTRUCTION OF THE TEMPLE OF TEMPLE OF ZEUS AT Nicos MAKRIS, Theodoros PSYCHOGIOS 386
NEMEA: RECENT PROGRESS AND FUTURE PERSPECTIVES
W2-1003 SEISMIC PERFORMANCE OF ROCK BLOCK STRUCTURES WITH Edmund MEDLEY, Dimitrios ZEKKOS 398
OBSERVATIONS FROM THE OCTOBER 2006 HAWAII EARTHQUAKE
W2-1005 REGIONAL SUBSIDENCE AND EARTHQUAKES AS THREATS TO Efrain OVANDO-SHELLEY, Marcia PINTO DE 410
ARCHITECTURAL MONUMENTS IN MEXICO CITY OLIVEIRA, Enrique SANTOYO
W2-1004 USING CLASSICAL MONUMENTS FOR THE ASSESSMENT OF PAST Ioannis PSYCHARIS 427
EARTHQUAKE SCENARIOS
W2-1008 SEISMIC PERFORMANCE OF THE 4TH CENTURY A.D., BYZANTINE Anastasios G. SEXTOS, Kosmas C.STYLIANIDIS 439
LAND WALLS OF THE CITY OF THESSALONIKI, GREECE
W2-1010 INFLUENCE OF ENGINEERING AND GEOLOGICAL ENVIRONMENT ON Erbol SHAIMERDENOV, Askar ZHUSUPBEKOV, 452
ARCHITECTURES MONUMENTS Tursun ZHUNISOV
W2-1009 THE INVERSE PROBLEM: MODELLING PAST EARTHQUAKES FROM Stathis C. STIROS, Villy A. KONTOGIANNI 457
THEIR EFFECTS ON ANCIENT CONSTRUCTIONS THE CASE OF THE
AD365 EAST MEDITERRANEAN EARTHQUAKE
W2-1014 EARTHQUAKE RESPONSE AND VULNERABILITY ASSESSMENT OF Costas SYRMAKEZIS, Athanasios ANTONOPOULOS, 465
MASONRY STRUCTURES Olga MAVROULI
W2-1006 GROUTING OF THREE-LEAF MASONRY: EXPERIMENTAL EVIDENCE Elizabeth VINTZILEOU 477

ON COMPRESSIVE AND SHEAR STRENGTH ENHANCEMENT
WORKSHOP 4: HOW CAN EARTHQUAKE GEOTECHNICAL ENGINEERING CONTRIBUTE TO

SAFER DESIGN OF STRUCTURES TO RESIST EARTHQUAKES?
W4-1001 KEY ISSUES IN THE ANALYSIS OF PILES IN LIQUEFYING SOILS Misko CUBRINOVSKI, Hayden BOWEN 489
W4-1004 STATE OF ART KNOWLEDGE VS. STATE OF PRACTICE IN SEISMIC Mauro DOLCE , Giacomo DI PASQUALE , Agostino 499
RISK MITIGATION THE ITALIAN EXPERIENCE AFTER THE 2002 S. GORETTI
GIULIANO EARTHQUAKE
W4-1003 THE CONTRIBUTION OF GEOTECHNICAL ENGINEERING Michele MAUGERI, Francesco CASTELLI and Maria 511
TO SAFER DESIGN OF EARTHQUAKE RESISTANT Rossella MASSIMINO
BUILDING FOUNDATION
W4-1002 INFLUENCE OF DYNAMIC LOADS (EARTHQUAKE LOADING AND Jost A. STUDER, Hansjrg GYSI 541
AIRPLANE CRASHES) ON THE BEARING CAPACITY OF PILES, A CASE
STUDY
4th International Conference on Earthquake Geotechnical Engineering
Invited Lectures Presentations

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28

29
30

31

32

33
34

35

36

37

38
39
40
41

42

43
44

45
46

47

48

49

50

51
52

53

54

55

56
57

58

59

60
Keynote Lecture Presentation

PILE RESPONSE TO LATERAL
SPREADING: FIELD OBSERVATIONS
AND CURRENT RESEARCH
61
Acknowledgments
62
Kobe Earthquake 1995
63
Outline
64
Fukaehama Island, Kobe 1995
65
Miwa et al. (2000)
Fukaehama Island, Kobe 1995
66
Usuoka et al. (2007)
Foundation System
67
Miwa et al. (2006)
Soil Profile
68
Miwa et al. (2006)
Ground Acceleration and Pore
Pressure Buildup in Liquefied Layer
69
Miwa et al. (2006)
Analytical Model for Dynamic
Soil-Pile-Structure Interaction
70
Miwa et al. (2006)
Pile Failures at Depth just before
Liquefaction (Kinematic Effect)
71
Miwa et al. (2006)
Response Computed by Dynamic Soil-
Pile-Structures Analysis
72
Miwa et al. (2000, 2006)
Lessons from this Case History
Inertial damage to superstructure and pile at

shallow depths may occur much before
liquefaction
Kinematic damage due to large cyclic ground
deformations associated with liquefaction tends
to concentrate at the two interfaces between
liquefied and nonliquefied layers; it occurs just
before or after soil liquefies
Kinematic damage due to large permanent
ground deformation associated with lateral
spreading also tends to concentrate at the top
and bottom of liquefied layer; it develops after
soil liquefies
73
Summaries of Case Histories
NCEER Case Studies of Japanese and U.S.
Earthquakes (Hamada and ORourke; ORourke
and Hamada, 1992)
Two Special Issues on Kobe Earthquake of Soils
and Foundations Journal, 1996 and 1998
Tokimatsu (1999)
Dobry and Abdoun (2001)
Oregon State University Report (Dickenson et al.,
2002)
Ishihara (2003)
U. California Davis Report (Boulanger et al., 2003)
Bhattacharya et al. (2004)
Proc. Workshop U. California Davis (Boulanger
and Tokimatsu, 2005)
74
Free field ground deformation D very important
parameter
Spatial variation of D under structure may
contribute to damage
Top of piles may move about same as D , or
much less if very stiff foundation (high EI, pile
groups, batter piles, superstructural constraints)
If shallow nonliquefied crust above liquefied
layer, passive thrust of that crust is key factor
Damaging maximum bending moments occur at
top / bottom boundaries of liquefied layer
Is pile buckling in liquefied layer a problem?
75
76
Some Unsolved Engineering Questions
77
Ongoing Research, Research Tools
In-depth studies of case histories (mostly in
Japan) using advanced technologies
Field tests with blasting (Rollins et al., 2005;
Ashford et al., 2006)
Large-scale 1 g shaking tests in Japan and U.S.
(6m tall inclined laminar boxes), use of advanced
sensors
Small-scale centrifuge testing (Japan, U.S. at UC
Davis and RPI, UK at Cambridge U.)
Use of advanced IT tools for data integration,
system identification and visualizations
Numerical simulations and analyses (DEM, FEM,
dynamic and static p-y, limit equilibrium /
pushover analyses)
78
Outline
79
80
81
82
Outline
83
RPI 150g-ton Centrifuge
84
85
Pile Bending Moment Profiles During
Shaking
0
Slightly
cemented sand
2 T=15sec
T=20sec
T=30sec
4
Nevada sand
(Dr=40%)
6
Soil depth (m)

8
Slightly
cemented sand
10
-300 -200 -100 0 100 200 300 400
86
87
88
89
90
Outline
91
92
Moment (kN-m) Disp. (cm) Disp. (cm)
Ground and Pile Response
93
94
95
96
Maximum bending moment profiles in single pile tests: 1g
test (water) and centrifuge tests (water and viscous fluid)
Height [m
97
Distance [m]
Distance [m]
98
Comparison between centrifuge and
full-scale shaking table test results
99
100
101
102
Maximum bending moment profiles in single pile tests: 1g
test (water) and centrifuge tests (water and viscous fluid)
Height [m
103
104
Outline
105
Recent Results from 1g Test on Lateral
Spreading
106
107
108
109
Depth (m)
110
111
Acceleration (g)
112
SAA vs Ring Accelerometer
3m Depth
113
SAA vs Potentiometer
Displacement at Surface
114
Input Base Excitation
(displacement of base)
Base Excitation
Deflection (cm)
115
Pore Pressure Development
Depth (m) 116

Lateral Spreading
Deflection (cm)
117
Lateral Spread Initiation - 1
Deflection (cm)
Depth (m)
118
Lateral Spread Initiation - 2
Deflection (cm)
119
Conclusions
We are starting to understand better detailed

mechanics of lateral spreading and interaction
with pile foundations
Some significant parameters:
Free field permanent ground deformation
Shallow nonliquefiable layer
Lateral stiffness (and strength) of pile
foundation system including superstructural
constraints
Areas of pile foundation exposed to soil lateral
pressures
Density, permeability of liquefiable layer
Active research in several countries, combining
field observations and testing, 1g and centrifuge
tests, numerical simulations, advanced sensors
120
Thank you!
121
4 th International Conference on
Earthquake Geotechnical Engineering
June 25-28, 2007
Keynote Lecture No.4
PILE RESPONSE TO LATERAL SPREADING: FIELD

OBSERVATIONS AND CURRENT RESEARCH
RICARDO DOBRY 1
CLAUDIA MEDINA
TAREK ABDOUN
SABANAYAGAM THEVANAYAGAM
LIST OF REFERENCES
Abdoun, T. (1997). Modeling of Seismically Induced Lateral Spreading of Multi-layered Soil and its
Effect on Pile Foundations, Ph.D. Thesis, Rensselaer Polytechnic Institute, Troy, NY.
Abdoun, T., and Dobry, R. (2002). Evaluation of Pile Foundation Response to Lateral Spreading,
Soil Dyn Earthq Eng, Vol. 22, No. 9, pp. 1051-1058.
Abdoun, T., Dobry, R., ORourke, T.D., and Goh, S.H. (2003). Pile Response to Lateral Spreads:
Centrifuge Modeling, J Geotech Geoenviron Eng, Vol. 129, No. 10, pp. 869-878.
Abdoun, T., Dobry, R., ORourke, T.D., and Goh, S.H. (2005). Closure to Pile Response to Lateral
Spreads: Centrifuge Modeling, J Geotech Geoenviron Eng, Vol. 131, No. 4, pp. 532-534.
Abdoun, T., Dobry, R., Zimmie, T., Zeghal, M., and Gallagher, P. (2005). Centrifuge Research of
Countermeasures to Protect Pile Foundations Against Liquefaction-induced Lateral Spreading,
Journal of Earthquake Engineering, Vol. 9, Special Issue 1, pp. 105-125.
Abdoun, T., Danisch, L., and Ha, D. (2005). Advanced Sensing for Real-Time Monitoring of
Geotechnical Systems, ASCE Geotech Special Public No. 138 Site Characterization and Modeling
(Ellen M. Rathje, ed.), Proc Geo-Frontiers 2005 Conference, January 24-26, 2005, Austin, TX.
Abdoun, T., Danisch, L., Ha, D., and Bennett, V. (2006). Advanced Sensing for Real-Time
Monitoring of Geotechnical Systems, TRB 85th Annual Meeting, January 22-26, 2006,
Washington, DC.
Abdoun, T., Abe, A., Bennett, V., Danisch, L., Sato, M., Tokimatsu, K., and Ubilla, J., (2007).
Wireless Real Time Monitoring of Soil and Soil-Structure Systems, Geotech Spec Publ No. 161
Embankments, Dams, and Slopes: Lessons From the New Orleans Levee Failures and Other
Current Issues (F. Silva-Tulla and P. Nicholson, eds.), Proc GeoDenver 2007 Conference: New
Peaks in Geotechnics, February 18-21, 2007, Denver, CO.
Adalier, K. (1996) Mitigation of Earthquake Induced Liquefaction Hazards, Ph.D. Thesis,

Rensselaer Polytechnic Institute, Troy, NY.
1
Professor, Department of Civil and Environmental Engineering, Rensselaer Polytechnic Institute,
Troy, NY, E-mail: dobryr@rpi.edu
122
Ashford, S.A., Juirnarongrit, T., Sugano, T., and Hamada, M. (2006) Soil-Pile Response to Blast-
Induced Lateral Spreading. I: Field Test, J Geotech Geoenviron Eng, Vol. 132, No. 2, pp. 152-
162.
Bartlett, S.F., and Youd, T.L. (1992). Case Histories of Lateral Spreads Caused by the 1964 Alaska
Earthquake, Ch. 2 of Case Studies of Liquefaction and Lifeline Performance During Past
Earthquakes, Vol. 2: United States Case Studies (ORourke and Hamada, eds.), NCEER, SUNY-
Buffalo, Buffalo, NY, Tech. Rept. NCEER-92-0002, pp. 2.1-2.127
Beavers, J.E., (ed.) (1991). Costa Rica Earthquake Reconnaissance Report, Earthquake Spectra,
EERI, Supplement B, Vol. 7, 127 pages.
Benuzca, L. (ed.) (1990). Loma Prieta Earthquake Reconnaissance Report, Earthquake Spectra,
EERI and NRC, Supplement to Vol. 6, 448 pages.
Berrill, J.B., Christensen, S.A., Keenan, R.J., Okada, W., and Pettinga, J.K. (1997). Lateral-
Spreading Loads on a Piled Bridge Foundation, Seismic Behavior of Ground and Geotechnical
Structures, (Seco E. Pinto, ed.), Balkema, Rotterdam, pp. 176-183.
Bhattacharya, S., Madabhushi, S.P.G., and Bolton, M.D. (2003). Pile Instability During Earthquake
Liquefaction, ASCE Engineering Mechanics Conference EM2003, July 16-18, 2003, Seattle, WA.
Bhattacharya, S., and Bolton, M.D. (2004). Errors in Design Leading to Pile Failures During Seismic
Liquefaction, Proc 5th International Conference on Case Histories in Geotechnical Engineering,
April 13-17, 2004, New York, NY.
Bhattacharya, S., Madabhushi, S.P.G., and Bolton, M.D. (2004). An Alternative Mechanism of Pile
Failure Deposits During Earthquakes, Geotechnique, Vol. 54, No. 3, pp. 203-213.
Bhattacharya, S., Madabhushi, S.P.G., and Bolton, M.D. (2005). Closure to An Alternative
Mechanism of Pile Failure Deposits During Earthquakes, Geotechnique, Vol. 55, No. 3, p. 263.
Boulanger, R.W., Mejia, L.H., and Idriss, I.M. (1997). Liquefaction at Moss Landing During Loma
Prieta Earthquake, Journal of Geotechnical and Geoenvironmental Engineering, Vol. 123, No.5,
pp.453-467.
Boulanger, R.W., Meyers, M.W., Mejia, L.H., and Idriss, I.M. (1998). Behavior of a Fine-grained
Soil During the Loma Prieta Earthquake, Canadian Geotechnical Journal, Vol. 35, pp.146-158.
Boulanger, R.W., Kutter, B.L, Brandenberg, S.J., Singh, P., and Chang, D. (2003). Pile Foundations
in Liquefied and Laterally Spreading Ground During Earthquakes: Centrifuge Experiments and
Analyses, Report No. UCD/CGM-03/01, Center for Geotechnical Modeling, University of
California, Davis, CA, 205 pages.
Boulanger, R.W. and Tokimatsu, K., eds. (2006). Proceedings Workshop on Seismic Performance
and Simulation of Pile Foundations in Liquefiable and Laterally Spreading Ground, March 16-18,
2005, University of California, Davis, CA, ASCE Geotech Special Public No. 145, pp. 50-60.
Brandenberg, S.J., Boulanger, R.W., and Kutter, B.L. (2005). Discussion of Pile Response to Lateral
Spreads: Centrifuge Modeling by Dobry et al., J Geotech Geoenviron Eng, Vol. 131, No. 4, pp.
529-531.
Brandenberg, S.J., Boulanger, R.W., Kutter, B.L., and Chang, D. (2005). Behavior of Pile
Foundations in Laterally Spreading Ground During Centrifuge Tests, J Geotech Geoenviron Eng,
Vol. 131, No. 11, pp. 1378-1391.
123
Brandenberg, S.J., Boulanger, R.W., Kutter, B.L., and Chang, D. (2007). Liquefaction-Induced
Softening of Load Transfer Between Pile Groups and Laterally Spreading Crusts, J Geotech
Geoenviron Eng, Vol. 133, No. 1, pp. 91-103.
Cubrinovski, M., Kokusho, T., and Ishihara, K. (2004). Interpretation From Large-Scale Shake Table
Tests on Piles Subjected to Spreading of Liquefied Soils, Proc 11th Intl Conf on Soil Dyn and
Earthq Eng, January 7-9, University of California, Berkeley, CA, Vol. 2, pp. 463-470.
Cubrinovski, M., Kokusho, T., and Ishihara, K. (2006). Interpretation From Large-Scale Shake Table
Tests on Piles Undergoing Lateral Spreading in Liquefied Soils, J of Soil Dyn and Earthq Eng,
Vol. 26, pp. 275-286.
Dale, G. (2002). Prediction of Pile Bending Response to Lateral Spreading by Elastoplastic Analysis
Validated by Centrifuge Tests, M.S. Thesis, Rensselaer Polytechnic Institute, Troy, NY.
Dickenson, S.E., McCullough, N.J., Barkau, M.G., and Wavra, B.J. (2002) Assessment and
Mitigation of Liquefaction Hazards to Bridge Approach Embankments in Oregon, Final Report
SPR 361, Oregon Department of Transportation and Federal Highway Administration, 272 pages.
Dobry, R., and Abdoun, T. (1998). Post-triggering Response of Liquefied Sand in the Free Field and
Near Foundations, Proc ASCE Geotechnical Earthquake Engineering and Soil Dynamics III,
Seattle, WA, August 3-6 1998, Vol. 1, pp. 270-300.
Dobry, R., and Abdoun, T. (2001). Recent Studies on Seismic Centrifuge Modeling of Liquefaction
and its Effect on Deep Foundation, State-of-the-Art Paper, Proc 4th Int Conf on Recent Advances
in Geotech Earthq Eng and Soil Dyn (S. Prakash, ed.), Paper SOAP 3, Vol. 2, March 26-31, 2001,
San Diego, CA, 30 pages.
Dobry, R., Taboada, V., and Liu, L. (1995). Centrifuge Modeling of Liquefaction Effects During
Earthquakes, Proc 1st Int Conf on Earthq Geotech Eng (IS-Tokyo), Tokyo, Japan, Keynote
Lecture, pp. 1291-1324.
Dobry, R., Abdoun, T., ORourke, T.D., and Goh, S.H. (2003). Single Piles in Lateral Spreads: Field
Bending Moment Evaluation, J Geotech Geoenviron Eng, Vol. 129, No. 10, pp. 879-889.
Dobry, R., Abdoun, T., ORourke, T.D., and Goh, S.H. (2005). Closure to Single Piles in Lateral
Spreads: Field Bending Moment Evaluation,. J Geotech Geoenviron Eng, Vol. 131, No. 4, pp.
532-534.
Dobry, R., Thevanayagam, S., Abdoun, T., Elgamal, A., El Shamy, U., Zeghal, M., and Medina, C.
(2007). Study of Pile Response to Lateral Spreading Using Physical Testing and Computational
Modeling, Proc. 4th Intl Conf Earthquake Geotechnical Engineering, Thessaloniki, Greece, June
25-28.
Dungca, J.R., Kuwano, J., Saruwatari, T., Izawa, J., Suzuki, H., and Tokimatsu, K. (2004). Shaking
Table Tests on the Lateral Response of a Pile Buried in Liquefied Sand, Proc 11th Intl Conf on
Soil Dyn and Earthq Eng, January 7-9, University of California, Berkeley, CA, Vol. 2, pp. 471-
477.
Dungca, J.R., Kuwano, J., Takahashi, A., Saruwatari, T., Izawa, J., Suzuki, H., and Tokimatsu, K.
(2006). Shaking Table Tests on the Lateral Response of a Pile Buried in Liquefiable Sand, J of
Soil Dyn and Earthq Eng, Vol. 26, pp. 287-295.
124
Elgamal, A., Zeghal, M., Taboada, V., and Dobry, R. (1996). Analysis of Site Liquefaction and
Lateral Spreading Using Centrifuge Testing Records, Soils and Foundations, JGS, Vol. 36, No. 2,
pp. 111-121.
Elgamal, A., Parra, E., Yang, Z. and Adalier, K. (2002). Numerical Analysis of Embankment
Foundation Liquefaction Countermeasures, J Earthq Eng, Vol. 6, No. 4, pp. 447-471.
Elgamal, A., Yang, Z., and Parra, E. (2002). Computational modeling of Cyclic Mobility and Post-
Liquefaction Site Response, J of Soil Dyn and Earthq Eng, Vol. 22, Issue 4, pp. 259-271.
Elgamal, A., Yang, Z., Parra, E., and Ragheb, A. (2003). Modeling of Cyclic Mobility in Saturated
Cohesionless Soils, Int J of Plast, Vol. 19, Issue 6, pp. 883-905.
Elmekati, A., and Zeghal, M. (2007). Systen Identification Analysis of Test LG-0, Internal Progress
Report, Rensselaer Polytechnic Institute, Troy, NY.
Finn, W.D.L. (2004). An Overview of the Behavior of Pile Foundations in Liquefiable and Non-
Liquefiable Soils during Earthquake Excitation, Ishihara Lecture, Proc 11th Intl Conf on Soil Dyn
and Earthq Eng, January 7-9, University of California, Berkeley, CA, Vol. 1, pp. 57-67.
Gonzalez, L., Abdoun, T., and Dobry, R. (2006). Effect of Soil Permeability on Centrifuge Modeling
of Pile Response to Lateral Spreading, ASCE Geotech Special Public No. 145 (Boulanger and
Tokimatsu, eds.), Proc Workshop on Seismic Performance and Simulation of Pile Foundations in
Liquefiable and Laterally Spreading Ground, March 16-18, 2005, University of California, Davis,
CA, pp. 50-60.
Gonzalez, L., Abdoun, T., and Dobry, R. (2007). Effect of Soil Permeability on Centrifuge Modeling
of Pile Response to Lateral Spreading, J Geotech Geoenviron Eng (accepted).
Guo, W.D., and Ghee, E.H. (2006). Behavior of Axially Loaded Pile Groups Subjected to Lateral
Soil Movement, ASCE Geotech Special Public No. 153 Foundation Analysis and Design:
Innovative Methods (Parsons, R.L., Zhang, L., Guo, W.D., Phoon, K.K., and Yang, M., eds.), Proc
Geoshanghai 2006 Conference, June 6-8, 2006, Shanghai, China.
Haigh, S.K., and Madabhushi, S.P.G. (2006). The Effect of Pile Flexibility on Pile-Loading in Lateral
Spreading Slopes, ASCE Geotech Special Public No. 145 (Boulanger and Tokimatsu, eds.), Proc
Workshop on Seismic Performance and Simulation of Pile Foundations in Liquefiable and
Laterally Spreading Ground, March 16-18, 2005, University of California, Davis, CA, pp. 24-37.
Hamada, M. (1992). Large Ground Deformations and Their Effects on Lifelines: 1964 Niigata
Earthquake, Ch. 3 of Case Studies of Liquefaction and Lifeline Performance During Past
Earthquake, Vol. 1: Japanese Case Studies (Hamada and ORourke, eds.), NCEER, SUNY-Buffalo,
Buffalo, NY, Tech. Rept. NCEER-92-0001, pp. 3.1-3.123.
Hamada, M. (1992). Large Ground Deformations and Their Effects on Lifelines: 1983 Nihonkai-
Chubu Earthquake, Ch. 4 of Case Studies of Liquefaction and Lifeline Performance During Past
Earthquake, Vol. 1: Japanese Case Studies (Hamada and ORourke, eds.), NCEER, SUNY-Buffalo,
Buffalo, NY, Tech. Rept. NCEER-92-0001, pp. 4.1-4.85.
Hamada, M., and ORourke, T.D. (eds.), (1992). Case Studies of Liquefaction and Lifeline
Performance During Past Earthquakes. Vol. 1: Japanese Case Studies, NCEER, SUNY-Buffalo,
Buffalo, NY, Tech. Rept. NCEER-92-0001.
125
Hamada, M., Yasuda, S., Isoyama, R., and Emoto, K. (1986). Study on Liquefaction Induced
Permanent Ground Displacement, Research Committee Report, Association for the Development
of Earthquake Prediction, Tokyo, Japan, 87 pages.
Hamada, M., R. Isoyama, K. Wakamatsu, (1995). The 1995 Hyogoken-Nanbu (Kobe) earthquake:
Liquefaction, Ground Displacement and Soil Condition in Hanshin Area, Assoc. for Development
of Earthquake Prediction, School of Science and Engineering, Waseda University, and the Japan
Engineering Consultants, Publ. by the Assoc. for Development of Earthquake Prediction, 194
pages.
Hamada, M., Wakamatsu, K., and Ando, T. (1996). Liquefaction-Induced Ground Deformation and
its Caused Damage During the 1995 Hyogeken-Nanbu Earthquake, Proc. 6th Japan-U.S.
Workshop on Earthquake Resistant Design of Lifeline Facilities and Countermeasures Against Soil
Liquefaction (Hamada and ORourke, eds.), NCEER, SUNY-Buffalo, Buffalo, NY, Tech. Rept.
NCEER-96-0012, Vol. 1, pp. 137-152.
Harada, N., Towhata, I., Takatsu, T., Tsunoda, S., and Sesov, V. (2004). Development of New Drain
Method for Protection of Existing Pile Foundations From Liquefaction Effects, Proc 11th Intl Conf
on Soil Dyn and Earthq Eng, January 7-9, University of California, Berkeley, CA, Vol. 2, pp. 498-
505.
Harada, N., Towhata, I., Takatsu, T., Tsunoda, S., and Sesov, V. (2006). Development of New Drain
Method for Protection of Existing Pile Foundations From Liquefaction Effects, J of Soil Dyn and
Earthq Eng, Vol. 26, pp. 297-312.
He, L., Elgamal, A., Abdoun, T., Abe, A., Dobry, Meneses, J., Sato, M., and Tokimatsu, K. (2006).
Lateral Load on Piles Due to Liquefaction-Induced Lateral Spreading During 1G Shake Table
Experiments, Proc. 8th U.S. National Conf on Earthq Eng, April 18-22, 2006, San Francisco, CA,
Paper No. 881.
Hwang, J.I., Kim, C.Y., Chung, C.K., and Kim, M.M. (2004). Behavior of a Single Pile Subjected to
Flow of Liquefied Soil in an Infinite Slope, Proc 11th Int Conf on Soil Dyn and Earthq Eng, Vol.
2, pp. 573-580.
Idriss, I.M., and Boulanger, R.W. (2004). Semi-empirical Procedures for Evaluating Liquefaction
Potential During Earthquakes, Proc 11th Intl Conf on Soil Dyn and Earthq Eng, January 7-9,
University of California, Berkeley, CA, Vol. 1, pp. 32-56.
Ishihara, K. (2003). Liquefaction-induced Lateral Flow and its Effects on Foundation Piles, Proc.
Fifth Natl Conf on Earthquake Engineering, Istanbul, Turkey, May 26-30, Invited Lecture, 28
pages.
Ishihara, K., Yasuda, S., and Nagase, H. (1996). Soil Characteristics and Ground Damage, Soils and
Foundations Special Issue on Geotechnical Aspects of the January 17 1995 Hyogoken-Nambu
Earthquake, JGS, Special Vol. January, pp. 109-118.
Iwasaki, T. (1986). Soil Liquefaction Studies in Japan, Technical Memorandum of PWRI No. 2239,
Public Works Research Institute, Ministry of Construction, Japan, 143 pages.
Juirnarongrit, T., and Ashford, S.A. (2006). Soil-Pile Response to Blast-Induced Lateral Spreading.
II: Analysis and Assessment of the p-y Method, J Geotech Geoenviron Eng, Vol. 132, No. 2, pp.
163-172.
Japanese Geotechnical Society (1998). Special Issue of Soils and Foundations on the Effects of the
1995 Hyogoke-Nambu Earthquake.
126
Kachadoorian, R. (1968). Effects of the Earthquake of March 27, 1964, on the Alaska Highway
System, U.S. Geological Survey Professional Paper 545-C, U.S. Geological Survey.
Kagawa, T., Sato, M., Minowa, C., Abe, A., and Tazoh, T. (2003) Centrifuge Simulations of Large-
Scale Shaking Table Tests: Case Studies, J Geotech Geoenviron Eng, Vol. 130, No. 7, pp. 663-
672.
Knappett, J.A., and Madabhushi, S.P.G. (2006). Modelling Effects of Axial Load on Behaviour of
Pile Groups in Laterally-Spreading Soil,. Proc 6th International Conference on Physical Modelling
in Geotechnics, August 4-6, 2006, Hong Kong, pp. 1007-1012.
Kawashima, K., Shimizu, K., Mori, S., Takagi, M., Susuki, N, and Nakamura, S. (1988). Analytical
Studies on Damage to Bridges and Foundation Piles Caused by Liquefaction-Induced Permanent
Ground Displacement, Proc. Japan-U.S. Workshop on Liquefaction, Large Ground Deformation
and Their Effect on Lifeline Facilities, Nov. 16-19, Tokyo, Japan, Association for the Development
of Earthquake Prediction, (Japan) and NCEER (U.S.), pp. 99-117.
Koyamada, K., Miyamato, Y., and Tokimatsu, K. (2006). Field Investigation and Analysis Study of
Damage Pile Foundation During the 2003 Tokachi-Oki Earthquake, ASCE Geotech Special Publ
No. 145 (Boulanger and Tokimatsu, eds.), pp. 97-108.
Lin S., Tseng, Y., Chiang, C., and Hung, C. (2006). Damage of Piles Caused by Lateral Spreading-
Back Study of Three Cases, ASCE Geotech Special Public No. 145 (Boulanger and Tokimatsu,
eds.), Proc Workshop on Seismic Performance and Simulation of Pile Foundations in Liquefiable
and Laterally Spreading Ground, March 16-18, 2005, University of California, Davis, CA, pp. 121-
133.
Matsui, T. and K. Oda. (1996). Foundation Damage to Structures, Soils and Foundations Special
Issue on Geotechnical Aspects of the January 17 1995 Hyogoken-Nambu Earthquake, JGS, Special
Vol., January, pp. 189-200.
McCulloch, D.S., and Bonilla, M.G. (1970). Effects of the Earthquake of March 27, 1964 on the
Alaska Railroad, U.S. Geological Survey Professional Paper 545-D, U.S. Geological Survey.
Miwa, S., Ikeda, T., and Sato, T. (2006). Damage Process of Pile Foundation in Liquefied Ground
During Strong Ground Motion, J of Soil Dyn and Earthq Eng, Vol. 26, pp. 325-336.
Mizuno, H. (1987). Pile Damage During Earthquakes in Japan (1923-1983), Proc ASCE Session on
Dynamic Response of Pile Foundations (T. Nogami ed.), April 27, 1987, Atlantic City, pp. 53-77.
ORourke, T.D., and Hamada, M. (eds.), (1992). Case Studies of Liquefaction and Lifeline
Performance During Past Earthquakes. Vol. 2: United States Case Studies, NCEER, SUNY-
Buffalo, Buffalo, NY, Tech. Rept. NCEER-92-0002.
O'Rourke, T.D., and Pease, J.W. (1992). Large Ground Deformations and Their Effects on Lifeline
Facilities: 1989 Loma Prieta Earthquake, Ch. 5 of Case Studies of Liquefaction and Lifeline
Performance During Past Earthquake, Vol. 2: United States Case Studies (ORourke and Hamada,
eds.), NCEER, SUNY-Buffalo, Buffalo, NY, Tech. Rept. NCEER-92-0002, pp. 5.1-5.85.
Parra, E. (1996). Numerical Modeling of Liquefaction and Lateral Ground Deformation Including
Cyclic Mobility and Dilation Response in Soil Systems, Ph.D. Thesis, Rensselaer Polytechnic
Institute, Troy, NY.
127
Pyke, R. (2003). Discussion of Liquefaction Resistance of Soils: Summary Report from the 1996
NCEER and 1998 NCEER/NSF Workshops on Evaluation of Liquefaction Resistance of Soils by
Youd et al., J Geotech Geoenviron Eng, Vol. 129, No. 3, pp. 283-284.
Radwan, H., and Abdoun, T. (2006). 3D Data Viewer of Centrifuge Physical Models, Geotech
Special Publ Proc GeoCongress 2006 Conference: Geotechnical Engineering in the Information
Technology Age (D.J. DeGroot, J.T. DeJong, J.D. Frost, and L.G. Baise, eds.), February 26-March
1, 2006, Atlanta, GA.
Ramos, R. (1999). Centrifuge Study of Bending Response of Pile Foundation to a Lateral Spread
Including Restraining Effect of Superstructure, Ph.D. Thesis, Rensselaer Polytechnic Institute,
Troy, NY.
Ross, G.A., Seed, H.B., and Migliaccio, R.R. (1973). Performance of Highway Bridge Foundations,
The Great Alaska Earthquake of 1964, Engineering Volume, National Academy of Sciences,
Washington, D.C., pp. 190-242.
Seed, R.B., Dickenson, S.E., Riemer, M.F., Bray, J.D., Sitar, N., Mitchell, J.K., Idriss, I.M., Kayen,
R.E., Kropp, A., Harder Jr., L.F., and Power, M.S. (1990). Preliminary Report on the Principal
Geotechnical Aspects of the October 17, 1989 Loma Prieta Earthquake, Report No. UCB/EERC-
90/05, Earthquake Engineering Research Center, University of California at Berkeley, Berkeley,
CA., 139 pages.
Shea, G.H., (ed.) (1991). Costa Rica Earthquake Reconnaissance Report, Earthquake Spectra, EERI,
Supplement B, Vol. 7, Chapter 6 Bridges, 127 pages.
Shengcong, F., and Tatsuoka, F. (1984). Soil Liquefaction During Haicheng and Tangshan
Earthquakes in China; A Review, Soils and Foundations, Vol. 24, No. 4, pp 11-29.
Shenthan T. (2001). Factors Affecting Liquefaction Miitigation in Silty Soils Using Stone Columns,
MS Thesis, University at Buffalo, SUNY, NY, p. 129.
Shibata, T., F. Oka, and Y. Ozawa. (1996). Characteristics of Ground Deformation due to
Liquefaction, Soils and Foundations Special Issue on Geotechnical Aspects of the January 17
1995 Hyogoken-Nambu Earthquake, JGS, Special Vol. January, pp. 65-80.
Stewart, H.E., Miura, F., and ORourke, T.D. (1988). Pile Damage Due to Large Ground
Displacement, Proc. Japan-U.S. Workshop on Liquefaction, Large Ground Deformation and Their
Effect on Lifeline Facilities, Nov. 16-19, Tokyo, Japan, Association for the Development of
Earthquake Prediction, (Japan) and NCEER (U.S.), pp. 173-182.
Susuki H., and Tokimatsu, K. (2004). Effect of Pore Water Pressure Response Around Pile on p-y
Relation During Liquefaction, Proc 11th Intl Conf on Soil Dyn and Earthq Eng, January 7-9,
Susuki H., Tokimatsu, K., Sato, M., and Abe, A. (2006) Factor Affecting Horizontal Subgrade
Reaction of Piles During Soil Liquefaction and Lateral Spreading, ASCE Geotech Special Public
No. 145 (Boulanger and Tokimatsu, eds.), Proc Workshop on Seismic Performance and Simulation
of Pile Foundations in Liquefiable and Laterally Spreading Ground, March 16-18, 2005, University
of California, Davis, CA, pp. 1-10.
Swan ,S.W., Flores, P.J., and Hooper, J.D. (1996). The Manzanillo Mexico Earthquake of October 9,
1995, NCEES Bulletin, The Quarterly Publication of NCEER, Vol. 10, No. 1, January.
128
Taboada, V.M. (1995). Centrifuge Modeling of Earthquake-Induced Lateral Spreading in Sand Using
a Laminar Box, J Geotech Geoenviron Eng, Vol. 129, No. 10, pp. 879-889.
Tamura, S., and Tokimatsu, K. (2006) Seismic Earth Pressure Acting on Embedded Footing Based on
Large-Scale Shaking Table Tests, ASCE Geotech Special Public No. 145 (Boulanger and
Tokimatsu, eds.), Proc Workshop on Seismic Performance and Simulation of Pile Foundations in
Liquefiable and Laterally Spreading Ground, March 16-18, 2005, University of California, Davis,
CA, pp. 83-96.
TGC (1995). Geotechnical Response of Quay and Loading Areas in Puerto Manzanillo During
October 1995 Earthquake, (in Spanish) Report TGC Geotecnia 95-1763-C, October, Mexico.
Tokimatsu, K. (1999). Performance of Pile Foundations in Laterally Spreading Soils, Proc 2nd Int
Conf on Earthq Geotech Eng (P. Seco e Pinto, ed.), June 21-25, 1999, Lisbon, Portugal, Vol. 3, pp.
957-964.
Tokimatsu K., Mizuno, H., and Kakurai, M. (1996). Building Damage Associated with Geotechnical
Problems, Soils and Foundations Special Issue on Geotechnical Aspects of the January 17 1995
Hyogoken-Nambu Earthquake, JGS, Special Vol. January, pp. 219-234.
Tokimatsu, K., H. Oh-Oka, Y. Shamoto, A. Nakazawa, and Y. Asaka. (1997). Failure and
Deformation Modes of Piles Caused by Liquefaction-Induced Lateral Spreading in 1995
Hyogoken-Nambu Earthquake, Proc 3rd Kansai International Geotechnical Forum on Comparative
Geotechnical Engineering (KIG-Forum '97), Kobe, January, Publ. by the Kansai Branch of the
Japanese Geotechnical Society, pp. 239-248.
Tokimatsu, K., H. Oh-Oka, K. Satake, Y. Shamoto, and Y. Asaka, Y. (1998). Effects of Lateral
Ground Movements on Failure Patterns of Piles in the 1995 Hyogoken-Nambu Earthquake, Proc
ASCE Geotechnical Earthquake Engineering and Soil Dynamics III, Seattle, WA, August 3-6,
1998, Vol. 2, pp. 1175-1186
Tokimatsu, K., Suzuki, H., and Sato, M. (2004). Influence of Inertial and Kinematic Components on
Pile Response during Earthquakes, Proc 11th Intl Conf on Soil Dyn and Earthq Eng, January 7-9,
Tokimatsu, K., Suzuki, H., and Sato, M. (2005). Effects of Inertial and Kinematic Interaction on
Seismic Behavior of Pile with Embedded Foundation, J of Soil Dyn and Earthq Eng, Vol. 25, pp.
753-762.
Towhata, I., Sesov, V., Motamed, R., and Gonzalez, M. (2006). Model Tests on Lateral Earth
Pressure on Large Group Pile Exerted by Horizontal Displacement of Liquefied Sandy Ground,
Proc. 8th U.S. National Conf on Earthq Eng, April 18-22, 2006, San Francisco, CA, Paper No.
1227.
Ubilla, J. (2007). Physical Modeling of the Effects of Natural Hazards on Soil-Structure Interaction,
Ph.D. Thesis, Rensselaer Polytechnic Institute, Troy, NY.
Uzuoka, R., Sento, N., Kazama, M., Zhang, F., Yashima, A., and Oka, F. (2007). Three-dimensional
Numerical Simulation of Earthquake Damage to Group-Piles in a Liquefied Ground, J of Soil Dyn
and Earthq Eng, Vol. 27, pp. 395-413.
Yao, S., Kobayashi, K., Yoshida, N., and Matsuo, H. (2004). Interactive Behavior of Soil-Pile-
Superstructure System in Transient State to Liquefaction by Means of Large Shake Table Test, J
of Soil Dyn and Earthq Eng, Vol. 24, pp. 397-409.
129
Yasuda, F., A. Nanjoh, and K. Kosa. (1997). Investigation and Checking of Load Bearing Capacity
of Damaged Piles, Proc. of the 3rd Kansai International Geotechnical Forum on Comparative
Geotechnical Engineering (KIG-Forum '97), Kobe, January, Publ. by the Kansai Branch of the
Japanese Geotechnical Society, pp. 249-257.
Yokoyama K., Tamura, K., and Matsuo, O. (1997). Design Methods of Bridge Foundations Against
Soil Liquefaction and Liquefaction-Induced Ground Flow, Proc 2nd Italy-Japan Workshop on
Seismic Design and Retrofit of Bridges, February 27-28, 1997, Rome, Italy, pp. 109-131.
Yoshida, N., and Hamada, M. (1991). Damage to Foundation Piles and Deformation Pattern of
Ground Due Liquefaction-Induced Permanent Ground Deformations, Proc. 3rd Japan-U.S.
Workshop on Earthquake Resistant Design of Lifeline Facilities and Countermeasures for Soil
Liquefaction, NCEER, SUNY-Buffalo, Buffalo, NY, Tech. Rept. NCEER-91-0001 , pp. 147-161.
Yoshida, N., Watanabe, H., Yasuda, S., and Mora, S. (1992). Liquefaction-Induced Ground Failure
and Related Damage to Structures During 1991 Telire-Limon, Costa Rica, Earthquake, Proc. 4th
Japan-U.S. Workshop on Earthquake Resistant Design of Lifeline Facilities and Countermeasures
for Soil Liquefaction (Hamada and ORourke, eds.), NCEER, SUNY-Buffalo, Buffalo, NY, Tech.
Rept. NCEER-92-0019, Vol. 1, pp. 37-52.
Yoshida, N., Towhata, I., Yasuda, S., and Kanatani, M. (2005). Discussion to An Alternative
Mechanism of Pile Failure Deposits During Earthquakes by Bhattacharya et al., Geotechnique,
Vol. 55, No. 3, pp. 259-263.
Youd, T.L. (1993), Liquefaction-Induced Damage to Bridges, Transportation Research Record,

published by the TRB and the National Research Council, Washington, DC., No. 1411, pp. 35-41.
Youd, T.L., Rollins, K.M., Salazar, A.F., and Wallace, R.M. (1992). Bridge Damage Caused by
Liquefaction During the 22 April 1991 Costa Rica Earthquake, Proc. 10th World Conference on
Earthquake Engineering, July 19-24, Madrid, Spain, Vol. 1, pp 153-158.
Youd, T.L., Idriss, I.M., Andrus, R.D., Arango, I., Castro, G., Christian, J.T., Dobry, R., Finn, W.D.L.,
Harder Jr., L.F., Hynes, M.E., Ishihara, K., Koester, J.P., Liao, S.S.C., Marcuson III, W.F., Martin,
G.R., Mitchell, J.K., Moriwaki, Y., Power, M.S., Robertson, P.K., Seed, R.B., and Stokoe II, K.H.
(2001). Liquefaction Resistance of Soils: Summary Report from the 1996 NCEER and 1998
NCEER/NSF Workshops on Evaluation of Liquefaction Resistance of Soils, J Geotech
Youd, T.L., Idriss, I.M., Andrus, R.D., Arango, I., Castro, G., Christian, J.T., Dobry, R., Finn, W.D.L.,
Harder Jr., L.F., Hynes, M.E., Ishihara, K., Koester, J.P., Liao, S.S.C., Marcuson III, W.F., Martin,
G.R., Mitchell, J.K., Moriwaki, Y., Power, M.S., Robertson, P.K., Seed, R.B., and Stokoe II, K.H.
(2003). Closure to Liquefaction Resistance of Soils: Summary Report from the 1996 NCEER and
1998 NCEER/NSF Workshops on Evaluation of Liquefaction Resistance of Soils, J Geotech
Zeghal, M., and El Shamy, U. (2004). A Continuum-Discrete Hydromechanical Analysis of Granular

Deposit Liquefaction, Int J Numer Anal Met Geomech, Vol. 28, No. 14, pp. 1361-1383.
Zeghal, M., and Oskay, C. (2003). A Local Identification Technique for Geotechnical and
Geophysical Systems, Int J Numer Anal Meth Geomech, Vol. 27, No. 11, pp. 967-987.
Zeghal, M., Elgamal, A.W., Tang, H.T., and Stepp, J.C. (1995). Lotung Downhole Array. II:
Evaluation of Soil Nonlinear Properties. J of Geotech Eng, Vol. 121, No. 4, pp. 363-378.
130
Delayed Papers

131

132

133

134

135

136

137

138

139

140

141

142

143

144

145
WORKSHOP 1
Large scale facilities, geotechnical strong
motion arrays and experimental sites
June 25-28, 2007
Paper No. W1-1005
REFLECTIONS ON THE IMPORTANCE OF THE QUALITY OF THE

INPUT MOTION IN SEISMIC CENTRIFUGE TESTS
Jean-Louis CHAZELAS 1, Gopal SP MADABHUSHI2
ABSTRACT
Dynamic centrifuge modelling is widely accepted as the experimental technique that can be used to
understand complex behaviour of soil-structure systems subjected to earthquake loading. In Europe,
the centrifuge facility at Schofield Centre in Cambridge has been active for more than 30 years in the
area of dynamic centrifuge modelling. The earthquake loading is simulated on the Cambridge
centrifuge using simple mechanical actuators that produce sinusoidal shaking. While such input
motions are simple and helpful in deciphering certain aspects of soil behaviour particularly while
assessing damaging effects of the earthquakes, the complex non-linear behaviour of soils requires
more sophisticated earthquake actuators that can simulate multi-frequency nature of real earthquakes.
Such a servo-hydraulic shaker has been established on the LCPC centrifuge in Nantes, similar to the
shaker at C-Core centrifuge facility in St Johns, Canada. In this keynote paper, the importance of the
quality of input motion is investigated. The difficulties in generating complex motion aboard
centrifuges are discussed. Another aim of this paper is to discuss some of the exciting developments
that are occurring in the modelling of earthquakes on centrifuges. The outline design of a 2-D
(horizontal and vertical) shaker being developed at Cambridge is presented. Similarly creation of
distributed testing facilities that are networked within UK under the UK-NEES project that is linked to
US-NEES and other similar networks opens up a new era of collaborative testing in earthquake
engineering.
Keywords: actuators, earthquakes, input motions, geotechnical engineering, centrifuge modelling
INTRODUCTION
Physical modeling in earthquake engineering with reduced scale experiments in the centrifuge is now
widely considered as the established experimental technique of obtaining data in controlled conditions
to help engineers and researchers to understand the mechanism involved in the response of soil
structure systems to seismic loading. This experimental approach recreates the stress state in soils
which is a fundamental condition to observe realistic soil behaviour. Of course, as any other
experimental method it has its limitations, among which the most evident is the boundary effect due to
the fact that the soil model mounted in the centrifuge is necessarily of limited dimensions. This is
classically treated by using laminar or shear stack box which allow the natural deformation of a soil
column. Another important limit is the ability to impose a realistic input at the base of the soil model.
Earthquakes generate a complex sequence of vibrations that can lead to 3D displacements with a broad
frequency content of very variable duration. Firstly, it is difficult to build a 3D shaking table in the
1
Senior Researcher, Laboratoire Central des Ponts et Chausses, Nantes, France, Email: jean-
louis.chazelas@lcpc.fr
2
Reader in Geotechnical Engineering, Department of Engineering, University of Cambridge,
Cambridge, UK, Email: mspg1@cam.ac.uk
146
limited volume of the basket of a centrifuge. Secondly, it is difficult to generate independent time
histories for each axis of shaking: it is of course simpler to generate harmonic inputs than to reproduce
broad band records of real earthquakes. Note that these two problems apply as well to full scale
shaking tables.
The purpose of this paper is to present recent and upcoming experimental facilities developed in
Europe and discuss the effort that led to an improvement in the quality of the simulations. Establishing
the quality of simulations essentially means to define and justify the choice of 1, 2 or 3D movements,
the ability to simulate single or multi frequency inputs, and the quality of the resulting inputs in
comparison with the target input. There is an on-going discussion amongst centrifuge modellers on the
best type of input motions that may be used in dynamic centrifuge modelling. The input motion that
has simple, sinusoidal tone bursts at different frequency will lend itself to easy analysis of the response
the soil and the superstructure. This is used extensively at Cambridge on a wide range of boundary
value problems in which the key mechanisms of failure are eloquently deciphered. This choice to
some extent is independent of the actuators available with RPI centrifuge facility using sinusoidal
inputs even though their servo-hydraulic shaker is capable of simulating realistic earthquakes.
Similarly, use of a more realistic input motion from a previous earthquake such as Kobe motion or
Northridge motion would be considered useful from the design of future structures point of view.
Further, the role of multi-frequency input motion on the dynamic behaviour of soils is not fully
understood. It is generally argued that for soil liquefaction problems use of simple input motions is
sufficient. Recently, finite element analyses were carried out by Ghosh and Madabhushi (2003) and
dynamic centrifuge modelling was carried out by Madabhushi, Ghosh and Kutter (2006) to investigate
the role of type of input motions in the generation of excess pore pressures. These investigations
revealed that the amount of excess pore pressure generated in loose, saturated sands may be not be
effected by sinusoidal input motions or more realistic input motions. However, the amount of lateral
spreading of sloping ground that can occur may be quite different if sinusoidal motions are used as the
dilation spikes that occur during strong shaking cycles are more pronounced compared to a more
realistic input motion with only a few strong cycles of shaking.
First attempts of centrifuge shaking tables were mechanical 1D harmonic devices based on leaf spring
device (Morris, 1979), bumpy road tracks (Kutter, 1982) or cams systems (Suzuki et al., 1991,
Kimura et al, 1998, ). Other technologies have been tested such as explosives (Zelikson et al., 1981),
piezo-electric jacks (Arulanandan, 1982), electromagnetic motors (Fujii, 1994) - but the majority of
the existing devices are now electro-hydraulic (Ketcham et al., 1988, Van Laak et al. 1994) because of
the ability of electric servo-valves to accept complex driving functions and the command hydraulic
jacks with a rich frequency content. Few 2 D devices have been developed either with two horizontal
shaking directions or one horizontal and one vertical. The difficulty of avoiding uncontrolled
frequency contents especially harmonics due to mechanical guidance and clearance in sine inputs
and spurious movements especially yawing and rocking is largely increased from 1D to 2D. Note
that these considerations apply as well to full scale shaking tables but with specific aspects due to the
fact that the device is embarked in the basket of a rotating machine.
Among main issues in a shaker design are how to shake the biggest mass i.e. the largest model and
how to control that the acceleration at the basis of the soil box with reduced spurious movements and
reduced transmissions of vibration to the centrifuge. Till recently, the basic option was to use the
basket of the centrifuge as the reaction mass for the shaking of the table. The heaviest payload was
then dependent on the basket weight and the level of vibration tolerated by the centrifuge. The
question of the spurious movements arises first from problems of symmetry in the case of a simple
actuator, and from the coordination in the case of multiple actuator. The type of bearings is also of
main importance in this design with two opposite schools; rigid guidance on rails and no guidance at
all on oil bearings.
C-Core Laboratory in Newfoundland, Canada, and Laboratoire Central des Ponts et Chausses,
France, have purchased a new concept 1D earthquake simulator (EQS) designed by ACTIDYN that
overcome these two basic difficulties with a highly sophisticated set of technological solutions
147
(Chazelas et al, 2006). On the other hand, Cambridge University has long operated 1D mechanical
shakers (ref of the bumpy road, Madabhushi et al, 1998), and is now developing a new 2D simulator.
We propose here to present the different aspects of the performances achieved at LCPC's facility, to
present the priorities assigned by Cambridge to its new design and then to open a debate on the interest
of achieving such quality of input and control.
LABORATORY SPECIFICATIONS OF A HIGH QUALITY 1D SHAKING TABLE

AT C-CORE AND LCPC
The specifications imposed to providers by C-Core and LCPC were very similar; they result from a
bibliographical review of existing device giving an insight on what was possible and what have been
the recent evolutions. Both Laboratories chose a 1D shaker, as very few 2D devices are now
operational as there is still much work to realize in 1D modelling. The direction of the shaking,
conventionally noted Y, was naturally fixed horizontal regarding the model in flight and parallel to the
axis of rotation of the centrifuge regarding the fix natural repair in order to limit the Coriolis forces.
The maximum g level of operation was fixed between 80 and 100 g as many models are achieved at
40, 50 or 80 g in the domain of foundations. Over 100 g, the models become very small and difficult
to be instrumented. The most important specifications were the level of the horizontal acceleration a
minimum of 0,4 g prototype scale this figure commonly admitted since Kobe 1995 earthquake. The
control on the acceleration an not on the displacement was considered as compulsory in relation to
the fact that the design of a reduced scale experiments in the centrifuge is controlled by the level of the
centrifuge acceleration. C-Core called for proposal in 2001 and finally negotiated with Actidyn the
performances recalled in table I. LCPC followed two year later and just increased certain values of the
specifications (see table I). It must be emphasized that these specifications correspond to sine inputs;
this inputs imply the highest power on the jacks shaking the table and the largest volume of hydraulic
storage. These specifications included upper bounds for the spurious moments, expressed as a
maximum 10% ratio between the X and Z acceleration on the table and the Y acceleration in the
direction of the shaking. Note that these accelerations were to be recorded at the extremity of the table,
the most severe position to evaluate these movements.
Table I : Specifications of C-Core and LCPC Earthquake simulators

Specifications C- Core LCPC
Maximum centrifuge acceleration of operation (MAO) 80 g 80 g
Maximum payload 400 kg 400 kg
Maximum horizontal acceleration (Y direction) 0.5 MAO 0.5 MAO
Duration of full power sine shake 1s 1s
Maximum velocity 1m/s 1m/s
Maximum displacement 2.5 mm 5 mm
Bandwidth of operation in sine 20 200 Hz 20 250 Hz
Maximum spurious accelerations in X and Z direction RMS X and RMS Z < RMS X and RMS Z <
recorded at the Y end of the shaking table 10% RMS Y 10% RMS Y
LCPC added a last category of specifications: the ability of the shaker to reproduce earthquakes with
very specific signatures supposed to solicit differently the machine. Four records from reference
earthquake were specified. They are recalled in table II with their respective selection criteria.
148
TECHNOLOGICAL RESPONSE AND PERFORMANCES
In addition to the fulfilment of the above mentioned specifications, the main Actidyn's ambitions were
to demonstrate its ability to isolate the centrifuge from the vibrations of the payload and to limit the
spurious movements applied to the model. The technological choices have been detailed in [Perdriat et
al, 2002] and [Hutin et al, 2004], and will only be described briefly here:
- the main innovation was the concept of a permanent dynamic equilibrium between the payload on
its shaking table and a counterweight embarked together in the basket of the centrifuge. Actidyn was
then able to increase largely the mass of the payload, provided the centrifuge was able to support the
total mass. This last condition was easily fulfilled in both laboratories,
- the second technological bet was to opt for oil bearings two superimposed, one for the
counterweight on the basis of the shaker and one for the payload table on the counterweight (see
fig.1). There is not any longitudinal guidance so as to avoid high frequencies due to micro shocks in
mechanical clearance,
Counterweight
Payload
Oil bearings
Basket Platform
Figure 1. Actityns EQS - Dynamic equilibrium payload counterweights
- another essential design option was to install one jack on each side of the shaking table between the
payload table and the counterweight table. These two jacks and their servo-valves are operated
through a multi-axis control system developed by Data Physics, Hutin et al., 2002.
In the simulation of the real earthquakes, specifications and technological limits of the machine
constraint the results: the control system computes the drive function sent to the servo-valves in a
bandwidth of 20 to 400 Hz. Actidyn considered that the jacks would probably have a flat response till
200 or 250 Hz and then fixed the bandwidth controlled by the Data Physic software controller to 20
400 Hz. In the domain of broad band signals, the records were first filtered in the 20 - 350 Hz
bandwidth because the power needed is much reduced as regard to sine tests. This means first that the
prototype record has to be filtered in 0,5 - 8.75 Hz for 1/40th tests or 0.25 - 4.75 Hz for 1/80th test, for
example before time scaling. It must then be pointed that the record from Kobe cannot be correctly
reproduced lower than 1/55 th because its Fourier spectrum still contain much energy at 0,4 Hz. At
higher reduction scales, the bandwidth of the system will accept such low frequency earthquakes.
The physical limits of the machines in terms of maximum accelerations, velocities and displacements
also impose constraints : the maximum horizontal acceleration is limited to 0.5 the centrifuge
acceleration, the velocity is limited to 1 m/s and the maximum displacement is limited to 5 mm. It is
necessary to control, by a double integration process, that none of these limits will be overcome by the
reference input. This leads to eventually apply a reduction factor on the filtered record as shown in the
last column of table II.
149
With the combination of the specifications and the physical limits of the machine, any earthquake
cannot be simulated without care but table II shows that a wide rage of characteristic earthquakes was
theoretically possible to run. Figures 2 and 3 show the quality of the fitting of the model earthquake to
the reference record for two of these references. All these figures are expressed in the prototype 1g
scale. The payload was in a first time a rigid concrete block of mass 400 kg. In the case of Mexico, the
high frequency content corresponds to non controlled high frequencies (around 360 Hz) at the time of
these tests. Since then, with a fine calibration of the accelerometers in the control loop, these spurious
frequencies have been largely reduced. This is proved by the comparison of the Landers reference to
records in the tree directions at the extremity of the shaking table supporting the payload (figure 5).
This test was run at 40 g and the payload was a shear stack box with sand, for a total mass of 350 kg.
Table II: Real earthquake records selected as references in order to evaluate the performance of
LCPCs earthquake simulator
Site Selection criteria Span Reduction ratio after
filtering
Landers- Lucerne Valley Short Strong amplitude in low 48 s
Station Component N09E - frequencies and important velocity 0 dB
28/06/1992 spike.
Kobe DAI8-G - Component Long span and high amplitudes of 120 s - 3 dB at 50 g
N43W - 17/01/95 acceleration - 4 dB at 80 g
Mexico - Sec. Com. Y Long span Rich spectrum in low 180 s
Transport Station frequencies 0 dB
Component 090 19/09/85
Northridge Tarzana Station Very impulsive. Very high 60 s - 8,5 dB at 50 g
Component 90 17/01/94 acceleration peaks - 5.6 dB at 80 g
0.3 0.4
Acceleration - g
Acceleration - g
0.2 0.3
0. 0
0 -0.2
-0.1 -0.4
-0.2 -0.6
32 33 34 35 36 36 36.5 37 37.5 38
time - s time - s
Y table
reference
0.4 0.2
Acceleration - g
Acceleration - g
0.2 0.1
0
0
-0.1
-0.2 -0.2
-0.4 -0.3
39.5 40 40.5 0.41 44 45 46 47 48
time - s time - s
Figure 2. Sequences of the Landers earthquake tested at 50 g centrifuge - 400 kg rigid payload
150
0.16 10-2
2.8
0.12 Mexico 50 g
Y table 2. Y table
0.08 refererence reference
Acceleration - g
Acceleration -
2
0.04
1.
0
1.
- 0.04
0.
- 0.08
- 0.12 0.
- 0.16 0
2 3 3 4 4 5 5 6 6 7 7 0 2 4 6 8
Frequency - hz
time - s
Figure 3. Principal sequence of Mexico earthquake(left) and Fourier spectrum (right) tested at
50 g centrifuge with a 400 kg rigid payload
10-2
1 reference Landers 40 g
X table
Y table
0.875 Z table
0.75
0.625
0.5
0.375
0.25
0 125
0
0 1.25 2 3.75 5 6.25 7.5 8.75 10
frequency - Hz
Figure 4. Fourier Spectrum of Landers earthquake simulation with spurious accelerations
This last figure shows that the system is able to face a certain lack of equilibrium of the dynamic
balance as the counterweights are not of tunable mass.
The above-mentioned tests verified the ability of the device to simulate true 1D broadband
earthquakes with different frequency content. Of course in the domain of sine inputs, known to be
much more demanding for the machine, the main problem was to control the ability to carry out full
power sine inputs under any centrifuge accelerations from 40 to 80 g and during 1 second (model
scale, which correspond respectively to 40 to 80 s at model scale). In brief, during the acceptance tests,
we could carry out very pure sine inputs from 40 Hz up to 100 Hz in this range of centrifuge
accelerations and with the rigid 400 kg payload. Pure, here, meant with no spurious movements
acceleration in the X (yawing) and Z (rocking) direction measured at the extremity of the shaking table
greater than 15 % (we had specified 10%). As indicated earlier, last tests with a fine calibration of
the accelerometers involved in the control loops largely widened this frequency range as we achieved
tests at 40 g centrifuge, 18 g horizontal acceleration (or 0.45 g prototype scale) from 32 to 200 Hz with
spurious moments in the X and Z direction limited to 10%.
Two other types of controls have been achieved during acceptance test: an analysis of the harmonic
content of the sine inputs and a verification of the vibration of the basket and the arms of the
centrifuge. The harmonic content had not been specified. The correction of the harmonic content is
realized by the Data Physic control software through an iterative fitting process applied to a dummy
payload. The most troublesome harmonics to be corrected were generally the third and the fifth and, at
the end of the fitting process, appeared to be limited to less than 10 % of the main frequency
151
component in the band 40 100 Hz. This should be improved with the recent corrections in the
control loops. Of course, when the sine input will be over 130 Hz, correcting the third harmonic will
be impossible due to the 400 Hz limit of the controller.
The control of the vibration of the basket is not yet completely checked: it has been only conducted at
40gs due to aging of the centrifuge itself. The isolation was determined at different frequencies with
sinusoidal motion. It was calculated as the ratio of the acceleration measured on the shaking table to
the one measured at the basket floor or on the arms. It is to be compared to the inertial isolation
ratio, the ratio of the basket and simulator mass to the payload mass. The inertial isolation ratio is 6.5
while from 40 to 100 Hz the real isolation is from 30 to 60, that is to say an improvement of a factor 5
to 9. At lower frequencies the isolation improvement is reduced to 3 and at higher frequencies, it
vanished progressively but these tests should be renewed next with the corrections introduced in the
control loops.
Globally speaking, the earthquake simulator at LCPC produces earthquakes with a very good fitting to
the reference signal, as well sinusoidal motion of up to 0.5 g (prototype scale) as a wide range of broad
band real records at 40 and 50 g centrifuge in the bandwidth 40 200 Hz for sinusoidal motion and 0
350 Hz for broad band record. At higher level of centrifuge acceleration up to 80 gc - the
bandwidth of acceptable response has been tentatively controlled to be narrower up to 100 Hz - but
should be once more controlled with recent improvements. Good fitting means limited spurious
movements yawing and rocking accelerations at the extremity of the table less that 10 to 15% - and
superior harmonics for sine inputs limited to 10% in amplitude.
EARTHQUAKE GEOTECHNICAL ENGINEERING RESEARCH AT CAMBRIDGE

UNIVERSITY
Current Facilities:
The success of earthquake geotechnical engineering at Cambridge depended to a large extent on the
simple mechanical actuators that have been used for more than 30 years. The current earthquake
actuator that relies on Stored Angular Momentum (SAM) to deliver powerful earthquakes at high
gravities was developed and is in operation for 12 years, Madabhushi et al (1998). In Fig.5 the front
view of the SAM actuator while in Fig.6 a view of the SAM actuator loaded onto the end of the 10m
diameter Turner Beam centrifuge is presented. The model seen in Fig.6 was from an investigation
carried out by Haigh and Madabhushi (2002), on lateral spreading of liquefied ground past square and
circular piles.
Figure 5. A view of SAM earthquake actuator at Cambridge
152
The SAM earthquake actuator is a mechanical device which stores the large amount of energy required
for the model earthquake event in a set of flywheels. At the desired moment this energy is transferred
to the soil model via a reciprocating rod and a fast acting clutch. When the clutch is closed through a
high pressure system to start the earthquake, the clutch grabs the reciprocating rod and shakes with an
amplitude of 2.5 mm. This is transferred to the soil model via a bell crank mechanism. The levering
distance can be adjusted to vary the strength of the earthquake. The duration of the earthquake can be
changed by determining the duration for which the clutch stays on. Earthquakes at different frequency
tone bursts can be obtained by selecting the angular frequency of the flywheels.
Recent modifications to the SAM actuator were carried out to further enhance its capabilities and to
improve the performance envelope. Early earthquakes using this device were non-symmetric as the
clutch migrated downwards to an end stop once the centrifugal acceleration was applied. This meant
that at the start of the earthquake the clutch body was hitting the end stop if it grabbed the
reciprocating rod during its downward motion. This problem has been rectified by incorporating a
pneumatic actuator that centralises the clutch prior to every earthquake. Logic controls automatically
turn the air to the pneumatic actuator off once the earthquake is fired and the clutch starts to move
with the reciprocating rod.
In its original conception the SAM actuator was mounted onto the end of the beam centrifuge and
shook a package on the special swing, reminiscent of the Bumpy Road actuator, Kutter (1982).
However this arrangement was modified and a self-contained swing platform was developed that
could house the SAM actuator as shown in Fig. 5 following a research grant (No:GR/L90415/01) from
EPSRC, UK. This has transformed the usage of the SAM actuator and since 1994 nearly PhD students
utilised this facility and several industrial, EPSRC and EU projects were successfully completed using
this actuator.
Figure 6. A view of SAM earthquake actuator loaded onto the end of the beam centrifuge
(Model seen is that of lateral spreading of soil past square and circular piles,
Haigh and Madabhushi (2002))
The technical specifications of the SAM actuator are listed in Table III. In its normal operational mode
the SAM actuator is used to deliver strong shaking to model packages either at particular but different
tone bursts that could recreate damaging cycles felt by the structure during an earthquake loading or to
apply a swept sine wave motion to detect the frequency response of the soil-structure system.
153
Table III: Specifications of the SAM actuator
Parameter Value
Maximum g-level of operation 100 g
Dimension of the soil models 56 m (L) 25 m (B) 22 m (H)
80 m (L) 25 m (B) 40 m (H)
Earthquake strength of choice Upto 0.4g of bed rock acceleration
Earthquake duration of choice From 0 s to 150 s
Earthquake frequency of choice From 0.5 Hz to 5 Hz
Swept sine wave capability
Note: All parameters above are in prototype scale
As mentioned earlier the SAM actuator was used in the investigation of several boundary value
problems. As an example of the input motions generated by the SAM actuator the following
investigation of liquefaction induced lateral spreading problem is presented. The dynamic behaviour
of the slope was studied using miniature instrumentation for the measurement of pore-pressures and
accelerations throughout the slope. Analysis of these signals has revealed interesting details about the
response of these slopes to earthquake loading. The accelerometer time-histories in Fig.7 show the
measured base (ACC 9082), mid layer acceleration (ACC 8076) and surface accelerations (ACC 8025)
in one of the models. It can be seen that whilst the base motion is approximately constant from cycle
to cycle, the surface response late in the earthquake shows alternate cycles having profoundly different
behaviour.
Figure 7. Acceleration time-histories
This shows itself as an amplified frequency component at half of the fundamental earthquake
frequency upon study of FFTs. Measurement of the phase lag of acceleration between base and
surface of the models, as could be achieved from the time-histories shown in Fig.8, allows estimates of
the shear wave velocity to be made at different times during the earthquake. From this data it can be
shown that as the soil liquefies and softens, the shear wave velocity falls to such a point that the
natural frequency of the soil column becomes approximately 25 Hz, half that of the earthquake
excitation. It is thus postulated that the soil column is resonating at this natural frequency, hence
giving the behaviour described above.
154
Figure 8. Upward propagation of S wave
It is also interesting to note the dilative response of the soil slope. All of the PPTs present in the
model show significant dilative behaviour occurring with the generation of a suction-spike once per
cycle. Examining the timing of these spikes with respect to depth illustrates that this suction pulse
propagates vertically from the base of the model to the surface at the shear wave velocity. This
behaviour is illustrated in Fig.9. It is postulated that this is due to the dynamic shear stress applied by
the wave, superimposed on the initial static shear stress causing the soil stress path to cross the
characteristic state threshold and hence the soil to dilate. This pore-pressure behaviour will cause a
slip-stick motion of the soil down the slope, with velocity and displacement being accumulated while
the base is accelerating upslope and then locking up on the other half-cycle when dilation occurs. As
described earlier using a sinusoidal input motion may lead to an under-estimate of the amount of
lateral spreading of sloping ground that can occur as strong cycles of shaking are applied throughout
the model earthquake, making the liquefied soil to dilate in each half cycle and hence stopping the
lateral displacement of the ground. If a more realistic earthquake motion is applied then the ground
liquefies during one or two strong cycles applied and stays liquefied during the smaller cycles that
follow allowing the ground to suffer much larger lateral displacement. This is one example where type
of input motion can have a bearing on the output from the centrifuge test.
Figure 9. Upward propagation of the suction spike
155
Future Developments:
At Cambridge University, preliminary design of a new 2-D earthquake actuator has been recently
completed. This actuator, when fabricated will be able to shake the soil models both horizontally and
vertically akin to the 2-D actuator recently established at UC Davis in the USA. The vertical ground
accelerations can play an important role in the ultimate performance of a structure. Recent earthquakes
have yielded many recordings of vertical accelerations which are quite large (in some cases up to 0.8g
to 1.0g). Current design codes only allow for a fraction of these as vertical accelerations. Also the
combination of vertical shaking followed by strong horizontal shaking can lead to unexpected and
interesting failure mechanisms in a wide range of civil engineering structures. With this in view the
Cambridge 2-D earthquake actuator project has been initiated and is currently at an early stage. A
schematic diagram of the 2-D earthquake actuator assembly is presented in Fig.10 below. A Pro-
Engineer CAD drawing is presented in Fig.11. The design of this 2-D actuator for the Cambridge
centrifuge is quite demanding as the payload capacity of the Turner beam centrifuge is limited to 1
tonne. In addition there are severe space constraints. Further, the Turner beam centrifuge is used
extensively for non-earthquake testing which means that the 2-D earthquake actuator needs to be
loaded and unloaded on and off the centrifuge quite frequently. These bring in additional complexities
such as breaks in high pressure hydraulic lines, contamination of the hydraulic fluid etc. Despite these
difficulties the design of this 2-D actuator is progressing well and with suitable funding should be
available for use in a few years time. This would become the only 2-D earthquake actuator to serve the
European Community.
The specifications of the 2-D shaker were drawn taking to consideration the special requirements of
the Turner beam centrifuge. Unlike the LCPC shaker the entire centrifuge is used as the reaction mass
as the swing platform on which the 2-D shaker is mounted is locked onto the centrifuge when the
centrifuge is speeded up beyond 10gs. Also the use of the 2-D shaker is expected to complement the
SAM earthquake actuator that can operate at high gravities and deliver powerful sinusoidal
earthquakes. The 2-D shaker will be used at relatively lower g levels but with more realistic
earthquake input motions in horizontal and vertical directions.
Figure 10. Schematic view of the 2-D actuator assembly
156
Table IV: Preliminary design specifications of the 2-D earthquake actuator
Parameter Value
Maximum g-level of operation 50 g ~ 80 g
Dimension of the soil models 56 m (L) 25 m (B) 22 m (H)
80 m (L) 25 m (B) 40 m (H)
Earthquake strength of choice Horizontal direction Up to 0.8g
Vertical direction Up to 0.6g
Earthquake duration of choice From 0 s to 150 s
Earthquake frequency of choice From 0.5 Hz to 5 Hz
Note: All parameters above are in prototype scale
There is a good collaboration between Cambridge group and the geotechnical centrifuge modellers at
UC Davis through the EPSRC funded UK-NEES project and at LCPC, Nantes through the funding
provided by British Council in France. The lessons learnt on the quality of input motions from servo-
hydraulic shakers at Davis and Nantes will be extremely valuable in the development of the 2-D
shaker at Cambridge.
UK NEES Project:
Another exciting development in the field of earthquake engineering research is the NEES project in
the USA that established the concept of distributed testing at geographically distributed sites. This
concept is extremely useful for Europe given the expertise in earthquake engineering in Europe and
the geographical distances between the centres of excellence. Having distributed experimental
facilities that are linked to a dedicated network will enable research workers in the whole of Europe to
not only access the experimental data but to actually have tele-observation and tele-participation
capabilities. The USA-NEES project has been well set up and a similar network in Europe benefits
from the technological advances already achieved in the USA. For example, network protocols for
data sharing and data archiving are already available.
Figure 11. Schematic view of the 2-D actuator
To complement the US-NEES, EPSRC funded a research project to develop a UK-NEES network
among Cambridge, Oxford and Bristol universities. This project is at an early stage and a further
opportunity arose to collaborate with NZ-NEES program in New Zealand. In Fig.12 a snapshot of one
of the meetings is presented which shows the teams from Cambridge, Oxford, Bristol and Auckland
are taking part with the Auckland team making a Power point presentation.
157
DISCUSSION
The importance of quality of input motion cannot be over-emphasised. The discussion on quality of
input motions can focus on three main aspects:
uncontrolled movements limit the capabilities of the facility,
the experimental work can be of limited quality,
given the non linearity of the soil behaviour, the realistic simulation of earthquake input motions is
a requirement.
Introducing vibrations in a rotating machine such a centrifuge is not very healthy for it. This a strong
contraindication in the case of the free top design of the Actidyn centrifuges where avoiding
unbalance is a priority. Even with more rigid designs of the centrifuge, the protection of the rotational
bearings impose to limit drastically the mass ratio between the payload and the basket when it is used
as a reaction mass. This induces a reduction of the size of the models, and then concerns arise with
respect to boundary effects. Any uncontrolled spurious movements increase this constraint, with the
risk of hitting a resonance frequency of the whole device that would bias the experimental observation
or, once more, limit the experimental possibilities. In this regard, one of the very positive aspects of
the Turner beam centrifuge at Cambridge where the swinging platform locks onto the end of the
centrifuge as the centrifuge speed is increased. This leads to the whole of the beam centrifuge acting as
a reaction mass.
Bristol New Oxford
Oxford Cambridge
Figure 12. UK-NEES INSORS Facility
Second, vibration analysis is complex enough to avoid any questions of the type: where does this
frequency comes from? With harmonic inputs, the problem is of limited importance but with broad
band tests it can be troublesome. So much effort is done for the consistency of experiments with scale
effects and scaling laws that it would not be consistent not to pay attention on the quality of the
seismic input especially in terms of harmonics: as pointed earlier, a first condition is that the simulator
is control in acceleration. If the acceleration amplitude of the fist harmonic is 10% of the fundamental
input of the test, the displacement generated by this harmonic is 40 time lower, which introduces an
acceptable error of 2,5%. With higher components, it can no longer be considered as negligible. With
the third and fifth harmonics, the problem is reduced due to the reduction ratio of 9 and 25 between
158
acceleration and displacement. Moreover, facing the non linear behaviour of the soil, adding
harmonics and spurious movements will skew the observations.
The role of experimental works in the scientific process is both to enable the observation of behaviour
mechanisms to confirm or to enrich the numerical models and to provide data for the validation and/or
for the fitting of parameters of these numerical models for design purposes. Both applications need
fully controlled experiments: being sure of the input motion, the fact it is not polluted by spurious
movements and unexpected harmonics is an imperative
The non linearity of the soil response to vibrations is the main justification of this reflection because it
conditions all the scientific method. Arguments are in favour of simplifications of experimental works:
it seems simpler to study the soil and soil-structure responses at discrete frequencies and at different
levels of amplitude to provide information on the vibration response, on the frequency dependence of
parameters and on the non linearity of the response with the amplitude of the input. The centrifuge is
adapted for such a parametric approach but so long as the data are not biased.
On the other hand, this approach does not give any information on the dependence of the soil-structure
response on multi-frequency input motion, the resonance of non-linear systems and the evaluation of
damping in such non-linear systems. The example of dependence of the generation of excess pore
pressure and the amount of lateral spreading obtained discussed earlier emphasises the importance of
input motion.
CONCLUSION
The European resources in experimental facilities present an important evolution: Schofield

Laboratory in Cambridge University has long been the only research centre able to achieve reduced
scale earthquake simulations, limited to 1D harmonic inputs. LCPC offers now a new 1D facility with
enlarged possibilities in terms on harmonic and broad band inputs. The quality of the input on this new
device is particularly well controlled due to technological innovations. Schofield Laboratory is
developing now the second generation of a simulator with 2D inputs. Similar efforts on the quality of
the input will drive its design.
1D and 2 D shaking is demonstrated here not to be opposite but complementary in the global approach
of the non linearity of the soil and soil structure response to vibrations.
REFERENCES
Arulanandan K., Canclini J., Anadarajah A., Similation of earthquake motions in the centrifuge,
Journal of the Geotechnical Engineering Division ASCE, Vol 8, nGT5, mai 1982, pp 730-742
Chazelas J-L, Derkx F., Thorel L., Escoffier S., Rault G., Buttigieg S., Cottineau L-M, Garnier J.,
Physical modelling of earthquakes in the LCPC centrifuge, First European Conference on
Earthquake Engineering and Seismology, Geneva, Switzerland, 3-8 September 2006, Paper
Number: 1064
Fujii N, Development of an electromagnetic centrifuge earthquake simulator, Ko H.Y., McLean F.G.
(eds), Centrifuge 91, Boulder, CO, 13-14 juin 1991, Balkema, Rotterdam, 1991, pp 351-354.
Ghosh, B. and Madabhushi, S.P.G., A numerical investigation into single and multiple frequency
earthquake input motions, Journal of Soil Dynamics and Earthquake Engineering, 2003, Vol. 23,
pp 691-704.
Haigh, S.K. and Madabhushi, S.P.G., Dynamic Earth Pressures on Piles due to Lateral Spreading.
Proc. International Conference on Physical Modelling in Geotechnics, St John's, Newfoundland,
Canada, July, Balkema, Rotterdam, 2002.
Hutin C., Perdriat J., Rames D., Dynamically balanced braod frequency earthquake simulation system,
in Phillips R., Guo P.J., Popescu R. (eds), Physical Modelling in Geotechnics, ICPMG'02, St John's
, NF - Canada, 10-12 july 2002, pp 175 178.
159
Ketcham S.A., Ko H.Y., Sture S., An electrohydraulic earthquake simulator for centrifuge testing,
Cort J.F. (ed), Centrifuge 88, Proceedings of the international conference on geotechnical
centrifuge modelling, Paris, France, 25-27 avr. 1988, Balkema, Rotterdam, 1988, pp 97-102.
Kutter, B. L., Centrifugal modelling of the response of clay embankments to earthquakes, PhD thesis,
Cambridge University, UK, 1982.
Kimura T, Takemura J., Saitoh K, Developpment of a simple mechanical shaker using cam shaft,
Cort J.F. (ed), Centrifuge 88, Proceedings of the international conference on geotechnical
centrifuge modelling, Paris, France, 25-27 avr. 1988, Balkema, Rotterdam, 1988, pp 107-110.
Madabhushi S.P.G., Schofield A.N., Lesley S., A new strored angular momentum (SAM) based
earthquake actuator, Kimura, Kusakabe & Takemura (eds), Centrifuge 98, Tokyo, Japon, 23-25
sept.98, Balkema, Rotterdam, 1998, pp 111-116.
Madabhushi S.P.G., Ghosh B., Kutter B. L., Role of input motion in excess pore pressure generation
in dynamic centrifuge modelling, Int. J. of Phys. Modelling in Goetech. 3, 2006, pp 25-34.
Morris, D.V., The centrifuge modelling of dynamic behaviour, PhD Thesis, Cambridge University,
UK, 1979.
Perdriat J., Phillips R., J. Nicolas Font J., C. Hutin, Dynamically balanced broad frequency earthquake
simulation system, in Phillips R., Guo P.J., Popescu R. (eds), Physical Modelling in Geotechnics,
ICPMG'02, St John's , NF - Canada, 10-12 july 2002, pp 169 174.
Suzuki K., Babasaki R., Suzuki Y., Centrifuge tests on liquefaction-proof foundation, Ko H.Y.,
McLean F.G. (eds), Centrifuge 91, Boulder, CO, 13-14 juin 1991, Balkema, Rotterdam, 1991, pp
409 415.
Van Laak P.A. , Elgamal A.W.,Dobry R., Design and performance of an electrohydraulic shaker for
the RPI centrifuge, Leung CF, F.H. Lee, Tan T.S. (eds), Centrifuge 94, Singapore, 31-08/2-09-94,
Balkema, Rotterdam, 1994, pp 139-144.
Zelikson, A., Devaure, B.,Badel, D., Scale Modeling of soils structure interaction during earthquakes
using a progammed series of explosions during centrifugation, Proceedings of the international
conference on recent advances in geotechnical earthquake engineering and soil dynamics , 1981,
pp 361 366.
160
June 25-28, 2007
Paper No. W1-1004
STUDY OF PILE RESPONSE TO LATERAL SPREADING USING

PHYSICAL TESTING AND COMPUTATIONAL MODELING
Ricardo DOBRY1, Sabanayagam THEVANAYAGAM2, Tarek ABDOUN3, Ahmed ELGAMAL4,

Usama EL SHAMY5, Mourad ZEGHAL6, and Claudia MEDINA 7
ABSTRACT
A progress report is presented on a 3-year, NEES research aimed at detailed clarification and
quantification of the mechanics of soil-pile interaction during lateral spreading, with emphasis on the
action of the liquefied soil on pile foundation. The project involves cooperative research between five
US universities. The centerpiece is 1g base shaking of 6-m tall geotechnical models using the laminar
box recently developed at the U. at Buffalo. Results are presented of a free field experiment of
liquefaction of a level loose sand deposit. Additional 1g lateral spreading experiments obtained by
inclining the laminar box are scheduled for 2007. They include one baseline free-field lateral
spreading test, and two more tests involving a stiff pile and a flexible pile, respectively. The 1g testing
involves use of state-of-the-art instrumentation techniques to measure soil and pile responses,
including a new advanced sensor array of MEMS accelerometers. The instrumentation aims at
recording the dynamic response as well as permanent changes in the free field, along the piles, and,
especially, in the near field soil around the pile where some of the most important and challenging
phenomena take place. Three other complementary efforts simulate these 1g full scale experiments
using: (i) small scale centrifuge model experiments at Rensselaer Polytechnic Institute (RPI); (ii)
Finite Element Modeling (FEM) computer simulations at the U. of California, San Diego; and (iii)
Discrete Element Modeling (DEM) computer simulations at RPI and Tulane U. System Identification
and Visualization techniques are systematically used to process the data. Compatible data bases from
the 1g and centrifuge experiments, as well as from the FEM and DEM simulations are fed to a newly
developed 3D Data Viewer for comparison and integration of the various results.
Keywords: Liquefaction, lateral spreading, piles, large scale testing, centrifuge testing, DEM
simulations, FEM simulations
1
Professor, Department of Civil and Environmental Engineering, Rensselaer Polytechnic Institute,
Troy, NY, E-mail: dobryr@rpi.edu
2
Associate Professor, Department of Civil, Structural and Environmental Engineering, University at
Buffalo, State University of New York, NY, E-mail: theva@eng.buffalo.edu
3
Associate Professor, Department of Civil and Environmental Engineering, Rensselaer Polytechnic
Institute, Troy, NY, E-mail: abdout@rpi.edu
4
Professor and Chair, Department of Structural Engineering, University of California, San
Diego, CA, E-mail: elgamal@ucsd.edu
5
Assistant Professor, Department of Civil and Environmental Engineering, Tulane University, New
Orleans, LA, E-mail: uelshamy@tulane.edu
6
Associate Professor, Department of Civil and Environmental Engineering, Rensselaer Polytechnic
Institute, Troy, NY, E-mail: zeghal@rpi.edu
7
Postdoctoral Research Associate, Department of Civil and Environmental Engineering, Rensselaer
Polytechnic Institute, Troy, NY, E-mail: medinc2@rpi.edu
161
INTRODUCTION
Liquefaction-induced lateral spreading of sloping ground and near waterfronts is a major cause of
earthquake damage to deep foundations. Earthquake case histories in the US, Japan, Mexico, Costa
Rica, Turkey and other countries, have shown damage to buildings, bridges, port facilities and other
pile-supported structures. Effects include cracking and rupture of piles at both shallow and deep
elevations, rupture of pile connections, and permanent lateral and vertical movements and rotations of
pile heads with corresponding effects on the superstructure (McCulloch and Bonilla, 1970; Hamada et
al., 1986; Mizuno, 1987; Hamada and ORourke, 1992; ORourke and Hamada, 1992; Youd, 1993;
Swan et al., 1996; Ishihara et al., 1996; Tokimatsu et al., 1996; Yokoyama et al., 1997; Tokimatsu,
1999; Dobry and Abdoun, 2001; Koyomada et al., 2005; Lin et al., 2005). Extensive experimental
research on the effects of lateral spreading on deep foundations in the last decade have included: full
scale soil-pile field experiments to blast-induced lateral spreading (Ashford et al., 2006); full scale as
well as model 1g tests in shaking tables (Suzuki et al., 2005; Tamura and Tokimatsu, 2005; Towhata et
al., 2006; He et al., 2006); and centrifuge model testing (Abdoun et al., 2003; Haig and Madabhushi,
2005; Brandenberg et al., 2005, 2007). A number of these centrifuge and 1g shaking tests of lateral
spreading are conducted in inclined laminar boxes attempting to approximate the spreading of an
infinite slope (Fig. 1).
Case histories as well as experiments suggest that the effect of lateral spreading on piles has the
character of a pseudostatic, kinematic soil-structure interaction phenomenon, driven by the permanent
lateral movement of the ground in the free field. Bhattacharya et al. (2004) have suggested that another
mechanism contributing to the observed failures is pile buckling under gravitational loads, due to
decreased lateral support provided by the surrounding liquefied soil. Various foundation analysis and
design methods have been proposed to account for the lateral loading effect, where the soil applies
static lateral forces to the pile foundation, either (i) as a function of the relative displacement between
the foundation and the free field (Beam-on-Winkler-Springs or p-y approach); or (ii) taking the
maximum possible values of these lateral soil static forces which depend on the soil strength within an
overall limit equilibrium method. When the lateral loads and resulting pile bending moments are
controlled by a strong shallow nonliquefied soil layer, the ground pressures can be reasonably related
to the passive failure of the soil against the foundation, and hence to the shear strength of the soil
(Berrill et al., 1997; Dobry et al., 2003). However, a much larger uncertainty appears when the lateral
pressures applied by the liquefied soil play an important role. One source for this uncertainty is the
vertical area over which these lateral pressures are applied in pile groups (only the areas of the piles,
or also the areas of soil between the piles?). Another source of uncertainty in this case is the
significance of soil permeability, not included in current design methods but which recent full scale
and centrifuge tests suggest may be extremely important, with less pervious soil such as fine or silty
sand tending to produce higher pile bending moments due to the lack of enough time for the negative
pore pressures near the pile to dissipate in the liquefied but still dilative soil (Dungca et al., 2004;
Hwang et al., 2004; Gonzalez et al., 2005; Suzuki et al., 2005).
This paper constitutes a progress report on a 3-year study (2005-08), aimed at detailed clarification
and quantification of the mechanics of soil-pile interaction during lateral spreading, with emphasis on
the action of the liquefied soil on the pile foundation. The effort is supported by the US National
Science Foundation through the NEES Consortium, NEES Inc. (NEES: George E. Brown, Jr. Network
for Earthquake Engineering Simulation). It involves cooperative research between five US universities
(University at Buffalo SUNY (UB), Rensselaer Polytechnic Institute (RPI), University of California at
San Diego (UCSD), and Tulane University (TU)), as well as Japanese collaborators at Tokyo Institute
of Technology (TIT) and the National Research Institute for Earth Science and Disaster Prevention
(NIED). The centerpiece of the effort is 1g full scale testing in the large laminar box recently
developed at UB, as sketched in Fig. 1. The figure shows two experiments of a stiff pile and a flexible
pile, respectively, subjected to lateral spreading induced by shaking at the base of the inclined laminar
box containing loose saturated sand. These two experiments, as well as a third baseline free-field
experiment of the inclined laminar box without a pile, are scheduled for mid 2007. Two successful
liquefaction free field experiments have already been conducted using this laminar box in a horizontal
162
position, simulating liquefaction of level sand deposits, and results are presented at the end of the
paper for one of them (Test LG-0). The video of Test LG-0 can be viewed at
https://central.nees.org/?projid=122&loc=Public#. Future experiments will be web cast live at
http://www.nees.buffalo.edu/projects/NEESPiles/.
This 1g testing involves the use of state-of-the-art instrumentation techniques, including an advanced,
dense sensor array of MEMS accelerometers, as well as more traditional piezometers, potentiometers,
and digital image analyzers used to monitor the time-history of accelerations, excess water pressures,
and vertical and lateral spreading within the ground and along the piles, with additional strain gages on
the piles monitoring bending moments. These sensors aim at recording dynamic fluctuations as well as
permanent changes in the free field, along the piles, and, especially, in the near field soil around the
pile where some of the most important and challenging phenomena take place. A 3D Data Viewer
originally developed at RPI for centrifuge testing has been adapted to display these responses on a
computer screen, allowing the researchers to visually capture the complex response of the soil and pile
foundation before and during liquefaction and lateral spreading.
Three other parallel efforts are underway to simulate these 1g full scale experiments at UB using: (i)
small scale centrifuge model experiments at RPI; (ii) Finite Element Modeling (FEM) computer
simulations at UCSD; and (iii) Discrete Element Modeling (DEM) computer simulations at RPI and
TU. Comparison between centrifuge and 1g tests should allow clarification of the applicable scaling
law(s); the DEM simulations can be considered to be numerical experiments providing additional
physical insights; and the FEM simulations should both provide additional insight and serve as a
calibrated platform for analysis and design of actual systems in engineering practice. Compatible data
bases from the 1g full scale and centrifuge experiments, as well as from the FEM and DEM
simulations are fed to the 3D Data Viewer, which is then used by the research team to compare and
integrate the various results.
Figure 1. Full scale testing of lateral spreading and its effect on pile foundations, scheduled in
2007 using the new geotechnical laminar box at the University at Buffalo. The laminar box sits
on rubber/sliding bearings on the strong floor and is shaken at the base by two fast actuators
connected to a reaction wall
RESEARCH TOOLS
The research combines two advanced NEES testing facilities at UB and RPI, and advanced
computational simulation facilities at UCSD, RPI and TU. Full details are available at three NEES
websites (www.nees.rpi.edu, www.nees.buffalo.edu, http://nees.ucsd.edu). Summaries of the tools
used in the research are presented below.
163
1g Full-Scale Prototype Testing at UB
1-g Full-scale prototype testing at the UB-NEES experimental site utilizes the new 2D modular
laminar box sketched in Fig. 1 (2.75 5 6.2m), two newly acquired 100 ton fast actuators, reaction
wall (9.1 m high, 12.5 m wide), strong-floor (close to 300 m2), image processor, and controllers. This
facility was developed in 2000-04 with NEES support. The tests utilize state-of-the-art sensors (pore
pressure, acceleration, LVDT, laser, fiber optic shape cables), CPT site characterization, and image
analysis tools acquired as part of the NEES development. The 2D laminar box is made of 24
laminates, separated by ball bearings to facilitate 2D motions, and has the ability to simulate sloping
ground subjected to large deformations. The laminar box is supported by the strong floor which rests
on a steel shaking base frame having high capacity ball bearings and shaken by two 100 ton fast
actuators mounted on the new UB-NEES reaction wall (Figs. 2a and 2b). The total weight of the box
filled with sand is about 150-170 ton, whereas the maximum horizontal dynamic actuator capacity is
90 ton in each horizontal direction simultaneously, or 180 ton in any one direction. Thus very large
shaking g levels are possible subjected to safety limitations. The actuators can be fed with an arbitrary
target motion, and the controllers compensate for compliance effects so as to accurately shake the base
of the soil to meet the desired target motion. Sand is deposited by hydraulic filling (Fig. 2c). Data
acquisition systems are available to monitor up to 256 channels at high frequencies. High resolution
imaging tools can be positioned to capture deformation patterns at selected zones in the soil box.
Interpretation software is available for accurately capturing deformation/strain characteristics at any
location. Figure 3a shows the laminar box filled with sand and fully instrumented, during a cone
penetration test (CPT) done before the shaking. Figure 3b shows front, side, and top views of the
laminar box during the shaking.
(a) (b) (c)

Figure 2. 1g full-scale geotechnical laminar box: (a) Base shaker of the geotechnical laminar
box; (b) Laminar ring stack; and (c) Hydraulic filling of soil and instrumentation
(a) (b)
Figure 3. 1g full-scale geotechnical laminar box liquefaction experiment:
(a) UB researchers setting up instruments and CPT testing before shaking; and
(b) Video of UB liquefaction experiment (https://central.nees.org/?projid=122&loc=Public#)
164
Centrifuge Model Testing at RPI
The RPI geotechnical centrifuge (Fig. 4) was commissioned in 1989 and started producing 1D
earthquake model simulation tests in 1991. Over the years it has produced about 600 documented
model experiments of soil and soil-structure systems, with most but not all being earthquake shaking
tests. In 2000-04, the centrifuge was upgraded and integrated into the NEES grid with NSF support.
This included an increase in capacity from 100 g-ton to 150 g-ton; a new 2D in-flight earthquake
shaker (Fig 5a), and 2D laminar box for shaking in the two prototype horizontal directions (Fig 5b);
and a 4-degree-of-freedom robot capable of performing in-flight operations such as construction and
excavation, pile driving, ground remediation, cone penetration (CPT), and model foundation loading
tests without stopping the centrifuge. Also, a model may now be instrumented with dense arrays of
advanced sensors (shape-acceleration-arrays and tactile pressure sensors) connected to RPIs new
web-based data acquisition system. Currently, the 3m-radius, 150 g-ton centrifuge facility has two 1D
shakers capable of providing in-flight base excitation to several available rigid and laminar (flexible
wall) model box containers; the new 2D shaker and 2D laminar box; split box model containers for
simulating the effect of large ground displacement on buried pipes; the new 4-degree-of-freedom
robot; LVDT, accelerometer, pore pressure, and earth pressure sensors as well as load cells and strain
gages, other advanced sensors and a high speed camera and image processing software to monitor
model response; data acquisition system with 128+ channels and Internet capability; client-server data
acquisition/control software for web-based observation and operation of centrifuge experiments; local
area network connected to the high speed RPI network connection making it possible to share
experimental data and other information in real time between the RPI research staff and outside
parties; a control/teleparticipation and a teleconference room; and model preparation, robot and
electronic rooms and other equipment and peripheral facilities.
(a) (b)
Figure 4. (a) RPI 150g-ton geotechnical centrifuge; and (b) Tele-control room
(a) (b)
Figure 5. RPI geotechnical centrifuge: (a) 2D in flight servohydraulic shaker (two prototype
horizontal directions); and
(b) 2D laminar container
165
Advanced Sensors
A main objective of this investigation is to monitor as closely as possible the 1g and centrifuge
shaking tests conducted at UB and RPI, by producing more and better data using advanced
instrumentation, thus providing a higher resolution picture (in space and time) of the model response
during and after shaking. Test data is collected using suites of traditional state-of-the-art sensors (pore
pressure, accelerometers, strain gages, etc.). In addition, advanced sensing techniques as described
below are also used for measuring soil and pile response. These new measurements are especially
suited to the needs of visualization, system identification, and comparisons with numerical
simulations. The two new advanced techniques used in the project are:
Shape-acceleration-arrays sensor (1g testing)

A novel field shape-acceleration-array (SAA) sensor has been developed, taking advantage of
promising new advances in the fiber optic and micro-machined electromechanical sensor (MEMS)
technologies. These sensors are capable of simultaneously measuring acceleration and permanent
ground deformation down to 30 m depth, at a cost that is 1/10th or 1/20th of existing inclinometer and
borehole technologies. In fact, the significance of the new wireless SAA sensor, reside in the type of
ground measurement that is recorded. The array is capable of measuring in situ, 3D ground
deformation as well as 2D soil acceleration every 0.25 m interval down to a soil depth of 30-40 m.
Each sensor array is connected to a wireless sensor node to enable real time monitoring as well as
remote sensor configuration (Abdoun and Danisch, 2005; Abdoun et al. 2006, 2007). Figure 6 shows
photos of the sensor array placed in the free field and on the pile foundation before placing the soil in
a 1g test in Japan. These new sensor arrays are being utilized in the 1g tests at UB.
(a) (b)
Figure 6. Shape-Acceleration-Arrays used in a 1g shaking table test in Japan: (a) Sensor array
placed in the free field; and (b) Sensor array attached to a pile foundation. Photos courtesy of
NIED (Abdoun et al., 2007)
High speed camera (centrifuge testing)

RPIs high speed camera is a Phantom 5 capable of recording full digital images at rate of 30,000
frames/sec. The use of high speed camera allows production of a high resolution picture of centrifuge
model response. This includes measurement of the ground deformation (and especially the permanent,
DC component of this deformation) at enough locations and times during and after shaking. The high
speed camera is used in the tests in conjunction with markers at the surface and sides of the model to
provide this information at the boundaries of the soil model.
Visualization tools and System Identification

The sensors used in this project are producing a growing wealth of data which must be integrated and
compared with other experimental and analytical data. System identification (SI), visualization (of
complete response of soil-foundation system), and a new 3D Data Viewer, are three main tools being
used in the project.
166
System Identification
System identification techniques provide tools to deduce a model or model parameters of a real system
based on its output (response) and input (excitation). Currently, these techniques are enjoying
increasing attention in earthquake engineering, in view of substantial enhancements of computational
resources as well as experimental data and earthquake records. For this research, the same simple
(Zeghal et al., 1995) as well as sophisticated (Zeghal and Oskay, 2003) procedures are being employed
to process data from both 1g and centrifuge experiments, so as to allow comparison and ensure
consistency of the results.
Visualization of Experiments
Visualization of Experiments is especially suited to test data measuring the earthquake response of a
system as done in this project. Such visualizations are essentially movies illustrating the actual
systems response to the base earthquake shaking. Visualizations of several centrifuge tests have
already been implemented at RPI and are being used systematically in the investigation. Figure 7a is a
frame taken from one of these "movies" which can be viewed at the RPI-NEES Web site
(www.nees.rpi.edu).
3D Data Viewer
The 3D Data Viewer (Fig. 7b) was developed at RPI initially for the processing and study of
centrifuge results (Radwan and Abdoun, 2006). It is a software especially designed to display both 3D
models as well as 2D plots side by side (http://nees.rpi.edu/3dviewer). This tool can be used for simple
3D visualization, or more specifically to organize data from various sources (sensors) which can be
represented in 3D space. To organize all these data, a virtual representation of the physical model is
displayed and the sensors became interactive elements so the user can select them and plot their data.
This aids greatly in organizing the data and getting a good high-level understanding of the test results.
As mentioned before, this same 3D Data Viewer has been adopted as a main integrative tool of the
whole research project. Compatible data bases are produced for the same 1g UB experiment at the
same sensor locations by the four tools of the project (1g tests, centrifuge tests, DEM simulations, and
FEM simulations), which are then uploaded to the 3D Data Viewer for side-by-side comparisons and
further integration of the results by the team researchers.
(a) (b)
Figure 7. Visualization tools: (a) Visualization of RPI liquefaction Model 3 centrifuge
experiment (see complete test animation at www.nees.rpi.edu); and
(b) 3-D Data Viewer software for visualization of the response of the ground and foundations
during strong shaking (download software from http://nees.rpi.edu/3dviewer)
167
Discrete Element Method (DEM) Micro-Mechanical Numerical Simulations
A coupled hydro-mechanical model is being employed to analyze the liquefaction and lateral
spreading of a saturated deposit of cohesionless particles (Zeghal and El Shamy, 2004), as well as the
soil-pile interaction when subjected to a dynamic base excitation. The fluid motion is idealized with
averaged Navier-Stokes equations, and the discrete element method (DEM) is used to model the
assemblage of solid particles (Fig. 8a), with appropriate fluid-particle interactions. The model is
employed to analyze liquefaction of both level and sloping saturated loose deposits of cohesionless
particles subjected to dynamic base excitation. Periodic boundaries and a high gravity field are used to
reduce the number of particles to a manageable value.
The DEM numerical experiments, validated by the physical tests, are starting to provide for the first
time detailed information at the level of individual soil grains and groups of grains, on the response of
the liquefied soil. It is expected that the upcoming DEM simulations of pile-soil response will be very
valuable in this respect, especially for the soil close to the pile foundation, where complicated
phenomena take place, including aspects such as dilative stress-strain response, rapid transient water
flows, strain localization, shear banding, and flow of the liquefied soil around the pile. New insights
provided by the DEM simulations that are validated by the 1g and/or centrifuge results will be used to
refine important unsolved aspects of the constitutive relations used in the FEM analyses.
Finite Element (FEM) Numerical Simulations

An effective-stress solid-fluid finite element computational framework (Program CYCLIC
http://cyclic.ucsd.edu/) together with an appropriate plasticity-based pressure-sensitive soil model
(Parra, 1996; Elgamal et al., 2003) is used to simulate pile foundations subjected to liquefaction-
induced lateral spreading in sloping ground (Fig. 8b). Centrifuge experimental programs conducted at
RPI have been a major source of calibration and validation for CYCLIC (Dobry et al., 1995; Taboada,
1995; Elgamal et al., 1996; Dobry and Abdoun, 1998; Elgamal et al., 2002a,b). The 1g large shake-
table models conducted at NIED in Tsukuba City, Japan (He et al., 2006), as part of a previous RPI-
UCSD-NIED US-Japan project supported by the National Science Foundation (NSF), were used for
further calibration of CYCLIC to simulate pile foundations. Recent developments funded by NSF in
collaboration with Prof. Kincho Law of Stanford University extend CYCLIC to perform Parallel
Computations (ParCYCLIC), allowing for high fidelity analyses of lateral spreading effects on piles.
RPI centrifuge tests (Abdoun, 1997; Abdoun and Dobry, 2002) have been used so far as a basis of
further calibration for displacement, pore pressure and acceleration.
(a) (b)
Figure 8. Numerical simulations: (a) DEM micromechanical computer model of sloping saturated
sand layer and pile; and (b) FEM mesh of pile in liquefiable soil model (different colors show mesh
partitioning for computation on 16 processors)
168
FULL SCALE LIQUEFACTION TEST AT UB (LG-0)
Following a successful shaking 3m demonstration test at UB, one full-scale laminar box shaking test
was conducted to evaluate the response of a level deposit of loose saturated sand in the free field when
subjected to strong shaking motion (Thevanayagam and Ecemis, 2007). The total depth of soil in the
laminar box in Test LG-0 was 4.9 m. The soil was fully instrumented with a MEMS shape
accelerometer array (SAA), traditional accelerometers, piezometers, and potentiometers.
Potentiometers measure displacement of a contact point, and are used to measure horizontal and
vertical displacement of selected locations on the top of the ground and along the laminar box side
wall. Saturated Ottawa sand with D10 = 0.161 mm and D50 = 0.230 mm was used. The average
permeability for the relevant range of void ratios, e = 0.61-0.66, is k = 110-5 m/s (Shenthan, 2001).
The base input acceleration was essentially sinusoidal with a frequency of 2 Hz, containing 4 phases
of increasing acceleration amplitude: (i) 0.01g (non-destructive phase) with a duration of 5 sec; (ii)
0.05g with a duration of 10 sec; (iii) 0.15g with a duration of 10 sec; and (iv) 0.30g with a duration of
10 sec. Figure 9a shows the accelerations measured during the non-destructive phase and beginning of
the second, 0.05g phase (when the soil liquefied in the first few cycles). Figure 9b presents the shear
stress-strain hysteretic response in the soil evaluated using system identification (Elmekati and Zeghal,
2007) for the non-destructive phase (i.e., first 5 seconds of shaking during which only a small pore
pressure buildup took place). Figure 10 shows the input shaking as well as the excess pore pressures
measured in the soil. A video of this experiment can be seen at
https://central.nees.org/?projid=122&loc=Public#.
(a) (b)
Figure 9. Results from full-scale laminar box 1g shaking Test LG-0: (a) Acceleration time
histories recorded at soil and rings during non-destructive phase (0-5 sec); and (b) 1D shear
stress-strain hysteretic response in the soil during non-destructive phase, obtained using SI
from ring acceleration time histories
The accelerations during Test LG-0 were recorded using three accelerometer systems. Two of these
systems recorded respectively outside the laminar box and inside the soil, using submersible
traditional accelerometers, while the third system consisted of the advanced SAA discussed before in
the paper. The comparison between the records from the three systems in Fig. 9a indicates good
169
agreement in the first 5 seconds, that is, during the non-destructive, 0.01g shaking. Shortly after the
second phase of 0.05g started at 5 seconds, the deposit liquefied and the acceleration records in Fig. 9a
diverged, probably due to rotation of the traditional accelerometers inside the soil following
liquefaction. This phenomenon of sensor rotation has been previously reported in the literature for
centrifuge experiments (Adalier, 1996). Thus, ring and soil accelerometers show different behavior
after liquefaction. The shear stress-strain loops in Fig. 9b, obtained by SI processing of the
acceleration records from the rings in the non-destructive phase, are being used to evaluate the
dynamic properties of the sand deposit. The loops become steeper at greater depths, indicating (as
expected) an increase in shear modulus with depth within the sand deposit. Figure 10 includes
comparisons between FEM predictions and measurements for Test LG-0 (Elgamal and Forcellini,
2007), with reasonable agreement. While the comparison of accelerations in Fig. 10a is restricted to
the first 7 seconds, the pore pressure comparisons in Fig. 10b cover the whole shaking. Both
measurements and FEM predictions indicate that liquefaction was induced by the second, 0.05g phase
of the shaking, but with the experimental data showing a higher rate of pore pressure build-up and an
earlier onset of liquefaction than predicted by the FEM simulation.
(a) (b)
Figure 10. Comparison between experimental recorded data and FEM simulation: (a)
Acceleration time histories during non-destructive phase (0-5 sec time window); and (b) Pore
pressure time histories during whole shaking
Efforts are currently under way in the project to complete the simulation of 1g level test LG-0 using
both DEM computer analysis (El Shamy and Zeghal, 2007), and centrifuge model testing (Abdoun
and Gonzalez, 2007).
CONCLUSIONS
Critical NEES research is under investigation on the problem of pile foundations subjected to
liquefaction-induced lateral spreading, with emphasis on the action of the liquefied soil on the piles.
The complex nature of the problem requires a fundamental approach using several available
experimental and analytical tools, The research includes use of the UB 1g 6m-tall laminar box
container NEES facility, and the RPI NEES centrifuge facility; use of novel advanced sensors to
170
measure soil accelerations, deformations and pressures during shaking; and micromechanical discrete
element modeling (DEM) and finite element modeling (FEM) analyses at RPI, U. of California, San
Diego, and Tulane University; all integrated by an appropriate identification and analysis framework
including system identification and visualization capabilities.
AKNOWLEDGEMENTS
This material is based upon work supported by the National Science Foundation under NEESR-SG
Grant No. 0529995. This support is gratefully acknowledged. The authors wish to thank the staff
members of the UB-NEES Site: Thomas Albrechcinski, Andrei Reinhorn, and Mark Pitman; and of
the RPI-NEES Site: Inthuorn Sasanakul, and Javier Ubilla, for their valuable support in the 1g and
centrifuge testing at these two NEES sites. The authors wish to thank NEES IT staff members Hassan
Radwan and Jason Hanley for their contribution to the visualization tools. Special thanks are extended
to graduate students Raghudeep Bethapudi, Victoria Bennett, Tewodros Dessalegn, Nurhan Ecemis,
Ahmed Elmekati, Davide Forcellini, Marcelo Gonzalez, Peng Hao, and Thangalingam Kanagalingam
for their dedication, creativity, and enthusiasm in conducting this research. Finally, previous 1g testing
shaking experiences in Japan and ideas for this research provided by Drs. Akio Abe, Masayoshi Sato,
and Kohji Tokimatsu, are also most gratefully acknowledged.
REFERENCES
Abdoun, T., Modeling of Seismically Induced Lateral Spreading of Multi-layered Soil and its Effect
on Pile Foundations, Ph.D. Thesis, Rensselaer Polytechnic Institute, Troy, NY, 1997.
Abdoun, T., and Danisch, L., Advanced Sensing for Real-Time Monitoring of Geotechnical
Systems, Proc Geo-Frontiers 2005 Conference, Austin, TX, January 24-26, 2005.
Abdoun, T., and Dobry, R., Evaluation of Pile Foundation Response to Lateral Spreading, Soil Dyn
Earthq Eng, Vol. 22, No. 9, pp. 1051-1058, 2002.
Abdoun, T., and Gonzalez, M., Centrifuge Model Test on Pile Foundation Subjected to Liquefied-
Induced Lateral Spreading: Calibration Factors, Internal Progress Report, Rensselaer Polytechnic
Institute, Troy, NY, 2007.
Abdoun, T., Dobry, R., ORourke, T.D., and Goh, S.H., Pile Response to Lateral Spreads: Centrifuge
Modeling, J Geotech Geoenviron Eng, Vol. 129, No. 10, pp. 869-878, 2003.
Abdoun, T., Danisch, L., Ha, D., and Bennett, V., Advanced Sensing for Real-Time Monitoring of
Geotechnical Systems, TRB 85th Annual Meeting, Washington, D.C., January 22-26, 2006.
Abdoun, T., Abe, A., Bennett, V., Danisch, L., Sato, M., Tokimatsu, K., and Ubilla, J., Wireless Real
Time Monitoring of Soil and Soil-Structure Systems, GSP Proc GeoDenver 2007 Conference,
Denver, CO, February 18-21, 2007.
Adalier, K., Mitigation of Earthquake Induced Liquefaction Hazards, Ph.D. Thesis, Rensselaer
Polytechnic Institute, Troy, NY, 1996.
Ashford, S.A., Juirnarongrit, T., Sugano, T., and Hamada, M., Soil-Pile Response to Blast-Induced
Lateral Spreading. I: Field Test, J Geotech Geoenviron Eng, Vol. 132, No. 2, pp. 152-162, 2006.
Berrill, J.B., Christensen, S.A., Keenan, R.J., Okada, W., and Pettinga, J.K., Lateral-Spreading Loads
on a Piled Bridge Foundation, Seismic Behavior of Ground and Geotechnical Structures, (Seco E.
Pinto, ed.), Balkema, Rotterdam, pp. 176-183, 1997.
Bhattacharya, S., Madabhushi, S.P.G., and Bolton, M.D., An Alternative Mechanism of Pile Failure
Deposits During Earthquakes, Geotechnique, Vol. 54, No. 3, pp. 203-213, 2004.
Brandenberg, S.J., Boulanger, R.W., Kutter, B.L., and Chang, D., Behavior of Pile Foundations in
Laterally Spreading Ground During Centrifuge Tests, J Geotech Geoenviron Eng, Vol. 131, No.
11, pp. 1378-1391, 2005.
Brandenberg, S.J., Boulanger, R.W., Kutter, B.L., and Chang, D., Liquefaction-Induced Softening of
Load Transfer Between Pile Groups and Laterally Spreading Crusts, J Geotech Geoenviron Eng,
Vol. 133, No. 1, pp. 91-103, 2007.
171
Dobry, R., and Abdoun, T., Post-triggering Response of Liquefied Sand in the Free Field and Near
Foundations, Proc ASCE GEESD, pp. 270-300, Seattle, WA, August 3-6, 1998.
Dobry, R., and Abdoun, T., Recent Studies on Seismic Centrifuge Modeling of Liquefaction and its
Effect on Deep Foundation, State-of-the-Art Paper, Proc 4th Int Conf on Recent Advances in
Geotech Earthq Eng and Soil Dyn (S. Prakash, ed.), Paper SOAP 3, Vol. 2, 30 pages, San Diego,
CA, March 26-31, 2001.
Dobry, R., Taboada, V., and Liu, L., Centrifuge Modeling of Liquefaction Effects During
Earthquakes, Proc 1st Int Conf on Earthq Geotech Eng (IS-Tokyo), Keynote Lecture, pp. 1291-
1324, Tokyo, Japan, 1995.
Dobry, R., Abdoun, T., ORourke, T.D., and Goh, S.H., Single Piles in Lateral Spreads: Field
Bending Moment Evaluation, J Geotech Geoenviron Eng, Vol. 129, No. 10, pp. 879-889, 2003.
Dungca, J.R., Kuwano, J., Saruwatari, T., Izawa, J., Suzuki, H., and Tokimatsu, K., Shaking Table
Tests on the Lateral Response of a Pile Buried in Liquefied Sand, Proc 11th Intl Conf on Soil Dyn
and Earthq Eng, Vol. 2, pp. 471-477, 2004.
El Shamy, U., and Zeghal, M., DEM Class A Prediction of UB Level Ground Shaking Table Tests,
Internal Progress Report, Tulane University, New Orleans, LA; and Rensselaer Polytechnic
Institute, Troy, NY, 2007.
Elgamal, A., and Forcellini, D., Experimental and Computational Study of Pile Foundations
Subjected to Liquefaction-Induced Lateral Spreading: Simulation of Test LG-0, Internal Progress
Report, University of California, San Diego, CA, 2007.
Elgamal, A., Zeghal, M., Taboada, V., and Dobry, R., Analysis of Site Liquefaction and Lateral
Spreading Using Centrifuge Testing Records, Soils and Foundations, JGS, Vol. 36, No. 2, pp.
111-121, 1996.
Elgamal, A., Parra, E., Yang, Z. and Adalier, K., Numerical Analysis of Embankment Foundation
Liquefaction Countermeasures, J Earthq Eng, Vol. 6, No. 4, pp. 447-471, 2002a.
Elgamal, A., Yang, Z., and Parra, E., Computational modeling of Cyclic Mobility and Post-
Liquefaction Site Response, J of Soil Dyn and Earthq Eng, Vol. 22, Issue 4, pp. 259-271, 2002b.
Elgamal, A., Yang, Z., Parra, E., and Ragheb, A., Modeling of Cyclic Mobility in Saturated
Cohesionless Soils, Int J of Plast, Vol. 19, Issue 6, pp. 883-905, 2003.
Elmekati, A., and Zeghal, M., Systen Identification Analysis of Test LG-0, Internal Progress Report,
Rensselaer Polytechnic Institute, Troy, NY, 2007.
Gonzalez, L., Abdoun, T., and Dobry, R., Effect of Soil Permeability on Centrifuge Modeling of Pile
Response to Lateral Spreading, ASCE GSP No. 145 (Boulanger and Tokimatsu, eds.), pp. 50-60,
2005.
Haigh, S.K., and Madabhushi, S.P.G., The Effect of Pile Flexibility on Pile-Loading in Lateral
Spreading Slopes, ASCE GSP No. 145 (Boulanger and Tokimatsu, eds.), pp. 24-37, 2005.
Hamada, M., and ORourke, T.D. (eds.), Case Studies of Liquefaction and Lifeline Performance
During Past Earthquakes. Vol. 1: Japanese Case Studies, NCEER, SUNY-Buffalo, Buffalo, NY,
Tech. Rept. NCEER-92-0001, February, 1992.
Hamada, M., Yasuda, S., Isoyama, R., and Emoto, K., Study on Liquefaction Induced Permanent
Ground Displacements, Research Rept., Assn. for Development of Earthquake Prediction, Japan,
87 pages, November, 1986.
He, L., Elgamal, A., Abdoun, T., Abe, A., Dobry, Meneses, J., Sato, M., and Tokimatsu, K., Lateral
Load on Piles Due to Liquefaction-Induced Lateral Spreading During 1G Shake Table
Experiments, Proc., 8th U.S. National Conf on Earthq Eng, Paper No. 881, San Francisco, CA,
April 18-22, 2006.
Hwang, J.I., Kim, C.Y., Chung, C.K., and Kim, M.M., Behavior of a Single Pile Subjected to Flow of
Liquefied Soil in an Infinite Slope, Proc 11th Int Conf on Soil Dyn and Earthq Eng, Vol. 2, pp.
573-580, 2004.
Ishihara, K., Yasuda, S., and Nagase, H., Soil Characteristics and Ground Damage, Soils and
Foundations, JGS, Special Issue: January, pp. 109-118, 1996.
Koyamada, K., Miyamato, Y., and Tokimatsu, K., Field Investigation and Analysis Study of Damage
Pile Foundation During the 2003 Tokachi-Oki Earthquake, ASCE GSP No. 145 (Boulanger and
Tokimatsu, eds.), pp. 97-108, 2005.
172
Lin S., Tseng, Y., Chiang, C., and Hung, C., Damage of Piles Caused by Lateral Spreading-Back
Study of Three Cases, ASCE GSP No. 145 (Boulanger and Tokimatsu, eds.), pp. 121-133, 2005.
McCulloch, D.S., and Bonilla, M.G., Effects of the Earthquake of March 27, 1964 on the Alaska
Railroad, Professional Paper 545-D, U.S. Geological Survey, 1970.
Mizuno, H., Pile Damage During Earthquakes in Japan (1923-1983), Proc ASCE Session on
Dynamic Response of Pile Foundations (T. Nogami ed.), pp. 53-77, Atlantic City, April 27, 1987.
ORourke, T.D., and Hamada, M. (eds.), Case Studies of Liquefaction and Lifeline Performance
During Past Earthquakes. Vol. 2: United States Case Studies, NCEER, SUNY-Buffalo, Buffalo,
NY, Tech. Rept. NCEER-92-0002, February, 1992.
Parra, E., Numerical Modeling of Liquefaction and Lateral Ground Deformation Including Cyclic
Mobility and Dilation Response in Soil Systems, Ph.D. Thesis, Rensselaer Polytechnic Institute,
Troy, NY, 1996.
Radwan, H., and Abdoun, T., 3D Data Viewer of Centrifuge Physical Models, Geotech Special Publ
Proc GeoCongress 2006 Conference, Atlanta, GA, February 26-March 1, 2006.
Shenthan T., Factors Affecting Liquefaction Miitigation in Silty Soils Using Stone Columns, MS
Thesis, University at Buffalo, SUNY, NY, p. 129, 2001.
Suzuki, H., Tokimatsu, K., Sato, M., and Abe, A., Factor Affecting Horizontal Subgrade Reaction of
Piles During Soil Liquefaction and Lateral Spreading, ASCE GSP No. 145 (Boulanger and
Tokimatsu, eds.), pp. 1-10, 2005.
Swan S.W., Flores, P.J., and Hooper, J.D., The Manzanillo Mexico Earthquake of October 9, 1995,
NCEES Bulletin, The Quarterly Publication of NCEER, Vol. 10, No. 1, January, 1996.
Taboada, V.M., Centrifuge Modeling of Earthquake-Induced Lateral Spreading in Sand Using a
Laminar Box, Ph.D. Thesis, Rensselaer Polytechnic Institute, Troy, NY, 1995.
Tamura, S., and Tokimatsu, K., Seismic Earth Pressure Acting on Embedded Footing Based on
Large-Scale Shaking Table Tests, ASCE GSP No. 145 (Boulanger and Tokimatsu, eds.), pp. 83-
96, 2005.
Thevanayagam, S., and Ecemis, N., NEESR-SG Project Experimental and Micromechanical
Computational Study of Pile Foundations Subjected to Liquefaction Induced Lateral Spreading,
Internal Progress Report, University at Buffalo, SUNY, Buffalo, NY, 2007.
Tokimatsu, K., Performance of Pile Foundations in Laterally Spreading Soils, Proc 2 nd Int Conf on
Earthq Geotech Eng (Seco E. Pinto, ed.), Vol. 3, pp. 957-964, Lisbon, Portugal, June 21-25, 1999.
Tokimatsu K., Mizuno, H., and Kakurai, M., Building Damage Associated with Geotechnical
Problems, Soils and Foundations, JGS, Special Issue: January, pp. 219-234, 1996.
Towhata, I., Sesov, V., Motamed, R., and Gonzalez, M., Model Tests on Lateral Earth Pressure on
Large Group Pile Exerted by Horizontal Displacement of Liquefied Sandy Ground, Proc 8th U.S.
National Conf on Earthq Eng, Paper No. 1227, San Francisco, CA, April 18-22, 2006.
Yokoyama K., Tamura, K., and Matsuo, O., Design Methods of Bridge Foundations Against Soil
Liquefaction and Liquefaction-Induced Ground Flow, Proc 2 nd Italy-Japan Workshop on Seismic
Design and Retrofit of Bridges, pp. 109-131, Rome, Italy, February 27-28, 1997.
Youd, T.L., Liquefaction-Induced Damage to Bridges, Transportation Research Record, published
by the TRB and the National Research Council, No. 1411, pp. 35-41, Washington, D.C., 1993.
Zeghal, M., and El Shamy, U., A Continuum-Discrete Hydromechanical Analysis of Granular
Deposit Liquefaction, Int J Numer Anal Meth Geomech, Vol. 28, No. 14, pp. 1361-1383, 2004.
Zeghal, M., and Oskay, C., A Local Identification Technique for Geotechnical and Geophysical
Systems, Int J Numer Anal Meth Geomech, Vol. 27, No. 11, pp. 967-987, 2003.
Zeghal, M., Elgamal, A.W., Tang, H.T., and Stepp, J.C., Lotung Downhole Array. II: Evaluation of
Soil Nonlinear Properties. J of Geotech Eng, Vol. 121, No. 4, pp. 363-378, 1995.
173
June 25-28, 2007
Paper No. W1-1010
TRENDS AND OPPORTUNITIES IN THE FURTHER USE AND

DEVELOPMENT OF THE EU LARGE RESEARCH
INFRASTRUCTURES FOR EARTHQUAKE ENGINEERING
Michel Gradin and Fabio Taucer
ABSTRACT
As stated in the European Strategic Research Agenda presented by the Earthquake Engineering
community, the research institutions operating the European large testing facilities for Earthquake
Engineering as well as their users community face the need to better integrate their research activities
and make a coherent use and development of infrastructures. A coordinated and collaborative research
strategy has to be adopted to take full advantage of the continuous advance in the knowledge of the
associated methodologies and in the technologies to implement them. In particular, information
technologies allow evolution towards global connectivity to support interoperability, hybrid and
distributed testing, to ensure remote access of the laboratories and to provide access to a common data
base of experiments and results.
If properly achieved, such evolution will greatly facilitate international collaboration as developed in
the context of the different working groups created under the umbrella of the World Forum on
Collaborative Research in Earthquake Engineering. It will also provide industry and SMEs easier
access to performing testing infrastructures. Following this strategy will imply making best use of the
specific instruments provided by the FP7 Capacities Programme.
.
174
4th International Conference on
June 25-28, 2007
Paper No.W1-1006
VISUALIZATION OF LARGE-SCALE SEISMIC DATA RECORDS
Falko Kuester 1, Tara C. Hutchinson2, Tung-Ju Hsieh 3
ABSTRACT
Visualization of data, whether it is measured or from numerical simulations, is key to allow a clear
interpretation of physical processes. Learning and understanding is substantiated by orders of
magnitude if crisp visualization of an event, data, or other information, is provided to the user. In this
paper, measured data from multiple scales (global, local, and the building-level) is used to demonstrate
a variety of visualization methods, each of which provide the user with the ability to maximize their
understanding of its underlying physical mechanisms. These scales encompass the use of field
acquired seismic measurements, digital elevation data, and building-level seismic sensors. The concept
of visual analytics is discussed in the context of these datasets, and demonstrated via data presentation
on a variety of hardware interfaces, including a massive high resolution display system and an
immersive virtual reality system.
Keywords: Visualization, display systems, virtual reality, seismic sensors, digital elevation maps
INTRODUCTION
Large-scale data acquisition, computing and networking initiatives are providing substantial resources
for the modeling, simulation, measurement, and analysis of complex geotechnical and structural
systems. The result is an overwhelming amount of scientific data being created at increasingly shorter
time intervals and ever-higher resolution. For example, massively networked, sensor grids provide
instantaneous access to time varying multivariate records capturing seismicity at the global and local
level. These data sources can then be further augmented with data from airborne imaging systems
delivering topology and topography data at centimeter resolution, forming geo-spatially referenced
records, capturing pre-event, event and post-event information. To cope with data of a complexity and
detail that greatly exceeds the capabilities of traditional 2D plots for 3D data expression and analysis,
a new generation of visual analytics techniques, tools and infrastructure are required that allow
researchers to collaboratively view, interrogate, correlate and manipulate data at a broad range of
temporal and spatial scales and resolutions. The objective of visual analytics is to enable data analysis
and hypothesis generation by means of interactive data visualization and fusion. The underlying idea
is that collaborative visual comparison of theory and practice, by teams of engineers, will allow new
insights to be derived more rapidly.
In this paper, a set of processing and visualization strategies for seismic records is described, that
leverage high-resolution display walls and virtual reality environments specifically designed to
support visual analytics. Application examples considering the global, local, and building-scales are
provided to demonstrate the potential of these tools.
1
Associate Professor, Departments of Structural Engineering and Computer Science and Engineering,
University of California, San Diego, CA, USA. Contact: fkuester@ucsd.edu
2
Associate Professor, Department of Structural Engineering, University of California, San Diego, CA,
USA.
3
Postdoctoral Researcher, Calit2, University of California, San Diego, CA, USA
175
RELATED WORK
A conventional approach for the analysis of field-collected data sets is to study earthquake induced
effects as time histories by plotting and comparing two-dimensional (2D) discrete waveforms. The
frequency characteristics can then be studied by overlaying different plots and evaluating ratios and
relationships. However, when properly geo-referenced and treated, seismic data sets can be presented
in a more natural and intuitive form that facilitates the understanding of governing events and
mechanisms. In particular, three-dimensional (3D) display can provide additional visual clues via an
intuitive representation of the time-varying wave propagation that is difficult to represent with
traditional 2D techniques. When combined with an immersive display environment, the spatial
relationships may be further strengthened. Nayak et al. (2003) used 3D glyphs, graphics primitives or
symbols with various geometric and color attributes, to represent the measured seismic data. These
glyphs were rendered in real time and combined with a 3D topography terrain map. An interactive 3D
visualization system was developed by (Hsieh et al., 2003b) to facilitate the intuitive analysis of field
measured seismic data sets. Field collected seismic data sets were rendered in 3D and geo-referenced
with Digital Elevation Models (DEMs), and subsequently displayed in a monoscopic or stereoscopic
display environment. In another example (Hutchinson et al., 2005) presented a visualization system
used to enhance the understanding of the time-varying movements of a case study building subjected
to the 1994 Northridge earthquake.
Simulation of seismic wave propagation characteristics of a region can further aid with estimating
potential damage caused by hypothetical or actual earthquakes. For example, the ground motion of the
1906 San Francisco earthquake was simulated by (Mavroeidis & Papageorgiou, 2001) and they
concluded that intense horizontal ground displacements were developed along the entire length of the
Saint Andreas fault plane that ruptured during the earthquake for the San Francisco Bay Area. Their
simulation results were plotted on a 2D map using vectors at each grid cell, pointing into the direction
of displacement. A series of grayscale images, taken at equal time intervals, was produced indicating
the evolution of the ground displacement field over time. These snapshots show the qualitative value,
yet reveal no other geological information. Another Simulation of wave propagation for the San
Francisco Bay Area using a numerical finite difference method was performed by (Antolik el al.
1997). One result was a 90-second video showing wave propagation overlapped with a 2D contour
map. The result of 3D simulation was displayed on a 2D surface plane with red color representing
positive amplitudes and blue color representing negative amplitudes. Another larger scale simulation
was conducted by (Komatitsch et al., 2003) to simulate wave propagation resulting from large
earthquakes. An earth-scale 5.5-billion-grid-point (average grid surface spacing of 2.9 km) crustal
model was constructed and simulated on a 1944-processor supercomputer. The obtained results were
visualized as seismic waveforms plotted on a 2D world map.
In order to fully explore the results of high-resolution simulation runs in combination with field
collected data, new rendering algorithms are needed that fuse different data sources to provide a
geospatially and temporally anchored, visual representation of the studied physical phenomenon.
Akcelik et a. (2003) used a 3000-processor supercomputer to simulate an earthquake in the Los
Angeles Basin using one billion grid points. This data was subsequently used to develop parallel
rendering algorithm capable of rendering time-varying large-scale earthquake simulation results in
real-time. Ma et al. (2003) rendered the earth crust volume data semi-transparent allowing viewers to
see through the volume, revealing 3D seismic wave propagation originating from the hypocenter of a
simulated earthquake event. Compared to conventional 2D approaches, volume rendering in this case
provides a perspective viewpoint, allowing the simulated data to be explored in the spatial and
temporal domain. These visual paradigms provide an enhanced understanding of the field measured
seismic data sets. Instead of just focusing on the local scale, this paper presents visualization
techniques that allow scientists to study seismic data across different spatial (global, local and
building-level) and temporal scales.
176
CASE STUDY: SEISMIC RECORDS AT THE GLOBAL, LOCAL, AND BUILDING LEVEL
Throughout the past century, millions of earthquakes have been recorded worldwide, processed (e.g.
(NEIC, 2007)) and studied to construct increasingly accurate analytical and numerical models. The
development of these improved models requires a careful and thorough understanding of historic
earthquake events and their effects. High-quality earthquake field data in combination with topological
and geological structures and new exploration techniques are of utmost importance for a better
understanding of the distribution of earthquakes in space, time, and magnitude.
Recent deployment of seismograph networks and the growth in number of installed digital seismic
measuring instruments has led to a rapid increase in the size of seismic data sets (Hudnut et al, 2001).
These massive field measured data sets, consisting of digitized seismic waveforms, are now readily
available to the public for download in near real-time. While visualization strategies for simulated data
have been explored in the past, real-time visualization of unstructured field collected earthquake data
is a largely unexplored research area.
Global-Scale
The United States Geological Survey (USGS) provides access to DEMs representing elevation
information in a raster grid of regularly spaced intervals. The DEMs are available from the USGS
Earth Resources Observation Systems (EROS, 2007) Data Center. The GTOPO30 (GTOPO30, 2007)
DEMs, with a horizontal grid spacing of approximately 1-kilometer, are used in this study. In addition,
the used historical earthquake records were obtained from NEIC. As part of the USGS Earthquake
Hazards Program, NEIC provides an extensive seismic database for scientific research compiled from
records of global seismograph networks. The obtained earthquake information includes position,
depth, magnitude, and date/time of a seismic event, i.e. the description of its hypocenter. For this
study, a total of 500,000 hypocenter records were used.
Local-Scale
In dense urban regions, depending of course upon the ground motion, site, and structure
characteristics, even a magnitude 4.0 earthquake may result in significant losses in widespread areas.
Beyond the tragic loss of life, important civil infrastructure such as buildings, dams, and bridges may
be damaged or destroyed and critical lifeline systems such as power grids, water and gas lines,
interrupted. California in particular is classified as one of the areas with the highest earthquake activity
in the United States.
For the past three decades, the California Strong Motion Instrumentation Program (CSMIP) has been
installing accelerometers at various representative locations throughout California, making it one of
the most extensively monitored states in the nation. Currently, more than 900 stations are operational,
including 650 ground-response, 170 building, 20 dam and 60 bridge stations. The corresponding data
records are maintained by the Consortium of Organizations for Strong-Motion Observation Systems
(COSMOS, 2007). The objective of COSMOS is to promote the advancement of strong-motion
measurement on the ground and in structures in densely populated urban areas likely to be struck by
future earthquakes and to develop national policies for the improvement of strong-motion earthquake
measurements and their application. Among the readily available records are those of stations that
recorded the characteristics of the 1994 Northridge earthquake.
The 1994 Northridge earthquake, with a magnitude 6.7, was the most damaging earthquake in the
United States since the San Francisco Earthquake of 1906 and its cumulative damage to public
property and California's infrastructure was estimated to be \$20 billion U.S. (Aurelius, 1994; EERI,
1995). However, the earthquake also generated a wealth of sensor records that when combined with
other historic events, provide statistical data that contributes towards the improvement and
development of seismic design guidelines. Augmenting these data sets with urban planning networks
such as telecommunication or power grids may then also assist in pre- and post-event planning and
hazard mitigation.
177
Building-Scale
Temporal loading plays a large role for the extreme loading conditions civil engineers must design for.
The studied building is a seven-story reinforced concrete structure located in central San Fernando
Valley, California. Numerous seismic sensors were placed throughout the building and recorded its
performance during the 1994 Northridge Earthquake. The current system of accelerometers includes
one measuring the vertical component of motion on the first floor, ten measuring the north-south
direction on floors one, two, three, six, and the roof, and five measuring the east-west direction on
floors one, two, three, six, and the roof level. Figure 1 shows the instrumentation layout throughout the
structure.
The Northridge earthquake caused significant structural and nonstructural damage to this building,
rendering it unsafe for occupancy (Trifunac et al., 1999). All exterior columns in the longitudinal
direction of the building on the fourth floor developed large shear cracks. On the South facing wall,
there were wide shear cracks in the columns. Analysis of response records concur with visual
observations, indicating that the building behaved roughly in the nonlinear range during the
earthquake.
Figure 1. Sensor locations throughout the studied building.
VISUALIZATION STRATEGIES
Even though these data records provide spatial and temporal details, which make them suitable for
fusion with high-resolution digital elevation models, satellite imagery, and inventory data to provide
geo-spatially and temporally anchored information, conventional data analysis techniques are
commonly still limited to 2D plots and maps. In the past, this limitation was a function of the
processing and rendering performance of commodity PCs. However, with the recent performance
increases of graphics processing units (GPUs) on today's graphics cards, new hardware accelerated
technique for data exploration are viable. Real-time interactive visualization of large-scale terrain
models tended to be challenging because of the huge amount of polygons that have to be rendered.
Utilizing adaptive out-of-core techniques (Duchaineau et al., 1997), which load data from external
storage as needed, in combination with hardware accelerated rendering, complex terrain meshes can
now be rendered at interactive frame rates and augmented with detailed texture maps. A variety of 2D
images, such as aerial photos, satellite images, and inventory data in image form, can be seamlessly
mapped onto the terrain surface and programmable graphics hardware used to move selected
geometric computations from the CPU to GPU, increasing the rendering speed (Amor et al., 2005).
Methodology
Interactive visualization is at the heart of any visual analytics environment, requiring a seamless data
processing, fusion, visualization and interrogation framework. For the presented visualization of
seismic records, pipeline stages include, (i) terrain mesh creation and adaptive refinement, (ii) data
178
augmentation and (iii) visualization of static and time-varying seismic records. Figure 2 illustrates the
building blocks of this visualization pipeline. These components are described in detail in the
following sections.
Figure 2. The components of the visualization pipeline.
Adaptive 3D Rendering and Terrain Mesh Refinement

Extensive high-resolution digital elevation models are available today, in combination with aerial
photographs and satellite imagery, providing a means for the creation of an accurate 3D model on a
global or local scale. In this study, USGS GTOPO30 DEMs, with a horizontal grid spacing of
approximately 1-kilometer and complexity of 43,200x21,600 pixels were used in combination with
satellite imagery and color-coded maps indicating terrain elevation. Interactive rendering of these
terrain meshes at full resolution is still beyond the capability of most commodity graphics hardware,
however, it is possible to create close approximations with good visual quality. One solution is to
approximate the original terrain mesh with a coarser mesh adaptively as the viewpoint changes by
selecting a subset of the vertices from the full resolution height field. In view-dependent mesh
refinement, the construction process is performed at run-time. A modified version of the SOAR terrain
rendering engine (Lindstrom & Pascucci, 2002) was used for the construction of the terrain. The mesh
is constructed adaptively at each frame such that the flat neighborhoods and distant areas are
triangulated more coarsely than close and high-detail neighborhoods. Figure 3 shows the snapshots of
an adaptively rendered terrain mesh. In order to articulate terrain variations on a global scale, terrain
elevation is exaggerated by a factor of 90.
(a) (b) (c)

Figure 3. Level-of-detail terrain model rendering and texture mapping. (a) A coarse terrain
mesh with 4,143 triangles and texture mapping enabled, (b) coarse terrain mesh without
textures, and (c) a refined terrain approximation using 133,984 triangles.
179
Terrain Texture Mapping
Textures can be defined as rectangular arrays of data, such as aerial photos, terrain height fields, and
ground motion data. Texture mapping is a computer graphics technique that applies an image to the
surface of a polygon. A rectangular texture can be mapped to nonrectangular polygons by indicating
how the texture is to be applied to the vertices that define the polygons. Figure 4 shows an example of
a 2D texture map using a cylindrical projection of the earth's surface obtained from NASA Goddard
Space Flight Center. At one kilometer resolution (43,200x21,600 pixels), it is one of the most detailed,
continuous and freely available, true-color image of the Earth to date.
(a) (b) (c)

Figure 4. Screenshots of EQVis: 3D rendering of the DEMs and NEIC earthquake data sets at
global scale, (a) global 3D seismicity map of 500,000 hypocenters, (b) shaded terrain model, (c)
3D terrain texture mapped with satellite image for North America plate, South American plate,
and Caribbean plate.
Rendering Static Seismic Data Sets

Hypocenters from the NEIC database are rendered as spherical proxies, and color-coded based on
depth. Different glyphs can also be used to represent other components of hypocenter records such as
event data/time, location, depth and magnitude in a geo-referenced context. In addition, detailed 3D
surface and surface level information can be shown, to provide important features to the viewer.
Figure 5 illustrates different data representation modes. Figure 5 (a) was provided by NEIC for
earthquake records from Alaska (Latitude = 60 to 66 North, longitude = 144 to 156 West). The total
number of earthquakes recorded in this area was 14,876. The difference with respect to visual
impression is significant when the same data set is studied in 3D as shown in Figure 5 (b). The
observed viewing volume in this example is 29.5 million km3. A spherical proxy is rendered at the
earthquake hypocenter location, defining the longitude, latitude and depth of the event. The proxy
radius represents the magnitude of the event and the color the earthquake year. In addition, detailed 3D
surface and aboveground level information is shown to provide context and orientation to the viewer
in respect to important surface features. Correlations with epicenters (hypocenters projected to the
terrain surface) are optionally encoded into the surface texture map, and geometrically connected.
Epicenter information alone would fail in revealing the spatial characteristics of the earthquake event
as shown in Figure 5(a). When viewed in an interactive environment, spatial correlation is straight
forward as users navigate through the data and view it from arbitrary viewpoints.
Figure 6 shows a 3D rendering of hypocenters in the greater Anchorage area. Interactive rendering of
the historic hypocenter data provides the user with spatial and temporal relationships and can aid in the
visual identification of fault plane locations.
180
(a) (b)
Figure 5. (a) 2D earthquake epicenter map for the greater Anchorage area. (Latitude: 60 to 66
North, Latitude: 144-156 West). Source NEIC. (b) Example of USGS NEIC earthquake data sets
for the Alaska area, combining terrain and static earthquake data sets. (Latitude: 60 to 66
North, Latitude: 144-156 West, 1973-2006).
(a) (b) (c) (d)

Figure 6. 3D renderings of Alaska DEMs and hypocenters viewed from different directions.
Imposters vs. 3D Proxies

In order to interactively render complex earthquake data sets in combination with high-resolution
digital elevation maps and auxiliary reference information, substantial resources are required. Since it
is desirable to have this technology available on standard computational platforms, optimized
algorithms and rendering techniques have to be employed to achieve the desired results. Since well-
defined geometric proxies are used to represent earthquake hypocenters, rendering techniques based
on imposters (billboards or depth sprites) can be used to achieve even better performance gains. These
techniques are particularly popular in flight simulation, in which static objects such as trees can be
approximated through simple images (textures) mapped onto a single viewpoint aligned polygon. The
challenge for this particular application is that the system has to handle tens to hundreds of thousands
of billboards. For the presented data, billboards and depth sprites are particularly suited to replace the
rotational symmetric spherical proxies of hypocenters, which allow rotation over an arbitrary axis
through the proxies' origin without the need to change the texture image representing it. Since all
hypocenter proxies have the same shape, only a single texture is needed.
181
Rendering Dynamic (Time-Varying) Seismic Data Sets
The time-varying nature of earthquake data records obtained from discretely positioned seismographs
is suitable for dynamic display, enabling intuitive correlation of the observed ground motion records.
Synchronized display of ground motion records can then reveal seismic wave propagation. This
knowledge could ultimately aid in developing refined numerical or analytical simulations of these
events. The horizontal orthogonal components of one-dimensional acceleration records can be
combined and displayed in 3D using billboarding techniques for seismograms and geometric proxies
to encode time varying information. When properly time stamped and geo-referenced, data from
multiple stations can be studied simultaneously. This study uses CSMIP free-field measured tri-
directional acceleration records of the 1994 Northridge earthquake, obtained via the COSMOS
database. The data of 72 representative stations was used for the visualization.
Figure 7(a) shows the acceleration amplitudes for the 1994 Northridge earthquake rendered as
billboards. Since the textures for the ground motion records are generated dynamically, additional
information such as time stamps or global reference markers can easily be augmented onto the data as
it is being rendered. This enables the correlation of the observed events and more intuitive analysis of
interconnected information. Figure 7(b) shows a visual representation of the peak ground accelerations
in the form of bounding boxes aligned with the respective data site. Bounding box dimensions
represent the peak accelerations observed during the event and are updated based on the seismogram
records. An additional proxy is rendered at each station location to visualize the waveform history
during playback. From these ground motion records it is easy to identify at which times particular
areas were subjected to the most significant shaking.
(a) (b)
Figure 7. (a) Rendering of 1994 Northridge seismogram data in east-west direction for the Los
Angeles basin. (b) Bounding box for maximum acceleration amplitude in east-west direction.
Visualizing the Seismic Response of a Building

A 3D model of the building was constructed for subsequent visualization for this case study. Major
components of the structure were subdivided into groups of solid objects, such as columns, beams,
floors (including the roof level), paneling, and facia. This allows groups of objects to have common
texture skins and to be independently manipulated. The field measured data sets are discretely mapped
onto the geometric model and used to generate temporal simulations reproducing the deformation
patterns observed.Records from the 1994 Northridge earthquake consist of 16 channels of
seismograms measuring acceleration in both horizontal directions, and select measurements in the
vertical direction. Accelerations can be double-integrated and processed to obtain displacement
records. Since sensors are not placed at each floor level, or at each extreme edge of the building, to
fully describe the 3D movement of the structure, interpolation was applied. Floor levels are
structurally rigid and no local damage to the members was observed, therefore, the assumption of
linear interpolation across the plan of each of the floors is reasonable.
182
Each seismogram consists of 3,000 sample points with a sample spacing of t=0.02 second, for a total
duration of the time-varying visualization of 60 s. Figure 8 shows a sequence of time steps and
snapshots of the fully deformed building model using shade (color) coding applied to the components
of the model. In the real simulation for the students, red is used to represent the maximum
deformations (over the entire record), while blue is used to represent minimum values. Interpolation is
applied across the color spectrum and mapped onto the model. Color provides an intuitive, rapid
assessment of the current state of the structure during time. It is based on the assumptions regarding
the unknown deformation points, as described. The ground floor level in this case is fixed and all
subsequent floors are shown relative to ground surface. It should be noted that deformations of the
building are amplified by 20 to enhance their visual perspective. It clearly shows the extreme in-floor
level torsion the structure undergoes during 3D oscillation, as well as the large (dominant) longitudinal
(long axis of the building) displacement demand. Displacement measurements indicate peak values in
the north-south direction of 17 cm on the east end and 23 cm on the west end. This significant
difference between the east and west-end displacements indicated that the building suffered large
torsional movements during the earthquake, which concurs with the clear depiction illustrated. As
observed in the field after this earthquake, the longitudinal load-resisting system was severely
damaged, thus significantly softening the structure in this direction, and the 3D interactive
visualization clearly shows this. In addition, between time steps shown, a sharp reversal in the
direction of movement can be observed (e.g., between parts i, j, and k of Figure 8. The visualization
provides a unique opportunity to show an overview of the structural response of the building for a
given ground motion history. In addition, the articulation of the spatial transitions assists in locating
the building components, which may have been damaged on a local level.
(a) 6396 sec (b) 7.02 (c) 7.08 (d) 7.14 (e) 7.20
(f) 7.26 (g) 7.32 (h) 7.38 (i) 7.44 (j) 7.50
(k) 7.56 (l) 7.62 (m) 7.68 (n) 7.74 (o) 7.80
Figure 8. Image sequence showing building deformations (color coded) between time t = 6.96 -
7.80 seconds (sub-sampled to a t = 0.06 seconds) - 1994 Northridge earthquake.
183
VISUAL ANALYTICS ENVIRONMENT
The presented visual analytics techniques can be further enhanced with a visualization environment
that matches demands imposed by the data to be analyzed. For the presented case studies, concurrent
visual data correlation is enabled through the use of HIPerWall, a wall-sized display. HIPerWall offers
a large-scale workspace for massive visual data correlation, providing the ability to display and study
multiple data sets in great detail. More precisely, HIPerWall supports the integration of multiple data
and analysis modalities, across different scales and information domains. The display consists of fifty
30" display tiles arranged in a ten-by-five matrix, each operating at a resolution of 2,560 by 1,600
pixels (4 mega-pixels), for an overall resolution of over 200 mega-pixels (25,600 by 8,000 pixels). A
combined display space of 7m x 2.2m is available for data visualization and collaborative analysis,
providing a visual, satisfying 20/20 vision across its entire surface at a distance of one meter from the
screen and outward. This effectively allows users to digitally and physically zoom into areas of
interest by either using an input device to adjust magnification levels or simply walking towards the
display. The display array is powered by 25 PowerMac G5s equipped with dual, dual-core processors
(quads), NVIDIA Quadro FX 4500 graphics cards, 2GB of on-board memory and gigabit Ethernet. All
nodes are interconnected via a dedicated Level 3 gigabit switch and data is served up by a dedicated
storage node, while a stand-alone control node functions as the front-end. Figure 9 illustrates power-
of-ten capabilities that are used to study data at different levels of resolution, seamlessly transitioning
between a global, local and building-level data representation.
In many cases, spatial and temporal data characteristics also benefit from a stereoscopic data
representation that allows the user to explore a truly three-dimensional representation of the data. A
wall-sized passive-stereo display, operating at 8 megapixel resolution (4 megapixel per eye) was used
to provide access to this additional dimension. Users access 3D space with a wand, a spatially tracked
3D pointing device, supporting unconstrained interaction. Figure 10 shows a series of models
spanning the global to local scale.
(a)
(b)
(c)
Figure 9. Visual Analytics on HIPerWall, leveraging an ultra-high resolution display wall to
explore data across different time and spatial scales, (a) global, (b) local and (c) building-level
visualization with powers of ten support, allowing for a seamless transition between levels.
184
(a) (b) (c)
Figure 10. Visual Analytics on CaveWall, (a) global, (b) local (c) building-level visualization in a
fully semi-immersive virtual reality environment.
CONCLUSIONS
An overwhelming amount of scientific and engineering data is being created at increasingly shorter
time intervals and at ever-higher resolution. This is particularly true in the seismic, geotechnical, and
structural domains, where sensing of the physical environment is essential to advancing understanding.
Visualization of this data, whether it is measured or from numerical simulations, is key to allow a clear
interpretation of physical processes. Learning and understanding is substantiated by orders of
magnitude if crisp visualization of an event, data, or other information, is provided to the user.
In this paper, measured data from multiple scales (global, local, and the building-level) is used to
demonstrate a variety of visualization methods, each of which provide the user with the ability to
maximize their understanding of its underlying physical mechanisms. These scales encompass the use
of field acquired seismic measurements, digital elevation data, and building-level seismic sensors. The
methods are simple, yet robust, as they rely upon out-of-core processing techniques, and simplistic
rendering strategies, to speed up the visualization activities for the user. Rendering methods include
spherical proxies for representing earthquake hypocenters, bounding boxes for representing maximum
structural deformations, and tracing elements to describe history. The concept of visual analytics is
discussed in the context of these datasets and their associated rendering methods, and demonstrated via
data presentation on a variety of hardware interfaces, including a massive high-resolution display
system and an immersive virtual reality system.
AKNOWLEDGEMENTS
This research was supported in part by the National Science Foundation Combined Research
Curriculum Development Program under Grant Number EIA-0203528 and the Major Research
Instrumentation Program under award number MRI-0421554, as well as the Holmes Fellowship
Foundation. Data for this study was provided by USGS, NEIC and COSMOS. Ms. Rebecca Chadwick
assisted with preparation of the waveform data sets and the structural model. The above support is
greatly appreciated.
REFERENCES
Akcelik V., Bielak J., Biros G., Epanomeritakis I., Fernandez A., Ghattas O., Kim EJ., Lopez J.,
OHallaron D., Tu T., and Urbanic J. High resolution forward and inverse earthquake modeling on
terascale computers, In SC 03: Proc. of the 2003 ACM/IEEE conference on Supercomputingpage
52, Washington, DC, USA, 2003. IEEE Computer Society.
185
Amor M., Boo M., Strasser W., Hirche J., and Doggett M. A meshing scheme for efficient hardware
implementation of butterfly subdivision using displacement mapping, IEEE Computer Graphics &
Application, 25(2):4659, 2005.
Antolik M., Stidham C., Dreger D., Larsen S., Lomax A., and Romanowicz B. 2-D and 3-D models
of broadband wave propagation in the san Francisco bay region and north coast ranges,
Seismological Research Letter, 68:328, 1997.
Aurelius E. The January 17, 1994 Northridge, CA earthquake: An EQE summary report. Technical
report, EQE International, Houston, TX, 1994.
COSMOS. Consortium of Organizations for Strong-Motion Observation Systems (COSMOS), 2007.
http://www.cosmos-eq.org/
CSMIP. California Strong Motion Instrumentation Program (CSMIP), 2007.
http://www.consrv.ca.gov/cgs/smip/about.htm
Duchaineau M., Wolinsky M., Sigeti DE., Miller MC., Aldrich C., and Mineev-Weinstein MB.
Roaming terrain: real-time optimally adapting meshes, Proc. VIS 97 the 8th conference on
Visualization 97, pages 8188, Los Alamitos, CA, USA, 1997. IEEE Computer Society Press.
EERI. Northridge earthquake of January 17, 1994 reconnaissance report vol.1, Earthquake Spectra,
Supplement C to vol.11, 1:377420, April 1995.
EROS. United States Geological Survey, National Center for Earth Resources Observation and
Science (EROS), 2007. http://edc.usgs.gov/
GTOPO30. United States Geological Survey, GTOPO30, 2007.
http://edcdaac.usgs.gov/gtopo30/gtopo30.asp
Hsieh TJ., Kuester F., and Hutchinson T. Terrain mapping for interactive visualization of earthquake
datasets, Proc. IASTED International Conference on Modeling and Simulation 2003, pages 489
496, Palm Springs, California, USA, February 2003.
Hsieh TJ., Kuester F., and Hutchinson T. Visualization of large-scale seismic field data, Proc. ISC
High Performance Computing Symposium (HPC 2003), pages 163170, Orlando, Florida, USA,
March 2003.
Hudnut KW., Bock Y., Galetzka JE., Webb FH., and Young WH. The southern California Integrated
GPS Network (SCIGN), Proc. International Symposium on Crustal Deformation Measurement,
pages 129148, 2001.
Hutchinson T., Kuester F., Hsieh TJ., and Chadwick R. A hybrid reality environment and its
application to the study of earthquake engineering, Virtual Reality, 9(1):1733, 2005.
Komatitsch D., Tsuboi S., Ji C., and Tromp J. A 14.6 billion degrees of freedom, 5 teraflops, 2.5
terabyte earthquake simulation on the earth simulator, Proc. SC 03 the 2003 ACM/IEEE
conference on Supercomputing, page 4, Washington, DC, USA, 2003. IEEE Computer Society.
Lindstrom P. and Pascucci V. Terrain simplification simplified: A general framework for view-
dependent out-of-core visualization, IEEE Transactions on Visualization and Computer Graphics,
8(3):239254, 2002.
Ma KL., Stompel A., Bielak J., Ghattas O., and Kim EJ. Visualizing very large-scale earthquake
simulations, Proc SC 03 the 2003 ACM/IEEE conference on Supercomputing, page 48,
Washington, DC, USA, 2003. IEEE Computer Society.
Mavroeidis GP. and Papageorgiou AS. Simulation of long-period near-field ground motion for the
great 1906 San Francisco earthquake, Seismological Research Letters, 72:227, 2001.
Nayak AM., Lindquist K., Newman R., Kilb D., Vernon FL., Johnson A., Leigh J., and Renambot L.
Using 3D glyph visualization to explore real-time seismic data on immersive and high-resolution
display systems, Earth-Oceans-Atmosphere (EOS) Transactions of American Geophysical Union
(AGU), 84(46): Fall 2003 AGU Meeting. Abstract ED32C1208, 2003.
NEIC. National Earthquake Information Center (NEIC), 2007. http://neic.usgs.gov/
Trifunac MD., Ivanovic SS., and Todorovska MI. Instrumented 7-storey reinforced concrete building
in Van Nuys, California: description of damage from the 1994 Northridge earthquake and strong
motion data, Technical Report CE 99-02, Univ. of Southern California, Los Angeles, California,
1999.
186
June 25-28, 2007
Paper No.W1-1007
NONLINEAR WAVE PROPAGATION AND TRENDS AT A LARGE-

SCALE CENTRIFUGE FACILITY
Bruce KUTTER 1 and Dan WILSON2
ABSTRACT
This paper describes the UC Davis centrifuge facilities, with an emphasis on recent developments in
data acquisition systems. An example application presents some centrifuge results that illustrate the
nonlinear propagation of waves through liquefiable soil. These waves are associated with the strain-
stiffening behavior of undrained dilatant soil. When dilation triggers negative pore pressures, the soil
stiffens and large acceleration spikes can develop. Similar behavior has been observed in the field
during real earthquakes. Shock wave theory can be used to describe many aspects of the propagation
of these spike waves. The paper closes with a summary of recent trends in large-scale modeling on
shared facilities. Trends in experimental hardware, instrumentation, research directions, and
networking are outlined, and the importance of collaborative and multidisciplinary research is
discussed.
Keywords: centrifuge, earthquake, wave propagation, instrumentation, equipment, collaboration
INTRODUCTION
The UC Davis centrifuge has the largest radius and largest platform area of any geotechnical
centrifuge in the US; it is one of the top few in these categories in the world. The centrifuge can carry
five-ton payloads and operate at 75 g (at the effective radius of 8.5 m). This centrifuge and its
associated equipment have capabilities for simulating the effects of earthquakes on geotechnical
models and models of soil-foundation-structure interaction. Figure 1 shows researchers installing
instrumented model piles into sand in one of the flexible shear beam model containers. The model
container will be placed on a shake table mounted on the swinging bucket at the end of the centrifuge
(shown in the background). The basic configuration of the shake table on the UC Davis centrifuge is
illustrated in Figure 2; the cylindrical accumulators shown in the diagram can be seen on the bucket in
the background of Figure 1.
Typical geotechnical model testing in the past used limited amounts of instrumentation, which yielded
observations at selected points in a physical model. Progress in earthquake engineering was hindered
by incomplete and ambiguous data sets that allowed for subjective and potentially mistaken
interpretation. The philosophy behind the development of the NEES Centrifuge at UC Davis
capitalizes on the size of the facility and innovations in instrumentation and information technology to
enable researchers to generate much higher resolution information (more control, sensors, and images)
and to create more realistic physical models that will provide unambiguous experimental data for
assessments of numerical simulations.
1
Professor, Department of Civil Engineering, University of California, Davis, USA, Email:
blkutter@ucdavis.edu
2
Associate Project Scientist, Department of Civil Engineering, University of California, Davis, USA
187
Figure 1. Researchers installing model bridge bents in a flexible shear beam container. The
end of the centrifuge arm is in the background. (Photo by Peter Essick)
The data acquisition (DAQ) infrastructure of the centrifuge includes about a dozen computers and
externally-mounted data acquisition chassis. The major components of the DAQ infrastructure are:
RESDAQ-Main: 128 channels of amplifiers connected to a high-speed A/D converter for

conventional sensors
RESDAQ-ERT: 48-channel electrode switching system useful for electrical resistivity
measurements, electro-location techniques, and electrical resistance tomography
RESDAQ-Aux: An auxiliary DAQ chassis dedicated to specialized systems such as the one that
drives eight channels of bender element sources and monitors 16-channels of bender element
receivers. It also contains a digital oscilloscope card useful for high-speed applications.
WiDAQ: Fifty-six 8-channel wireless data acquisition systems that can be used for conventional
sensors and MEMS accelerometers.
VIDDAQ: Six computers, each dedicated to one high speed (up to 200 frames per second)
camera.
Figure 2. Basic configuration of the shaking table mounted on the centrifuge bucket (note the
corresponding cylindrical accumulators in Fig. 1)
188
Figure 3 shows a model, similar to that being prepared in Figure 1, mounted on the centrifuge with all
instrumentation installed just prior to spinning. Visible in the image are cameras, strobe lights and
wireless DAQ units, all monitoring the behavior of six individual pile supported bent structures
(Ilankatharan et al. 2006). Aluminum blocks representing a bridge deck mass were supported on two
piles for each bridge bent model. The blocks were oriented in varying directions to enable a study of
the effect of shaking in different directions relative to the alignment of the piles. The size of the
centrifuge enabled approximately 200 sensors to be used to monitor a simultaneous test of the six
structures under identical soil conditions and identical sequences of ground shaking.
At UC Davis we have developed miniature data acquisition systems that can be mounted in a
centrifuge soil model to reduce the number and complexity of signal wires emanating from highly
instrumented experiments. The WiDAQ Wireless Data Acquisition System was developed as part of
the NEES construction award at UC Davis. It is now in continuous use by researchers at UC Davis
(Wilson et al. 2007). Figure 4 shows a collage of images associated with the WiDAQ system. In the
upper left, 4 two-channel MEMS accelerometers are shown connected to one of the WiDAQ sensor
managers. In the upper right, one of our low-cost, tiny packaging systems is shown. An Analog
Devices ADXL sensor is attached to a miniature circuit board, the sensor cable is stress relieved to the
circuit board, and then the system is dipped in epoxy to seal it. Just below that, the WiDAQ radio
board, DAQ board, and a AAA battery are assembled outside of their housing. On the bottom left of
Figure 4, individual Silicon Designs, Inc., MEMS accelerometers are shown. The image on the
bottom right is a depiction of how the sensor managers can be deployed around a model container to
reduce the number of wires emanating from the experiment. The sensor managers can be buried in the
soil or mounted at convenient locations on the model container.
Pile strain gauges

routed to WCM/SM
MEMS accelerometers
on superstructure
routed to WCM/SM
Pile strain gauges bundled and

routed to DAQ conditioning
ICP accelerometers on
superstructure bundled and
routed to DAQ conditioning
Figure 3. Piles with bent caps skewed relative to the direction of shaking. Visible
instrumentation includes strobe lights, video cameras, displacement potentiometers,
accelerometers and strain gauges.
189
Figure 4. Clockwise from top left: MEMS accelerometers in an open sealable plastic housing;
a MEMS accelerometer with custom packaging, side view of Radio, DAQ card and battery;
model container with embedded sensor managers; and MEMS accelerometers.
The capabilities of the equipment to support research at the UC Davis Center for Geotechnical
Modeling are illustrated in the example study on nonlinear wave propagation described below.
Additional information about the centrifuge is available at http://nees.ucdavis.edu. This paper
concludes with a discussion of future directions for shared major equipment and collaborative
research.
RESEARCH ON NONLINEAR WAVE PROPAGATION IN LIQUEFIABLE SOIL
An example of ongoing research is presented to provide one realization of an experiment on the large
centrifuge. The example chosen is on the study of nonlinear wave propagation in liquefiable soil.
Observations from ground motion instrumentation indicate that liquefaction generally results in a
softening of the soil and an increase in the predominant period of the motion. But at the same time,
high frequency spikes or cusps of acceleration have been observed. The magnitude of these spikes can
be quite large (see data from Frankel et al. (2002) shown in Figure 5), producing the peak ground
acceleration at a site, and hence they should be accounted for in evaluation of shaking hazards.
Youd and Holzer (1994), after analyzing the pore pressure and acceleration data from the Wildlife
Site, explained that as the ground oscillated back and forth displacement was arrested by dilatancy that
caused a sudden drop of pore pressure accompanied by a sudden deceleration. They explained that the
concurrent dips in acceleration and pore-pressure records indicate that the liquefied soil was in a state
of cyclic mobility. Youd and Holzer (1994) explained that the spikes in surface acceleration occurred
later in the record because ground oscillation actually increased in amplitude for about 20 s after the
primary train of acceleration pulses had passed through the site. More recently, Holzer and Youd
(2007) explained that Love waves were responsible for the long period motions. The spiky
accelerations that occurred as a result of long period excitation are clearly a result of highly nonlinear
soil behavior.
190
Figure 5. Ground surface accelerations at a site a few kilometers south of downtown Seattle in
the M 6.8 Nisqually, Washington earthquake. Note the large one-sided spikes between t = 24 s
and 29 s. (Data from Frankel et al. 2002)
At this point, it is useful to distinguish two types of nonlinearity:

1) Strain-softening nonlinearity: shear modulus degrades as strain increases (e.g., hyperbolic stress-
strain law and conventional modulus degradation relationships that show the shear modulus
monotonically decreases as shear strain increases)
2) Strain-stiffening nonlinearity: shear modulus increases as strain increases (e.g., effect of dilatancy
on saturated sand which produces stress-strain behavior indicated in Figure 6).
Strain-Stiffening Nonlinearity and Shock Waves

Zeghal et al. (1994), Youd and Holzer (1994), Bonilla et al. (2005), and Holzer and Youd (2007)
analyzed acceleration and pore pressure data from the Wildlife Site obtained during the Superstition
Hills Earthquake and found that pulses of negative pore water pressure coincided with large shear
strains. They also observed stiffening of the soil at the site associated with the reduction in pore water
pressure and the increase in effective stress. They were able to back-calculate stress-strain and stress
path relationships for the soil at the Wildlife Site and observed that the patterns were similar to those
observed in laboratory element tests (e.g., Figure 7).
Several researchers have demonstrated that acceleration spikes, similar to those observed in the field,
can be reproduced in models tested on geotechnical centrifuges with shaking tables (e.g., Fiegel and
Kutter 1994, Coelho et al. 2004, Kutter and Wilson 1999, Taboada and Dobry 1998). Some results of
centrifuge experiments to study the formation, evolution, and propagation of spiky acceleration pulses
is introduced here.
Figure 6. Stress-strain relationship for a strain-stiffening material.
191
Figure 7. (Top) Undrained torsional shear behavior of Fuji River Sand at a relative density of
75% (Ishihara 1985). Note the strain-stiffening behavior in all except the first stress-strain
loop. The stress-strain curve at the bottom (from Holzer and Youd 2007) was obtained by
processing accelerometer data at the Wildlife Site according to the procedures developed by
Zeghal and Elgamal (1994). Labels on the circled points on the curves correspond to the
times at which peak pore pressures were observed.
Kutter and Wilson (1999) recognized that many aspects of the propagation of acceleration spikes in
centrifuge model tests could be explained using shock wave theories. Shock phenomena occur in
other media (e.g., breaking water waves and sonic booms) due to nonlinearities that cause the peak of
a wave to travel faster than the front. In laboratory tests, the tangent shear modulus of dilatant sand
under undrained conditions has been observed to increase as shear stress increases. This is analogous
to the increase in the tangent bulk modulus of air as pressure increases a necessary condition for
formation of shock waves in air. The extent to which the acceleration spikes can be described by
shock-wave theory and understanding the factors that control the width and amplitude of acceleration
spikes is the subject of ongoing research.
Nonlinear wave-propagation centrifuge results

Figure 8 shows data from Kutter et al. (2004) obtained from a dense vertical array in a layered soil
profile. The left side of Fig. 8 shows a schematic of the soil layering, and the accelerometer locations
are indicated by arrows. The base layer (with H2 = 9 m) consisted of sand with relative density of
80%, the layer with H1 = 4.5 m consisted of sand with relative density of 50%, and the top layer is an
approximately 2-m thick sloping clay layer. The entire box is sloping to the left so that lateral
spreading could be studied. From these data it was clear to see that pulses travel at a variety of speeds,
some of them disappear, and some of them focus into spikes. The shape of the acceleration spikes is
clearly a function of the soil type; the spikes are sharper in dense sand than loose sand, and their
192
Figure 8. Data from Kutter et al. (2004) showing location of sensors in a centrifuge model
container, and the acceleration records in a large shaking event. Time and depth have been
scaled by a factor of 30 to prototype scale.
amplitude decreases and width increases in the clayey surface layer. The spikes are unsymmetrical
due to the existence of a static shear stress in the direction of the sloping ground. Some pulses were
observed to travel as slow as 11 m/s. Figure 9 presents a closer look at the data from this vertical
array, and also presents velocity records, obtained by integration and baseline correction of the
acceleration data. In Fig. 9, fine lines have been added to enable visualization of the propagation of
various wave features. Wave velocities are obviously not constant, some small waves coalesce into
sharper waves, sometimes forming a sharp spike and then the spike waves sometimes grow and
sometimes disperse as they propagate through the soil. Interestingly, the inelastic waves seem to die
as they reach the surface without significant reflection.
The nonlinear wave propagation problem is just one application for which dense arrays of sensors are
desirable to capture the phenomenon. The multi-component experiment shown in Figure 3 is another
example where the performance of a variety of structures can be compared in a single model test,
provided that sufficient instrumentation is available.
TRENDS IN CENTRIFUGE MODELING OF EARTHQUAKE ENGINEERING PROBLEMS
Facilities (experimental hardware)

Over recent years there has been a significant growth in ability for robotic manipulation of
experiments while they spin on the centrifuge. Robotics may be used for inspection with cameras, site
investigation with penetrometers or resistivity probes, and for simulation of construction processes
such as excavation or pile installation. Another area where progress has been significant is in the
development of flexible shear beam model containers that permit waves to propagate through the
model without over-constraining the behavior of the experiment.
In shake table testing on the centrifuge and elsewhere, there is an interaction between the shaking
apparatus and the experiment, the interaction alters resonant frequencies of the system and results in
energy transfer between the experiment and the apparatus. In centrifuges the ratio of mass-of-
experiment to the mass of the shaking table or to the reaction mass is typically larger than it is in 1-g
shaking tables. This significance of the interaction between the apparatus and the experiment should
be accounted for.
193
Figure 9. A closer look (12 to 19 s time window) of the data in Figure 6. The top graph shows
acceleration, the bottom shows velocity obtained by integration of acceleration record. Light
lines were added by eye to help visualize the evolution of nonlinear waves. Spacing of traces
is in proportion to the spacing between sensors in the vertical array. Time is scaled by a
factor of 30 to convert model data to prototype scale.
Instrumentation and data acquisition

The trend in instrumentation can be summarized by: more, faster, and different. Commercial high
speed video cameras are becoming more and more capable, reliable, and have better resolution. In
selecting video cameras, one is faced by the decision of using one very fast (1000s of frames per
second) and expensive camera or opting for use of a larger number of inexpensive intermediate speed
cameras (200 frames/second). Wireless data acquisition systems are now being routinely used at
Davis and we invite collaborators to help maintain and develop this platform. The system enables
100s of sensors to be economically added to a facility with minor increases in wire congestion in the
experiments. The WiDAQ system is able to record data from MEMS accelerometers, which promises
to reduce the cost associated with increasing instrumentation.
Geophysical testing and tomography are really promising tools for better monitoring of experiments.
In this category we include arrays of bender elements to enable monitoring of the evolution of shear
wave velocity in the experiments and p-wave ultrasound imaging which has been shown to hold
promise for detecting interfaces and pockets of unsaturated soils in the centrifuge models. Electrical
194
methods also have been demonstrated to hold promise for measuring volumetric and shear strains in
soil, for detection of movements of buried electrodes, and producing tomographic images of resistivity
distribution (Li et al. 2006). Fiber-optic sensors, as they become more and more economical, have
great potential to revolutionize the ability to measure strain distributions on structural components
such as piles or retaining walls. Laser scanners and close range photogrammetry could potentially
provide accurate 3-D maps of experiments.
There are so many exciting instrumentation technologies, and many of them require different
equipment and software to make them work. In dynamic centrifuge testing, the speed of
measurements is frequently an issue. The cost of increasing instrumentation can be measured in
purchasing cost or in the time required to install the sensors, measure their locations, mount the
equipment on the centrifuge and learn or develop the software to process the resulting data.
Research topics
With the development of shared major experimental facilities in the US, interdisciplinary research is
likely to continue to increase. This increases the breadth of knowledge of researchers, and forces them
to figure out how to communicate the results of their research to members of another discipline. This
must be beneficial for growing the impact of the research. Recent efforts to model more complex
systems in the large centrifuge have included the dynamic behavior of a nonlinear building frame
supported on shallow foundations and multi-bent sections of a bridge structure. The complex
experiments are both stimulated and made possible through collaboration. (e.g. Kutter and Wilson
2006).
Networking, data sharing, simulation and visualization tools, and collaboration

Due to the increase in complexity of the experiments being performed, the planning of large scale
experiments and construction of the models is best done by teams of researchers. With the advent of
teleconferencing tools provided by NEESit (the information technology provider for the Network for
Earthquake Engineering Simulation), cross country teleconferences have become quite commonplace
in the US. One of the common justifications for model tests is to provide data for verification or
development of numerical procedures. This is benefited by the provision of open source community
simulation tools (e.g., the OpenSees platform http://it.nees.org/software/opensees/).
An oft cited benefit of model testing is that a valuable data set will be produced and archived for use
by others. Data sharing is a great idea because it has the potential to reduce the need for repeating
experiments that have already been done. Sharing data does not mean posting a file to a web site. The
file needs to be accompanied by the metadata necessary for others to understand the formats of the
data (e.g. Kutter et al. 2002). The complexity of data from large scale experiments is an additional
non-technical barrier to data sharing. Visualization tools could play an important role in enabling
complex data sets to be understood by others. A visualization tool called N3DV is intended to be
useful for presenting the data from a large number of sensors in an intuitive format (Weber et al.
2003). A screenshot from N3DV showing sensor metadata and sensor locations is presented in
Figure 10. Dynamic data may be presented by animating the behavior or icons that represent different
sensors in an experiment. Visualization tools such as N3DV are hoped to be a carrot that
encourages people to archive there data in a standard format that can be imported by N3DV and
understood by others. The added benefit is that the tool can hopefully be used on-the-fly during an
experiment enabling researchers to quickly peruse their data and identify sensors that produce
anomalous data.
195
Figure 10. Data and Metadata from centrifuge experiments are shown using a tool called
N3DV. Each sensor is represented by an icon for which the color, size, and location may be
animated in proportion to the data recorded.
DISCUSSION
Research progress in civil engineering has been steady but slow over the last century. Research has
historically been accomplished by individuals with small groups of researchers. A small fraction of
these individuals have been successful in closing the loop from identifying a problem, researching a
solution and being successful in getting the solution applied in engineering practice.
The slow but steady progress in civil engineering contrasts with relatively rapid progress that has been
made in the computing industry. New computer chips are designed by large teams of engineers, and
parallel advances in data storage, displays, networks, batteries, and software are essential for the
computing revolution that has occurred. Individuals play a minor role in the computing revolution; the
progress is driven by highly coordinated organizations.
There is great potential to revolutionize research progress in civil engineering through collaboration,
sharing, visualization, and knowledge generation. IT infrastructure will eventually be established to
facilitate seamless collaboration and shared access to data. Access should be considered in the
broadest sense. It includes ability to obtain and analyze the data as well as visualization tools that
enable efficient understanding of the data.
Changing the emphasis toward collaborative interdisciplinary research requires a change in research
culture. We cannot underestimate the difficulty of changing a culture that is well-established and has
a successful history, even when there is broad consensus that the change is for the better.
There is a lot of incompletely understood physics associated with dynamics of particulate, multiphase
materials. The nonlinear wave propagation issue presented in this paper presents just one example. A
goal of NEES is to provide equipment and an environment in which we can make the fundamental
advances to make a real impact on understanding of the complex physics.
196
CONCLUSIONS
There are many problems involving nonlinear dynamic soil behavior that remain to be resolved. The
fact that 1-D nonlinear wave propagation in strain-stiffening soil can produce large acceleration
spikes, which can be modeled as shock waves, is an example of a basic phenomenon that has become
well understood by studying the results of model tests.
The NEES centrifuge at UC Davis, with its large size and versatile data acquisition network, provides
a facility for production of comprehensive data sets that enable description of model behavior as
opposed to sparse sampling of data at selected points. Hundreds of sensors can be used to enable
significantly improved resolution of the behavior and mechanisms that occur during shaking.
Geophysical tools are being developed to enable researchers to measure internal deformations and
strains.
A large centrifuge experiment using all of the resources described herein will be more than simply a
test. With the advanced instrumentation and large size of the experiments, the experiments should
be viewed as repeatable case histories. Due to the large quantity of information to be captured from
one model test sequence, theories and design procedures can be improved and tested. There is still
work to be done to train the profession to fully appreciate and to understand what may and what may
not be learned from the large-scale centrifuge model tests. Increased collaboration with practicing
engineers and development of software tools for simulation and intuitive visualizations of the data are
needed to improve the technology transfer from university to practice.
The Center for Geotechnical Modeling facilities are intended for use by researchers from academia
and industry, from both the US and abroad, as a shared-use resource. We have an all hands on deck
philosophy; one person cannot quickly operate a shaking table, all the sensors, tomography tools, as
well as the sequence of events being tested. Future experiments will be performed by teams of
researchers; with remote researchers able to control their experiments at UC Davis as easily as local
researchers.
ACKNOWLEDGEMENTS
Support for operations and maintenance of the centrifuge is provided by National Science Foundation
Award No. CMS - 0402490 provided via the Network for Earthquake Engineering Simulation
(NEES). The experimental work was made possible by funding of the Earthquake Engineering
Research Centers Program of the National Science Foundation, under award number EEC-9701568
through the Pacific Earthquake Engineering Research Center (PEER). Any opinions, findings, and
conclusions or recommendations expressed in this material are those of the author(s) and do not
necessarily reflect those of the National Science Foundation.
The facility development and work presented in this paper were made possible by the staff and
colleagues in the UC Davis Center for Geotechnical Modeling: Ross Boulanger, Lars Pedersen, Chad
Justice, Cypress Winters, Ray Gerhard, Peter Rojas, and Nick Sinikas.
REFERENCES
Bonilla, L. F., Archuleta, R. J. and Lavallee, D. (2005). Hysteretic and Dilatant Behavior of
Cohesionless Soils and Their Effects on Nonlinear Site Response: Field Data Observations and
Modeling, Bull. Seism. Soc. Am., 95, pp. 2373-2395.
Coelho, P. A. L. F., Haigh, S. K. and Gopal Madabhushi, S. P., (2004). Centrifuge modelling of
earthquake effects in uniform deposits of saturated sand, Fifth International Conference on Case
Histories in Geotechnical Engineering, April 13-17, 2004, New York, Vol. 1, pp. 8
197
Fiegel , G. L. and Kutter, B. L. (1994). Liquefaction-Induced Lateral Spreading of Mildly Sloping
Ground, J. Geotech. Engrg. 120, pp. 2236.
Frankel, A. D., D. L. Carver, and R. A. Williams (2002). Nonlinear and linear site response and basin
effects in Seattle for the M 6.8 Nisqually, Washington earthquake, Bull. Seism. Soc. Am., 92, pp.
20902109.
Holzer, T,L., and T.L. Youd (2007) Liquefaction. Ground Oscillation, and Soil Deformation at the
Wildlife Array, California, Bulletin of the Seismological Society of America, In Press.
Ilankatharan, M., Kutter, B.L., H. Shin, P. Arduino, S.L. Kramer, N. Johnson, and T. Sasaki. (2006).
Comparison of Centrifuge and 1g Shaking Table Models of Pile Supported Bridge Structure,
International Conference on Physical Modeling in Geotechnics, ISSMGE, Hong Kong, pp. 165 -
170
Ishihara, K. (1985). Stability of natural deposits during earth-quakes, Proc., 11th Int. Conf. on Soil
Mechanics and Foundation Engineering, Vol. 2, San Francisco, pp. 321376.
Kutter, B. L. (2006). Phenomena associated with undrained and partly drained dilatant soil,
International Conference on Physical Modeling in Geotechnics, ISSMGE, Hong Kong, pp. 165-
170.
Kutter B.L., Gajan, S.M., Manda, K.K. and Balakrishnan, A., (2004), Effects of layer thickness and
density on settlement and lateral spreading, ASCE Journal of Geotechnical and Geo-
environmental Engineering, Vol. 130, No.6, pp. 603-614.
Kutter B.L & Wilson D.W. (1999): De-liquefaction shock waves, Proceedings 7th U.S Japan
Workshop on earthquake resistant design of life line facilities and countermeasures against
liquefaction, pp. 295-310. http://cgm.engineering.ucdavis.edu/Publications/reports/shockpaper.pdf
Kutter, B.L. & Wilson, D.W. (2006) Physical Modelling of Dynamic Behavior of Soil-Foundation-
Superstructure Systems, International Journal for Physical Modeling in Geotechnics, Volume 6,
No. 1, pp. 1-12, 2006.
Kutter, B.L., Wilson, D.W., and Bardet, J.P. (2002). Metadata Structure for Geotechnical Physical
Model Tests, Proc. Int. Conf. On Physical Modelling in Geotechnics, St. Johns, Canada (Phillips,
Guo, Popescu, eds.), Balkema, Lisse, Netherlands, pp. 137-142.
Li, Z., Kutter, B. L., LaBrecque, D. and Versteeg, R., (2006). A New Electrode Switching System
(ESS) and a Scheme for Measurement of the Movement of Buried Objects, 6th International
Conference on Physical Modeling in Geotechnics 2006, Hong Kong, China.
Taboada, V.M. and Dobry, R. (1998). Centrifuge modeling of earthquake-induced lateral spreading,
J. Geotech. and Geoenvir. Engrg., 124, pp. 1195.
Weber, G.H., Schneider, M., Wilson, D.W., Hagen, H., Hamann, B., and Kutter, B.L. (2003).
Visualization of Experimental Earthquake Data. Visualization and Data Analysis 2003, Proc.
SPIE, (Boerner, K., Chen, P.C., Erbacher, R.F., Groehn, M. and Roberts, J.C., eds.), August, Vol.
5009, Bellingham, Washington.
Wilson, D.W., Boulanger, R.W., Feng, X., Hamann, B., Jeremic, B., Kutter, B.L., Ma, K.-L.,
Santamarina, C., Sprott, K.S., Velinsky, S.A., Weber, G., and Yoo, S.J.B. (2004). The NEES
Geotechnical Centrifuge at UC Davis, 13th World Conference on Earthquake Engineering,
Vancouver, B.C., Canada, August 1-6, Paper No. 2497.
Wilson, D.W. and Kutter, B.L. (2002). Dense Instrumentation Arrays, Proc. Int. Conf. On Physical
Modelling in Geotechnics, St. Johns, Canada (Phillips, Guo, Popescu, eds.), Balkema, Lisse,
Netherlands, pp. 131-136.
Wilson, D.W., Kutter, B., Ilankatharan, M., Robidart, C. (2007) The UC Davis High-Speed Wireless
Data Acquisition System, in press, International Symposium on Field Measurements in
Geomechanics, ASCE Geoinstitute, Boston, September.
Youd, T. L., and T. L. Holzer (1994). Piezometer performance at the wildlife liquefaction site,
California, J. Geotech. Eng., 120, no. 6, pp. 975995.
Zeghal, M., and A. W. Elgamal (1994). Analyses of site liquefaction using earthquake records, J.
Geotech. Eng., 120, no. 6, pp. 996-1017.
198

W1-1011
199

200

201

202

203

204

205

206

207

208

209

210
4 th INTERNATIONAL CONFERENCE ON EARTHQUAKE GEOTECHNICAL ENGINEERING
Thessaloniki, Greece 25-28 June 2007 4th INTERNATIONAL CONFERENCE ON EARTHQUAKE GEOTECHNICAL
ENGINEERING
Thessaloniki, Greece 25-28 June 2007
15 YEARS of EUROSEISTEST
K. PITILAKIS (Project Coordinator)
D. RAPTAKIS, K. MAKRA, F. CHAVEZ-GARCIA (Site Effects)
M. MANAKOU, P. APOSTOLIDIS (3D Soil Model)
G. MANOS (Structural Studies)

Thessaloniki, Greece 25-28 June 2007 GENERAL PRESENTATION
European Projects:
EUROSEIS-TEST (EV5V-CT.93-0281)
EUROSEISTEST EUROSEIS-MOD (ENV4-CT.96-0255)
EUROSEIS-RISK (EVG1-CT-2001-00040)
Euroseis-Test site
Thessaloniki
Urban Area
N
Area: 60 km2 Volvi Lake
Stivos
Profitis EuroseisTest site
211
Thessaloniki, Greece 25-28 June 2007 SEISMICITY STRESS FIELD
Recording period 1973-2002, M>3 Geodetical surveys
(Geophysical laboratory AUTH)
Permanent Seismological network
Focal mechanism and stress field Recording period 1981-2003

(GALANIS et al., 2004)
seismotectonic model
(VAMVAKARIS et al., 2004)

Thessaloniki, Greece 25-28 June 2007 VOLVI EARTHQUAKE 20/6/1978, M w=6.5
The only record until the 1993

Volvi earthquake 20/6/1978, Mw=6.5
200 MN' 78 (L) AF '78 (L)

100
0
-100
-200 A single record for the time period 1978-1993
200 40
MN '78 (T) AF '78 (T)
100
0
20
-100
-200
200 0
MN '78 (V) AF '78 (V)
100
0
-20
-100
-200
-40
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 0 4 8 12 16 20
TIME (sec) TIME (sec)
Time (sec)
212
Thessaloniki, Greece 25-28 June 2007 GEOPHYSICAL GEOTECHNICAL EXPERIMENTS
In situ measurements
Sampling boreholes
In situ geotechnical surveys (NSPT, CPT)
Seismic surveys for P-wave refraction and Seismic Wave Inversion (SWI)
Geoelectrical tomography and soundings
Cross-hole and Down-hole surveys
Logging measurements
Microtremor array measurements (SPAC method)
Laboratory measurements
Conventional geotechnical tests
Resonant column
Triaxial
Oedometer
Cyclic triaxial tests

Thessaloniki, Greece 25-28 June 2007 FIRST EXPERIMENTS 1993 - 1995
EUROSEIS-TEST project (1993-1995)
12 km surface measurements -1 km borehole measurements

Refraction - Surface wave inversion microtremors Cross-hole Down-hole
before starting project

250 STI
200
150 PRO
100 TST
50
0
-50
-100
-150
-200
-16500 -16000 -15500 -15000 -14500 -14000 -13500

DISTANCE (m)
250 after the end STI

200
150 PRO
100 TST A
50 B
C
0 D
-50 E
-100 F
-150 G*
G
-200
-16500 -16000 -15500 -15000 -14500 -14000 -13500
DISTANCE (m)
213
Thessaloniki, Greece 25-28 June 2007 GEOPHYSICAL GEOTECHNICAL RESULTS 1993-1995
S14 S7 S6 S5 S4 S3 S2 S1 S8 S9 S10 S11 S13 S15 S12
0 2 2 4 4 4 4 2. 5 2. 5 2
5 5 5 5 4. 5 4. 5 5 5 5
8
10 10 10 10 10 10 10 10
SOIL 12 12 12 12 12
USCS SOIL BODY WAVES N30-SPT 15 15 15 15 17

21
15 15 15 15
20 20
15
20
TYPE DESCRIPTION VELOCITIES (km/sec) 25 25 25 25
30 30
0.0 0.5 1.0 1.5 0 20 40 60 80 100 35 35 37
0 SM-SC 40 40 40 40 40 40 40
A Silty Clay Sand 370 m/s 50

45 45 45 45 45
50
45 45
CL
ML-CL Vp 55 55 55 55 55 55 55 5 5 5 5 55 55
Silty Sand
10 and 1450 m/s 70
65 65 65 65 65 6 5 65 6655 65
65 6655
SM B 75 75 75 75 75 75 75
SC Sandy Clay Vp 80 80 80
85 85 85 85 8 5 8 5
20 CL Grey - Black 90 90
95 95
100 Vs (m/ s)
ML 105/25 cm
30 Soi l Type
115 11 5 115 115 95-150
SM 50/7cm 130
125
150-250
CL Silt - Silty Sand SM . SM -SC, M L 135 13 5 1 35
250-400
40 C Sandy Clay Qs 70/15cm
70/15cm CL 145 145
SM 70/15cm 150
Light Brown 106 SM 400-550
160 160
SM -SG 560-650
50 CL 165 165
17 0
165
170
16 5
170
SM -M L 600-900
70/ 13cm 180 180
CL-M L 185 18 5
1000-1700
60 ML-CL CH
190190 190
195 195
190
19 5 19 5
190190 190
Clayey Silt 200 20 0 200 200 200 200 200

2300-2600
CL s tiff
D and 1650 m/s 70/13cm
70 Marly Clay 118
SC-CL g ravel s
CH Grey - Green Wheat her Rock
Rock
80
250
300
90 SM-SC
FORM ATIONS A B C D E F G* G
Silty Clay Vs 250 Vp (m /s) 330 450 550 - - - - -
100 with sand Vpw (m/ s) - 1500 1600 2000 2500 2600 3500 4500
E 130 200 300 450 650 800 1250
and NNW Vs (m/s) 2600 SSE
Qs 15 20 30 40 200
110 CL gravels 200
Brown to Grey d (t/ m 3) 2.05 2.15 2.0, 2.15 2.10 2.15 2.20 2.5 2.6
93/25cm
120 SM-CL Vp 150
94/25cm C
130 ML-SM Silty-clayey Sand
B G*
100
G
140 CL C
Sandy Clay 90/15cm C B A
or 50 E C B D
150 CH Silty Clay F C E
2700 m/s G* D
with E F
F gravels 0 G
160 ML D G*
F
100/8cm
Clayey-silty Sand G
170 SM-ML -50 E
G*
180 SC CPT F4
G
-100 F F1
Vp
190 SCHIST G* Vs (RC) 3100 m/s G* WEATHERED ROCK
Weathered schist *V : V P BELOW WATER TABLE ( ) G*
Vp -150 PW
Qs (RC) ** 2.0 t/m : CENTRAL PART
3 G F2
200 GNEISS G Gneiss 4900 m/s F3
2.15 t/m3 : EDGES
0 10 20 30 40 50 60 0 60 120 180 240 300
3D ACCEL EROMETERS qc-CPT (kgr/cm 2)
QUALITY FACTOR (Qs) -200 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 5.5
DISTAN CE (Km)
(RAPTAKIS et al., 2000)

Thessaloniki, Greece 25-28 June 2007 FIELD EXPERIMENTS 1996 - 2005
EUROSEIS-MOD project (1996-1999)
sensor
EUROSEIS-RISK project (2002-2005)
214
Thessaloniki, Greece 25-28 June 2007 GEOPHYSICAL RESULTS 1996-2005
R3
2 R2
R1
1
o
12 0
3 4
1.1
Experimental dispersion curves GER1

GER2
GER3
1
Phase velocity Rayleigh waves SCH
LAG1
LAG2
0.9
BRG
LIM2 Inverted Vs profile
1
2
EVA
0.8 VATH
PER2
DA B2
MAN
0.7 ANA
THE
FTE
TRI
FRM
0.6
TOPO
VOL
FID
VRO
0.5 KAP
MES
0.4
0.3
0.2
0.1
1 2 3 4 5 6 7 8
(Hz)
MANAKOU, 2007 (PhD THESIS)

Thessaloniki, Greece 25-28 June 2007 GEOPHYSICAL RESULTS 1996-2005
60 m
130 m
I
S P2
120 m
S P1 BRG
120 m
II
SP3
Body
4
Su
wave s
rfa
ce
wa
ve
s
(RAPTAKIS et al., 2005)
215
+
Vs: 130-275m/sec
Thickness (m)
0-30
30-60
N
400
BA
400
60-90
350 350 90-120
300 300
250
200
250
200
C+D 120-150
150-180
75 76
150 150
NIK1
79 S1L 91
90 DD2
Vs: 280-450m/sec 180-210
De pth (m )
100 100
TST TSTT OPO VRO
TRI
50 50
A+B
0 C+D 0 210-240
-5 0 Rock E+F Rock -5 0
-1 0 0 -1 0 0 240-270
-1 5 0 -1 5 0
-2 0 0
Rock
-2 0 0
270-300
-2 5 0 -2 5 0
F- GNSP F- Sx
-3 0 0 -3 0 0
-2 4 00 0 -2 3 00 0 -2 2 00 0 -2 10 0 0 -20 0 0 0 - 19 0 0 0 -1 8 0 0 0 -1 7 0 00 -1 6 0 00 -1 5 00 0 -1 40 0 0 -13 0 0 0 -12 0 0 0 - 11 0 0 0 -1 0 0 0 0
Dista nce (m )
ABA
2 00
1 50
geo -2003
20 0
15 0
E+F
MES
1 00 L AG BRG -
BRG S3 a rtBRG exp- 2003 ANA THE DAB2 geo -2004 10 0
KO K TOPO TSE PED ATHATH exp- 2004
L EF ANA ANA L OS THE S1LTST 85 FID EXP VOL
Vs: 500-800m/sec
TST TRI 8 4 DA B G GEO
50 A+B 50
0 C+D 0
-5 0 -50
E+F
-10 0 -10 0
-15 0 -15 0
-20 0 -20 0
Rock
-25 0 -25 0
-30 0 -30 0
-35 0 -35 0
F-Sx
-40 0 -40 0
-2 15 00 -21 00 0 -2 05 00 -20 00 0 -1 95 00 -19 00 0 -1 85 00 -18 00 0 -1 75 00 -17 00 0 -1 65 00 -16 00 0 -1 55 0 0 -15 0 00 -1 4 50 0 -1 40 00 -13 50 0 -1 30 00 -12 50 0 -1 20 00 -11 50 0 -1 10 00 -10 50 0 -1 00 00
Distance (m )

3D GEOLOGICAL STRUCTURE FOR SITE RESPONSE STUDIES
RELIEF
AB/CD INTERFACE
CD/EF INTERFACE
EF/BEDROCK INTERFACE
BEDROCK TOP SURFACE
216
Thessaloniki, Greece 25-28 June 2007 SITE RESPONSE SITE EFFECTS STUDIES
SITE RESPONSE SITE EFFECTS STUDIES
To investigate complex site effects (2D and 3D soil structures, differential

motion due to edges effects, lateral variation)
To develop models and tools for ground motion estimates combining
source, path and site effects. Experimental study on azimuthal
dependencies and directivity effect
Experimental evaluation of soil non-linearities using records from DH array
To study liquefaction phenomena (if any). Evaluation of existing
methodologies for liquefaction susceptibility and risk
To study the spatial variability of ground motion, to identify the relative
importance of the parameters involved and to validate the existing methods
for loss of coherency estimation.
To evaluate the reliability of the empirical techniques used for site
amplification estimation
To improve the existing numerical and analytical modeling techniques as
well as to develop new simulation methods and software.
To contribute to the more explicit definition of site effects for the design
earthquakes and spectra in Eurocode 8.

Thessaloniki, Greece 25-28 June 2007 INSTRUMENTATION PERMANENT ARRAY
Guralp PRO
K2 ROA
Guralp!
K2 GRA
WT3 WT2 WT1 TST ET1 ET2 ET3
Etna Etna Etna Etna Etna Etna
Etna FRM
Guralp STH
Guralp STC
Etna STE
217
Thessaloniki, Greece 25-28 June 2007 INSTRUMENTATION PERMANENT ARRAY
NNW
PRO
GRA
PRO-33m
BRG KOK ANA TST TSE ATH DAB
W E
FRM
TST-21m
BUT
TST-40m
STC
STI
TST-72m
TST-136m SSE
TST-197m Out o f sca le
MORE THAN 300 EVENTS

No EventDate EventTime Magnitude EcoordN EcoordE Depth
68 27/12/1998 10:14:25 3.1 40.723 23.277 9.4
69 17/8/1999 0:01:37 7.4 40.76 29.97 18
70 12/12/1999 19:25:58 4.5 40.554 23.646 10
71 31/1/2000 9:56:12 3.9 40.636 23.638 10
72 12/3/2000 8:32:42 3.3 40.625 23.153 2
73 10/6/2000 18:35:56 2.7 40.713 23.349 9
74 16/12/2000 2:49:52 2.8 40.595 23.428
75 13/1/2001 0:42:24 3.9 40.67 23.47 1
76 26/7/2001 0:21:39 6.4 39.05 24.35 19
77 10/8/2001 21:49:39 4.8 40.71 23.45 5
78 8/10/2001 4:50:21 4.4 40.64 23.16 11
79 8/10/2001 5:32:17 4.4 40.45 23.08 3
80 8/10/2001 6:55:30 3.5 40.52 23.08 5
81 8/10/2001 8:00:27 3.7 40.54 23.13 5
82 29/10/2003 21:15:48 4.7 40.71 22.83 25
83 27/3/2004 10:39:26 * * * *
84 8/6/2004 22:38:23 * * * *
85 15/7/2004 0:40:31 * * * *
86 15/7/2004 4:12:15 * * * *
* No information yet available

STE STC
10
ARNE AS' EARTHQUAKE 1997, MAY 4, M=5.8
1 SSR HORIZ
SSR VERT
HVS R
FRM TST
10
1
300
250 NN W FORMATION A B C D E F G* G SS E
Vp (m/s) 330 450 550 - - - - -
V pw (m/s) - 1500 1600 200025002600 35004500 STE
GRB GRA
Vs (m/s) 130 200 300 450 650 800 12502600 ST2 2
200 10
D ensity (t/m 3) 2.05 2.15 2.0, 2.15 2.10 2.15 2.20 2.5 2.6
S TC
150
PR O
C
F RM B D
100
G 1
C
G RA G RB
C B TST
A
50 E C D
TS17 B
F E
G D C
0 E F 1 10
TS72 PRO
F D G* 10 F REQ UEN CY (Hz)
-50 G E G
F4
F Figure 5. Comparison of the results
-100 F1
obtained from the HVSR
3D ACCELEROM ETERS technique (thin line), SSR
*V 1 for horizontal (thick line)
PW : V P BELOW WA TER TABLE ( ) G*
-150 and for vertical components
** 2.0 t/m 3 : CENTRAL PA RT F2 (dashed line)
G
2.15 t/m 3 : EDGES F3
G * WEATHERED RO CK
-200
0.00 0 .25 0.5 0 0.75 1.00 1.2 5 1.50 1.75 2.00 2.25 2.50 2.75 3.00 3.25 3 .50 3.7 5 4.00 4 .25 4.5 0 4.75 5.00 5.2 5 5.50 5.75
D ISTA NCE (Km) 1 10
F REQ UEN CY (Hz)
RAPTAKIS et al., 1998
218
RADIAL COMPONENT TRANSVERSAL COMPONENT

40 40
36 36
32 32
28 28
24 24
20 20
16 16
12 12
8 8
4 4
0 0
3.0 3.0
2.5 2.5
2.0 2.0
1.5 1.5
1.0 1.0
0.5 0.5
0.0 0.0
300 300
250 250
200 200
150 150
100 100
50 50
0 0
-50 -50
-100 -100
-150 -150
-200 -200
-250 -250
-300 -300
0 500 1 000 1500 2000 2500 3 00 0 3500 4000 4500 50 00 5500 6000 0 500 1 000 1500 2000 2500 3 00 0 3500 4000 4500 50 00 5500 6000
DISTANCE (m) DISTANCE (m)
RAPTAKIS et al., 2000

PRO GRA GRB TST FRM ST C STE
0.0 1.0 2 .0 3.0 4.0 5.0

Distance (km)
CHAVEZ-GARCIA et al., 2000
219
2D PARAMETRIC ANALYSES
Our results suggest that, in order

to improve current schemes to take
into account site effects in building
codes, the more to be gained
comes from consideration of lateral
heterogeneity.
MAKRA et al., 2005

T1 0.40 sec T2 2. 0sec

10.0 3 T2
3 2D/1D Aggravation factor 3.720 synthetic (2D & 1D)
T A 1 2 accelerograms and their
A 1 2 T
T1 response spectra
1.0
Sa(2D) A Sa(1D)
0.1
0.1 T1 1.0 T2
period (sec)
STE STC 6.0
10.0 10.0
5.3
4.5
1.0 1.0 3.8
REC/1D
3.0
2D/1D
0.1 0.1 2.3
FRM TST 1.5
10.0 10.0
0.8
0.0
1.0 1.0
0.1 0.1
GRB GRA
10.0 10.0
1.0 1.0
MAKRA, 2000 (PhD THESIS)

0.1 0.1
0.1 1.0 0.1 1.0
Period (sec) Period (sec)
220

Thessaloniki, Greece 25-28 June 2007 STRUCTURAL BEHAVIOUR & SSI EFFECTS
EXPERIMENTAL AND NUMERICAL

INVESTIGATION OF SSI EFFECTS
Prof. George C. Manos
221
Thessaloniki, Greece 25-28 June 2007 FACILITIES FOR TESTING OF STRUCTURAL MODELS
Permanently instrumented field site for SSI effects
React ion Mass

Store 3.5
House
1.5
6.46 3.6
25,0m
3,38
4.2
Moving Crane Facility

2.52
A A
7,0
Max. Capacity 20 KNt React ion Mass

2,0
2.95
9,80
2.46
9.2
3.15 1.5 6,0 3.5

PIER M ODEL
1.52
2,0
3.36
2,00
5,82
3.5
3.5
4,30
5-Story M odel
7,54m 5,00 5,00m
10m 8,25m
7,0
2 00 3 6 My
4 .0 OVERTURNING MOMENT EURO SEIS- TEST
3 .5
3 .0
LAY O UT OF EUROSEIS T EST SITE
Ut ili ty
Room
2 .5
2 .0
1 .5
1 .0
0 .5
Movin g Crane Facility
0 .0
Max. Capacity 20 KNt SECTION A - A -0 .5 0 .0 0 .5 1 .0 1 .5 2 .0 2 .5 3 .0 3 .5 4 .0 4 .5 5 .0 5 .5 6 .0
-1 .0 t [se c]
Store House 5-Story M ode l -1 .5
Total weight = 153,6t
PIER M ODEL -2 .0
0.4
4,30m -2 .5
9,00m
-3 .0
6,50m
4,00m
-3 .5
3,70
3,30
-4 .0
React ion Mass
0,35
7,10 3,35 2,5
10,0 8,25 6,75

Study of SSI effects on model structures at Volvi European Test Site

Euro-Risk Project
Coordinator: Prof. George C. Manos
http://euroseis.civil.auth.gr
6th Storey Model
222
STRUCTURES OF TEST SITE
5storey building
Pier bridge
model
me an
9.524E-03

TYPES OF
ANALYSES
2. Time domain
ADINA 1.67 2.68 3.16
3. Various special [Hz]
FFTrec 1.65 2.77 3.28
analysis
[Hz]
Pier material: elastic linear

Vertical component
E=32GPa , v=0.3 , =2.5t/m3
Soil material: elastic linear

E=97.14MPa , v=0.3 ,
=2.05t/m3 , Vs=135m/sec
223
Thessaloniki, Greece 25-28 June 2007 REFFERENCES
Chvez-Garca, F.J., D. Raptakis, K. Makra, and K. Pitilakis (2000). Site effects at Euroseistest II. Results from 2D numerical
modeling and comparison with observations, Soil Dyn. Earthq. Engrg. 19, 23-39.
Kudo K., T..Kanno, H. Okada, T. Sasatani, N. Morikawa, P. Apostollidis, K. Pitilakis, D.Raptakis, M. Takahashi, S. Ling,
H.Nagumo, K. Irikura, S. Higashi & K. Yoshida, 2002. S-Wave Velocity Structure at EURO-SEISTEST, Volvi, Greece Determined by
the Spatial Auto-Correlation Method applied for Array Records of Microtremors, Proc. 11th Japan Earthquake Engineering Symposium,
Paper No. 62.
Makra, K., D. Raptakis, F.J. Chvez-Garca, and K. Pitilakis (2001). Site effects and design provisions: the case of Euroseistest,
PAGEOPH 158, 2349-2367.
Makra, K., D. Raptakis, F.J. Chavez-Garcia & K. Pitilakis, 2002. How important is the detailed knowledge of a 2D soil structure for
site response evaluation, Proc. 12th ECEE, Paper No 682.
Makra K., F.J. Chvez-Garca, D. Raptakis & K. Pitilakis, 2005. Parametric analysis of the seismic response of a 2D sedimentary
valley: Implications for code implementations of complex site effects. SDEE 25, pp. 303-315.
Manakou (2007). Contribution to the determination of a three dimensional soil model for studying seismic response: application to the
Mygdonian sedimentary basin. Ph.D. Thesis (in Greek), Civil Engineering Department, Aristotle University of Thessaloniki.
Manos C. (Presentation by V. Renda on behalf of Prof. G. Manos), 2004. Experimental and Numerical Studies of Model Structures at the
Volvi European Test Site. OECD/NEA International Workshop "Seismic Input Motions, Incorporating Recent Geological Studies,
OECD/NEA Workshop proceedings.
Manos, G., Kourtidis, V., Soulis, V., Sextos, A., Renault, P. and Yassin, B., 2005. Numerical and experimental soil-structure-
interaction of a bridge pier model at the Volvi-Greece European Test Site. 6th European Conference on Structural Dynamics, Eurodyn
2005, Paris, II, pp. 1335-1340.
Manos, G.C., Renault, P. & Sextos, A.G., 2005. Investigation of the interaction between neighboring model structures at the
Euroseistest site. Proceedings: EURODYN, pp. 1297-1302.
Pitilakis, K., D. Raptakis, K. Lontzetidis, Th. Tika-Vassilikou, and D. Jongmans (1999). Geotechnical and geophysical description of
EURO-SEISTEST, using field, laboratory tests and moderate strong motion recordings. J. Earthq. Engrg. 3, 381-409.
Raptakis, D., N. Theodulidis, and K. Pitilakis (1998). Data analysis of the EURO-SEISTEST strong motion array in Volvi (Greece):
Standard and horizontal to vertical spectral ratio techniques. Earthq. Spectra 14, 203-224.
Raptakis, D., F.J. Chvez-Garca, K. Makra, and K. Pitilakis (2000). Site effects at Euroseistest I. Determination of the valley
structure and confrontation of observations with 1D analysis, Soil Dyn. Earthq. Engrg. 19, 1-22.
Raptakis, D., M. Manakou, F.J.Chvez-Garca, K.Makra, and K. Pitilakis, 2005. 3D Configuration of Mygdonian basin and
preliminary estimate of its site response, SDEE 25, pp. 871-887.
224

W1-1008
225

226

227

228

229

230

231

232

233
June 25-28, 2007
Paper No. W1-1009
INSTRUMENTED GEOTECHNICAL SITES: CURRENT AND FUTURE

TRENDS
Jamison H. Steidl 1
ABSTRACT
A goal of engineering seismology research is to generate analytical and empirical models for accurate
prediction of ground shaking, pore water pressure generation, ground deformation and soil-
foundation-structure interaction (SFSI), and to help engineers understand how these predictions will
affect the built environment. The development of simulation capabilities that can reproduce these
effects at various strain levels requires well-instrumented test sites where actual ground response, pore
pressure, and deformation can be monitored during earthquake shaking to provide benchmark case
histories for verification of the simulation models. In the U.S. alone, there are many extensive
geotechnical strong motion array facilities available for use in calibration and validation of our
modeling techniques. An update on these facilities including recently deployed arrays, and a summary
of the current research activities using these facilities will be presented. In particular, the experimental
field site facility that is part of the National Science Foundations George E. Brown Jr. Network for
Earthquake Engineering Simulation (NEES) will be highlighted. This facility includes two
permanently instrumented field sites for the study of ground response, ground failure, soil-foundation-
structure interaction, and liquefaction. The current and future trend for instrumented sites seems to be
moving to the simultaneous monitoring of both the geotechnical and structural components at a given
site. This integration of the two sub-disciplines within earthquake engineering provides opportunities
for new collaborations. The performance analysis of instrumented structures incorporating
geotechnical and structural aspects should provide advances in our ability to predict the effects of
earthquakes on the built environment.
Keywords: geotechnical strong motion array, engineering seismology, site response, test sites.
INTRODUCTION
Over the last two decades there has been significant activity in terms of the construction and operation
of instrumented geotechnical sites. This summary paper is a continuation of the work presented almost
decade ago (Archuleta and Steidl, 1998) that outlined the results of borehole array studies in the
United States. Two issues come to mind when thinking about the last two decades of activity with
instrumented geotechnical sites. First, the need for increased coordination among the agencies and
organizations that install, maintain, and disseminate the data form these instrumented sites. This has
been repeatedly expressed in the proceedings and reports from various workshops and conferences.
Second is the lack of collaboration between the structural and geotechnical engineering communities
to improve the design and performance of structures by integrating the components of the problem
from both disciplines. Until recently, these components had been treated separately.
This workshop contribution will begin with a discussion of the geotechnical strong motion array
(GSMA) activities in the United States, including a list of what is thought to be the current operational
sites broken into categories. Then some examples of the current experimental activities using these
1
Institute for Crustal Studies, University of California at Santa Barbara, California, USA, Email:
steidl@crustal.ucsb.edu
234
sites will be provided. Lastly, some discussion of the future opportunities and challenges that the
community faces in trying to deal with the two issues of coordination and collaboration mentioned
above.
EXISTING GEOTECHNICAL STRONG MOTION ARRAYS IN THE UNITED STATES
In taking a look at the Geotechnical Strong Motion Array (GSMA) sites in the United States, the first
question to ask is how to define a GSMA in the first place. In this case, any strong motion station that
contains both a surface and a sub-surface borehole sensor qualifies. However, the GSMAs are further
sub-divided into three categories.
Extensive Arrays (4+ borehole sensors):

An extensive GSMA is a site with four or more borehole accelerometers in addition to the surface
sensor. Using this definition, 14 such sites are found. These sites are operated by a mix of agencies
and institutions, including; the United States Geological Survey (USGS), the California Geological
Survey (CGS/CSMIP), California Department of Transportation (Caltrans), the National Science
Foundation (NSF), and the University of California at Santa Barbara (UCSB).
The site name, region, and responsible agencies are listed for each of these extensive array sites in
Table 1. Note that these sites are primarily in California, with the exception of one site in Anchorage,
Alaska. This is primarily reflective of the high cost of installing and operating extensive GSMAs and
the probability of obtaining useful strong motion data over the life span of the array. The high
seismicity regions of the United States have historically been the most heavily instrumented, though
this is changing somewhat with the new Advanced National Seismic System (ANSS). This is not to
say that having strong motion records in the lower seismicity regions like the Central and Eastern
United States is not important, however, the potential to catch a significant event does play into the
funding agencies decisions regarding where to allocate resources for these extensive facilities.
Table 1. Extensive Array sites in the United States

Site Name Location Agency(s)
Bessie Charmichael School Bay Area Northern California USGS
Borrego Valley Southern California UCSB
Delaney Park Anchorage, Alaska ANSS/USGS/UAF
Embarcadero Plaza Bay Area Northern California USGS
Eureka Array Northern California CGS/CSMIP/Caltrans
Garner Valley Southern California UCSB/NEES
Hayward, San Mateo Bridge Northern California CGS/CSMIP/Caltrans
Hollister Observatory Central California UCSB
Levi Plaza Northern California USGS
Melloland Array El Centro Southern California CGS/CSMIP/Caltrans
San Diego Coronado Bridge Southern California CGS/CSMIP/Caltrans
Treasure Island NGES site Bay Area Northern California CGS/CSMIP/NSF
Vincent Thomas Array Southern California CGS/CSMIP/Caltrans
Wildlife Liquefaction Array Southern California UCSB/NEES
As an example of an extensive array we show the recently deployed Anchorage ANSS Delaney Park
Array and instrumented structure, the Atwood Building, located 500 meters from the geotechnical
array. Figure 1 shows the location of Delaney Park in relation to the Atwood building in the skyline
behind the park. Figure 2 shows a schematic of the vertical array with seven 3-component
accelerometers located at depths from the surface down to 61 meters. The instrumentation of the
Atwood building is also shown in Figure 2. This is an ideal case where the input motion to a well-
instrumented structure is provided through a GSMA located so as to provide the free field input
motions to that structure. This instrumented site represents the new trend in collaborative geotechnical
and structural engineering monitoring sites.
235
Figure 1. The Delaney Park GSMA with the instrumented Atwood Building in the background
Figure 2. Schematic representation of the GSMA and Structural Array monitoring at Delaney
Park and the Atwood Building in Anchorage, Alaska.
236
Moderate Arrays (2-3 borehole sensors):
The second category of GSMA facilities in the United States is moderate arrays, or sites with two or
three borehole accelerometers in addition to a surface sensor. The list of 15 such arrays provided in
Table 2 is again heavily weighted towards California. However, looking at these less extensive (and
therefore less expensive) arrays, the lower seismicity regions begin to be represented here, with three
moderate array sites located in the Central and Eastern United States, in this case operated by the
University of Kentucky and Northeastern University, respectively. The same agencies and institutions
that operate the previously discussed extensive arrays operate the remaining twelve sites.
An example of the moderate class of GSMA is the Southern California Earthquake Centers (SCEC)
Long Beach Water Replenishment District (WRD) site in Southern California. This site has a
relatively deep borehole accelerometer at 350 meters, an intermediate depth accelerometer at 30
meters, and the surface accelerometer. This is a typical soft deep-basin site in Los Angeles with
engineering rock just barely being encountered at 350 meters depth.
Table 2. Moderate Array sites in the United States

Carquinez Bridge Northern California CGS/CSMIP/Caltrans
Corona Array Southern California CGS/CSMIP/Caltrans
Foster City, San Mateo Bridge Bay Area Northern California CGS/CSMIP/Caltrans
Half Moon Bay Northern California CGS/CSMIP/Caltrans
La Cienega Array Southern California CGS/CSMIP/Caltrans
Long Beach Water District Southern California UCSB/SCEC
Northeastern University Eastern US Northeastern University
Olmstead Locks and Dam Central US University of Kentucky/ACOE
Paducah, KY Central US University of Kentucky
Parkfield, Turkey Flat Central California CGS/CSMIP/Industry
Rohnert Park Northern California CGS/CSMIP/Caltrans
San Francisco Bay Bridge Bay Area Northern California CGS/CSMIP/Caltrans
Sassafras Ridge, KY Central US University of Kentucky
University of California, Riverside Southern California UC/SCEC
Winfield Scott School Northern California USGS
Surface Borehole Pair (1 borehole sensor):

The third category of GSMA sites in the United States is the Surface Borehole Pair. The list below of
17 such sites are again a combination of Central US and California sites, weighted heavily toward
Southern California (Table 3). The majority of these sites were targets of opportunity where an
existing strong motion station was being drilled and logged to collect geotechnical site
characterization data, and another agency (SCEC for example) would leverage the drilling and logging
costs to justify funding the installation of well casing in the hole and the installation of a borehole
sensor. Since many of these sites exist because site characterization was occurring at the station,
information regarding the soil profiles at these sites is available. Having good site characterization data
is critical for future analysis of strong motion records obtained at these GSMA sites.
Many of these sites are also collaborative with existing monitoring networks where much of the
infrastructure (power and communications) is already in place at the site for the existing strong motion
instrument. While these sites will not provide the extensive array multi-level detail within the soil
column needed to calibrate complex nonlinear simulation models, these sites are still extremely useful.
They provide the input motion at a depth where the material behavior during strong shaking in
significant earthquakes is expected to remain in the linear stress-strain regime. Thus, they are the
control motion that can at a minimum be used to evaluate the degree of nonlinear soil behavior as
observations are made at different excitation levels. Assuming that the behavior during small
earthquakes is linear throughout the soil column, changes in the site response transfer function from
237
borehole to surface as input motions increase with larger earthquakes can be interpreted as changes in
the material behavior (Tsuda and Steidl, 2006; Assimaki and Steidl, 2007). With enough of these
observations at various sites, over a range of input motions and site classifications, average site
response correction factors can be determined and used in seismic code provisions. Currently, these
factors are based primarily on theory, not empirical data. So while these surface-borehole pair sites
may not be extensive enough for some research applications, they do make an important contribution.
Table 3. Surface Borehole Pair Sites in the United States

Central Fire Station Southern California USGS/UCSB
Cerritos College Southern California USGS
Griffith Park Southern California UCSB/SCEC
Jensen Filtration Plant Southern California UCSB/SCEC/USGS
Kentucky Bend, KY Central US University of Kentucky
Mira Catalina School Southern California UCSB/SCEC
Obregon Park, Los Angeles Southern California CGS/CSMIP/UCSB
Pacific Park Plaza, Emeryville Bay Area Northern California USGS
Ridgely, TN Central US University of Kentucky
Rinaldi Substation Southern California UCSB/SCEC
Stone Canyon Southern California UCSB/SCEC
Superstition Mountain Southern California UCSB/SCEC/USGS
Tarzana, Ceder Hill Southern California CGS/CSMIP
University of California, Los Angeles Southern California USGS/ANSS
University of California, Santa Barbara Southern California UC/SCEC
University of California, San Diego Southern California UC
Wonderland Avenue School Southern California UCSB/SCEC
HIGHLIGHTS OF ACTIVITIES AT THE NEES SITES
The U.S. National Science Foundation George E. Brown Jr. Network for Earthquake Engineering
Simulation (NEES) program provides an unprecedented infrastructure for research and education,
consisting of networked and geographically distributed resources for experimentation, computation,
model-based simulation, data management, and communication. Rather than placing all of these
resources at a single location, NSF has distributed its resources among 15 equipment sites throughout
the US. To insure that all earthquake engineering researchers can effectively use this equipment, the
equipment sites are operated as shared-use facilities, and NEES will be implemented under a new
paradigm, as a network-enabled collaboratory. As such, members of the earthquake engineering
community are able to interact with one another, access unique, state-of-the-art instruments and
equipment, share data and computational resources, and retrieve information from digital libraries
without regard to geographical location.
One of these 15 NEES equipment sites is the NEES@UCSB facility, which includes two permanently
instrumented field sites constructed for the study of ground response, ground failure, soil-foundation-
structure interaction, and liquefaction. The two sites are located in Southern California. Both sites are
located close to major faults and have previous histories of recording ground motions and pore-water
pressures. They also have a history of site characterization studies, and both sites are underlain by soft,
liquefiable ground. These field sites are well suited for ambient noise studies, passive earthquake
monitoring, and active testing using mobile shakers.
Garner Valley Downhole Array

The NEES Garner Valley Downhole Array (GVDA) is located in southern California at a latitude of
33 40.127 north, and a longitude of 116 40.427 west. The instrumented site is located in a narrow
valley within the peninsular ranges batholith east of Hemet and southwest of Palm Springs, California.
This seismically active location is 7km from the San Jacinto Fault and 40 km from the San Andreas
238
Fault. The valley is 4-5 km wide at its widest and about 10 km long. The valley trends northwest-
southeast parallel to the major faults of southern California. The valley floor is at an elevation of 1310
m and the surrounding mountains reach heights slightly greater than 3,000 m. A panoramic view of the
GVDA field site is shown in Figure 4, taken at the completion of the NEES construction in Fall of
2004. The details of the geotechnical site conditions and instrumentation at the GVDA facility can be
found at the NEES@UCSB website (http://nees.ucsb.edu/), and in previous studies of the observations
from this site (Archuleta et al., 1992; Steidl et al., 1996; Bonilla et al., 2002).
Figure 4. The GVDA site in 2004 after the NEES program upgrade.
The NEES GVDA facility exemplifies the trend of instrumented sites moving to multi-disciplinary
collaborations between seismologists, geotechnical, and structural engineers. The reconfigurable
structure (Figure 4) constructed at the GVDA site is instrumented with pressure cells under the four
corners of the foundation, vertical displacement transducers on the four corners, accelerometers on the
corners, bottom slab, and top slab, and a rotational sensor on the bottom slab. In addition, a downhole
accelerometer and pore pressure transducer are installed below the foundation. The new structure is
intended for improving our understanding of soil-foundation-structure interaction (SFSI). Figure 5 is a
schematic of the structure and the different input forces that can be used in response testing.
Figure 5. The various input forces used to study soil-foundation-structure interaction at GVDA.
239
In figure 6, the locations of the various sensors installed on and beneath the SFSI test structure are
shown. In addition to the instrumented structure, the soil column at GVDA is heavily instrumented
with 6 additional downhole accelerometers (6, 15, 22, 50, 150, and 501 meter depths) and 4 additional
pore pressure transducers (6.1, 8.8, 10.1, and 12.4 meter depths).
Figure 5. Instrumentation at the GVDA SFSI Facility
The SFSI test structure instrumentation is designed to easily capture both rocking and torsional modes
of the structure. It was also designed to be re-configurable, so that the stiffness could be modified by
adding or removing bracing on any of the 4 sides. The mass of the structure can also be modified
through the addition of weight on the roof slab, or even the addition of a second story. A permanent
shaker is mounted under the roof slab, and can be operated remotely, providing an excellent tool for
teaching SFSI and structural dynamics concepts. The shaker is also used in research by exciting the
structure on a regular basis and comparing the response with environmental factors like soil saturation
and temperature. A weather station is installed at the GVDA site to provide rainfall data and
temperature data, and soil moisture probes are installed below the foundation of the structure.
Liquefaction array monitoring at GVDA

In addition to the SFSI test facility and ground response sensors at GVDA, pore pressure monitoring
in the near surface soil layers is providing new observations to help better understand the liquefaction
240
phenomena and nonlinear material behavior in general. The largest motions recorded so far at the
GVDA site are just at the level where the onset of nonlinear soil behavior might be expected, around
0.1 g peak ground acceleration. Observations from the liquefaction array sensors are beginning to
show the build up of pore pressure at this level, and then show the slow decay back to the background
level. A recent M5.1 earthquake near Anza, CA produced a quality set of observations showing this
behavior (Figure 7). Interestingly, the shallow transducers show increases in pore pressure during the
strongest shaking, while the deeper transducer seems to show an opposite effect. It is expected that
these observations will be used for many years to come as simulation techniques are tested and new
physics based models are proposed to model this behavior.
Surface Acceleration (PGA ~
Pore Pressure increase with S-wave and

slow recovery back to pre-event level
Figure 7. 150 seconds of surface ground acceleration and sub-surface pore pressure observations
from the 2005 M5.1 Anza event recorded at GVDA
Wildlife Liquefaction Array
The Wildlife Liquefaction Array (WLA) is located on the west bank of the Alamo River 13 km due
north of Brawley, California and 160 km due east of San Diego. The site is located in the Imperial
Wildlife Area, a California State game refuge. This region has been frequently shaken by earthquakes
with six earthquakes in the past 75 years generating liquefaction effects at or within 10 km of the
WLA site. Based on this history, there is high expectation that additional liquefaction-producing
earthquakes will shake the WLA site during the 10-year operational phase (2004-2014) of the NEES
program. Figure 8 is a view of the WLA site after construction was completed in Fall 2004.
Figure 8. The NEES WLA facility just after construction was completed in 2004.
241
The extensive instrumentation at this site includes 4 surface accelerometers, 6 downhole
accelerometers, 11 sub-surface pore pressure transducers, and numerous benchmarks and inclinometer
casings for monitoring lateral ground displacements. In this paper, we will focus on some of the
experimental activities that have taken place since its inception in 2004, and refer the reader to other
publications which contain detailed information on the instrumentation and geotechnical properties of
the site (Youd et. al., 2004; Youd et. al., 2007).
Both earthquakes and active testing using the NEES@UTA T-Rex mobile shaker have been used to
examine the response of the WLA site. In the late summer of 2005 the T-Rex shaker excited the
WLA site and provided a useful test of the system, as well as some provocative observations of pore
pressure during local shaking. Figure 9 shows the location of the shaker relative to the accelerometer
and pore pressure instruments at the WLA site.
Figure 9. The WLA instrumentation and T-Rex shaker location during testing in 2005.
The active shaking from the NEES@UTA shaker only lasts for approximately 10 seconds, however
the pore pressure signals that are generated by the active source last for minutes. Pore pressures at the
two transducers located closest to the shaker (P7 and P8 in Figure 10) show an immediate increase in
pore pressure and slow decay back to the pre-shake level that takes more than 20 minutes. Another
interesting observation is that just following the initial pulse, the deeper of the two closest transducers
(P8) also shows another slight pressure increase, which can be explained by a pressure wave generated
at the source and spreading out and propagating through the sand layer as it dissipates. All of the
transducers that are located at or near the top of the sand layer at the site (see Youd et al., 2004, 2007
for details of the soil layers) show an impulsive initial arrival, including the two located further from
the shaker (P4 and P1), with the amplitude of the pulse decreasing with both lateral and vertical
distance from the source. The time histories shown in Figure 10 are sorted by distance from the shaker
source with the top trace being the closest and bottom trace the furthest. Pore pressure continues to rise
for as much as three minutes after the shaking has stopped at the deeper and further transducers. Even
242
at 10 minutes after the source has stopped the pore pressures have still not dissipated back to the pre-
shake level.
Figure 10. The WLA instrumentation and T-Rex shaker location during testing in 2005.
These active source observations of pore pressure generation and slow dissipation are also seen when
earthquakes shake the site. The largest ground motions observed to data at the WLA site are from a
local magnitude 5.1 event located approximately 10 km from the site. In addition to the M5.1 event,
numerous M3 and M4 level earthquakes also generated very interesting observations at the site. The
largest ground accelerations from this swarm of events was only about 10% g, however even these
modest levels of shaking, the pore pressure response can clearly be seen in the observations. The
observations from the eight pore pressure transducers for a 1-hour period during this swarm of
earthquakes is plotted in Figure 11 from shallowest to deepest (top to bottom). Similar to the active
source testing, all of the transducers show a clear response to the earthquake activity.
In the earthquake swarm observations, the deeper transducers have a larger dynamic response to the
passage of the body waves from the events. All transducers show slight pore pressure increase and
slow recovery back to the pre-swarm level. The deeper transducers also have this increased pore
pressure, but its harder to see due to their larger dynamic response. The pore pressure increase during
the swarm is approximately 0.2-0.3 kPa, while the dynamic response of the deeper transducers is
almost an order of magnitude larger, at approximately 3-5 kPa peak to peak (Figure 11).
243
P4 - LQ1 2.7m
M4.4 M4.6
P1 - LQ2 2.9m
P7 - LQ3 3.4m
Shallow Pore Pressure Increase

and Slow Recovery
P3 - LQ4 3.7m
P4 - LQ5 4.4m
P8 - LQ6 4.4m
P2 - LQ7 6.4m
P5 - LQ8 6.4m
Figure 11. The WLA liquefaction array. Sixty seconds of data showing both dynamic response of
the pore pressure sensors and static pressure increase, with slow dissipation when excited by the
2005 Obsidian Buttes earthquake swarm.
In addition to the active source and earthquake monitoring at the WLA site, the benchmarks and
inclinometer casings are re-surveyed approximately once per year to obtain a baseline for lateral
displacements. The free face of the Alamo river bank located just meters from some of the
instrumentation, benchmarks, and casings should provide an excellent source for lateral spread activity
in the next large earthquake.
CONCLUDING REMARKS
In order to maintain a vital monitoring and research program at instrumented geotechnical sites, often
many different funding agencies must be involved, and a very broad scope of work must be
envisioned. Coordination and collaboration among seismologists, geotechnical and structural
engineers to establish instrumented sites that serve the engineering and scientific goals of each of these
disciplines represents the current trend for increasing the number of these facilities. It would be useful
to also get the practitioners involved, in both defining the scientific and engineering goals for
instrumented sites, and also to get instrumentation planning as part of the design phase for new
structures.
At present there is a lack of coordination among the various responsible organizations, both at the
domestic level in the United States as well as internationally, when it comes to the operations and
dissemination of data from instrumented geotechnical array sites worldwide. The maintenance and
operations of the existing instrumented geotechnical site resources and dissemination of their data
should be considered a high priority. There is a need for an umbrella organization or working group
244
with representatives from both the national and international agencies with monitoring programs to
facilitate the coordination and collaboration between these agencies.
The operations and maintenance of instrumented geotechnical sites, especially in regions of relatively
low seismicity, is often a difficult task for the responsible agency or organizations. The lack of
recordings of significant events over long time periods can make the funding agencies question the
benefit of maintaining these resources. An international organization that regularly re-evaluates the
scientific and engineering needs in conjunction with the current inventory of instrumented sites could
provide justification needed for agencies to obtain funding for maintaining these facilities.
AKNOWLEDGEMENTS
The Author would like to acknowledge the many agencies and individuals who have provided
information regarding the operational status of existing GSMA resources, both in the United States
and Internationally. The dedication and patience of the organizations that maintain these array sites
that serve the geotechnical earthquake engineering and seismological communities is very much
appreciated. The author would also like to acknowledge the many funding agencies and collaborators
that have been involved with GSMA monitoring at the University of California, Santa Barbara over
the last two decades. While many have come and gone, the longevity of these facilities would not have
been possible without them. These include the U.S. Nuclear Regulatory Commission, the U.S.
Geological Survey, the U.S. National Science Foundation, the French Commissariat lEnergie
Atomic, the Electrical Power Research Institute, Agbabian Associates, Kinemetrics Inc., the California
Department of Transportation, Kajima Corporation of Japan, and the Nuclear Power Engineering
Corporation of Japan. The GVDA and WLA sites are currently operated under contract to the National
Science Foundation as part of the George E. Brown Jr., Network for Earthquake engineering
Simulation, award number CMS-0402490. Without the support and cooperation of the Lake Hemet
Municipal Water District and the California Department of Fish and Game, the GSMA monitoring at
the GVDA and WLA sites would not be possible.
REFERENCES
Archuleta, R. J., S. H. Seale, P. V. Sangas, L. M. Baker, and S. T. Swain (1992). Garner Valley
downhole array of accelerometers: instrumentation and preliminary data analysis, Bull. Seism. Soc.
Am., 82, 1592-1621 (Correction, Bull. Seism. Soc. Am., 83, 2039).
Archuleta, R. J. and J. H. Steidl (1998). ESG studies in the United States: Results from borehole
arrays, The Effects of Surface Geology on Seismic Motion, Irikura, Kudo, Okada, and Sasatani
(eds), Balkema, Rotterdam, v 1, p.3-14.
Assimaki, D. and J. H. Steidl (2007). Inverse analysis of weak and strong motion borehole array data
from the M w7.0 Sanriku-Minami earthquake, Soil Dynamics and Earthquake Engineering, 27, p.
73-92.
Bonilla, L. F., J. H. Steidl, J-C. Gariel, and R. J. Archuleta (2002). Borehole response studies at the
Garner Valley downhole array, southern California, Bulletin of the Seismological Society of
America, 92, p. 3165-3179.
Steidl, J. H., A. G. Tumarkin, and R. J. Archuleta (1996). What is a reference site? Bulletin of the
Seismological Society of America, 86, pp.1733-1748
Tsuda, K., and J. H. Steidl (2006). Nonlinear site response from the 2003 and 2005 Miyagi-Oki
earthquakes, Earth, Planets, and Space, 58, 1593-1597.
Youd, T.L., J. H. Steidl, and R. L. Nigbor (2004), Lessons learned and need for instrumented
liquefaction sites, Soil Dynamics and Earthquake Engineering, vol. 24, Issues 9-10, p 639-646.
Youd, T.L., J. H. Steidl, and R. A. Steller (2007), Instrumentation of the Wildlife Liquefaction
Array, Proceedings of the 4th International Conference on Earthquake Geotechnical Engineering,
June 25-28 th, 2007, Thessaloniki, Greece.
245
June 25-28, 2007
Paper No. W1-1001
DENSE SEISMIC INSTRUMENTATION OF SMALL SOFT BASINS.
Bill (W.R.) STEPHENSON 1
ABSTRACT
The paper describes the evolution of dense arrays of seismographs on soft sites in New Zealand,
detailing the overall seismic response of each of the three most valuable of these arrays. The basins
had lateral dimensions of around 400 m and depths of a few tens of metres, and were densely
instrumented with three-component velocity seismographs in order to investigate extreme site
response to earthquake shaking. At each site a month of recording resulted in several tens of useful
earthquake records. These recordings were supplemented with geophysical and geotechnical
investigations.
Although all three basins were roughly equivalent in their size, depth and flexibility, their responses to
earthquakes were strikingly different. The Parkway basin shows several valley-wide frequencies,
which correlate with peaks in the Horizontal-to-Vertical Spectral Ratios of microtremors. These
valley-wide frequencies are associated with large amplifications. The methods used to elucidate the
wave-propagation basis of these amplifications are described, together with comments on international
cooperation, data sharing and future experiments.
Keywords: Basin Array, Resonance
INTRODUCTION
Recent rapid advances in technology have made it economically possible to operate dense
seismograph arrays in areas of soft soil where both earthquake-related damage, and the amplitude and
duration of earthquake shaking are known to be abnormally great, or are suspected to be abnormally
great. Various agencies have taken advantage of these advances, and have installed seismograph arrays
of various types in such soft areas. The arrays have varied in the type of instrumentation used, the
inter-station distance, and the duration of the field deployment. Both surface arrays and three-
dimensional arrays have been implemented, some being strictly of a temporary nature, and some being
more permanent. A summary of array locations and properties was produced by Zerva et al (2004).
Since the time of that summary other arrays have been implemented, notably the Cavola array
(Bordoni et al, 2006) in Italy.
The majority of the arrays discussed by Zerva et al (2004) do not fall within the compass of the
densely instrumented small soft basins that this paper will treat. Examples of the sort of experiment
which this paper will not treat are the Roma array in Mexico City (only three locations, and not within
a confined basin), the Grenoble array (a small part of a large basin) and the San Jose array in the
United States (wide sensor spacing, varying site conditions across the basin).
1
Senior Scientist, Hazards Group, GNS Science, Lower Hutt, New Zealand, Email:
Bill.Stephenson@gns.cri.nz
246
The use of the term small soft basins in the title of this paper is an attempt to restrict the scope of the
paper to situations where abrupt discontinuities in material properties in both the lateral and vertical
sense define a structure in which energy may be trapped. Small is intended to describe areas of less
than a few hundred metres diameter, and soft is intended to describe shear wave velocities of less
than a few hundred metres per second.
The prime purpose of this paper is to disseminate information about three seismometer arrays that
have been operated on small soft basins in New Zealand, and to encourage further use of the seismic
data that they have generated. The paper will also describe the analyses applied to the New Zealand
data. The approach to these analyses was conditioned by early observations that suggested that the
type behaviour to look for would be the coherent response at a single frequency of a limited area of
ground.
New Zealand workers have had a long involvement with small arrays, improving their performance as
better instrumentation became available and as experience was accumulated. Stephenson (1974)
described an early initiative set up in the Hutt Valley in which a trigger system sensing rock motion
sent a radio signal to frequency modulated tape recorders, commanding them to record the output of
one-second geophones. This system was short-lived because subsequent to preliminary recordings of
earthquakes being made, the minicomputer used to digitise the recordings became wholly dedicated to
another task and was hence unavailable. At the same time funding to expand the system was diverted
to other projects. Nevertheless this first array gave useful insights, stimulating thoughts that a north-
south rock signal could give rise to an east-west resonant soil response. This was formalised by
Stephenson (1972) with the suggestion that an appropriate impulse response for the soil would consist
of a sequence of two-by-two matrices, thus foreshadowing the current use of matrix transfer functions
(Benites, pers. comm.)
A second attempted array operation carried out at Pukehou in New Zealands North Island
(Stephenson & Barker, 1991) was of limited value because strong motion accelerographs were used,
with a consequent limitation on the size of useful earthquake. As with the earlier array the system was
triggered by a rock-derived radio command signal. Although the Pukehou array was in an area of high
seismic activity it was also in a flood-prone region, so that only two stations were operational (the
others having been withdrawn in response to rising water) when useable ground motion eventually
transpired. The single useable earthquake was recorded at one soil station and one rock station and it
provided a clear example of a directional resonance. With modern instrumentation and in the
absence of flooding the Pukehou site would be an ideal location because it is a small-scale limestone
basin filled with peat, thus being bounded on all sides and having a high impedance contrast. The two
peat layers had shear-wave velocities of 17 m/s and 34 m/s, while the limestone shear-wave velocity
probably was in excess of 1000 m/s and possibly was as much as 2000 m/s. The landowner has
expressed reluctance to allow further work because it would interfere with a cash crop.
Three subsequent arrays on other basins in New Zealand all yielded useful data. In each of these three
cases a basin with a strong valley-wide resonance was sought, but the Alfredton array (Stephenson,
1996) proved non-resonant and as a consequence the Parkway array (Stephenson, 2000) was
implemented. A casual inspection of initial seismograms from Parkway and their spectra suggested
that valley-wide resonance was absent, and planning therefore started for the Wainuiomata array
(Stephenson et al., 2002). The Wainuiomata valley was found to consist of several sub-systems, each
showing valley-wide resonance. In the meantime a more sophisticated analysis of Parkway data
suggested that valley-wide resonance was present, and that this justified a re-appraisal of the Parkway
seismograms. The Parkway array data has proven extraordinarily rich.
The locations and responses of these three arrays are discussed more fully in paper 1726 of this
conference. The behaviour of each basin was distinct, and was caused by details in the manner in
which shear wave velocity varied with depth. If conventional site evaluation criteria had been applied,
the three basins would have been deemed equivalent, having similar depths and similar shear wave
velocities at the surface.
247
COMPARATIVE RESPONSES OF NEW ZEALAND BASINS
Mean Standard Spectral Ratios (SSR) for all well recorded earthquakes at a site near the centre of each
of the three basin arrays are shown in Fig. 1. Averaging of spectra over all rock sites, and many
earthquakes, ensured that the SSR curves were robust.
This evaluation of SSR gives a clear idea of the different responses of the three basins upon which the
seismograph arrays were laid out. Standard Spectral Ratios are the ratios of spectra of earthquake
motions recorded on soil and on rock. In the case of the basins it was considered best to choose
representative stations near the basin centres so as to avoid basin edge effects.
30 10
(c) Wainuiomata (a) Alfredton
W03 A02
5
20 0
0 2 4 6 8 10
Frequency (Hz)
15
(b) Parkway
P06
10 10
0 0
0 2 4 6 8 10 0 2 4 6 8 10
Frequency (Hz) Frequency (Hz)
Figure 1. Mean values of Standard Spectral Ratio for a site near the centre of each basin array.
The dashed lines specify an SSR of 1 and the vertical scales are the same for each plot.
Alfredton basin amplifies earthquake shaking to a small extent, over a broad range of frequencies. This
is consistent with the shear wave velocities of the soil being low, but becoming progressively higher
with depth until an eventual low contrast with bedrock velocity occurs.
Parkway basin amplifies earthquake shaking in a complex resonant fashion, with shaking at several
frequencies being amplified to various extents. The frequencies being amplified are broadly consistent
with what is known about the shear wave velocity profile.
Each Wainuiomata sub-basin amplifies earthquake shaking in a resonant fashion, with extreme
amplifications at around 1 Hz. At these frequencies the shaking is greatly prolonged. At frequencies
above 7 Hz, shaking is attenuated, with the layer of soft soil functioning as a low pass filter. The
surface layer of soft soil, having an abrupt impedance contrast with deeper material is responsible for
this behaviour.
248
POSSIBLE FUTURE WORK
Although recent progress has been very encouraging, the lessons learned from existing arrays could be
applied when designing new arrays, and these new arrays could be improved even more by adopting
better technology. In addition, the implementation of proven techniques used for the analysis of data
from other existing arrays, or from future arrays, could be fruitful. For instance coherency studies were
carried out on Parkway array data (Liao et al, 2007) and could be carried out on data from other
arrays. The significance of coherency is that it can reveal relative displacements. The motion at
Parkway is dominated by fast-travelling waves which should maintain high coherency, whereas the
motion between sites at Alfredton is less related, and lower coherency (meaning greater relative
displacement) might be expected.
Other possible projects might implement the technique of filtering followed by progressive component
rotation and the calculation of wavenumber spectra, and the technique of forming real-time movies of
particle motion. Both of these techniques have been demonstrated at Parkway and proved most
fruitful.
The underlying reason for studying the earthquake-induced motion in small basins is to understand
why buildings founded on such basins may be subject to much greater damage than buildings on
nearby firm ground. It is noteworthy that on the one hand the most comprehensive arrays have been
operated in areas where no excess damage has been observed, and on the other hand that arrays in
areas associated with excess damage have been limited by small size or lack of dense coverage. Thus
to fully understand situations with excess damage it is necessary to properly instrument a restricted
location where such damage is known to occur.
An opportunity therefore exists to implement a new temporary dense seismograph array in the La
Molina alluvial basin near Lima, Peru. Successive damaging earthquakes in 1940, 1966 and 1974
showed that local soil conditions increased intensities from MMVI on nearby stiff soil to as much as
MMIX in La Molina, (Kuroiwa, 2002) and a recent as-yet unpublished analysis of a 1984 temporary
array deployment in La Molina shows that four of the seven soil stations show a peak in ground
motion at 1.2 Hz; the frequency that dominated the velocity response spectrum of an aftershock of the
1974 earthquake. This observation of a single frequency in a restricted area is reminiscent of the initial
observations at Parkway (Stephenson, 2000) that eventually led to the conclusion that a superposition
of modified Rayleigh waves and modified Love waves travelled down valley, guided by the valley
walls. At Parkway, classic Rayleigh waves were modified by a variation in amplitude across the
valley. For Love waves, the modifications likewise involved a variation across the valley, but also the
adoption of a vertical component near the valley walls in order to accommodate pure shear, while
simultaneously allowing the in-plane displacement to fall to zero at the valley walls. A dense array of
modern seismographs covering an area of 1 km diameter could resolve the behaviour of the soil within
the resonant area of La Molina, perhaps implying behaviour equivalent or similar to that at Parkway.
Arrays to date have often used recorders that proved to have inadequate resolution, or to be fixed gain
rather than gain-ranging, or to rely on triggered operation. In the case of the 1984 La Molina
deployment all three of these deficiencies were present. Any future deployment would best use at least
16-bit gain-ranging recorders and would record ground motion continuously, with useful data being
extracted later. This was the procedure followed in the 2004 Cavola array deployment. The Cavola
instrumentation would be an excellent bechmark to keep in mind when designing new arrays.
A feature of all three New Zealand array deployments was that they were insufficiently supported by
geotechnical observations. Because this was not an intended outcome there is a lesson to be learned. In
each case it was thought on reasonable grounds that the site consisted of horizontal layers, and that it
would be sufficient to characterise each layer using SCPT, and then to determine the extent of each
layer with sparse CPT probes. However the Parkway site had an impenetrable thin gravel layer at
depth, the Alfredton site was not built up from horizontal layers, and the Wainuiomata site was
actually a series of independent sub-basins. It is usual for the costs of an array project to be estimated
249
beforehand, and for the costs of additional work to be difficult to obtain. This was the case for these
projects, and only a special grant enabled characterisation at Parkway below the gravel layer. It would
be wise when pricing future arrays, to assume an absolute worst-case scenario when planning
geotechnical aspects.
A strength of all three New Zealand array deployments was that their planning included four rock
stations with the intent that the incoming wavefield would be adequately characterised. Yu and Haines
(2003) reported that the use of the average motion from four rock sites provided a usefully improved
reference signal.
DATA ANALYSIS
To date, approaches to the analysis of dense array data recorded by seismographs on soft basins have
mainly used techniques adapted from those used for arrays installed at homogeneous sites, where
simple phases, well separated in time, are expected. These analyses are often carried out within a
wavenumber space and address horizontal propagation. It is usually assumed that Love waves will be
associated with solely horizontal motion, Rayleigh waves with a preponderance of vertical motion,
and bulk waves with both vertical and horizontal motion. These assumptions may be true for wave
motion in a halfspace, but they are not borne out for a layered site in a horizontally-restricted context.
The response of a restricted volume of soil, often with its extent being greater than its depth, can be
very complex. Monochromatic waves at different frequencies, with different polarisations, and
travelling in different directions, will be superposed. The task of unravelling these contributions has
been relatively neglected, but the start that has been made at Parkway by filtering followed by a
sequence of component rotation and computation of wavenumber spectra, shows what can be done.
Summarising the procedures used at Parkway, Fig 2 shows the result of summing Parkway spectra
over all soil-based stations and all earthquakes. The peaks at 1.58 Hz and 1.68 Hz show the
importance of these frequencies, and subsequent attention was focussed on 1.58 Hz. A long duration
earthquake was then chosen so that meaningful narrow band filtering could be used.
Figure 2. Summed normalised spectra of all horizontal seismograms at Parkway, including both
components, for all soil sites and all earthquakes. There are site-wide responses at 1.58 Hz, 1.68
Hz and 2.1 Hz.
For this long duration earthquake, Figs 3a and 3b show wavenumber spectra for north components and
for vertical components, for motion between 1.53 Hz and 1.63 Hz. It is noteworthy that the implied
propagation for the north component is neither along nor orthogonal to, the north axis. Furthermore
250
the velocities for the vertical components and for the north components bear no relationship to each
other.
(a) (b)
Figure 3. Wavenumber spectra at Parkway for a long-duration earthquake. (a) This does not
correspond to either a transverse or a longitudinal wave. (b) Vertical components of soil stations
at 1.58 Hz. This bears no relationship to part (a).
However when the coordinate system is rotated to 30 and 120 (the axes natural to the basin), the
result is much more reasonable as shown in Fig 4a and Fig 4b. These figures show a transverse wave
travelling down valley at 1.3 km/s, and a longitudinal wave travelling down valley at 2.7 km/s. The
wavenumber spectrum of the vertical motion, however, does not correspond to the longitudinal wave
(imagined to be a Rayleigh wave).
(a) (b)
Figure 4. Wavenumber spectrum at Parkway for a long-duration earthquake. (a) 30

components of soil stations at 1.58 Hz; 2.7 km/s longitudinal wave. (b) 120 components of soil
stations at 1.58 Hz; 1.3 km/s transverse wave. Both waves travel down-valley (210).
When the vertical motion is partitioned between the longitudinal wave and the transverse wave as
described by Stephenson (2002) it is clear that the longitudinal wave is a Rayleigh wave guided by the
valley and that the transverse wave is similar to a Love wave except that in accommodating the
boundary conditions corresponding to a large impedance contrast at the valley wall, some vertical
251
particle motion is involved as indicated by Fig 5. This is further elaborated in paper 1726 of this
conference, and in Stephenson (2007).
Figure 5. Freeze-frame from a movie of postulated transverse wave, showing how the constraint
of the valley wall enforces some vertical motion. This cross-sectional view illustrates the
sloshing nature of the wave.
The most recent analysis technique applied to Parkway data used real-time movies of particle motion
as shown by Stephenson (2007). It clearly shows the nature of the propagating waves and also points
to other patterns of ground motion.
MODEL VERIFICATION
Workers who devote their efforts to understanding site response tend to fall into one of the two broad
camps of data analysis and of modelling, although there may sometimes be overlaps. Both approaches
are valid provided that the ultimate goal is attaining a level of understanding which allows future site
response to be predicted accurately, independent of the source earthquake. This understanding could
be considered complete when (given the incident wavefield) every nuance of every component, at
every station, for every earthquake, was reproduced in the modelling. At the moment this is not so. It
is common to correctly predict frequencies, but much harder to correctly predict durations.
The different analytical techniques described in the previous section all have a part to play in model
verification. The situation may be likened to viewing a scene through various colours of glass, with
each technique giving emphasis to a different aspect of ground motion. Only when a given model
agrees with the data as seen from all these viewpoints should it be considered as being accurate.
INTERNATIONAL COLLABORATION
The planning, funding and implementation of a temporary seismograph array of more than 20
recorders is a non-trivial task. Experience in New Zealand suggests that a five year gap between the
decision to run a dense seismograph array, and the final availability of both ground motion and
geotechnical data, is normal. This time includes initial data reduction plus geotechnical and
geophysical studies. The result is typically gigabytes of data which require considerable time and
effort to interpret.
It is easy to understand the perspective of a researcher who has invested five years of effort in
acquiring a great deal of data and who wants to reap some eventual rewards from analysing it, without
being forestalled by somebody who has not been involved in the project. At the same time it is easy to
understand the perspective of a funder who might take the point of view that public money should
result in publicly available data. The funder could correctly assert that the full value of data is only
achieved when it is thoroughly analysed from multiple points of view. Perhaps one answer lies in
making data freely available on condition that the original team members are sufficiently involved in
each use of the data that they become co-authors. This approach has certainly worked in the case of
252
the Parkway array data, and it has the advantage that a people who have a close knowledge of the site
can prevent mistakes being made. It is for instance very easy from the other side of the world to insert
the wrong valley in a diagram. Funding for travel should be incorporated in collaborative work both to
build a true team and to enable non-local workers to get a hands-on feel for the local situation.
Another way of ensuring that the people who did the work got some credit would be to establish a
foundation paper for each dataset listing all contributors as authors, and to provide copies of the data
only on the understanding that the foundation paper be cited in all resultant publications.
Finally, the possibility of totally international projects should be considered. This would involve
international funding and international operation as well as international data sharing. That would
require a considerable organisational effort and would be subject to issues of parochialism.
LESSONS LEARNED
Dense temporary arrays of velocity seismographs can provide valuable insights into the responses of
small basins filled with soft soil. In the short deployment times of a few weeks the earthquakes
recorded by the New Zealand arrays were seldom felt none for the Alfredton array, one for the
Parkway array and one for the Wainuiomata array. Therefore the recordings only apply to low
amplitude shaking, where the soil is expected to behave in a linear manner. However an understanding
of linear response is a prerequisite to a full understanding.
As technology has improved it has become more feasible to record ground motion continuously. At
the times that the three New Zealand arrays of this paper were set up however, data storage was at a
premium, and it became necessary to allow each recorder to operate autonomously, using preset
criteria to determine whether a given data stream was worth recording. We used the STA/LTA method
(ratio of short term average to long term average), with an overarching amplitude criterion in case of
strong shaking of an emergent nature. This was because during the earlier phases of recording at
Alfredton it was found that the signal on soft soil caused by a distant earthquake often rises slowly,
and the STA/LTA approach can fail. Our approach today would be to record everything, to examine
the recordings on a daily basis, and to reject non-earthquake data quite early on.
The establishment and operation of the Alfredton array, together with later recordings of
microtremors, confirmed that microtremor recordings can be used to distinguish between resonant and
non-resonant basins. Had microtremors been recorded and analysed prior to the Alfredton project, that
array would not have been installed because a valley-wide response was being sought. In the case of
the Parkway and Wainuiomata arrays, microtremor recording provided insights that both these areas
would undergo valley-wide response. Microtremors played an important role in choosing these areas
in preference to other areas simultaneously under consideration.
In terms of the analytical techniques employed, the varied inter-station spacings employed for the
Parkway array were preferable to the regular grid of the Alfredton array or the large inter-station
spacings of the Wainuiomata array. When a selection of inter-station spacings is used, spatial aliasing
is avoided while at the same time both high and low wave speeds can be accommodated. In the case of
studying the behaviour of coherency with distance a spread of inter-station distances enables a better
spread of data points, better enabling relevant curves to be plotted. In respect of array layout, Parkway
was superior to both Alfredton and Wainuiomata.
At Parkway four stations on the rock surrounding the basin were used. The same was not possible in
Wainuiomata because one site owner withdrew consent at short notice a few days after recording had
started. Four stations allowed better averaging of the rock signal, and potentially can be used to
reconstruct the fronts of incident plane waves.
253
Seismic CPT proved an invaluable tool for characterising the basins, especially when the velocity
inversion that is present at Parkway is considered. It was our intent for each basin to carry out a very
few SCPT probes, and then to correlate any layers identified by CPT with known velocities. At
Alfredton this failed because of the inhomogeneity of the sediments; at Parkway it failed because CPT
probes were unable to penetrate a hard layer; and at Wainuiomata severe problems were encountered
with groundwater because a deep aquifer carried water at high pressure and penetrometry to this depth
resulted in flooding in urban parks, with attendant problems of sealing a high pressure leak.
The computation of spectra in the wavenumber domain, combined with frequency filtering and
component rotation allowed complex propagation phenomena to be untangled, and simplex
optimisation allowed vertical motion at Parkway to be partitioned between two waves. When this
approach failed at one frequency, displaying the motion as a real-time movie was very instructive,
showing the reality (but insignificance) of torsional types of response.
Much has been learned from studies of the three basins from the perspectives of both understanding
basin response, and instrumenting basins. Nevertheless much still remains to be gained from further
studies. At Alfredton the behaviour of coherency as a function of frequency and station separation
remains unstudied, and a simple technique to predict the lack of valley-wide response needs to be
devised. At Parkway a valley-wide resonant response at 2.05 Hz remains unexplained. For
Wainuiomata there still exists a possibility of valley-wide response for the whole valley system
including sub-basins. Summing normalised spectra over all soil sites and all earthquakes as
accomplished for Parkway (Stephenson & Chvez-Garca, 1998) could be a useful way of
investigating this.
ACKNOWLEDGEMENTS
It is difficult to name all the individuals who have influenced the writing of this paper. Many staff at
GNS Science, both technicians and scientists, have made the paper possible, and they are hereby
thanked. Likewise my overseas colleagues have provided endless ideas and guidance and they are
thanked. It is appropriate however to specifically thank Carolyn Hume who drafted the diagrams, and
Nick Perrin and Rafael Benites who critically appraised the text.
REFERENCES
Bordoni P, Cara F, Cercato M, Di Giulio G, Haines AJ, Milana G, A. Rovelli A, Ruso S. The use of a
very dense seismic array to characterize the Cavola (northern Italy) active landslide body. Poster
44, Third ESG International Symposium, August 2006.
Kuroiwa, J Reduccin de desastres: viviendo en armona con la naturaleza. Ediciones del PNUD,
Per, Lima, January 2002.
Liao S, Zerva A and Stephenson WR "Seismic spatial coherency at a site with irregular subsurface
topography Proceedings, ASCE Conference New Peaks in Geotechnics, Denver, Colorado,
February, 2007.
Stephenson, WR An experimental study of normal modes of vibration of saturated alluvium. In:
Proceedings : Symposium on Earthquake Engineering, 5th, Roorkee, India, Nov. 9-11. 119-126,
1974.
Stephenson, WR Microzoning for design in Structural design for earthquakes University of
Auckland, New Zealand, August 1972.
Stephenson, WR Basin response to earthquakes : the Alfredton soft soil site. Institute of Geological
& Nuclear Sciences science report 96/03. Institute of Geological & Nuclear Sciences Lower Hutt,
New Zealand, 1996.
Stephenson, WR The dominant resonance response of Parkway Basin. In: 12WCEE 2000 : 12th
World Conference on Earthquake Engineering. Upper Hutt, NZ: New Zealand Society for
Earthquake Engineering. Proceedings of the World Conference on Earthquake Engineering, 2000.
254
Stephenson, WR Guided Love- and Rayleigh-waves in Parkway Valley, Wainuiomata, N.Z. Bulletin
of the New Zealand Society for Earthquake Engineering, 35(4), 255-265, 2002.
Stephenson, WR "Visualisation of Resonant Basin Response at the Parkway Array, New Zealand" Soil
Dynamics and Earthquake Engineering, 27(5), 487-496, 2007.
Stephenson, WR and Barker, PR Results from the Pukehou array. In: Pacific Conference on
Earthquake Engineering, Auckland, New Zealand, 20-23 November 1991, Proceedings, Volume 3.
229-238 Wellington: New Zealand National Society for Earthquake Engineering. 1991.
Stephenson, WR, Barker, PR, and Yu, J. The Wainuiomata "local effects" seismograph network.
Bulletin of the New Zealand Society for Earthquake Engineering, 35(4) 243-254, 2002.
Stephenson, WR, and Chvez-Garca, FJ Preliminary assessment of resonant phenomena recorded by
the Parkway Network. Institute of Geological & Nuclear Sciences science report 98/18 Institute of
Geological & Nuclear Sciences Lower Hutt: New Zealand 1998.
Yu, J and Haines, AJ The choice of reference sites for seismic ground motion amplification analyses:
case study at Parkway, New Zealand. Bulletin of the Seismological Society of America, 93(2):
713-723, 2003.
Zerva, A, Piasecki, M, Liao, S and Yargici, V. Spatial (ground surface) array data: observations and
needs. http://www.cosmos-eq.org/Projects/GSMA/Paper/3.2_Zerva_et_al.pdf. 2004
255
June 25-28, 2007
Paper No. W1-1003
RECENT DEVELOPMENTS OF GEOSYNTHETIC-REINFORCED

SOIL STRUCTURES TO SURVIVE STRONG EARTHQUAKES
Fumio TATSUOKA1, Junichi KOSEKI2, Masaru TATEYAMA3, Daiki HIRAKAWA4
ABSTRACT
A geosynthetic-reinforced soil (GRS) retaining wall (RW) technology has been developed to replace
conventional-type soil RWs, aiming at higher cost-effectiveness as well as higher seismic stability.
With this technology, full-height rigid (FHR) facing is staged-constructed so that the geosynthetic
reinforcement layers are firmly connected to the back of the facing. The following is reported: a)
lessons from the failure of conventional RWs on level ground during the 1995 Kobe Earthquake in
Japan and those on slopes during the 1999 ChiChi Earthquake in Taiwan and 2004 Niigata-ken-
Chuetsu Earthquake in Japan; b) reconstruction of damaged conventional-type slopes and RWs to
GRS RWs with FHR facing; and c) related researches by model shaking table tests. Three new bridge
types with geosynthetic-reinforced backfill (i.e., vertically preloading and prestressing the
geosynthetic-reinforced backfill, using lightly cemented gravel reinforced with geosynthetic
reinforcement and reinforcing the backfill of integral bridge with geosynthetic reinforcement layers
connected to the parapet) developed also aiming at higher cost-effectiveness as well as higher seismic
stability are described.
Keywords: Bridge abutment, full-height rigid facing, geosynthetic reinforcement, soil retaining wall
INTRODUCTION
Due to vast and serious damage to a great number of civil engineering structures during the 1995
Hyogo-ken Nambu Earthquake (i.e., the 1995 Kobe Earthquake), the seismic design codes for RC and
steel structures in Japan were revised substantially so that the structures can survive such high level
seismic load (i.e., the so-called level II design seismic load). A great number of embankments,
retaining walls (RWs) and bridge abutments with backfill for railways, highways and residential areas
were seriously damaged during not only the 1995 Kobe Earthquake (Fig. 1), but also many other
major earthquakes in Japan, including the 2004 Niigata-ken Chuestu and 2007 Noto Hanto
Earthquakes. After the 1995 Kobe Earthquake, the seismic design codes for new construction of soil
structures have also been revised. However, the revision was generally not as substantial and sufficient
as those for RC and steel structures. Correspondingly, soil structures that failed by earthquakes have
been usually reconstructed to those of original structural types, although it is possible that the
reconstructed soil structures would not survive a similar level of seismic load by which the concerned
soil structures collapsed. It seems that this situation reflects such a conventional design concept that
collapse of soil structures should be accepted for the following reasons. Firstly, it is most important to
as soon as possible recover the original function of failed soil structures even if reconstructed soil
structures may not have the same level of structural stability as the original. Secondly, the
1
Professor, Department of Civil Engineering, Tokyo University of Science, Japan, Email:
tatsuoka@rs.noda.tus.ac.jp
2
Professor, Institute of Industrial Science, University of Tokyo, Japan.
3
Chief Research Engineer, Railway Technical Research Institute, Japan
4
Assistant Professor, Department of Civil Engineering, Tokyo University of Science, Japan
256
reconstruction of failed soil structures to those having a seismic stability higher than the original level,
preferably to the level withstanding level II design seismic load could be overly costly. Thirdly, as the
number of existing important soil structures is so large, it is financially not feasible to rehabilitate all
of them to have sufficiently high seismic stability. It is also the case with important soil structures that
are to be newly constructed. Lastly, it is generally much easier and faster to reconstruct failed soil
structures when compared with the case of RC and steel structures and, so, soil structures could be
allowed to fail or collapse.
a) b)
Figure 1. a) Gravity type RW without a pile foundation (Ishiyagawa, Hanshin Railway); and b)
leaning type unreinforced concrete RW without a pile foundation (Sumiyoshi, JR West) that
collapsed during the 1995 Kobe Earthquake (Tatsuoka et al., 1997).
Nowadays, however, these rationales listed above are not relevant with many soil structures that are to
be constructed as important infrastructures. Firstly, with line structures, such as highways and
railways, collapse of soil structures at a limited number of places results in a long closure of the whole
system. Secondly, a great number of huge soil structures, such as massive high embankment, have
been constructed for residential and industrial areas, railways, highways and so on by means of
modern construction machines, which was not feasible previously. It is usually highly time-consuming
and costly to reconstruct these soil structures if they collapse. Thirdly, the reconstruction of damaged
and failed conventional type soil structures (i.e., embankments with gentle slopes and gravity- or
cantilever-type RC soil RWs) to geosynthetic-reinforced steepened slopes or GRS soil retaining walls
could be much more cost-effective with a higher seismic stability than reconstruction to the original
structural types. The high cost-effectiveness of GRS soil structures in spite of increased structural
stability results from smaller earthwork, a higher constructability under restricted conditions (e.g., in a
narrow space, on a steep slope or at a remote place) and a higher construction speed. Moreover, it
becomes easier to arrange a drainage system in geosynthetic-reinforced backfill than in conventional
type soil structures. For the same reasons, the geosynthetic-reinforced soil technology is also useful to
newly construct soil structures in many cases. Therefore, there is no strong reason to reconstruct failed
conventional type soil structures to those of the original structure types in usual cases.
In this report, firstly the geosynthetic-reinforced soil retaining wall (GRS-RW) technology that is now
becoming popular is described. Then, several case histories in which conventional embankments and
soil RWs that fully collapsed by seismic loads were reconstructed to geosynthetic-reinforced
steepened slopes and GRS-RWs having FHR facing are reported. Finally, several new type bridge
abutments taking advantages of the geosynthetic-reinforced soil technology are introduced.
GEOSYNTHETIC-REINFORCED SOIL RETAINING WALL WITH FHR FACING
Fig. 2 shows a procedure to construct permanent GRS RWs, which is now widely used in Japan
(Tatsuoka et al., 1997). This technology has three major features. Firstly, GRS RWs are constructed by
the following staged construction procedure: 1) the backfill is constructed to the full wall height with
a help of gravel gabions placed at the shoulder of each soil layer; 2) geosynthetic reinforcement layers
are arranged with a vertical spacing of 30 cm to ensure a high compaction of the backfill; and 3) after
257
sufficient deformation of the completed geosynthetic-reinforced backfill and the supporting ground
has taken place, a lightly steel-reinforced concrete full-height rigid (FHR) facing is constructed by
casting-in-place concrete directly on the wrapped-around wall face ensuring a strong connection
between the facing and the reinforced backfill. The advantages of the staged construction method are:
1) the damage to the connection between the reinforcement and the facing due to in- and post-
construction deformation of the backfill and supporting ground can be avoided, which makes possible
the construction of GRS RWs on compressive subsoil of thick clayey soil on the premise that the
stability of wall can be ensured; 2) the backfill immediately back of the wall face can be compacted
dense with better mobilization of reinforcement tensile force; and 3) the alignment of completed wall
face becomes easy. The second feature is the use of a polymer geogrid with cohesionless soil to ensure
a good interlocking with the backfill and a composite of non-woven and woven geotextiles with high-
water content cohesive soils to facilitate both drainage and tensile reinforcement of the backfill so that
low-quality on-site soil can be used as the backfill, if necessary. The third feature is the use of
relatively short reinforcement, which becomes possible by using FHR facing connected to the
reinforcement.
Drain hole Gravel gabion
Geosynthetic
1) Leveling pad for facing 2) Placing geosynthetic &

gravel gabions
3) Backfilling & compaction 4) Second layer
5) Completion of 6) Casting-in-place
a) wrapped-around wall RC facing b)
Figure 2. a) Staged construction of GRS RW with FHR facing; and b) a typical wall at Nagoya.
Earth Very small force in the

facing, resulting to a
pressure
Large force in the wall; simple facing structure
Earth and
pressure large overturning moment Reinforce-
& large horizontal load at ment
the bottom of the wall.
Stress concentration
Needs for Very small overturning moment

Backfill
a massive or strong & very small lateral force at the
wall structure; and bottom of facing, resulting in no
a pile foundation need for a pile foundation!
a) b)
Figure 3. Mechanisms of: a) cantilever RW; and b) GRS RW with FHR facing.
Roles of full-height rigid facing

The facing structure of a conventional type RW is a cantilever structure that resists against the active
earth pressure, PA, from the unreinforced backfill activated on the facing by reaction force at the base
of the facing (Fig. 3a). Therefore, large internal moment and shear force are activated inside the facing
while large overturning moment and sliding force develop at the base of the facing. With reinforced
soil RWs, on the other hand, the backfill is retained by the tensile force in the reinforcement. The
conventional explanation of the function of the facing, which is actually misleading, is that, because of
tensile-reinforcement effects, only very small earth pressure is activated on the back of the facing,
accordingly, only a light and flexible facing that can prevent the backfill soil from spilling out is
sufficient. However, if the earth pressure activated on the back of the facing is very small, the tensile
force at the connection between the reinforcement and the back of the facing becomes very small (Fig.
258
4a) and therefore the soil retaining capability of the reinforcement is largely reduced (Tatsuoka, 1992).
That is, the maximum available tensile force, T max, in the reinforcement is obtained by Eq. 1 (Fig. 4c):
Tmax= Minimum of [TR, T anchor, Tretain + TWmax] (1)
where TR= tensile rupture strength of each reinforcement layer; Tanchor= anchorage capacity, which is
approximately proportional to the anchorage length, La; Tretain= available retaining strength, which is
approximately proportional to the retaining length, L r; and, TWmax= available tensile force at the
connection between the reinforcement and the back of the facing, which increases with an increase in
the earth pressure activated on the back of the facing and with an increase in the connection strength.
When T Wmax decreases to zero, the distribution of the reinforcement tensile force, T, becomes one as
those presented in Fig. 4a. In this case, at lower levels in the wall, Tmax cannot become large enough.
That is, low connection force results in low tensile force in the reinforcement, which means low lateral
confining pressure in the active zone. Then, the active zone becomes more deformable and less stable,
particularly when the backfill is a cohesionless soil. On the other hand, large TWmax results in large Tmax
(i.e., Fig. 4b). With this distribution pattern, the active zone is effectively confined without exhibiting
large strains, and therefore, large bond stresses are not mobilized at the surface of the reinforcement.
Consequently, the reinforcement tensile force, T, becomes rather constant in the active zone. When the
reinforcement layers are connected to a FHR facing, large earth pressure, which is similar to the active
earth pressure in the unreinforced backfill or even higher earth pressure, can be activated on the back
of the facing (Fig. 4b), resulting into larger connection force and higher reinforcement force. Then, the
confining pressure in the active zone is kept high, which results in higher stiffness and small
deformation as well as higher strength and stability of the active zone (i.e., preferred stable
conditions). The FHR facing for a GRS-RW behaves like a continuous beam with a large number of
supports with a small span (Fig. 4b). Then, even with high earth pressure activated on the back of the
facing, only small force is activated inside the facing, resulting in a simple facing structure, and
insignificant overturning moment and lateral force activated at the base of the facing, resulting in no
need for a pile foundation in usual cases.
Active Active
zone zone
Reinforcement
Connected
a) b)
c)
Figure 4. Distributions of tensile force in the reinforcement; a) when no facing is used or when
the facing and reinforcement are not connected; and b) when a rigid facing and reinforcement
are connected; and c) available maximum tensile force in the reinforcement.
In summary, the GRS RW system described in Fig. 2 is much more stable while the facing structure
and foundation structure are much simpler than the conventional type RWs, which results in lower
construction cost, a higher construction speed and the use of lighter construction machines. Despite the
above, the wall performance can be equivalent to, or even better than, conventional type RWs.
Some case histories

The elevated transportation structures in Japan have gradually changed from embankment having a
gentle slope towards embankments supported with conventional type RWs (initially masonry RWs,
then RC cantilever RWs with a pile foundation) and RC frame structures for higher ones (Fig. 5a).
About 35 years ago, the Terre Armee RW technique was introduced to construct railway and highway
RWs and had dominated in the permanent reinforced-soil RW market in Japan. More recently, the
GRS RW system described in Fig. 2 was introduced and now has become one of the standardized RW
259
technologies for railways in Japan while becoming popular also in other fields. Today, the use of the
Terre Armee RWs for railways has declined to nearly zero. Because of a higher cost-effectiveness and
a sufficiently high stability and low deformability, the GRS-RW system (Fig. 2) is routinely adopted
instead of conventional RC RWs and the Terre Armee RWs. By the end of 2006, GRS RWs (Fig. 2)
were constructed at 518 sites with a total wall length more than 80 km (Fig. 6). The stability of GRS-
RW systems has been validated by excellent post-construction performance of the constructed walls. It
should be emphasized that any problematic case, for example with too large wall deformation, has not
been reported.
Gentle slope
could be unstable;
could be too deformable; (basically Some cases
and
with piles)
occupies too large space.
Higher cost-effectiveness
Sufficiently stable and stiff (basically no piles)
a) b)
Figure 5. a) History of elevated railway and highway structures in Japan; and b) a typical GRS
RW in Nagoya, presented in Fig. 2b (Tatsuoka et al., 1997).
80 80 km
1995 Kobe Earthquake:
a very high-seismic stability of
GRS RWs with a FHR facing
60 was proven, which resulted in
Wall length km
its more popular use

subsequently. 2004
40 Niigataken-
23 Chuetsu
11 Earthquake
20 Total
53 Annual
0
1990 1995 2000 2005
16 Since 1982: research at the University of Tokyo
140 96 82 97 & the Railway Technical Research Institute
a) b)
Figure 6. a) Locations of GRS RWs with FHR facing constructed by the staged construction
procedure (as of June 2006); and b) annual and total constructed wall length.
a) b)
Figure 7. GRS RW with a FHR facing for railway at Tanata; a) immediately after
construction; and b) one week after the earthquake (Tatsuoka et al., 1997).
One of the important case histories with respect to the seismic stability of GRS RW with FHR facing
is the one at Tanata site that survived the 1995 Kobe Earthquake (Fig. 7). The damage to the Tanata
GRS RW was substantially less serious when compared to many conventional types of soil RWs
including those described in Fig. 1. The Tanata wall was completed in February 1992 on the southern
260
slope of an existing embankment for the JR Kobe Line to increase the number of railway tracks from
four to five. The total length of the wall is 305 m and the greatest height is 6.2 m. The surface layer in
the subsoil supporting the wall consists of relatively stiff terrace soils. The backfill soil is basically a
cohesionless soil with a small amount of fines. The reinforcement is a geogrid made of PVA
(polyvinyl alcohol) coated with soft PVC (polyvinyl chloride) for protection, having a nearly
rectangular cross section of 2 mm by 1 mm and an aperture size of 20 mm with a nominal tensile
rupture strength TR = 30.4 kN/m. The wall deformed to a limited extent and moved slightly. The
largest outward displacement occurred at the crest of the wall at the location in contact with a RC box
culvert structure crossing the railway embankment. The outward lateral displacement was 26 cm and
10 cm at the crest of the wall and at ground level. Despite the above, the performance of the GRS-RW
was judged to be very satisfactory because of the following facts. Firstly, the peak ground horizontal
acceleration at the site was estimated to be more than 700 gals (i.e., 700 cm/sec2; or 0.7 g, where g is
the gravitational acceleration). This is consistent with a high collapse rate of wooden houses at the site
(Fig. 7b). Many of the collapsed houses were relatively new, constructed less than approximately ten
years before. Secondly, on the opposite side along the railway of the RC box structure, a RC retaining
wall with a maximum height of approximately 5.4 m had been constructed at the same time as the
GRS-RW. This wall is supported by a row of bored piles despite the similar subsoil conditions as the
GRS-RW. Therefore, the construction cost per wall length of the RC retaining wall was approximately
double to triple of that of the GRS-RW. Despite these differences, the RC retaining wall displaced
similarity to the GRS-RW: i.e. at the interface with the RC box structure, the outward lateral
displacement of the retaining wall was 21.5 cm at the top and 10 cm at ground level. Thirdly, the
length of geogrid reinforcement used in GRS-RWs with a FHR facing is generally much shorter than
that for most metal strip-reinforced soil retaining walls and other types of GRS retaining walls with a
deformable facing. For conservatism, most of the GRS-RWs with FHR facing that have been
constructed to date have several longer reinforcement layers at higher levels of the wall, as typically
shown in Fig. 5b. With the Tanata wall, the length of all reinforcement layers was truncated to
approximately the same length due to construction restraints at the site. This arrangement may have
reduced the seismic stability of the wall; i.e., the wall would have tilted less if the top geogrid layers
had been made longer.
Model shaking table tests (1g) to evaluate the seismic stability of RWs on level ground
To understand the different performances between the conventional RW types (Fig .1) and the GRS
RW with FHR facing at Tanata (Fig. 7) and also to evaluate the effects of different arrangements of
reinforcement layers on the stability of GRS RW, a series of model shaking table tests (in 1 g) were
performed (Tatsuoka et al., 1998; Koseki et al., 2003, 2006).
155 140 140

155
SuSurchrge
rcharge (1kPa
1 kPa )
140
SuSurchrge
rch arge (1 kPa )
140
SuSurchrge
rch arge 1kPa
(1 kPa )
1kPa
53
53 53
53 53
53
Air-dried Toyoura Air-dried Toyoura
Air-dried Toyoura
ModelBackfill ModelBackfill sand (D r = 90 %)
ModelBackfill
sand (Dr=(Dr=90%)
90 %) sand (D r= 90 %)
20 23 20 23 20 18
18
a.Model
Cantilever type(C)
C: cantilever b. Gravity
Model G: type(G)
gravity c.Model
Leaning type(L)
L: leaning
140
140 140
140 140
140
20 20 Su rch arge1kPa
Surchrge (1 kPa ) 20 SuSur
rchchrge
arge 1kPa
(1 kPa ) 35
35 Su Sur
rcharge
chr ge( 11kPa
kPa )
20
80
80
Air-dried 45
45
50
50 50 50
50
50 Ext ended Air-dried
Toyoura sand
ModelBackfill Extended
Reinforcement ModelBackfill
Toyoura sand
(Dr= 90 %) ModelBackfill
reinforcement (Dr= 90 %)
20 20 20
d. Reinforced-soil type 1(R1) e. Reinforced-soil type 2(R2) f. Reinforced-soil

Model R3: GRS-RW type 3(R3)
type 3
Model R1: GRS-RW type 1 Model R2: GRS-RW type 2
Figure 8. Six models of RW on level ground for shaking table tests (all units in cm: Tatsuoka et
al., 1998; Koseki et al., 2003, 2006)
Fig. 8 shows six different models of RWs without a pile foundation on level ground. The models are
50 53 cm high. Models G and L simulate the collapsed prototype gravity type and cantilever RWs
described in Fig. 1, while model R1 simulates the Tanata GRS RW (Fig. 7). The performance of model
261
R1 was compared with those of models R2 and R3 to evaluate the effects of reinforcement
arrangement. The backfill was air-pluviated dense Toyoura sand (emax=0.977; emin=0.609; Gs= 2.64;
and D50=0.23 mm). Each model RW was constructed on a level 20 cm-thick sand layer. The model
reinforcement was a grid made of phosphor-bronze strips. A surcharge of 1.0 kPa was applied on the
backfill to simulate the railway ballast. The models were subjected to several shaking steps of
horizontal excitation with an irregular time history of acceleration shown in Fig. 9a, made from the
time history of ground acceleration recorded during the 1995 Kobe Earthquake. The maximum
amplitude of the base acceleration amax was initially set 100 gals and increased at an increment of about
100 gals. The time scale was reduced by a factor of 0.3 so that the relationship between the
predominant frequency of the original acceleration record and the natural frequency of the prototype
structures is as much as possible maintained in the model tests.
100
dtop G
Wall top displacement, dtop(mm)

5 cm Backfill 50 or L
-1.0 80 53 cm R1 R3
amax
-0.8
-0.6 Predoninant frequency: Subsoil 20 cm
adjusted to 5 Hz 60 C
Base acceleration (g)
-0.4 R2
-0.2
0.0 40
0.2
0.4 20
GRS-RW
0.6
Modified from N-S component at Kobe Marine Meteorological with a FHR facing
0.8 Observation Station during the 1995 Kobe EQ. 0
1.0 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4
1 2 3 4 5 6 7 8 Seismic coefficient, k h=amax/g
Time (sec)
a) b)
Figure 9. a) Table acceleration; and b) relationships between the residual wall displacement and
the seismic coefficient from the model shaking table tests (see Fig. 8 for model names (Tatsuoka
et al., 1998; Koseki et al., 1999, 2006).
Level
Total normal force, P (kN/m)
Leaning Slope
Seismic
force
Level
Smaller P in the
a)
Slope slope
Bearing
capacity, P Seismic coefficient
Smaller T by
Seismic Force Shear band failure in the
Tensile force, T (N)
slope
Level
Tensile b)
force, T Slope
Reinforced
G
Seismic coefficient
Figure 10. a) Total normal force P at the footing base of leaning RW; and b) total tensile
reinforcement force T of GRS RW type 2 versus input load, level ground and slope.
Fig. 9b shows the relationships between the residual lateral outward displacements 5 cm down
from the crest of the facing observed after the respective shaking stages and the seismic
coefficient, kh, defined as amax /g, where amax is the maximum acceleration at the shaking table.
Fig. 10 shows the total normal force at the bottom face of the footing of model L (leaning
type RW) measured with a set of local load cells and the total tensile force at the connection between
the reinforcement layers and the facing of model R2 (GRS RW type 2). The results from the tests on
the RWs on slope presented in this figure are explained later. The following trends of behaviour may
be seen from Figs. 9 and 10:
262
1. GRS RW model R1 is much more stable than model G (gravity type RW), model L (leaning type
RW) and model C (cantilever RW). This trend is consistent with the performances of the prototype
RWs during the 1995 Kobe Earthquake (Figs. 1 and 7).
2. The failure of model R1 (i.e., GRS RW with FHR facing) was much more ductile than models G, L
and C (i.e., conventional type RWs). This is because, as seen from Fig. 10, the tensile force in the
reinforcement of the GRS RW increased with an increase in the input acceleration, while, with
model L (conventional leaning type RW without a pile foundation) as well as models G and C, the
bearing capacity was suddenly lost, which resulted in brittle failure of these RWs.
3. The GRS RW becomes more stable by increasing the length of reinforcement. The effects of
increasing a limited number of layers at high levels of the wall are equivalent to those by equally
increasing the length of all the reinforcement layers from 20 cm to 35 cm. The former arrangement
using a less total amount of reinforcement is more cost-effective.
SOIL RETAINING WALLS ON SLOPE
1999 ChiChi Earthquake (Taiwan)

A great number of RWs among those that had been constructed on slope in the mountain areas
affected by the earthquake collapsed as typically shown in Fig. 11a (e.g., Huang, 2005). All the
collapsed RWs were a kind of gravity type leaning on the backfill (i.e., leaning type). The inferred
major cause for the failure is the bearing capacity failure of the slope supporting the RWs (Fig. 11b).
In comparison, a number of gravity RWs leaning on the cut slope did not exhibit any noticeable
damage (Fig. 11c), which is due seemingly to a high bearing capacity of the supporting ground.
Asphalt pavement
Excavation
5.0m
Original slope
0.6m 3.6m
2.5m
0.7m Backfill
about 7.0 m
Nearly no
sandy gravel damage
Estimated failure surface

Serious damage
Bearing capacity failure
a) b) c)
Figure 11 a) Typical collapsed gravity RW on slope; b) failure of RW on slope; and c) different
performances of RWs, 1999 ChiChi Earthquake, Taiwan (Tatsuoka et al., 2007a).
Model shaking table tests

Another series of shaking table tests were performed on model RWs described in Fig. 12, similar to
those on RWs on level ground described in Figs. 8 and 9. The differences are; 1) the supporting
ground in front of the respective model RWs was sloped at 1:2 (V:H); and 2) a different model
combination was adopted. Namely, model L (leaning RW), model C (cantilever RW) and model R2
(GRS RW with FHR facing) in Fig. 12 are the same as models L, C and R2 in Fig. 8, while model G
(gravity) in Fig. 8 was not tested in this series. The short reinforcement layers at the lower levels of
model R2 is realistic when constructed on slope: i.e., if long reinforcement layers are to be arranged at
the lower levels of the wall, too much excavation of the slope becomes necessary. In this series, model
L-N (leaning type RW with nails) and model R2-N (GRS RW with nails) were also prepared to
evaluate the effects of reinforcing the backfill with nails of RWs on slope. Four model large-diameter
nails, which were 4 cm-in diameter and 40 cm-long mortar columns with a 3 mm-in diameter steel
reinforcement at the center, were placed at a horizontal center-to-center spacing of about 10 cm in
each layer.
Fig. 13 shows the test results. It may be seen that the seismic stability of the three RW types, the
leaning and cantilever RWs as well as the GRS RW, constructed on slope is significantly lower than
the respective RW types constructed on level ground. With the leaning and cantilever RWs, the
deformation of the supporting slope and the normal load measured at the foundation base of the facing
263
showed that the bearing capacity failure took place in the slope below the RWs at lower input
acceleration than in the level ground supporting the respective RWs. This trend can be seen also from
Fig. 10a. With the GRS RW on slope, the reinforcement force suddenly decreased when the input
seismic load reached some level (Fig. 10b). This was due to the development of a shear band starting
from below the bottom of the facing extending toward back of the reinforced backfill zone as
illustrated in Fig. 10. It may also be seen from Fig. 13 that nailing the backfill can effectively increase
the seismic stability of RWs on slope. With the leaning RW, the upper nails anchored the top part of
the facing preventing it from overturning failure while the lower nails increased the bearing capacity
of the slope supporting the facing. Such nailing as above can effectively increase the seismic stability
of leaning RWs already existing on slope. With the GRS RW, the nails prevented the development of
a shear band starting from below the facing. When a RW is to be newly constructed on slope, a GRS
RW with nails is very cost-effective with very high seismic stability.
L)
C)
R2)
Horizontal accelerometer
Horizontal & vertical acc.
Disp. Displacement transducer
Electric-resistance strain gauge
L-N) R2-N) (all units in mm)
Figure 12. RW models on slope: L) leaning RW; C) cantilever RW; R2) GRS RW; L-N) leaning
RW with nails; and R2-N) GRS RW with nails (Kato et al., 2002).
200
C: Cantilever (slope)
45 cm
Wall top displacement. dtop (mm)
C: Cantilever (level)
150
R2-N:
R2: Reinfoced Reinforced
(slope) with nails
(slope)
100
R2: Reinforced
L: Leaning (slope) (level)
L: Leaning (level
50
L-N: Leaning with nails (slope)

0
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8
Seismic coefficient, kh =amax/g
Figure 13. Relationships between residual wall displacement and seismic coefficient from the
model shaking table tests on RWs on slope (see Fig. 12 for model names) (Kato et al., 2002).
2004 Niigataken-Chuetsu Earthquake

A great number of embankments, RWs and bridge abutments with backfill for railways, highways and
residential areas were seriously damaged during the mainshock with a magnitude of 6.8 and several
aftershocks (M= 6.5 and less) of the 2004 Niigata-ken Chuetsu Earthquake. Compared to the damage
to RC structures, the scale and extent of the damage to soil structures and its effect on the
transportation system and associated civil life were much more extensive. In particular, a number of
embankments that had been constructed in narrow valleys and on slopes, where ground and surface
264
water tends to concentrate, were most seriously failed. The effects of heavy rainfall a couple of days
before the earthquake had made the backfill saturated more than usual conditions.
After remedy work

GRS-RW with a full-height rigid facing
Slope: 1:0.3 (V:H); height= 13.2 m
Vertical spacing of geogrid= 30 cm
Joetsu line
Before failure:
13.18 m
backfill of sand including round-shaped gravel on
sedimentary soft rock (weathered, more at
shallow places)
After remedy
After work .
Failed failure 1:2
gravity wall 0
Silt rock
Shinano riiver
Sand rock Rock blots
Gravel-filled steel
wire mesh basket
)
(V:H
a) 1:4
b) c) d)
Figure 14. a) Slope profiles before and after failure and after reconstruction to a GRS RW; b)
failed embankment; and c) & d) the GRS RW under and after construction. Jo-etsu Line, JR
East, 2004 Niigata-ken Chuetsu Earthquake (Tatsuoka et al., 2007a).
Three totally failed embankments for the Jo-etsu line of the East Japan Railway Company that had
been constructed on the slope on the right bank of Shinano River were typical of the above. These
were reconstructed to three GRS RWs with FHR facing (Fig. 2), as typically presented in Fig. 14, for
the following reasons. Firstly, several GRS RWs of this type performed very satisfactorily during the
1995 Kobe Earthquake (e.g., Fig. 7). On the other hand, a number of conventional type RWs (gravity,
leaning, masonry and cantilever RC) were failed and may of them were reconstructed to this type of
GRS RWs within a short duration (Tatsuoka et al., 1997, 1998). Secondly, it was considered that this
type of GRS RW is superior over other types of soil structures (i.e., conventional sloped embankment
supported by a RC cantilever RW with a pile foundation; bridges and so on) in terms of construction
cost, construction period and performance in terms of deformation and ultimate stability. In fact, these
three railway embankments were reconstructed within two months after the failure. This new case
history validated again that the GRS RW (Fig. 2) could be very competitive to construct wall
structures for such important infrastructures (i.e., railway and highway).
BRIDGE ABUTMENTS
Several problems with the conventional type bridge abutment with backfill
Despite a wide use over the world, such conventional type bridge abutment as described in Fig. 15
have several drawbacks as listed below, which are due mostly to that the backfill is not reinforced and
the abutment is a cantilever structure on which a girder is placed via movable and fixed supports:
1) As the abutment is a cantilever structure, a pile foundation is usually necessary (Fig. 3a).
265
2) Despite that the RC abutment is not allowed to noticeably displace once constructed, associated
with the backfill construction, it is subjected to earth pressure and effects of the settlement and
lateral flow in the subsoil via its effects on the piles. To alleviate these negative effects, it may
become necessary to increase the number and size of piles.
3) The construction and long-term maintenance of girder-supports is generally costly.
4) The seismic stability of the unreinforced backfill as well as the abutment supporting the girder via
a fixed-support is generally low, as observed in many previous major earthquakes. Watanabe et al.
(2002) and Aizawa et al. (2007) confirmed this trend by model shaking table tests.
5) A bump may be formed behind the abutment by residual deformation of the backfill due to its self-
weight as well as traffic and seismic loads.
To develop bridge systems that are more cost-effective than the conventional bridge types while
alleviating these problems described above, several solutions have been proposed and attempted as
described in Fig. 16.
(7) Long-term settlement by self-weight, traffic load & seismic load
(6) Girder
(5) Girder support (3) Backfill
(fixed or movable);
(7) Long-term maintenance
Displacement due to (2) RC (4) Earth pressure
earth pressure abutment
Ground
The numbers, (1) (7)

indicate the event Ground settlement and lateral
sequence. flow due to the weight of
backfill, and associated
(1) Piles negative effects (i.e., negative
friction & bending) on the
piles.
Figure 15. Several technical problems with conventional type bridge abutments
g p p
Well-graded Cement-
a1) gravel a2) mixed gravel
Ordinary backfill
without improvement
Improvement of the backfill (already adopted)
Conventional type
bridge abutment Connected
Connected
Geosynthetic
b1) reinforcement b2)
Reinforcing of the backfill

(b1: already adopted)
Connected
Connected G. R. Gravel G. R.
G. R. with
c1) c2) c3) PL/PS
Cement-mixed gravel Connected

Combined measures
Figure 16. Different proposed new bridge abutments (Tatsuoka et al., 2005).
Improving the backfill

One of the earliest attempts employed by the Japanese railway engineers is to construct a trapezoidal
zone of well-compacted well-graded gravelly soil behind the abutment (type a1 in Fig. 16). However,
the performance of this abutment type during several previous earthquakes in Japan was
unsatisfactory. Watanabe et al. (2002) and Tatsuoka et al. (2005) confirmed the above by model
266
shaking table tests. They also showed that the seismic stability of another similar type constructing a
trapezoidal zone of cement-mixed gravel (type a2, Fig. 16) is also low.
Reinforcing the backfill

A number of bridges comprising of a pair of GRS RWs with FHR facing (described in Fig. 2) that
support a girder (type b1 in Fig. 16) were constructed (Tatsuoka et al., 1997). This bridge type (Fig.
17) is herein called the GRS RW bridge. Although this bridge type is structurally simpler and more
cost-effective than the conventional type, it has the following limitations. Firstly, the girder cannot be
very long due to low stiffness and potential large residual deformation of the backfill supporting the
girder. Secondly, the construction and long-term maintenance of movable and fixed girder-supports is
costly. This is the common problem with all of the bridge types presented in Fig. 16. Lastly, despite
that the seismic stability of GRS RWs with FHR facing is generally very high (e.g., Tatsuoka et al.,
1998; Koseki et al., 2003), the seismic stability of the sill beam on which a fixed girder-support is
placed is not (Aizawa et al., 2007; Tatsuoka et al., 2007b). This is because the mass of the sill beam is
much smaller than the inertia force of the girder while the anchorage capacity of the reinforcement
layers connected to its back is small due to their shallow depths.
3 3: Sill beam
5. Girder
4. Movable and fixed

girder-supports
1: GRS RW
1 2: FHR facing
- The numbers indicate construction sequence.

- Not to scale
Figure 17. GRS RW bridge (type b1 in Fig. 16)
Type b2 (Fig. 16), placing a girder on the crest of the FHR facing, is more stable than type b1
(Watanabe et al., 2002; Tatsuoka et al., 2005). However, they also showed that, with type b2, the
reinforced backfill behind the facing supporting the girder via a fixed-support would exhibit too large
deformation when subjected to level II design seismic load.
Combining multiple-measures
To substantially decrease long-term residual deformation of the backfill, it is very effective to
vertically preload the reinforced backfill and then maintain some vertical prestress that is about a half
of the preload in the backfill during long-term service (i.e., the PL & PS technology). The above was
validated by laboratory model tests (Shinoda et al., 2003a&b) and long-term performance of a
prototype railway bridge pier (Uchimura et al., 2003). Moreover, Uchimura et al. (2003) and Tatsuoka
et al. (2005) showed that the seismic stability of PL & PS reinforced bridge pier and abutment is very
high. It is in particular the case if high prestress is maintained during dynamic loading and this can be
ensured by using a ratchet mechanism as shown by model shaking table tests (Shinoda et al.,
2003a&b). Type c3 in Fig. 16 consists of a PL & PS GRS RW with a ratchet system supporting a
girder via a fixed-support. Its high seismic stability was validated by laboratory shaking table tests
(i.e., Nakarai et al., 2002). Despite the above, any prototype bridge of this type has not been
constructed, because possible long-term maintenance works of the ratchet system were not preferred
by practicing engineers.
Types c1 and c2 were then proposed, which are combining types b2 and b1 with type a2. The first
prototype of type c1 was constructed for a new bullet train line in Kyushu (Fig. 18: Tatsuoka, 2004;
Tatsuoka et al., 2005). The conventional RC abutment (Fig. 16) supports the backfill with the earth
pressure activated on its back. In comparison, with type c1 as well as type b2, the reinforced backfill
zone laterally supports the RC parapet (i.e., facing) that is supporting a girder without dynamic earth
pressure activated on its back. Type c1 abutments are constructed by the staged procedure presented in
267
Fig. 18b, which is basically the same as the one described in Fig. 2a. A number of similar bridge
abutments of type c1 are now at the design stage.
Soil backfill 3 Soil backfill

. .
b) 1 2
Cement-mixed gravel Cement-mixed gravel
c)
a)
Figure 18. a) New type bridge abutment using cement-mixed gravel constructed at Takada
(type c1): b) staged construction; and c) the completed new type bridge abutment, mid 2003.
Settlement by Distance (cm) from

T: Seasonal cyclic expansion long-term
Settlement and contraction by thermal the back of the facing
due to T traffic load and Lateral
effects seismic effects Settlement of the crest
displacement
5 10 20 35
Displacement-controlled Box width: 40
loading system
4. Integration 3. Girder
8 polyester
4. Backfill
Load cell reinforcement
50.5
Increase in the 2. RC parapet layers
earth pressure Hinge support
due to T Air-dried Toyoura sand
(Dr = about 90%)
8.5
9 separated load cells: LCs
- The numbers indicate
1. Pile construction sequence.
- Not to scale 129.5
[unit : cm]
a) b)
Figure 19. Conventional integral bridge; and b) model test setup (Tatsuoka et al., 2007b).
Integral bridge
A pair of RC parapets (or facing) is integrated with a bridge girder, without using girder supports (Fig.
19a). This type is very popular in the UK and the USA due mainly to low construction and
maintenance cost resulting from no use of girder supports. However, the backfill may exhibit large
residual settlements by self-weight as well as traffic and seismic loads, while the seismic stability of
both girder-parapet system and backfill is relatively low (Aizawa et al., 2007; Hirakawa et al., 2007b;
Tatsuoka et al., 2007b). Moreover, as the girder is integrated with the parapets, seasonal thermal
expansion and contraction of the girder results into cyclic lateral displacements at the top of the
facings, which results in a gradual increase in the earth pressure on the facing and residual settlements
in the backfill as shown below (Hirakawa et al., 2006, 2007a).
Effects of cyclic displacements of the facing

Small-scale model tests (Fig. 19b) were performed in the laboratory to evaluate the problems
described above and also to examine whether they can be alleviated by reinforcing the backfill. The
backfill was air-dried Toyoura sand produced by air-pluviation for the unreinforced backfill while by
hand-tamping for the reinforced backfill. The reinforcement was a Polyester grid (strand diameter= 1
mm; spacing between the adjacent strands= 18 mm; covering ratio= 9.5 %; and rupture tensile strength
at an axial strain rate of 1.0 %/min.= 19.6 kN/m). The FHR facing was cyclically displaced about the
bottom hinge at a rotational displacement rate of 0.00053 degree/min. Fig. 20a summarizes the peak
268
earth-pressure coefficients in the respective cycles, Kpeak= 2Qpeak/H2 , where Qpeak is the peak total
earth pressure per wall width in each cycle; H is the wall height (50.5 cm); and is the dry unit weight
of the backfill (16 kN/m3), plotted against the ratio of the double amplitude of cyclic displacement at
the facing top to the facing height, (DA)/H, at selected numbers of loading cycle, N. The facing top
was allowed to move up to about 0.2 mm ( /H= 0.04 %) toward the active direction associated with an
increase in the earth pressure. The solid squares represent the cycles when the active failure plane
developed in the backfill. The earth pressure increased with an increase in (DA)/H and N. These test
results are consistent with previous laboratory model tests (Ng et al., 1998; England et al., 2000) as
well as full-scale behaviour of a prototype RW for three seasons (i.e., N= 3). This earth pressure
increase may result in structural damage to the facing and may push out the bottom of the facing. By
reinforcing the backfill, this earth pressure increase does not reduce, but the facing is not structurally
damaged and not pushed out at the bottom, as the FHR facing behaves like a continuous beam
supported by a number of reinforcement layers at a small spacing (Fig. 3b).
2.0
Backfill in the model tests: 26 Active failure in
1.8 the model backfill
unreinforced Toyoura sand
7
Earth pressure coefficient, K and K peak
1.6
158 N: number of
Likely behaviour
1.4 loading cycles
of prototype wall
N=
1.2 for three seasons:
10
seasonal (DA) vs. 2,200
1.0 estimated Kprototype N=1
0.8 K peak during

10 1
respective
0.6 N= 2,000
10 cyclics
1 1
0.4 1
3,000 0 0 0
0 0
0.2 0
K0 in the model walls
0.0
1E-3 0.01 0.1 1 10
Lateral displacement ratio, (DA) /H [%]
a)
-0.5
at 5 cm back of the facing, Sg/H (%)
R & C: D/H= 0.6 %

Residual settlement of the backfill
0.0
Reinforced & Connected (R & C):
0.5 D/H= 0.2 %
1.0 Reinforced & No Connected

(R & NoC): D/H= 0.2 %
1.5 NR:
D/H= 0.6 % R & NoC:
2.0 S5 D/H= 0.6 %
NR (no reinforcement):
5 cm
2.5 D/H= 0.2 %
3.0
0 50 100 150 200
Number of loading cycles, N (cycles)
b)
Figure 20. a) Peak earth pressure coefficients in the model tests and a field full-scale case; and
b) residual settlements of the backfill (when =0) by cyclic displacements of the facing and
effects of reinforcing the backfill (Hirakawa et al., 2006, 2007a).
The other detrimental effect of cyclic displacement of the facing with unreinforced backfill is gradual
but eventually large settlements in the backfill associated with the development of an active failure
plane in the backfill (case NR in Fig. 20b). In the experiments, the backfill settlement increased with
an increase in the cyclic facing displacement, (DA)/H. On the other hand, the backfill settlement
became nearly null when the backfill was reinforced with reinforcement layers connected to the back
of the facing (case R&C). Even slight heaving of the backfill took place, which was due seemingly to
dilatation of the backfill associated with repeated passive movements of the facing. The benefits of
reinforcing the backfill with reinforcement layers connected to the facing are as follow. Firstly, for the
269
same thermal thrust from the girder, the displacements of the facing become smaller due to higher
stiffness of the reinforced backfill. Secondly, for the same cyclic facing displacement, the residual
settlement in the backfill decreases drastically due to higher confining pressure in the backfill and
membrane effects of reinforcement layers connected to the facing. It may also be seen from Fig.20b
that these positive effects of reinforcing the backfill are largely lost when the reinforcement layers are
not connected to the facing (case R & NoC). This is because the deformation of the active zone cannot
be effectively restrained by the reinforcement layers.
5. Integration 4. Girder
2: GRS RW
2 3: FHR facing
- The numbers indicate

construction sequence.
a) 1. Pile - Not to scale
Bridge type Cost & period Maintenance Seismic

of construction cost stability Total
Conventional (gravity) A, B C, D 252 gal*
GRS RW C 589 gal*
Integral D, E 641 gal*
GRS Integral 1,048 gal*
b) (* Acceleration at failure in model shaking table test)

Figure 21. a) GRS integral bridge; and b) features of four different bridge types.
GRS integral bridge

Fig. 21a shows a new bridge type (called the GRS integral bridge), which is most cost-effective and
most dynamically stable among those described in this paper. This type combines the GRS RW bridge
(Fig. 17) and the integral bridge (Fig. 19) taking their advantages: i.e., the backfill and the facing are
stabilized by reinforcing the backfill with geosynthetic reinforcement connected to the facing (the
GRS RW bridge); and a simple and cost-effective RC bridge structure without using girder-supports
(the integral bridge), while alleviating their inherent problems. A GRS integral bridge may also need a
pile foundation to support the girder, but a lighter one than the integral bridge may be sufficient, as
needs for a pile foundation are usually low with GRS RWs. In particular, as seen from Fig. 20b, the
residual settlement of the backfill reinforced with reinforcement layers connected to the facing is very
small. Moreover, a high seismic stability with small deformation and displacements can be expected
because of integrated performance of the whole bridge system, as shown below.
Fig. 21b compares the advantages and disadvantages in the three factors listed in the top line of the
four bridge types: i.e., conventional gravity type, GRS RW, integral and GRS integral. The
accelerations shown in the second column from the right are those at which the respective bridge
models collapsed in the shaking table tests described below. Letters A through E denote negative
factors other than seismic stability with the respective bridge types: A= heavy abutment structure
because of a cantilever structure; B= need for a pile foundation; C= high cost for construction and
long-term maintenance of girder-supports; D= bump due to settlement of backfill by self-weight,
270
traffic load and seismic load; and E= settlement of the backfill and structural damage to the facing by
cyclic lateral displacements of facing due to seasonal thermal expansion and contraction of the girder.
The full point assigned to each factor is equal to three, which is reduced one by one when these
negative factors are relevant. In addition, the seismic stability is classified to three levels with points
equal to 1, 2 and 3. So, the total full point is equal to nine when free from all these negative factors A
E with the highest seismic stability, which is assigned only to the GRS integral bridge.
Model shaking table tests

Shaking table tests of the four bridge types listed in Fig. 21b were performed to validate a high-
seismic stability of the GRS integral bridge (Aizawa et al., 2007; Hirakawa et al., 2007b; Tatsuoka et
al., 2007b). Fig. 22a shows the GRS integral bridge model before shaking. Assuming a length
similitude ratio equal to 1/10, the facings were 51 cm-high and the girder was 61 cm-long. By adding a
mass of 200 kg at the center of the girder, the equivalent length became 2 m (i.e., 20 m in the assumed
prototype). Twenty sinusoidal waves with a frequency of 5 Hz was applied at the table while step by
step increasing the maximum acceleration max with an increment of 100 gal. Fig. 22b shows the
backfill settlements at 5 cm back of either the sill beam supporting the girder via a fixed support with
the GRS RW type (Fig. 17) or the facing with the other three types. Fig. 22c shows the lateral
displacements at the top and bottom of the facing. In Fig. 22c, with the GRS RW type, dt is the
displacement of the sill beam and in Figs. 22b and c, with the gravity and GRS RW types, the
displacements on the side supporting the girder via a fixed-support are presented. It can be readily
seen that the GRS integral bridge is much more stable than the other types, while the gravity type is
least stable. Moreover, the pushing out of the facing bottom is the major failure mode with the integral
and GRS integral bridges.
Girder 200 kg (an equivalent length = 2 m)
Box width: 60 cm 60 Dislodging of girder

Lateral displacement at top,
61 cm
Conventional
40 (gravity) GRS RW (displacement
dT (mm)
at the sill beam)

51 cm
GRS
20
Integral
Grid reinforcement L= 35 cm
35 cm Air-dried Toyoura sand Integral
(D r = 90 % ) 0
a)
Settlement of the backfill, S5 (mm)
0 60 dT Integral
Lateral displacement at bottom,
GRS
GRS RW Integral GRS
40 dB Integral
20
dB (mm)
Integral
S5 Conventional
40 Out of 20
measument (gravity)
Conventional range 5 cm
(gravity)
60 0
GRS RW
b) 0 200 400 600 800 1000 1200 0 200 400 600 800 1000 1200
Base acceleration, (gal) Base acceleration, (gal)
max
c) max
Figure 22. a) GRS integral bridge model; b) backfill settlement; and c) outward lateral displacements of
the facing in laboratory shaking table tests.
5 CONCLUSIONS
The recent developments of new types of soil retaining structures and bridge with backfill made in
Japan aiming at not only a high cost-effectiveness but also a high seismic stability were reviewed.
Several case histories of prototype soil structures, including those during earthquakes, and results from
271
model shaking table tests that validate high advantages of geosynthetic-reinforced soil retaining walls
with staged-constructed full-height rigid facing are presented. Among a couple of new bridge systems
with reinforced backfill, the advantageous features of the GRS integral bridge, which comprises of an
integral bridge and backfill geosynthetic-reinforced with geosynthetic reinforcement layers connected
to the facing are described.
AKNOWLEDGEMENTS
The authors express their sincere thanks to their past and present colleagues for their help in
performing this long-term research program at the University of Tokyo, Railway Technical Research
Institute and the Tokyo University of Science. The financial support from The Ministry of Education,
Culture, Sports, Science and Technology, the Japanese Government, and Japan Railway
Construction, Transport and Technology Agency is deeply appreciated.
REFERENCES
Aizawa, H., Nojiri, M., Hirakawa, D., Nishikiori, H., Tatsuoka, F., Tateyama, M. and Watanabe, K.
Validation of high seismic stability of A new type integral bridge consisting of geosynthetic-
reinforced soil walls, Proc. of 5th Int. Sym. on Earth Reinforcement (IS Kyushu 2007), 2007.
England, G, L., Neil, C, M. and Bush, D, I. 2000, Integral Bridges, A fundamental approach to the
time-temperature loading problem, Thomas Telford.
Hirakawa, D., Nojiri, M., Aizawa, H., Tatsuoka, F., Sumiyoshi, T. and Uchimura, T. Behaviour of
geosynthetic-reinforced soil retaining wall subjected to forced cyclic horizontal displacement at
wall face, Proc. 8th Int. Conf. on Geosynthetics, Yokohama, 3, 1075-1078, 2006
Hirakawa, D., Nojiri, M., Aizawa, H., Tatsuoka, F., Sumiyoshi, T. and Uchimura, T. Residual earth
pressure on a retaining wall with sand backfill subjected to forced cyclic lateral displacements, Soil
Stress-Strain Behavior: Measurement, Modeling and Analysis, Geotech. Symposium in Roma 2006
(Ling et al., eds.), 2007a.
Hirakawa, D., Nojiri, M., Aizawa, H. , Nishikiori, H., Tatsuoka, F., Tateyama, M. and Watanabe, K.
Effects of the tensile resistance of reinforcement embedded in the backfill on the seismic stability
of GRS integral bridge, Proc. of 5th Int. Sym. on Earth Reinforcement (IS Kyushu 2007), 2007b.
Huang, C.C. Seismic displacements of soil retaining walls situated on slope, Journal of Geotechnical
and Geoenvironmental Engineering, ASCE, Vol.131, No. 9, pp.1108-1117, 2005
Koseki, J., Tatsuoka, F., Watanabe, K., Tateyama. M., Kojima, K. and Munaf, Y. Model tests of
seismic stability of several types of retaining walls, Reinforced soil engineering, Advances in
Research and Practice, Marcel Dekker, Inc. (Ling et al. eds.), pp. 317-358, 2003.
Koseki, J., Bathurst, R.J., Guler, E., Kuwano, J. and Maugeri, M. Seismic Stability of Reinforced Soil
Walls, Proc. of 8th International Conference on Geosynthetics, Yokohama, Vol.1, pp. 51-77, 2006.
Nakarai, K., Uchimura, T., Tatsuoka, F., Shinoda, M., Watanabe, K. and Tateyama, M. Seismic
stability of geosynthetic-reinforced soil bridge abutment, Proc. of 7th Int. Conf. on Geosynthetics,
Nice, 1, 249-252, 2002.
Ng, C., Springman, S. and Norrish, A. Soil-structure interaction of spread-base integral bridge
abutments, Soils and Foundations, Vol.38, No.1, 145-162, 1998.
Shinoda, M., Uchimura, T. and Tatsuoka, F. Increasing the stiffness of mechanically reinforced
backfill by preloading and prestressing, Soils and Foundations, Vol.43, No.1, 75-92, 2003a.
Shinoda, M., Uchimura, T and Tatsuoka, F. Improving the dynamic performance of preloaded and
prestressed mechanically reinforced backfill by using a ratchet connection, Soils and Foundations,
Vol.43, No.2, 33-54, 2003b.
Tatsuoka, F. Roles of facing rigidity in soil reinforcing, Keynote Lecture, Proc. Earth
Reinforcement Practice, IS-Kyushu 92 (Ochiai et al. eds.), II, 831-870, 1992.
Tatsuoka, F., Tateyama, M, Uchimura, T. and Koseki, J. Geosynthetic-Reinforced Soil Retaining
Walls as Important Permanent Structures, Geosynthetic International, Vol.4, No.2, 81-136, 1997.
272
Tatsuoka, F., Koseki, J., Tateyama, M., Munaf, Y. and Horii, N. Seismic stability against high
seismic loads of geosynthetic-reinforced soil retaining structures Keynote Lecture, Proc. 6th Int.
Conf. on Geosynthetics, Atlanta, 1, 103-142, 1998.
Tatsuoka, F. Cement-mixed soil for Trans-Tokyo Bay Highway and railway bridge abutments,
Geotechnical Engineering for Transportation Projects, Proc. of GeoTrans 04, GI, Los Angels,
ASCE GSP No. 126 (Yegian & Kavazanjian eds.), pp.18-76, 2004.
Tatsuoka, F., Tateyama, M., Aoki, H. and Watanabe, K. Bridge abutment made of cement-mixed
gravel backfill, Ground Improvement, Case Histories, Elesevier Geo-Engineering Book Series,
Vol. 3 (Indradratna & Chu eds.), pp.829-873, 2005.
Tatsuoka, F., Tateyama, M., Mohri, Y. and Matsushima, K. Remedial treatment of soil structures
using geosynthetic-reinforcing technology, Geotextiles and Geomembranes, Vol.25, 2007a.
Tatsuoka, F., Hirakawa, D., Nojiri, M. & Aizawa, H., Tateyama, M. and Watanabe, K. A New Type
Integral Bridge Comprising of Geosynthetic-Reinforced Soil Walls, Proc. of 5th Int. Sym. on Earth
Reinforcement (IS Kyushu 2007), 2007b.
Uchimura, T., Tateyama, M., Koga, T. and Tatsuoka, F. Performance of a preloaded-prestressed
geogrid-reinforced soil pier for a railway bridge, Soils and Foundations, Vol.43, No.6, 33-50,
2003.
Watanabe, K., Tateyama, M., Yonezawa, T., Aoki, H., Tatsuoka, F. and Koseki, J. Shaking table tests
on a new type bridge abutment with geogrid-reinforced cement treated backfill, Proc. of 7th Int.
Conf. on Geosynthetics, Nice, 1, 119-122, 2002.
273
June 25-28, 2007
Paper No W1-1002
.
LARGE-SCALE SHAKE TABLE TESTS FOR EARTHQUAKE

GEOTECHNICAL ENGINEERING AT NCREE, TAIWAN
Tzou-Shin UENG 1, Meei-Ling LIN 1, Wen-Jong CHANG2, Chia-Han CHEN3, and Kuo-Lung
WANG4
ABSTRACT
This paper describes the experiments and researches performed in the area of geotechnical earthquake
engineering using the large 5 m 5 m shake table at the National Research Center for Earthquake
Engineering (NCREE) in Taiwan. A biaxial laminar shear box was designed and manufactured for
tests on a soil specimen of 1.880 m (W) 1.880 m (L) 1.520 m (H) under one- and multi-directional
earthquake shakings. With densely placed instruments, the soil responses, including pore pressure
changes, acceleration, strains, and settlements could be observed spatially and temporally. Researches
on soil-structure interaction and development of new in situ measurement devices are under way using
this laminar shear box. Another model box, 4.4 m (L) 1.3 m (W) 1.2 m (H), with fixed but
transparent boundaries was used to contain a model slope of a size of 0.5 m in height, 1.3 m in width,
and with a slope angle of 30 to study the slope stability under seismic shaking. The responses and
failure of the slope were observed and analyzed. Image analyses of the slope test results are to be
performed. Applications of these research results and future developments are also discussed.
Keywords: Shaking table tests, Earthquake, Sand, Liquefaction, Model slope.
INTRODUCTION
In order to study the soil behavior, such as liquefaction and soil-structure interaction, large soil
specimens have been placed on a shaking table that can reproduce the actual seismic ground shaking
according to the earthquake recording under either 1 g or centrifugal conditions. Thus, the responding
phenomena and behavior of a soil stratum as a whole, instead of a small soil element, under the more
realistic seismic loading conditions can be observed and analyzed. For a large size specimen, possibly
denser sensor placements with respect to the size of soil specimen provide well-distributed
measurements, spatially and temporally, for a better understanding of the behaviors of the modeled
soil and structures, especially during soil liquefaction and slope failure under shakings.
Large scale 1 g shaking table tests were performed on a 5 m 5 m shake table at the National Center
for Research on Earthquake Engineering (NCREE), Taiwan for researches in earthquake geotechnical
engineering. Two types of soil model containers were developed and used for different types of
experiments. One is a biaxial laminar shear box (1.880 m 1.880 m in plan and 1.520 m in height)
which was designed and manufactured for modeling a soil stratum of about 1.5 m in thickness under
one- and multi-directional earthquake shakings on a horizontal plane. The shear box is used for the
studies of soil liquefaction and soil-structure interaction during earthquakes. Another one is a fixed
1
Professors, Department of Civil Engineering, National Taiwan University, Taiwan, Email:
ueng@ntu.edu.tw
2
Assistant Professor, Department of Civil Engineering, National Chi Nan University, Puli, Taiwan.
3
Assistant research fellow, National Center for Research on Earthquake Engineering, Taiwan.
4
Postdoctoral research, Department of Civil Engineering, National Taiwan University, Taiwan.
274
boundary box with a size of 4.4 m (L) 1.3 m (W) 1.2 m (H) to be used for tests on models of
slopes and quay walls. The performances of these two boxes were tested and verified, and they were
found satisfactory (Ueng, et al., 2006, Lin and Wang, 2006). Experiments which had been performed
using these boxes on the shake table are described and the results of the tests and their analyses are
presented briefly. Applications of the results, future research activities, and possible collaborations
with other testing facilities are also discussed in this paper.
BIAXIAL LAMINAR SHEAR BOX
Figure 1 is the biaxial laminar shear box on the shake table at NCREE. It is composed of 15 layers of
sliding frames. Each layer consists of two nested frames, an inner frame (1880 mm 1880 mm) and
an outer frame (1940 mm 2340 mm). Both frames are made of a special aluminum alloy with 30
mm in thickness and 80 mm in height, except the uppermost layer that has a height of 100 mm. These
15 layers of frames are separately supported on the surrounding rigid steel walls with a gap of 20 mm
between adjacent layers. The 20-mm gap is provided to avoid rupture of the rubber membrane inside
the box during a large relative deformation between layers. Thus, a sand specimen of 1880 mm
1880 mm 1520 mm can be placed inside the inner frames. A 2-mm thick silicone rubber membrane
was placed inside the box to obtain a watertight container.
Figure 1. The biaxial laminar shear box on the shake table at NCREE
Linear guideways consisting of sliding rails and bearing blocks are used to allow an almost
frictionless horizontal movement without vertical motions. Each outer frame is supported by the
sliding rails built on two opposite sides of the outer rigid walls. The bearing blocks on the outer frame
allow its movement in the X direction. Similarly, sliding rails are also provided for each outer frame
to support the inner frame of the same layer such that the inner frame can move in the Y direction
with respect to the outer frame. With these 15 nested layers of inner and outer frames supported
independently on the rigid walls, the soil at each depth can move multidirectionally in the horizontal
plane without torsion. Figure 2 shows the schematic drawings of the biaxial laminar shear box.
275
1
2
3
4
5
sliding rails 6
inner frames 7
8
9
10
11
soil specimen 12
13
14
15
rigid base
outer frames Z
Y
shake table
X rigid walls X
(a) plan view (b) side view
Figure 2. Schematic drawings of the biaxial laminar shear box
LIQUEFACTION TESTS
Tests on clean sand

Series of shaking table tests on clean sand in the shear box have been conducted at NCREE since
August 2002. A fine silica sand from Vietnam was used to prepare the sand specimen for the shaking
table tests at NCREE. The maximum void ratio, emax, and the minimum void ratio, emin, range from
0.887 to 0.912 and 0.569 to 0.610, respectively, for different batches of sand used in the tests. The
permeability of the sand ranges from 0.04 cm/s to 0.1 cm/s for relative densities from 35% to 90%,
respectively. One- and multi-directional shakings including sinusoidal waves (with frequencies from 1
to 8 Hz and amplitudes, A max, from 0.03 to 0.15 g) and recorded accelerations, full and reduced
amplitudes, from various earthquakes were applied. Miniature piezometers and accelerometers were
installed within the sand specimen for pore water pressure and acceleration measurements at different
locations and depths in the soil during shaking. Transducers for displacements and accelerations were
also placed on different layers of the frames to record the movements of the specimen. Figure 3 are
the layout of instrumentation inside the sand specimen and on the inner frames. Large amount of data
were obtained from the densely placed sensors inside and outside the sand specimen.
Sand Surface 1
2
3
20 4
30 5
40 20 6
30 7
8
30 9
30 30 10
60 11
30 12
13
50 60 50 14
30 15
30
(a) within soil specimen (b) on inner frame
Accelerometer Piezometer Settlement Plate Coupled Sensor LDT
Figure 3. Layout of instrumentation within the specimen and on the frames (unit = cm)
The test results showed that a multidirectional shaking induced a higher excess pore water pressure
and it took a longer time to dissipate in a multidirectional shaking test than under a one-directional
shaking. Figure 4 shows the excess pore water pressure distribution along the depth of the specimen at
various time during 1-D and 2-D shaking tests. It indicates that one-directional shaking induced less
276
excess pore pressure and probably caused only a shallower liquefied layer, while the multidirectional
shaking caused a deeper liquefaction zone. A clear liquefaction depth can be seen in the large
specimen from the rather dense piezometer measurements. The depth of the liquefied sand can be
better determined to a degree of accuracy based on the measurements of mini-piezometers and
accelerometers on the inner frames as describe in Ueng et al. 2007b.
Exce ss p ore pressure head (mm) Excess pore pressure head (mm)
0 200 400 600 800 1000 0 200 400 600 800 1000
0 0
Time Tim e
200 1s 200 1s
3s 3s
5s 5s
400 400
7s 7s
10 s 10 s
13 s 13 s
600 600
16 s 16 s
800 800
1000 1000
1200 1200
(a) 1D shaking, A max = 0.075g (b) 2D shaking, A max = 0.075g
Figure 4. Excess pressure heads inside the specimen during shaking test, October 2004
According to the settlement measurements during and after the shaking tests, the settlements during
ose
when there is liquefaction of the soil. The settlements resulted from the multidirectional shaking were
larger than those under one-directional shaking in both cases of liquefaction and non-liquefaction of
the soil. The volumetric strain of the sand after liquefaction caused by shaking was calculated
considering the depth of liquefaction depth. With consideration of thickness of the liquefied sand, the
test results show that the volumetric strain after liquefaction, under sinusoidal shakings decreases with
the relative density of the sand regardless of the amplitude, frequency and directions of shaking.
Figure 5 shows volumetric strains after liquefaction under sinusoidal shakings with durations of 5, 10,
20 and 30 seconds in this study (Ueng et al., 2007b). It can be seen that the volumetric strain after
liquefaction increases with the shaking duration. The results show a good comparison with those
given by Tokimatsu & Seed (1987).
10
9 Shaking duration
5s
8 10 s
20 s
7 30 s
D = 10 line
6 D=5s
D = 20 s
Duration = 10 s
D = 30 s
5
Duratio n = 20 s
4
3 Duration = 30 s
1
Duration = 5 s
0
0 20 40 60 80 100
Dr (%)
Figure 5. Volumetric strain after liquefaction versus relative density of sand under various
shaking durations
277
Preparation and testing of Mailiao silty sand
Mailiao silty sand, a typical soil of the reclaimed lands in the western coastal area in Taiwan, was
used in this study. The particle shape of the Mailiao silty sand is mainly sub-angular and flaky. The
sand is more compressible and less dilatable than silica sand. Sample preparation method of the
Mailiao silty sand specimen in a large scale laminar shear box was evaluated. The fines content and
the large size of the sand specimen are the main concerns for obtaining a reasonably homogeneous
and representative sample of the in situ soil. A series of experiments in a smaller glass container were
conducted using various preparation methods, including wet sedimentation method using dry and
moist sand, staged sedimentation method and pre-mixture filled method. This study used the staged
sedimentation method that rather uniformly distributed fines contents at different depths except a top
thin layer of sand with a higher fines content. The moist Mailiao silty sand with a mass of 1.0 Mg and
an average water content of 7.2% were dumped down evenly through a specially designed pluviator
into the laminar box filled with water to a pre-calculated depth at first. After an interval of about 30
minutes for the primary sedimentation, the laminar box was filled with water slowly to the calculated
depth based on the measurements of water level and elevation of the sand specimen after the first step.
Then another stage of soil placement proceeded in the same manner. It took seven stages to complete
a Mailiao silty sand specimen of about 1.350 m in height. The uniformity, density and saturation of
the sand were checked by the thin-walled tube sampling and P-wave velocity measurements. The
results showed that both uniformity and saturation of the specimen were satisfactory by this
preparation method (Ueng et al., 2007a). Two series of one- and multi-directional shaking table tests
were performed on the saturated Mailiao silty sand in the laminar shear box. The preliminary results
showed that the relatively low permeability layer of the Mailiao silty sand resulted in a longer
duration of dissipation of generated pore water pressure and surface settlement compared with those
of the clean Vietnam silica sand. The sand boils were observed when there was liquefaction of the
Mailiao silty sand. Further analyses of the test results and more shaking table tests are under way.
SOIL-PILE INTERACTION
For the study of the soil-structure interaction in a liquefiable soil during one- and two-dimensional
shaking, physical pile model tests in the large biaxial laminar shear box is under way at NCREE. The
model pile is made of stainless steel with an outer diameter of 100 mm and the thickness of 3 mm.
The pile tip was fixed at the laminar shear box to simulate the condition of the pile foundation
embedded in the rock. Strain gauges and accelerometers will also be installed on pile surface to obtain
data relating to soil-pile interaction. The soil responses, including pore pressure changes, acceleration,
strains, and settlements will also be measured.
The first stage of tests is the lateral loading tests from reaction walls. The lateral loads will be applied
on the pile head, monotonically and cyclically, with and without soil in the shear box. The same
physical models will also be placed on the shaking table under one- and multi-directional shakings to
study the soil-pile interaction under 1D and 2D shakings. A mass or a multi-degree-of-freedom
structure will be added on top of a single pile or a pile group to evaluate the possible inertial effect on
pile performances. An inclined shear box will be set up to simulate the possible lateral spreading
effect on the pile under 2-D shaking in order to clarify mechanisms of soil-pile interaction in liquefied
and laterally spreading ground.
DEVELOPMENT OF COUPLED EMBEDDED SENSOR
The spatial variations of soil motions raise concerns of the shear strains evaluated of the ground under
seismic loading. In addition, pore pressures and shear strains measured at different locations can not
precisely capture the coupled responses between the induced shear strain and the generated pore
pressure due to time lag and/or non-uniformity of the soil. Therefore, development of an instrument
278
that can capture the coupled shear strain-pore pressure responses locally will be beneficial to
understanding of the process of pore water pressure generation and its relation to the shear strain
induced by earthquake loading.
To capture the coupled shear strain-pore pressure responses of saturated soils subjected to horizontal
shaking, a coupled sensor, which integrates a triaxial accelerometer and a miniature pore pressure
transducer in an acrylic case, is fabricated as shown in Figure 6 (Chang et al., 2007). The small size
(12 cm long, 5.5 cm in diameter) of the coupled sensor enables it to simultaneously measure both
particle accelerations and water pressure changes at the same location. Four coupled sensors are
embedded in the soil specimen in the laminar box to form a 0.6m by 0.6m square array on the vertical
plane parallel to the wave propagation direction, as shown in Figure 3(a). This array of sensors is used
to evaluate the shear strain within the array. The densely embedded single-function sensors and the
external instrumentation on the frames of the shear box at NCREE provide a testing platform to verify
the performance of the developed coupled sensors.
Figure 6. Coupled sensor
Soil accelerations and frame movements were processed to calculate shear strains. The 2D
displacement-based method (2DBM) uses the isoparametric formulation in the finite element
framework to calculate strains at any point within the instrument array (Rathje et al., 2005). The nodal
displacement components are computed by numerically integrating the vertical and horizontal
accelerations obtained by the coupled sensors. In a 1D shear wave propagation condition as the test
condition in this study, a soil deposit can be modelled as a stack of shear beams and the average shear
strain between two depths ( 12) can be evaluated by:
u y1 u y 2
12 (1)
z2 z1
where uyi represents the horizontal displacement of point i at depth of zi. It was found that the
accelerations measured by two coupled sensors at the same depth are essentially the same. That is, the
1D shear wave condition did occur as intended in the laminar shear box. The horizontal displacements
are computed by double integrations of accelerations measured by embedded accelerometers and the
strains within the sensor array can be obtained. The results were compared with the shear strains
calculated using the direct measurements of frame displacements by LDTs. The comparisons revealed
that two methods agree well for cases with shear strain level less than 2 10-2 %, when only small
excess pore pressure was generated. Two methods gave different strains when there is liquefaction.
Pore pressure data were processed to highlight characteristics of pore water pressure changes during
and after the shaking. The shear strains and pore pressure changes at the same location were
279
compared to observe their coupling behavior in the liquefiable sand. Time histories of excess pore
pressure ratio (ru = u/ v', where u = excess pore pressure and v' = effective vertical stress) at one
of the sensors and the shear strain computed according to the measurements by the coupled sensors in
a liquefied case are shown in Figure 7.
Excess Pore pressure ratio, ru
Stage 1 Stage 2 Stage 3 Stage 4
(sec)
Shear strain (%)
Note: Calculated by 2DBM
Time (sec)
Figure 7. Coupled response of a liquefied case measured by a coupled sensor
The developed embedded coupled sensors are being used in the large field soil-structure interaction
tests in the previously liquefied area at Taichung harbor. A pilot test that used the embedded sensors
to measure the free-field responses due to surface shaking had been conducted in 2006. The coupled
sensors were deployed in a 1D fashion and an eccentric vibrator with maximum output of 17 tons at a
frequency of 7 Hz is used to generate downward propagating shear waves. Although only about 20%
of excess pore pressure was generated at a depth of 6 m, the measured coupled behaviors during
shaking and subsequent dissipations are observed. Further tests in the future will include the pile-soil
interaction measurements in the liquefied reconstituted soils and field dynamic responses of wharf
due to surface dynamic sources.
SEISMIC SLOPE STABILITY
Figure 8 illustrates the fixed boundary box for the model slope on the shaking table at NCREE. The
inner dimension of the box is 120cm (H) x 130cm (W) x 440cm (L) and the weight of the box is
39,800 kN. The box has fixed boundaries with reinforced glass panels mounted in the longitudinal
axis. Sensors were installed at the four corners of the box to observe the performance of the box
including acceleration, velocity and displacement.
280
Figure 8. Fixed boundary box for the model slope
The specimen was prepared by two methods: compaction and pluviation. For the compaction
methods, the soil was prepared by mixing with water to 8 % water content, and then cured for 24
hours. The soil specimen was compacted as shown in Figure 9(a) into the model box using the
controlled-volume method to obtain a unit weight of 16.6 kN/m3. The slope surface was compacted by
a modeling tool to keep the slope angle at the designated value. For the pluviation method, the soil
was pluviated through the pluviation box (Figure 9(b)), which contains several layers of sieves to
assure that the granular particles drop vertically and evenly into the model box.
(a) Compaction method (b) Dry pluviation method
Figure 9. Construction of the slope specimen
Significant nonlinear responses and amplifications were observed with acceleration amplitude up to
0.6g in the specimen prepared with compaction method. The observed slope failure surface appeared
to be quite shallow and in the upper part of the slope. Although the specimen was carefully excavated
after the test, the shear surface inside the specimen could not be well-defined. For the specimens
prepared by the pluviation method, the amplification effect was not obvious and the slip surface is
shallower than the specimen prepared by the compaction method. However, the measured
accelerations inside the specimens showed that slip surface could reach deeper than that observed
through the transparent glass.
The testing processes were recorded by digital video (DV) and charged-couple device (CCD) during
experiment. In order to study the displacement behavior of slope under seismic force, the Particle
Image Velocimetry (PIV) analysis is adopted in this research. PIV is based on the processed images
recorded during the experiment. The general concept of PIV is to find the displacement vector of
281
particle from one image to another. The displacement vector can be transferred to velocity when the
duration between two images is known.
The computer program used in this research is PIVview 2C. The purpose of this analysis is to
understand the displacement behavior and the landslip initiation of slope during seismic loading. The
recorded images by digital video (DV) and charged-couple device (CCD) are 740 480 pixels and
640 480 pixels, respectively.
The images recorded of specimen prepared by compaction method are 740 by 480 pixels. The PIV
result for 32-33 seconds of a loading sequence is illustrated in Figure 10. The down-slope direction is
to the right hand side and the distance between dots is 10 cm. The slope starts moving from the slope
surface on the bottom of the image.
Figure 10. PIV results of specimen from 32 to 33 seconds (compaction method)
CONCLUDING REMARKS
Various types of large scale shaking table test for earthquake geotechnical engineering were
conducted using the biaxial laminar shear box and the fixed boundary box at NCREE. The large-scale
tests with larger soil specimens simulate closely the in situ situations under earthquake loading. With
proper instrumentation, these tests give important information for the understanding of the phenomena
occurring in the field. Appropriate analyses can then be performed according to the findings from the
large-scale tests.
Future studies at NCREE will include:

Studies of the liquefaction characteristics of soils with high fines contents, especially the local
soils in Taiwan;
Effect of multidirectional shakings on pile foundations, including bridge foundations;
Development of sensors including optical devices to be used in the field utilizing the shear
box on the shaking table to simulate the ground conditions for the purposes of verification and
calibration;
Further development of image analysis techniques, hardware and software, for large scale
tests.
Collaborations with other institutes can be achieved by:

Conducting specially designed experiments at NCREE for comparisons with the results
obtained by the centrifuge tests and full scale tests. It is hoped that the scale effects in these
282
tests can be better understood via these comparisons so that the testing and interpretation of
physical model tests can be improved;
Comparing the results of experiments using the biaxial laminar shear box at NCREE with
those from large one-dimensional shaking tests at other institutes to study the effect of
multidirectional shaking on geotechnical structures.
ACKNOWLEDGEMENTS
These researches are partly supported by the National Science Council, Taiwan and NCREE. The
technical supports and operational assistances in the shaking table testing including large specimen
preparation by the engineers at NCREE are gratefully acknowledged.
REFERENCES
Chang WJ, Ueng TS, Chen CH and Yang CW. Embedded instrumentation for coupled shear strain-
pore pressure response in multidirectional shaking table test, Proc. 4th International Conference on
Earthquake Geotechnical Engineering, Paper 1213, 2007.
Iai S. Similutude for shaking table tests on soil-structure-fluid model in 1-g gravitational field , Soils
and Foundations, 29(1), 105-118, 1989.
Lin ML and Wang KL. Seismic slope behavior in a large-scale shaking table model test,
Engineering Geology, 86, 118-133, 2006.
Rathje EM, Chang WJ, and Stokoe KHII. in situ dynamic liquefaction test
ASTM Geotechnical Testing Journal, 28(1), 65-76, 2005.
Ueng TS, Chen CH and Tsou CF. Preparation of a large Mailiao silty sand specimen for shaking
table test, Proc. 4 th International Conference on Earthquake Geotechnical Engineering, Paper
1339, 2007a.
Ueng TS, Wang MH, Chen MH, Chen CH and Peng LH. A large biaxial shear box for shaking table
tests on saturated sand, Geotechnical Testing Journal, ASTM International, 29(1), 1-8, 2006.
Ueng TS, Wu CW, Cheng HW and Chen CH. Settlements of Saturated Clean Sand Deposits in
Shaking Table Tests, in progress, 2007b.
283
WORKSHOP 2
Geotechnical earthquake engineering
related to monuments and historical centres
June 25-28, 2007
Paper No. W2-1012
SEISMIC RESPONSE ANALYSIS OF ANCIENT COLUMNS
Nikolaos ARGYRIOU 1 , Olga-Joan KTENIDOU2, Maria MANAKOU3 , Pashalis

APOSTOLIDIS4, Francisco J. CHAVEZ GARCIA5 and Kyriazis PITILAKIS6
ABSTRACT
This paper presents a numerical and an experimental study of the seismic response of ancient
columns. The multi-drum column analysed here corresponds to the Hellenistic portico of
Lindos acropolis. This structure was modelled using a finite element model. The simulations
were made in three dimensions. Extensive time domain parametric analyses were performed
in order to examine the response of the column subjected to seismic motion having different
values of peak ground acceleration and frequency content. The seismic input used consisted
of horizontal components of three real earthquakes with different frequency content. Three
different systems connecting the drums of the column are analysed in order to examine their
influence on the seismic response of the column. The analysis takes into account the complex
behaviour of the structure with the aim to determine the threshold PGA value before its
collapse. In a second stage we discuss available procedures to estimate the fundamental
period of a monolithic, a multi-drum, and two multidrum columns connected with an
architrave using microtremors measurements. The results of the microtremor measurements
are compared with the numerical simulations to assess the effectiveness of the procedure. It is
shown that microtremor measurements may be a useful tool to estimate the modal shapes of
ancient columns.
Keywords: ancient columns, numerical analysis, seismic response, microtremors, HVSR, SSR
INTRODUCTION
Ancient Roman and Greek temples used to be systems made of curved stones (drums) laid one on top
of the other without any connecting mortar. Mostly, gravity loads were carried to the ground by means
of monolithic and multi-block columns and colonnades, connected at the top with architraves.
Recently, these systems have received increasing scientific attention. The reason is their interesting
static and dynamic behaviour which derived from the ductility of the whole system. In addition,
shear transfer is achieved through friction and through wooden or metal pins shear keys- (called
polos) between the drums. In fact, this element should have a finite shear strength that does not
exceed the shear strength of the interface of the drums; it is a sort of early version of the capacity
1
Civil Engineer MSc, Department of Civil Engineering, University of Thessaloniki, Greece, Email:
nargi@teemail.gr
2
PhD Candidate, Department of Civil Engineering, University of Thessaloniki, Greece, Email:
ktenidou@civil.auth.gr
3
Dr Geologist, Department of Civil Engineering, University of Thessaloniki, Greece, Email:
manakou@civil.auth.gr
4
Dr Geologist, Department of Civil Engineering, University of Thessaloniki, Greece, Email:
pasapostolidis@gmail.com
5
Professor, Instituto de Ingeniera, Universidad Nacional Autnoma de Mxico, Mxico D.F. 04510
Mxico, Email: paco@pumas.iingen.unam.mx
6
Professor, Department of Civil Engineering, University of Thessaloniki, Greece, Email:
kpitilak@civil.auth.gr
284
design process currently used for the seismic protection of structures. The effectiveness of this process
is demonstrated by the fact that a large number of these monuments have survived historic earthquakes
which must have had undergone large peak ground accelerations.
The seismic response of these systems is a multi-parametric problem involving both material and
geometrical non-linearities. The structure response may go through several modes during the same
excitation, thus no modes exist with the classical meaning that is used in continuum mechanics
medium (Papantonopoulos et al., 2002). As a matter of fact, the modes of such systems depend on the
nature of contact existing between the interfaces of adjacent drums (after Dr. Kostas Zambas, personal
communication). As a result, the response is highly non-linear including relative sliding of the drums
and rocking. This complex response is caused by the polos behaviour and strength, the non-uniform
friction and contact surface and the dispersion of axial load developed during excitation at the contact
surface between the drums.
Over the past two decades many researchers have approached the phenomenon by studying the
behaviour of a rigid 2D block resting on a horizontal plane either experimentally, analytically and
numerically (Zambas, 1994; Shenton, 1996; Sinopoli, 1989; Yim et al., 1980, Zhang and Makris,
2001). In addition, complex numerical models simulating real cases were examined (Mouzakis et al.,
2002; Papantonopoulos et al., 2002; Manos and Demosthenous, 1997, Psycharis et al. 2000; Psycharis
et al., 2003; Makris and Konstantinidis, 2005). The main conclusions of these researches can be
summarized as follows:
The response of the system is very sensitive to very small changes in the parameters of the
system or the parameters of the excitation.
The results of a 2D study of the problem can be more conservative than those of a 3D study
Pure sliding, pure rocking, as well as a slide-rock mode may be observed during the response
of these systems during dynamic excitation.
An in-plane excitation can produce large displacements in the out-of-plane direction
The predominant period of the ground motion and the size of the structure are the main
parameters for stability, with low-frequency pulses being more dangerous as they increase the
possibility of collapse. Moreover smaller columns, while having the same aspect ratio as
larger columns, can be more unstable.
Free standing columns are more susceptible to earthquakes than colonnades connected with
architraves
Another interesting feature of the detailed study of the damage patterns and displacements of ancient
columns is that these analyses may improve the existing knowledge on the seismic history of an area
(Bottari, 2003) and through back analysis may allow estimation of the monuments current seismic
resistance level.
In this paper a comprehensive dynamic analysis of a multidrum column is presented using the finite
element method. As previously mentioned, it is extremely difficult to simulate the exact geometry of
the interfaces between the drums; thus, any attempt to study such a case with numerical or even
experimental approaches, remains a rough approximation. But still, these methods are always useful in
order to estimate the response even in a qualitative way. We examined the differences in the response
of the column to earthquakes having different frequency content and different PGA values. We also
investigated the differences in the response of the column due to the presence or not of several polos-
empolio systems. In addition, we made an effort to determine the threshold PGA value that would
cause the column to topple. Finally, microtremor measurements were conducted in three different
cases: a multidrum column, a monolithic column and a system of three multidrum columns connected
with an architrave. We propose a procedure for estimating the modal periods of the columns by
comparing numerical results with the results from the analysis of the microtremor measurements. The
aim of this paper is to contribute to our understanding of the seismic response of ancient columns,
which is an exciting topic of mechanics and dynamics.
285
NUMERICAL STUDY
Selected case study

The multidrum column of the Hellenistic Portico at Lindos Acropolis, Rhodes was selected for the
analyses. The column is 5.0 m high and is composed of seven drums and a capital. The diameters of
the drums range from 0.78m at the base to 0.62m at the top and each drum is around 0.68m high.
In ancient times, drums were connected with wooden or metal pins, a system called polos-empolio.
This system was used as a guide in order to achieve perfect contact between the drums. Recent
researches (Manos, 2006; Psycharis et al., 2003) have highlighted the role of these pins as shear keys,
able to dissipate seismic energy through fracture. The connection between drums is also important for
modern restoration works. In the present study two different polos systems were selected for the
analyses assuming different materials, either wood or titanium. The latter is used in current restoration
projects. In addition, the case without any polos system was also studied.
Analysis overview
Most of the numerical studies examining the seismic behaviour of ancient columns have used the
discrete element method. The reasons are the small computing cost and the possibility to efficiently
simulate large displacements. Under this approximation, the drums of each column are assumed to
behave as rigid bodies (Psycharis, 2003).
In this paper, all the analyses were conducted using the finite element program Ansys (2004) noting
that advances in computing software and hardware have enormously decreased the computing costs
associated to non-linear analyses. Contact elements were used to simulate the interfaces between the
drums. Static and dynamic coefficients of friction and cohesion can be used while the Mohr Coulomb
criterion was adopted for the analyses. Contact elements can also simulate the rocking response of the
drums, making it possible to describe accurately all possible motions of these blocks (Argyriou et al.,
2006). In addition, friction stress can be evaluated at each step of the computation. Table 1 shows the
parameters used to simulate the interfaces between drums and the mechanical parameters of the drums
themselves for the multidrum column of Lindos Acropolis (Eleftheriou, 2002). For the dynamic
coefficient of friction, a generally accepted value equal to 0.7 was considered.
Table 1. Mechanical properties of the interface and the stone

Shear Shear
Dynamic Modulus of
Density resistance resistance for
Cohesion coefficient of elasticity E for
(ton/m3) for wood titanium
friction stone (MPa)
(KN/cm2) (KN/cm2)
0 0.7 1044 2 0.7 31
The time step for the numerical integration was set to 10 -3s. Larger values led to numerical instabilities
(extremely large displacements or overlap between bodies). It was verified that the spatial
discretization step affected only the computing cost and not the accuracy of the results. The numerical
analysis performed in two stages. In the first one, the model was subjected to the acceleration of
gravity. This allowed for modelling the weight of the system, imposing the friction forces at each
contact element. In the second stage, the model was subjected to the input seismic signal, incident
from below, and the seismic response of the system was computed. The influence of the polos
system was included indirectly in the models by the artificial increase of the static coefficient of
friction s to a value equal to the shear resistance of the poles q (1).
s w( z) q
's (1)
w( z)
w(z) is the weight of the column (in KN) (according to the analysed interface) and q (in KN) is the
shear strength of the pole. A diameter of 4.0cm was considered for hard pine wood and of 1.0cm for
titanium. This simplification reduces the cost of the computation and allows a qualitative evaluation of
286
the effect of each system (Makris and Konstantinidis, 2005). Clearly, the physical phenomenon is
more complex (even fracture may occur) due to inelastic behaviour under recycling loading and the
possible appearance of gaps between the polos and its stone enclosure. The influence of the polos
system case to the shear resistance of each interface is shown in Figure 1 through the ratios of static
to dynamic coefficient of friction.
The damping was set equal to zero based partially to experimental results (Mouzakis et al. 2002) and
partially to the observed oscillation of the drum column.
Without polos Wooden polos Titanium polos

1.0 8.72 22.63
1.0 5.81 14.50
1.0 4.44 10.65
1.0 3.60 8.28
1.0 3.08 6.83
1.0 2.70 5.76
1.0 2.41 4.95
1.0 2.20 4.37
Figure 1. Ratios of static to dynamic coefficient of friction for each interface
Selection of the seismic input

The seismic input applied at the base of the model consisted of the horizontal component records of
selected earthquakes. Three earthquakes were chosen:
Aigion: 15/6/1995 00:15:49, Ms=6.2, Mw=6.5, Ml=5.8. Station: Telecommunication building,
Re=15km
Kalamata: 13/9/1986 17:24:35, Ms=5.8, Mw=5.9, Ml=5.5. Station: Old Telecommunication
building, Re=10km
Erzincan: 13/3/1992, Mw 6.9, Ms=6.8. Station: 95 Erzincan, Rrup=2km, Re=1km
The criterion for this choice was their different frequency content, as they will give information about
the response of the column in different period ranges. Figure 2 shows one of the horizontal
components of acceleration for each record. The records were retrieved from PEER and ESMD online
databases.
The accelerogram selected from the Aigion, Greece 1995, was recorded 15km from the epicentre. Its
PGA is approximately 0.5g, it was recorded on medium rigid soils and it is dominated by a 0.5sec
period pulse. The second record was obtained during the Kalamata, Greece, 1986 earthquake, was
recorded on medium to loose ground at a distance of 10km and has PGA of 0.27g. Its largest
amplitudes occur at three periods: 0.4, 0.7 and 1.3 sec. This earthquake caused considerable damage to
the buildings of Kalamata city. The final record was obtained during the Erzincan, Turkey, 1992,
earthquake. The record, obtained at 1km from the epicentre, has a PGA of 0.42g and its largest
amplitudes occur at periods at 0.3, 1 and 2 sec.
In order to estimate the threshold PGA value to collapse, a set of parametric analyses was carried out.
The column was subjected to increasing levels of horizontal acceleration.
0,5 0,3
0,3 0,2
0,1
0,1
0
-0,1 0 5 10 15 20 25 30 0 5 10 15 2
-0,1
time (sec)
time (sec) -0,2
-0,3
-0,3
-0,5
(a) (b)
287
0,6
0,4
0,2
0
0 5 10 15 20 25
-0,2
time (sec)
-0,4
-0,6
(c)
Figure 2. One of the two horizontal components of the earthquakes used as input:
(a) Aigion 1995, (b) Kalamata 1986, (c) Erzincan 1992
Analysis results
Because the response of a multidrum column can vary between several modes during excitation, a
reliable index of damage should be introduced in order to describe the final mode of the column. The
maximum permanent displacement relative to the base, as calculated at the end of each analysis, was
chosen as such an index. Figure 3 summarizes final displacements estimated for the three earthquake
records. For cases where the column did not collapse (stable modes), the drums where maximum
displacement occurred are shown. For cases where it collapsed (collapse mode), the drums from which
collapse begins are shown. In order to classify the possible response of a column for increasing PGA
values and thus define levels of damage, the column is subdivided along its height into two parts.
The first part contains all drums from the 6th up to the capital (green circles), while the second part
contains all drums from the 3rd up to 6th (red circles). It is noted that this response refers either to the
permanent displacement of a drum (for stable mode cases) or to the drum that collapsed (collapse
mode cases).
As can be seen for the Erzincan case (figure 3a), the column is stable for PGA values up to 0.2g, with
cases of failure beginning to occur for higher levels of excitation. The same is observed for all polos
system cases. The wooden polos system seems to have minor influence on the response of the
column, as the total displacements are not affected by it. The titanium polos system affects results by
reducing permanent displacements in most PGA levels. Moreover, for small values of PGA (scale
factor f=0.20-0.35) displacements are only noticed in the upper drums, namely the capital and 7th
drum, while the lower drums are excited only for higher values of horizontal acceleration (f=0.40-
0.80). In general it can be said that this second level of damage (red circles) is introduced when PGA
values are increased.
In Figure 4, the final response modes of the column are depicted for the three system cases when
excited by the Erzincan record. It is shown that for low PGA values all drums enter the rocking mode,
while for larger PGA values rocking occurs only in one drum, whereas the upper drums behave as a
monolithic body. Moreover, small twisting rotation of the capital was observed during the first
seconds of excitation for PGA values equal to 0.15g. The rotation was due to rocking. During the next
seconds little rocking of the 4th drum was noticed, which induced sliding and twisting rotation of the
3rd and 5th drum. For PGA values higher than 0.2g, the failure mechanism in the case of the no polos
system is due to extensive rocking of the 3 rd or 4 th drum. It must be noted that in this case failure
occurred at the drum above the one exhibiting the extensive rocking. Contrary to that, for the titanium
polos system, the failure mechanism started from the drum with the extensive rocking. In general it
is observed that the induced seismic force was dissipated between the 3 rd and 5th drum for most levels
of PGA.
288
capital Response concern
7th 3rd -5th drum
6th Response concern
Collapse modes 5th 6th drum - capital
4th
3rd
2nd
1st
base
Stable modes
Displacements (mm)
(a)
capital
7th
6th
5th
4th
3rd
2nd
1st
base
(b)
capital
7th
6th
5th
4th
3rd
2nd
1st
base
(c)
Figure 3. Comparative results of collapse and stable modes for
(a) Erzincan 1992, (b) Aigion 1995, (c) Kalamata 1986
289
PGA : 0.15g
PGA : 0.20g PGA : 0.26g
Rotation of the capital at
Intense rocking of the Tilting begins at 3rd
first seconds of the
2nd drum at first. drum at first.
excitation.
Intense rocking of the The motion transmits
Rotation of the 4th drum
3rd drum after a while fast to the 4th and 5th
produce sliding and which produce its drum and produce
Without polos
rotation of the 3rd and 5th

collapse. intense rocking and
drum in the next few
their collapse
seconds
PGA : 0.20g PGA : 0.41g

PGA : 0.15g
The response is almost Intense rocking of the
The response is almost
the same as with no 3rd drum. The motion
the same as with no
polos system case. doesnt transmit to the
polos system case.
The rocking is more upper drums.
Sliding of the 5th drum is
intense and produces A small bound of the
not produced here.
Wooden polos
the collapse of the 3rd drum was also

column earlier. produced.
PGA : 0.18g
PGA : 0.15g
PGA : 0.20g Intense rocking of the
The response is almost
Same response with 3rd drum. Contrary to
the same as with no
wooden polos system PGA=0.17g the
polos system case.
case. The 3rd drum motion transmits to
Few centimeters sliding
observes the maximum upper drum and
of the 5th drum is
energy of the produces the collapse
Titanium polos
produced in this case.

excitation. of the 4th drum.
Figure 4. Response shapes of the column for several PGA values for Erzincan case
290
PGA : 0.40g PGA : 0.45g PGA : 0.60g
Rocking of the 6th and 7th Almost the same Intense rocking of the
drum produces rotation mechanism with the 7th drum drives to its
of the capital. previous PGA=0.40g collapse. This motion
case. Small differences produces severe
In the next seconds are due to the fact that rocking of the 5th and
rocking transmits to the capital dissipates the 6th drum and finally
Without polos
4th drum and produces energy and not the 7th their collapse too.
small displacement. drum.
PGA : 0.40g PGA : 0.45g PGA : 0.60g

Contrary to no polos The rocking of the 7th Intense rocking of the
system case, more drum produces sliding 6th drum.
severe rocking of the 6th of the 6th drum and the Sliding of the lower
drum produces the capital. drums
collapse of the 7th drum The rocking of lower Collapse of the 7th
and the capital. drums coming drum and the capital.
Wooden polos
afterwards produce a
sinusoidal mode shape.
PGA : 0.35g PGA : 0.45g PGA : 0.50g
Intense rocking of 6th Severe rocking of 5th Rocking and rotation
drum that produces drum that produces of the 6th drum.
rocking and sliding of collapse of the 6th. Sliding due to rocking
the lower drums. The rocking of lower is produced to lower
drums coming drums.
afterwards produce
Titanium polos
smaller displacements
contrary to wooden
polos case.
Figure 5. Response shapes of the column for several PGA values for Aigion case
291
PGA : 0.29g PGA : 0.36g PGA : 0.40g
Rocking begins at 5th Sliding and rocking of Sliding and rocking of the 6th
drum and transmits to the 6th drum at first produced small displacement
the upper drums seconds of motion. of the 5th drum and the
Intense rocking of the capital.
It produces rotation of 2nd drum after a while Rocking of 2nd drum is not
the 7th drum and of the which produce rocking so intense (contrary to
Without polos
capital due to small of the 4th drum and PGA=0.36 case) and shared
tilting finally the collapse. also to the upper drums. The
column not collapses.
PGA : 0.33g PGA : 0.36g PGA : 0.55g
Rocking and sliding of Sliding and rocking of Intense rocking of the 6th
the 6th drum but with the 6th drum. drum produces the collapse
higher displacements Intense rocking of the of the 7th drum and the
(see no polos case). 2nd drum produces capital.
Rocking of the 2nd rocking of the 3rd drum The rocking of the lower
drum which follows (contrary to no polos drums that follows is softer
Wooden polos
doesnt produce many case) and finally the and produces a sinusoidal
displacements to the collapse. shape response of drums.
lower drums.
PGA : 0.33g PGA : 0.40g PGA : 0.55g
Rocking, sliding and Rocking, sliding and Intense rocking of the 6th and
rotation of 6th drum. rotation of 6th drum. the 7th drums produces their
Soft rocking is The final collapse.
produced also to the 4th displacements that The rocking of the lower
drum produced are 110% drums that follows collapse
bigger than those of is softer and produces a
no polos case.
Titanium polos
sinusoidal shape response of

drums.
Figure 6. Response shapes of the column for several PGA values for Kalamata case
292
In the case of Aigion earthquake, it is observed that the threshold PGA value for collapse cannot be
precisely defined. Collapse of the column is observed for a PGA value around 0.5g, while for higher
values the column could also prove stable (Figure 3b). The titanium poles reduced the displacements
for most PGA levels. From Figure 5 it is clear that a great part of the seismic energy is dissipated at
the capital and upper drums even for low values of PGA. The mode response shows rotation of the
capital by about 45 o due to rocking after the end of the excitation. Because of the intense rocking, the
7th and 6th drums also exhibit some displacement, of the order of 10cm (Figure 5). For greater values
of PGA (>0.40g), rocking is induced to lower drums, in particular at the 3rd one Finally, it was also
observed as in the previous case (Erzincan) that the 2 nd level of damage was reached for increasing
values of PGA. But here, the higher frequency content of the earthquake, contrary to the previous case
of Erzincan, distributed the seismic energy across each drum of the column. This is indicated by the
final displacement pattern which is also distributed over all drums.
Finally, neither for the Kalamata earthquake is it possible to determine a specific value of PGA for
which the column collapses. Collapse in most cases happens due to rocking of the lower drums,
namely the 5th and 6th, as is presented in Figure 3c. It has been also observed that upper drums are
more excited for PGA values lower than 0.35g.On the contrary for higher PGA values the lower drums
are excited the most. In the case of a system without polos, extensive rocking is observed at the 6th
drum during the first seconds of strong motion. During the next few seconds deflections are
transmitted to lower drums. In many cases that produced a sort of sinusoidal shape of the column. The
same was observed for the wooden polos system case (Figure 6). The titanium poles reduced
permanent displacements for most PGA levels. Extensive rocking was observed for PGA values
greater than 0.40g and more precisely to the lower drums, namely the 4th and 5th, which finally
produced the collapse of the column. For the other two polos cases the failure process started with
the collapse of the capital, followed by the 7 th or 6th drum. Distribution of energy over all drums was
also observed, probably due to the high frequency content of the input motion. That is probably the
reason for the sinusoidal shape observed.
EXPERIMENTAL ESTIMATION OF COLUMN RESPONSE THROUGH MICROTREMOR

MEASUREMENTS AND COMPARISON WITH NUMERICAL RESULTS
Selection of the case study

Ancient columns located at two different sites were used as case studies for microtremor
measurements. The first one is a free multidrum column and the second one is a system of three
multidrum columns connected with architraves. Both are located in the Acropolis of Lindos, Rhodes
(Figure 7a). In addition to these two cases, a monolithic column located in the ancient Agora of
Thessaloniki is the third case investigated (Figure 7b).
Geological setting of the investigated sites

Lindos Acropolis is situated near the middle of the west coast of Rhodes Island. It is laid on a
limestone rock outcrop hill of 80m high. The Ancient agora of Thessaloniki is extended on stiff clay
with average Vs of 450 to 550 m/sec. This layer overlies the schist rock basement at relatively low
depth, around 50m (Apostolidis 2002).
Ambient noise tests

Single station ambient noise measurements is an interesting method traditionally used for site
response analysis. It has been recently applied in structure response analysis as well. Microtremors
include random vibration of the ground caused by physical phenomena (wind, ocean waves, etc.) or
human activities (traffic, industrial noise and other ambient disturbances). It was found that it can
accurately describe a few basic vibration characteristics of ground and structures. (Aki, 1957; 1965).
The ratio of the horizontal to vertical Fourier spectra (HVSR) is a usual way to estimate the
fundamental frequency of a soil profile (Nogoshi and Igarashi, 1971). Although the significance of the
HVSR peak amplitude is still a subject of debate (Nakamura, 1989, proposed that it gave reliable
293
estimates of soil amplification, something which was later questioned by Kudo, 1995), the frequency
at which the peaks appear is considered reliable (Bard, 1999). Lermo and Chvez-Garca (1993)
introduced the use of HVSR computed for earthquake records to study site response and site effects.
More recently, noise measurements have been used for estimating the seismic response of structures,
investigating site-structure interaction and evaluating damage and vulnerability (Chvez-Garca and
Crdenas, 2002; Mucciarelli et al., 2004; Gallipoli et al., 2003; Irie and Nakamura, 2000; Volant et al.,
2002; Nakamura et al., 1995; Nakamura, 1997). Galli et al. (2006) used noise at a historical site in
Calabria, to estimate site effects and monument response. Noise measurement HVSR has been also
performed to study the response of historical monuments, such as the Coliseum in Rome (Nakamura et
al., 2000) and the leaning tower of Pisa (Nakamura et al.,1999).
Another empirical technique used here, and which also initiated as a site response study tool, is the
standard spectral ratio (SSR). This consists of the spectral ratio relative to a reference station. SSR has
been used to evaluate site effects since Borcherdt (1970). It is assumed that a record on the bedrock or
a nearby rock outcrop can be considered representative of the input motion below the sediments and
thus the ratio is related to site response of the soil profile. Again, this methods scope is extended in
this case so as to study the response of the structure, rather than the sediments. The sites used as
reference sites lay on rock outcrop, or almost.
5.70 m
6.70 m
(a) (b)
Figure 7.(a) Archeological site of Lindos in Rhodes and column case studies. (b)
Archeological site of Ancient Agora in Thessaloniki and column case study. The
column heights and locations of the recording instruments are also shown.
In the present study, ambient noise was recorded at the top of the columns under study and at a
reference site nearby. In the case of the Acropolis of Lindos in Rhodes, instruments were placed in
the configurations shown in Figure 8a and 8b: one at the top of a single column (a), one on top of the
architrave of the three-column-system (b) and one on the ground, rock outcrop (reference instrument),
at a distance roughly equal to the height of the column. In the case of the Agora in Thessaloniki,
instruments were placed as shown in Figure 8a: one at the top of a single column and one on the
ground, at a distance again roughly equal to the column height. The exact locations of the instruments
are shown with red arrows in Figure 7. It should be noted here that in the case of the architrave,
294
because only one instrument was placed on the structure, there is no way of estimating rotational
motion. Thus, all results can only be interpreted in terms of translation.
In both cases, ambient noise was recorded continuously for 30 to 35 min using three component
broad-band seismometers Guralp (CMG-40T of 30 sec natural frequency), coupled to a high resolution
Reftek recording system (DAS-130) and a GPS unit. The sampling interval of the records was
0.008sec.
(a) (b)
Figure 8. Schematic representation of the place of instruments
H/V Spectra Ratios have been computed using SAC (Goldstein et al., 2003). First, the records were
corrected for linear trend and constant offset. A band-pass filter with frequency corners from 0.1Hz to
25 Hz was applied and a representative noise window of 400sec time duration was chosen, avoiding
the obvious parasites and spikes in the records. This window was divided into 19 windows of 40sec
duration with a 50% overlap. Each window was cosine-tapered (10%) and Fourier transformed and its
amplitude spectrum smoothed with a running Hanning filter. Finally, the ratio between the two
horizontal components (EW, NS) relative to the vertical one was computed for every time window.
In agreement with previous results (Mucciarelli et al., 2005), it was proved that wind does not affect
the frequencies indicated by the HVSR while it increases the amplitude on all components. Thus the
results can be considered unaffected by any wind present during the measurements, at least regarding
the resonant frequencies. For all calculations related to the architrave case in Lindos, the NS and
EW components of noise recorded were rotated parallel and perpendicular to the architrave
direction.
We applied both HVSR and the classic SSR technique, where the fundamental response characteristics
are estimated through the ratio of the two horizontal components recorded on the structure and at the
reference site. Figure 9 shows the HVSR for the reference sites. Theoretically, a reference site lying
on rock outcrop is unaffected by any site effects which is reflected by an HVSR near unity over a
significant frequency range. However this is not always the case, as it is observed in the case of Lindos
outcrop site where the mean ratio is about 2.
10 10
1 1
0.1 0.1
0.1 1 10 0.1 1 10
f [hz] f [Hz]
(a) (b)
Figure 9. HVSR in case of (a) Agora Thessaloniki , (b)Lindos
Numerical response
295
Modal analyses were performed to investigate the eigen characteristics of the examined columns
which in both case studies were considered as monolithic Figures 10b and 11b present the results of
the modal analysis of the single column and the columns with architrave in Lindos. The range of
values of the fundamental frequencies is due to the range of values assumed for the elastic modulus
and the mass density of the column material. The parametric analyses considered different
combinations of the limestones elastic modulus (E) (considered to take any value between 1.40 and
1.55 GPa) and its density ( ) (taking values between 1.7 and 2.0 ton/m3). .
Comparison of experimental and numerical results: Lindos

Single column
Figure 10a shows the results derived from noise measurements using SSR for both horizontal
components. As expected, the two components are almost identical. Three distinctive peaks appear at
frequencies 4.0Hz, 6.5Hz and 7.0Hz.
On the same figure the results from numerical simulation (violet shaded rectangles) are shown,
indicating two distinguished frequencies range between 3.8-4.2Hz for the first mode and between 17-
21Hz for the second. The results for HVSR are shown in Figure 10c.
SSR and HVSR agree well for the peaks that appear at 6.5Hz and around 18Hz, while the HVSR
method completely misses the 4Hz peak which is visible only through SSR. This frequency is also
depicted numerically as the equivalent fundamental frequency (3.4Hz). One possible explanation is
that, for frequencies smaller than 5Hz, the vertical component is so well correlated to the horizontal
ones due to flexibility of the column that the ratio is not useful. Moreover it should be expected that
the fundamental mode and overtones should follow the fundamental expression fn=fo*(2n+1).
However the two peaks at 4.0 Hz and at 6.5 Hz do not follow this relation, suggesting that they
correspond to different vibration mechanisms. SSR is perhaps more appropriate to capture these
complex response compared to HVSR which has some inherent shortcomings.
f 1=3.25 4.25 Hz
100
90 HVSR(2)
80
HVSR(3)
70
60
50
40
30
20
10
0
0 5 10 15 20
f (Hz)
f 2=17.0 19.8Hz
(c) (b)
Figure 10. (a) SSR results (b) mode shapes of the free column (c) HVSR results of the free
column
Columns with an architrave
296
Figure 11a shows the SSR results obtained with ambient noise measurements for the architrave;
motion is oriented in parallel and transversely to the architrave (N32E, N58W respectively). The peaks
appear at frequencies of 4.0 and 7.0 Hz. On the same graph the numerical results are shown as ranges
for the first three modal shapes. The red shaded rectangle indicates the range of the frequency for the
1st mode. The green shaded rectangle indicates the range of the frequency for the 2nd mode. Finally,
the possible range for the frequency of the 3 rd mode is indicated with the blue shaded rectangle. The
parametric analyses accounted for the same combinations of the limestones mechanical properties as
for the single column case. l. Figure 11c, finally, shows the results from HVSR for both components
oriented, like the SSR, parallel and transversely to the architrave (N32E, N58W respectively). The
peaks appear at frequencies 4.0, 7.0, 11.0 and 17.0 Hz. It is of interest to note that the peak at 7.0Hz
disappears completely for the transverse component.
1200
SSR(2)
1000
SSR(3)
800
600
400
200
0
0 5 10 15 20
1st mode shape 2nd mode shape
f (Hz) f1=2.73Hz f2 =2.77 Hz
(a)
35
30 HVSR(2)-N58W
HVSR(3)-N32E
25
20
15
10
5
3rd mode shape 4th mode shape
0 F3=4.38Hz f2 =10.0 Hz
0 5 10 15 20
f (Hz)
(b)
(c)
Figure 11. (a) SSR results (b) mode shapes of the free column (c) HVSR results of the three
column case
Comparison of experimental and numerical results: Thessaloniki ancient Agora site

Figure 12a shows the results derived from noise measurements using SSR ratios for both components,
for the single monolithic column. As expected, the two components (2 and 3) are almost identical. As
has been already mentioned, the range of amplitudes is not significant; and we rely only on the
recorded frequency values. Two clear peaks appear at frequencies of 5.2 Hz and 6.8Hz.
In the same figure the computed results shown a range of frequency between 7.0-10.0Hz (green
shaded rectangle). This frequency range is reached through several parametric analyses, for different
combinations of the marbles mechanical properties. Namely, the elastic modulus (E) ranges from 45
to 60GPa while mass density ( ) ranges between 2.5 and 5.0ton/m3.
The results using HVSR are shown in Figure 12b. The experimental frequency of the first peak lies
around 6.4Hz, a value close to the second one given by SSR. Finally it must also remarked that the
experimental ambient noise measurements might include certain site effects due to the fact that the site
taken as reference does not show a flat response for HVSR in a frequency range of 8-10Hz.
297
100
SSR(2) Top-HVSR(2)
SSR(3) Top-HVSR(3)
10
10
1 1
0,1 0.1
0,1 1 f (Hz) 10 0.1 1 10
(a) (b)
Figure 12. (a) SSR results (b) HVSR results of the free column
CONCLUSIONS
The paper presents a set of numerical and experimental analyses in order to investigate the seismic
response of monolithic and multidrum ancient columns. The main conclusions are summarised as
follows:
From the three different input signals used, two of them did not allow to determine a specific
yielding value of PGA that would lead to the collapse of the columns.
Rocking occurs at all drum interfaces for PGA values around 0.2g-0.4g for rather high-
frequency, near-field earthquakes (Aigion and Kalamata case). Larger PGA values lead to
extensive rocking of the lower drums while the response of the upper part is rather monolithic.
The seismic energy is dissipated by the lower drums for increasing values of PGA.
For Erzincan earthquake record, showing an energy distribution over a wide range of periods,
the maximum dissipation of seismic energy through extensive rocking occurs between the 2nd
and 4th drum for most PGA values. The collapse mode that was observed (between the 2nd and
the 3rd drum) indicates a monolithic behaviour of the column. For the Kalamata record, whose
maximum energy occurs in the period range between 0.3 and 0.7 sec, the 4th and 5th drums
also seem to be affected by the excitation. However, the higher frequency content of this input
motion distributes the energy along the height of the column by dislocating more drums.
Finally, for the Aegion record, with a fundamental period of 0.5 sec, it was observed that the
6th drum suffered the largest dislocation and its rocking affected the motion of the other
drums.
The titanium poles reduced the displacements in most of the cases. For PGA values larger than
0.5g, however, the use of titanium produced a monolithic collapse of the column which is
generally undesired.
The observed twisting rotation was mainly due to rocking. It is observed that the twisting
rotations of the lower drums were enhanced by the extensive rocking of the upper ones.
Finally, a set of microtremor measurements were performed at two different archaeological sites for
three different column configurations. The goal was to estimate the eigen-properties of the column and
compare it with the numerical modal analysis. Both HVSR and SSR methods were applied. Distinct
peaks were clearly visible over specific frequency values which were quite close to the theoretical
modal analysis at least for the SSR. With HVSR it is more difficult to capture the probable
fundamental mode frequency. This is probably due to the fact that the vertical motion at the top of the
column is very similar to the horizontal one because of the large flexibility of the column. The results
derived from the numerical analysis were very similar to the results from SSR, especially in the case
of Lindos. Some discrepancies were noted, possibly due to site effects at the sites taken as reference.
In general it is concluded that noise measurements may give interesting tool to study the seismic
response of monuments and in specific for ancient columns. Given the low cost of this method, as well
298
as the fact that it does not cause any damage to the monuments, it may be a very useful tool to
investigate the seismic response of monuments and complement numerical calculations.
AKNOWLEDGEMENTS
The field surveys described in this paper were financially supported through the Pythagoras project
(Contract no. 97436). The authors express their thanks to Prof. Michalis Tiverios (Department of
History and Archaeology, AUTH), to archaeologist Lilian Acheilara (16 th Ephorate of Prehistoric and
Classical Antiquities, Thessaloniki) and to civil engineer Maria Pikoula (Committee of Restoration
Works Progress in Lindos Acropolis, Rhodes) for the measurements in the archaeological sites of
Thessaloniki and Rhodes. The first author (NA) would like to express his gratitude to Dr. Anastasios
Sextos (Lecturer, Department of Civil Engineering, AUTH) for his overall contribution; His initial
inspiration and the fruitful discussions proved very helpful in producing this paper. The first author
also thanks Prof. Marco Mucciarelli (Universita della Basilicata, Italy) for supplying relevant
literature. The second author (OJK) wishes to thank the Propondis Foundation for its financial support
through scholarship.
REFERENCES
Aki K., 1957. Space and Time Spectra of Stationary Stochastic Waves, with Special Reference to
Microtremors. Bull. Earthquake Res. Inst. Tokyo Univ. 25, pp. 415-457.
Aki K., A note on the use of microseismic in determining the shallow structure of the earths crust,
Geophysics Vol. 30, pp. 665-666, 1965.
ANSYS 8.1 (2004). Users Manual
Apostolidis P. Determination of ground structure with the use of microtremors
Application:Estimation of the dynamic properties and geometry of Thessaloniki soil formations.
Phd Thesis, Department of Civil Engineering, AUTH, 2002 (in Greek),
Argyriou N, Pitilakis K. Sextos A., Numerical study of the seismic behavior of ancient columns,
Proceedings of 1st Hellenic conference on Scientific Restoration Works, Thessaloniki 14-17 June
2006. (in Greek),.
Bard P.-Y. "Local effects on strong ground motion: Physical basis and estimation methods in view of
microzoning studies." Proc. of Advanced Study Course in 'Seismotectonic and Microzonation
techniques in Earthquake engineering', Kefalonia, Greece 1999.
Borcherdt, R.D. Effects of local geology on ground motion near San Francisco Bay, Bull. Seis. Soc.
Am. 60, 29-61, 1970.
Bottari C. Ancient constructions as markers of tectonic deformation and of strong seismic motions,
Proceedings of 11th FIG Symposium on Deformation Measurements, Santorini, Greece, 2003.
Chvez-Garca FJ and Crdenas M, The contribution of the build environment to the free-field
ground motion, Soil Dyn & Earthq. Engrg., 22, 773-780, 2002.
Eleftheriou V. Committee for the restoration of monuments at Lindos Acropolis, Rhodes.
Reconstruction works on the Hellenistic Portico. Athens 2002 (in Greek)
European Strong Motion Database (ESMD): http://www.isesd.hi.is/ESD_Local/frameset.htm, Mar.
2007
Galli P., Ruga A., Scionti V., Spadea R., Archaeoseismic evidence for a Late Roman earthquake in
the Crotone area (Ionian Calabria, Southern Italy):Seismotectonic implications. J. Seismol. 10, pp.
443-458, 2006
Gallipoli M.R., Mucciarelli M., Gallicchio S., Tropeano M., Lizza C. Structure, soil-structure
response and effects of damage based on observations of horizontal-to-vertical spectral ratio of
microtremors, Soil Dyn. Earthquake Eng. 24, pp. 487-495, 2003.
Goldstein P, Dodge D, Firpo M, Lee Minner "SAC2000: Signal processing and analysis tools for
seismologists and engineers. Invited contribution to "The IASPEI International Handbook of
Earthquake and Engineering Seismology", Edited by WHK Lee, H. Kanamori, P.C. Jennings, and
C. Kisslinger, Academic Press, London, 2003.
299
Irie Y., Nakamura Y. (). Dynamic characteristics of a r/c building of five stories based on
microtremor measurements and earthquake observations, Proc. of the 12th World Conf. on
Earthquake Engineering, Wellington (NZ), pp. 500-508, 2000.
Kudo K., Practical estimates of site response, State of the art report, in: Proc. of the 5th intern. conf.
on seismic zonation, Nice, Ouest editions Nantes, 3, pp 1878-1907, 1995
Lermo, J., and F.J. Chavez-Garcia. Site effects evaluation using spectral ratios with only one station,
Bull. Seis. Soc. Am. 83, 1574-1594, 1993.
Makris N Konstantinidis D,. Seismic response analysis of multidrum classical columns. Earthquake
Engng Struct. Dyn.,DOI:10.1002/eqe.378, 2005
Manos G., Demosthenous M., Models of ancient columns and colonnades subjected to horizontal base
motions Study of their dynamic and earthquake behaviour. Transactions on the Built
Environment vol 26, 1997 WIT Press
Manos G. The recycling behavior of polos-empolio, as a joint of sliding drums under earthquake
excitations, Proceedings of 1st Hellenic conference on Scientific Restoration Works, Thessaloniki
14-17 June 2006. (in Greek)
Mucciarelly M., Masi A., Rosaria Gallipoli M., Harabaglia P., Vona M., Ponzo F., Dolce M. Analysis
of RC building dynamic response and soil-building resonance based on data recorded during a
damaging earthquake (Molise, Italy, 2002). BSSA 94(5), pp. 1943-1953, 2004.
Mucciarelli M., Gallipoli M.R., Giacomo D., Nota F., Nino E., The influence of wind on
measurements of seimic noise, Geophysical Journal International, Vol. 61, issue 2, pages 303-308,
2005
Mouzakis HP, Psycharis IN, Papastamatiou DY, Carydis PG, Papantonopoulos C, Zambas C.
Experimental investigation of the earthquake response of the model of a marble classical column.
Earthquake Engineering and Structural Dynamics; 31(9):1681-1698, 2002.
Nakamura Y. A method for dynamic characteristics estimation of subsurface using microtremors on
the ground surface, QR Railway Technical Resource Institute 30, 1, 1989.
Nakamura Y., Hidaka K., Sato S., Tachibana M. Porposition of a method for pier inspection using
microtremor, Quarterly Report of RTRI, Vol. 36, No. 1, ISSN 0033-9008, 1995.
Nakamura, Y., Seismic vulnerability indices for ground and structures using microtremor, World
Congress on Railway Research in Florence, Italy, November 1997.
Nakamura, Y., Gurler, E.D. and Saita, J., Dynamic Characteristics of Leaning Tower of Pisa Using
Microtremor-Preliminary Results, Proc. 25th JSCE Earthquake Eng. Symposium, Vol. 2, 921-924,
1999
Nakamura, Y., Gurler, E.D., Saita, J., Rovelli A. and Donati S., Vulnerability investigation of Roman
Colliseum using microtremor, Proc. 12th WCEE, paper no. 2660, 2000
Nogoshi M and Igarashi T. On the amplitude characteristics of microtremor (Part 2), Journal of
Seismological Society of Japan, 24, 2640, 1971 (in Japanese with English abstract).
Orlandos AK. Construction Materials of the Ancient Greeks, vol.2. Archeological Society of
Athens: Greece, 1958 (in Greek)
Papantonopoulos C, Psycharis IN, Papastamatiou DY, Lemos JV, Mouzakis H., Numerical prediction
of the earthquake response of classical columns using the distinct element method. Earthquake
Engineering and Structural Dynamics; 31(9):1699 1717, 2002.
PEER Strong Motion Database (http://peer.berkeley.edu/smcat/search.html), Mar. 2007
Psycharis IN, Papastamatiou DY, Alexandris AP. Parametric investigation of the stability of classical
columns under harmonic and earthquake excitations. Earthquake Engineering and Structural
Dynamics; 29(8):10931109, 2000.
Psycharis I, Lemos J, Papastamatiou D, Zambas C, Papantonopoulos C, Numerical study of the
seismic behaviour of a part of the Parthenon Pronaos. Earthquake Engng Struct. 32:2063-2084,
2003
Shenton III HW. Criteria for initiation of slide, rock, and slide-rock rigid body modes. Journal of
EngineeringMechanics (ASCE); 122:690693, 1996.
Sinopoli A. Kinematic approach in the impact problem of rigid bodies. Applied Mechanics Reviews
(ASME); 42(11):S233S244, 1989.
300
Volant P.H., Orbovic N., Dunand F., Seismic evaluation of existing nuclear facility using ambient
vibration test to characterise dynamic behaviour of the structure and microtremor measurements to
characterise the soil: a case study, Soil Dyn. Earthquake Eng. 22, pp. 1159-1167, 2002.
Yim CK, Chopra AK, Penzien J Rocking response of rigid blocks to earthquakes. Earthquake
engineering and structural dynamics; 8: 565-587, 1980
Zambas K. Study for the restoration of the Parthenon. Committee for the preservation of the
acropolis monuments, Ministry of Culture, Athens 1994
Zhang J, Makris N., Rocking response of free-standing blocks under cycloidal pulses. Journal of
Engineering Mechanics (ASCE); 127(5):473483, 2001.
301
June 25-28, 2007
Paper No. W2-1015
Design and Implementation of Engineering Measures for the

Protection of a Historical Monument
at the Seismic Area of Mount Athos Peninsula, Greece
Stavros C Bandis1, Christos Schinas2, Elias Bakasis3
ABSTRACT
The paper refers to a major project of geotechnical studies, design and implementation of engineering
measures for the protection of the understructure of an important historical monument under
restoration, the Stavronikita Monastery at Mount Athos, Greece dating back to the 10 th century. The
area is part of a highly seismic zone of the Greek territory and earthquakes of intensity >7 on the
Richter scale are known to have occurred over the last century. The superstructure of the monument
has suffered considerable damages due to extensive cracking associated with slope movements. Site
investigation works were carried out for the recognition of the principal causes of instability and the
design of appropriate and adequate engineering measures. The implementation of the project presented
significant challenges because of the extent of the works, the construction limitations associated with
the site and the requirement to minimize environmental impact. The project was completed in 2000
and is being monitored by an appropriate network of instrumentation.
Keywords : seismic slope stability, Stavronikita Monastery
INTRODUCTION
Integrated engineering schemes for long-term protection of historical monuments invariably

encompass design issues associated with the effects of time and processes on the natural structural
materials of the superstructure and the ground forming the monument understructure.
The monastery of Stavronikita is located at the eastern coast of the Athos peninsula in northern Greece
and is an important historical monument dating back to the 10th century (Figure 1). It is a masonry
wall compound occupying an area 60x35m and is founded on the edges of a 35-45m high rock cliff
forming slopes 30o-40 o to the east and 50o-70o to the north.
FIGURE 1. Views of the northern side of the cliff of the Stavronikita Monastery, Mount Athos,
1
Professor, Department of Civil Engineering, Aristotle University, Thessaloniki, GR <sbandis@civil.auth.gr>
2
Lecturer, Department of Civil Engineering, Aristotle University, Thessaloniki, GR <schinass@civil.auth.gr>
3
Lecturer, Department of Civil Engineering, Aristotle University, Thessaloniki, GR <bacasis@civil.auth.gr>
302
The superstructure of the monument has suffered considerable damages due to extensive cracking
associated with slope movements both towards the northern and the eastern directions. Crack widths
were 50-350mm and exhibiting periodical widening. The cracks had a preferential orientation in two
principal directions namely E-W and N-S as shown in Figure 2, which includes typical views of major
cracking in the masonry walls and the foundation bedrock. Such effects were attributed to earthquake
damages, permanent deformations of the rock cliffs, natural wear of construction materials and the
construction history of the complex.
(a)
<
(b)
10m
(a)
(b)
(c)
FIGURE 2 . Plan view of the ground floor of the Stavronikita Monastery compound with
network of major cracks. Photographs illustrate cracks in (a) the northern wall
(b) the western wall and (c) in the rock mass at foundation level of the
north wall basement.
GEOTECHNICAL INVESTIGATIONS
The site investigation program comprised exploratory boreholes with orientated core recovery, joint
surveying, piezometric monitoring and extensive laboratory tests. Special in-situ tests were also
executed to derive shear strength parameters for critical discontinuities in the rock mass.
The rock cliff materials are schists and gneisses, with well-developed schistosity planes and joints.
Attributes of engineering significance were the variable degree of weathering / alteration, the intensive
jointing and the presence of continuous, open planes of schistosity.
Typical strength parameters for the rock mass components (rock material and discontinuities) were
measured and assessed as summarized in Table 1.
303
Table 1. Summary of rock mechanical properties of rock materials and discontinuities
Rock materials Rock discontinuities

Depth WG UCS E Description peak res IN-SITU TESTS (Fig. 3a
(m) (MPa) (MPa) (deg) (deg) &b)
0-5 IV 1-5 2000 Schistosity in 40o 30o Roughness
Grade II-III rock coefficient of schistosity
5-10 III 5-20 12000 Schistosity in 35o 25o JRC=6-7
Grade II-III rock Weathered schist
10-20 II-III 40-70 40000 Joints 50 o 30o c=60 kPa =35o-40 o
WG : weathering grade UCS : Uniaxial Compressive Strength, E: Modulus of Elasticity
peak : Angle of peak friction, res : Angle of residual friction
(a)
Application of normal load

by hydraulic jack resting
against an anchored
reaction beam
Anchors retaining
the reaction beam
Hydraulic jack
for application
of shear load
MULTI-STAGE TEST
Normal stress Shear stress
(b) 73 kPa 27 kPa
115 kPa 45 kPa
155 kPa 71 kPA
Figure 3. (a) In situ shear test on a schistosity plane for the determination of large scale
roughness and shear strength (block size 1000x600 mm)
(b) In situ shear test on 500x600 mm block for measurement of the shear strength of
weathered schist
304
SLOPE STABILITY MECHANISMS
From evaluation of all information the following conclusions were drawn relating to the principal
stability mechanisms:
The stability of the cliff slopes was controlled predominantly by the rock mass structure.
A major unstable wedge was identified at the eastern slope formed by schistosity and two major
structural features acting as release planes. The base plane (schistosity) of the wedge was below
sea level.
Multi-planar kinematic sliding mechanisms controlled the stability of the northern slope.
The structure of the rock mass has a dominant control of the overall deformation modes of the rock
cliff. Several normal faults running N-S and E-W intersect the rock mass and three faults were
identified cutting across the foundation bedrock of the monastery. Jointing is well-developed and
comprises three main sets, namely J1:0 o020 o/65o-75o, J2:0o-020o/45o-55 o, J3:320o/30o-40o. The
network of discontinuities combined with the slope faces creates different kinematic mechanisms
towards the north and the east sides.
Eastern slope
Orientated core recovery revealed several schistocity planes (dip direction / dip : 140 /25 ) and joints
(225 / 80 ) with slickensided surfaces at depth of ca 20 m (Figure 4c). The orientation of the
slickensides was 100 -130 /15 -25 and almost coincided with the line of intersection between the two
sets of discontinuities. The geometrical combination of the planes 140 /25 , 225 /80 and a known
fault isolated a 5-plane wedge volume sliding along the line of intersection of the above planes. The
outline of the unstable wedge at the eastern slope is illustrated in Figure 3.
The Factor of Safety (FoS) of the wedge with the toe terminating below sea level was calculated to be
close to limiting state (FoS=0.9) for seismic loading.
225 /80
(b)
FAULT
PLANE
N-S FAULT
DIRECTION
OF WEDGE Schistosity
MOVEMENT 140 /25
Direction of slickensides
coincides with the line o
intersection of schistosity
SCHISTOSITY with joint set 225 /80
140 /25
(a)
FIGURE 4
(a) Diagrammatic illustration of major wedge at eastern slope Scistocity plane
(b) Kinematic mechanism of slope instability (140 /25 )
(c) Slickensided schistosity plane
(c)
305
Northern slope
The stability is controlled by the joint sets J1 and J2 which promote planar or multi-planar sliding
mechanisms. The schistocity planes daylighting at the northern slope face, form the lowermost plane
of sliding with a shallow opposite dip (i.e. towards the slope), thus achieving a marginal static
equilibrium.
However, the extensive sliding failures that have occurred in the past (evident from the slope face
overbreaks as seen on the photograph of Figure 1) imply that seismic loadings and local toe
overstressing under static load may activate instability. The latter can be envisaged to develop rapidly
as a result of the structure controlled and gravity driven mechanism influencing a slope zone that
penetrates 3-4 m from the face and endangering the largest part of the northern wall of the compound.
Permanent deformation of the rock mass under past seismic events has resulted in notable loosening
(loss of interlocking) of the rock blocks close to the surface with associated significant reduction of the
shear strength and stiffness. A view of the disturbed rock mass is provided by the photograph in
Figure 2c, where the open cracks are dislocated joints of the set J1. The disturbed rock mass zone
extends at least down to the level of the slope undercutting (see Fig. 1) created by a local zone of poor
materials and wave erosion. The illustrations in Figure 5 present the distributions of the rock materials
and structure as modelled in the UDEC analyses, which will be referred to later.
Weathered
Highly schist
disturbed rock Schistosity
Gneissic
mass zone Schist
moderately
Slope weathered
undercut
J1
Schist J2
Slightly-
moderately
weathered
(a) (b)
Figure 5. Sections of the northern slope

(a) Distribution of geological materials and extent of disturbed rock mass zones
(b) Rock mass structure model simulated in the UDEC analyses.
STUDY OF AREAL SEISMICITY
The particular area is highly seismic as shown by the accumulation of epicentres during the last 100
years in Figure 6. Studies of the areal seismicity assuming a radius of 150 km yielded the seismic
parameters summarized in Table 2.
306
ZONE 17
Figure 6. Zonal map of seismicity and earthquake epicentres in the area of study
Table 2. Seismic characteristics of the area within 150 km from the site
Seismic 5 years 10 years 20 years 50 years
characteristics
I (MM) V V-VI VI-VII VII-VIII
amax 0.04 0.08 0.12 1.17
Vmax (cm/sec) 2.60 5.50 8.40 12.20
SHALLOW EARTHQUAKES
Time period Epicentre Magnitude
1901-1985 6.5-7.7
1911-1985 5.5.- 6.4
1911-1985 5.2 5.4
1950-1985 4.8 5.1
1964-1985 4.5 4.7
The seismic design parameters were selected as follows:
a = 0.12
Vhorizontal = 10 cm/sec (1932 earthquake)
Cs( shear wave velocity)= 2000 m/sec
STABILITY ANALYSES
Stability calculations were carried out on representative sections of the slopes in parallel with wedge
analyses. Factors of safety under static/ seismic loadings (a=0.12) were derived for the northern slope
as summarized in Table 3.
307
Table 3. Factors of safety of northern slope against multi-planar sliding failure
Initial state (undisturbed1 rock mass):

100t/m 100t/m 100t/m 100t/m 100t/m
FoS (static) = 1.2 - 1.9 (ave 1.35)
FoS (seismic)= 0.9 - 1.4 (ave 1.04) 600
600 0 1300 600
130
2 700
Current state (disturbed rock mass): 0
60 580
600 700
FoS (static) = 0.95 - 1.6 (ave 1.1) 1300
1300
FoS (seismic3) < 1.0 20 0
1000
00
100
100
500
1000
600
200 00
1000
1
The term undisturbed rock mass implies an initial inferred condition prior to the permanent deformation state and
associated loosening as currently observed.
2
The disturbedcondition was represented by assigning reduced shear strength and stiffness parameters to the rock mass.
3
The design stability calulations were carried out on the basis of Fhor =2 W Fvert =(2 W)/3 as suggested by the
Code for Antiseismic Design valid at the time (1991). At construction stage all calculations were carried out on the basis of
the revised Code (NEAK 1995) where Fhor =0.5 W and Fvert= 0.25 W.
At the eastern slope the 2-D and 3-D analyses demonstrated that the Factors of Safety were:
Static FoS = 1.20

Seismic FoS 1.00
The engineering assessment of all analytical studies converged to the conclusion that the stability state
and geotechnical conditions as the northern slope presented high risk of a catastrophic failure. At the
eastern slope, although the geotechnical conditions did not favour sudden destabilization, it was
evident that the periodical mobilization of slope movement due to earthquake activity and creeping
could prompt critical conditions to the structural integrity of the eastern wing of the compound.
DESIGN OF STABILIZATION MEASURES
The following design criteria were set for the selection of the stabilization scheme:
FoS (static) >2.0

FoS (seismic+ground water pressure) > 1.4
Control deformations under seismic loadings to < 20mm (ca 5o/oo x H)
The FoS were considerably higher than would be considered for a typical civil engineering project.
However, it was deemed necessary to upgrade the safety reserves due to the special nature of the
structure, the adverse seismic loading environment and the highly disturbed nature of the rock mass.
The stabilization measures selected after a rigorous investigation of different options comprised:
308
Perimetrical retention system consisting of rows of high capacity pre-tensioned anchors.
Reinforcement of the foundation rock mass with micropiles and grouting.
Underpinning of masonry structures and part of the wing walls.
Special retaining structures to support critical slope zones, especially at the northern slope
Figure 7 illustrates the general layout of the slope anchoring system.
NORTHERN SLOPE EASTERN SLOPE
Figure 7. General layout of the perimetrical anchoring and micro-piling systems
309
(a) Layout of micropiles for strengthening of the north wing foundation subgrade
(b) Section at basement of compound (c) Section of micropile founded

with details of micro-piling system pedestal structure at undercutting of
the northern slope
Figure 8. Details of engineering measures for improvement of the rock mass strength and
stiffness and provision of local support.
DESIGN VALIDATION AND IMPLEMENTATION
Due to the limited access to various sectors of the site during the SI and subsequent design stages,
systematic monitoring and interpretation of the conditions was undertaken during construction to
validate the design assumptions and engineering solutions. It was thus possible to optimize the
engineering measures at local scale by adapting the design to the conditions as exposed, particularly
for strengthening the foundation subgrade at the previously unaccessible basements of the northern
310
wing of the monastery. However, no modifications were required for the principal stabilization
scheme of slope anchoring.
The validated geotechnical models provided a reliable basis for further computational verification of
the adequacy of the engineering measures using state-of-art numerical modelling techniques.
The Distinct Element Method (DEM) was applied for rigorous seismic stability analyses of the slope-
stabilization system. The DEM is implemented in the Finite Difference Software UDEC-BB
(Universal Distinct Element Code coupled with the BartonBandis constitutive model for interface
shear and normal stiffness behavior.
A set of characteristic plots from the UDEC analyses demonstrating the predicted stabilization effects
of the remedial measures under seismic loading of the northern slope is presented in Figures 9 and 10.
Instant of seismic
excitation of model
at displacement
stage illustrated
0.6g below
20 mm
FAULT
ANCHORS
SLIDING PLANE 22 mm
OF WEDGE AT
EASTERN SLOPE
Figure 9. UDEC model predicted displacement state of eastern slope with fully deployed
Stabilization measures.
311
(a) (b) (c)
Figure 10. UDEC analyses of northern slope responses under the design earthquake:
(a) The model predicted slope collapse without stabilization measures

(b) Stable state was predicted (albeit below displacement threshold of <20mm) with partial application of the stabilization scheme (ex micropiles
and grouting)
(c) Stable state fulfilling the displacement requirements was predicted for the full stabilization scheme.
312
The implementation of the project presented significant challenges because of the extent of the works,
the construction limitations associated with the site and the requirement to minimize environmental
impact. Brief technical details of the construction stage are provided in the next section.
CONSTRUCTION STAGE
The stabilizations scheme involved the following quantities of the applied component measures:
Prestressed anchors of 400-800 kN capacity of permanent type (DCP: Double Corrosion

Protected), length of units 25-40m, total length 6200m.
Micropiles of 100-250mm diameter, length 5-12m, total length 7100m.
R/C structures of total concrete volume 1500 m3 and steel reinforcement 180000 kg.
Grouting of total volume 2000t.
Access and working platforms
The transport of plant and materials was carried out via sea till the Mont Athos port of Daphni and
thereafter by road to the jobsite. In order to avoid obstruction to the community of the monastery all
installations were independently installed at locations selected to minimize obstruction and impact to
the surround.
A significant technical problem was the access to the location of works. At the eastern slope access
was partly attained by limited earthworks but the main access and work platforms were provided by
scafoldings specially designed both to be adaptable to the constantly changing topography and to carry
safely the rig plant for the construction of the anchors.
The bearing platform of the scaffolds was constructed by using horizontal H beams which were fixed
to the rock face through an articulating joint and supported at the outer end by a retractable
(telescopic) prop connected to the beam through a joint (Figure 11). Thus a 2-hinged triangular frame
was formed, which could easily be adapted to the variable geometry of the slope face. The platforms
allowed safe access and operation of equipment of weight up to 6 t (Figure 12).
Hinged
connection
Telescopic
props
Horizontal H beams fixed

to the rock face through a
pinned joint
Figure 11 Deployment of scaffolding platforms at the northern slope
313
Figure 12 Installation of anchor in the northern slope
Anchors
Drilling of the holes for installation of the anchors was carried out by percussion drilling and using
lining tubes through the upper weak and loose materials. The drilling equipment used were a hydraulic
rig type CMV 420 and a heavier type Casagrande C6.
The anchors consisted of steel tendons (4-7No) and were manufactured as permanent support units
under double corrosion protection.
Micropiles
Those included four types with correspondingly different fuction:

(a) Vertical micropiles of 150mm diameter and 12m length installed perimetricallly close to the
foundation of the outer wall for local stitching of the dislocated rock blocks close to surface.
(b) Horizontal micropiles (150 mm) for distribution of the anchor loads where the rock mass was
particularly disturbed or weathered.
(c) Inclined micropiles for stitching of the highly disturbed rock mass at the basement level of the
northern wing (100mm diameter, 4-8m long)
(d) Vertical piles with structural reinforcement for foundation of the pedestal structure at the
undercutting of the northern slope.
Concrete structures
These included the anchor beams (csa 1000x800mm) with circa 100 kg reinforcement per linear
metre. A special arched pedestal structure was constructed at the undercutting of the northern slope.
The pedestal structure was designed to sustain ground loading even in case of loss of up to 50% of the
nerth slope anchors and hence the structural reinforcement requirements were significant (900 kg /
linear metre of column). Due to the adverse corrosion environment protection admixtures were used in
the concrete such as microsilica for reduction of permeability (50 kg/m3) and additional Ferrogard 901
(20 kg/m3) for protection against corrosion.
314
Grouting
Extensive grouting was undertaken in the northern slope for filling of the wide cracks across the
highly disturbed rock mass zones penetrating 2-3m into the rock mass. Grouting was carried out prior
to the prestressing of the anchors to work load levels. The grout material consisted of cement:water =
2:1 admixture with sand where the width of the cracks was large.
Bolting
A large number of bolts was installed for the protection of the staff and local stabilization. The bolts
were non-permanent 25mm diameter and 3-5m long.
Drainage measures
Those included installation of 10No drainage holes each of 15 m length at the northern slope and re-
construction of the drainage system of the monastery to avoid infiltration of waste water in the rock
mass.
Instrumentation
The following instruments were installed at selected locations:
21No electrical load cells for monitoring the anchor loads

6No extensometers
1No inclinometer for monitoring the overall deformation state of the north wall foundation zone
1No tiltmeter (north wall0
The instruments were connected to a data logger allowing tele-controlled transmission of

measurements.
The photographs in Figures 13-16 illustrate different stages of construction.
FIGURE 13 View of eastern slope with anchor system in place.
315
Figure 14. Strengthening of the foundation rock mass by micropiles and grouting and
underpinning of bearing columns and other masonry elements of the monastery compound.
316
Figure 15 Construction of pedestal structure at the undercutting of the northern slope
Figure 16. General view of stabilization measures at the northern slope

of the Stavronikita Monastery.
317
CONCLUSIONS
Historical monuments are invariably exposed to the influence of the geological and geotechnical
environment. Given the lifespan of such structures, several dynamic geological processes
(weathering / erosion, surface movements and earthquakes) usually have a dramatic impact on
the integrity of the monuments. Geotechnical Engineering has a dominant role in the study of
the natural causes of wear and damages of the structures and in the design of appropriate
interventions for effective and sustainable technical solutions.
The protection of monuments requires special approaches in terms of adaptation of the

engineering interventions to the historical environment and the lifetime of such interventions.
The significant cost and implicit long-term effectiveness of engineering schemes for the
protection of historical monuments necessitates integrated approaches requiring on-going
validation of the Design. The co-operation between the Designer and the Contractor during
construction and long-term performance monitoring are key components for the success of such
undertakings.
REFERENCES
Bandis, S.C., C.A. Schinas, G. Rizopoulos, K. Mantas (2001) : Stabilization of the Stavronikita
Monastery Cliffs Part A : Project Design Studies, Proceedings 4 th Hellenic Conference in
Geotechnical Engineering, June 2001(in Greek)
Rizopoulos, G., K. Mantas, S.C. Bandis and C.A. Schinas (2001) : Stabilization of the Stavronikita
Monastery Cliffs Part B : Project Construction, Proceedings 4th Hellenic Conference in Geotechnical
Engineering, June 2001(in Greek)
Note: Several unpublished references (in Greek) which contributed to the study of the project are included in the
reference list of the above publications and will not be repeated here.
318
June 25-28, 2007
Paper No. W2-1007
SEISMIC RESPONSE OF HISTORICAL CENTERS IN ITALY:

SELECTED CASE STUDIES
Antonio COSTANZO 1, Anna DONOFRIO2, Giuseppe LANZO3, Alessandro PAGLIAROLI4,

Augusto PENNA5, Rodolfo PUGLIA6, Filippo SANTUCCI DE MAGISTRIS 7, Stefania SICA8
Francesco SILVESTRI9, Paolo TOMMASI10
ABSTRACT
The paper is a comprehensive collection of up-to-date case studies of seismic response of historical
towns, located in dominant positions and affected by either ancient or recent strong-motion
earthquakes. The centers of Benevento, Orvieto, Nicastro, Gerace, San Giuliano di Puglia have been
characterised in terms of seismic hazard, subsoil model, ground motion simulation for maximum
historical earthquakes. The results of 1D and 2D simulations point out the specific influence and the
interplay of different factors, in particular the seismic input motion, the bedrock geometry, the soil
layering and the surface topography.
Keywords: Historical centers, Strong-motion earthquakes, Seismic response analysis, Site effects
INTRODUCTION
Much of the cultural heritage in Italy is kept in small to medium urban centers, rather than in big
cities. Since pre-Roman period to Middle Age, great part of such towns was edified on dominant
topographic locations, mainly for strategic reasons, because this choice gave more chance to safeguard
populations from the danger of barbarian attacks and invasions. However, the position of these urban
centers did not result so safe with respect to the vulnerability towards natural hazards, because they
often fall in orogenetic contexts prone to both hydrogeological and seismic actions (e.g. Fenelli et al.,
1998). Nowadays, the above factors claim for an enhanced degree of protection with reference to the
high exposition of monuments, historical buildings and people activities developed in these towns to
the damage by ground shaking and/or deformation induced by strong-motion earthquakes.
The Italian territory is fairly unstable and with frequent seismic activity, especially along the Apennine
chain. The map in Fig. 1 represents the peak ground acceleration, amax, expected with 10% probability
of exceedance in 50 yrs (i.e. with 475 yrs return period), as predicted by the most recent official
seismic hazard zonation in Italy (MPS Working Group, 2004).
1
PhD, DDS, University of Calabria, Italy, Email: acostanzo@dds.unical.it
2
Professor, DIG, University of Naples Federico II, Italy, Email: donofrio@unina.it
3
Professor, DISG, University of Rome La Sapienza, Italy. Email: giuseppe.lanzo@uniroma1.it
4
PhD, DISG, University of Rome La Sapienza, Italy, Email: alessandro.pagliaroli@uniroma1.it
5
Research Assistant, DIG, University of Naples Federico II, Italy. Email: aupenna@unina.it
6
PhD Student, DDS, University of Calabria, Italy, Email: rpuglia@dds.unical.it
7
Research Associate, SAVA, University of Molise, Italy. Email: filippo.santucci@unimol.it
8
Research Associate, DING, University of Sannio, Italy. Email: stefsica@unisannio.it
9
Professor, DDS, University of Calabria, Italy. Email: f.silvestri@unical.it
10
Researcher, National Research Council, Italy. Email: paolo.tommasi@uniroma1.it
319
Figure 1. Seismic hazard map of Italy (after MPS Working Group, 2004)
and landscapes of the historical centers selected.
The most hazardous areas (amax up to 0.3 g) are concentrated along the Central-Southern Apennine
chain, which is struck by strong motion earthquakes about once a decade. Since early 80s, after the
Irpinia earthquake (23.XI.1980, ML=6.9), a wide effort was dedicated by the Authors institutions to
applied research on earthquake geotechnical engineering problems related to the historical towns lying
in the most active seismic Apennine regions.
This paper summarizes and compares five examples of centers located on hilly areas (see Fig. 1),
which were struck by strong motion earthquakes in the past (Table 1), always with widespread damage
to buildings, and often with significant ground deformation or failure. For each of them, the seismic
response was seen to depend on the specific combination between earthquake characteristics and
subsoil morphology and stratigraphy. Accurate predictions or assessment of the seismic response and,
eventually, microzonation studies could take profit from advanced analysis of:
- the geological and geotechnical models, to characterize the subsoil vulnerability;
- the reference input motions, to represent the seismic hazard;
- the seismic response, to predict or assess the ground motion in the urban center.
Table 1. Seismic hazard data relevant to the analysed centers.

max historical earthquake
town (population)
Date MS IS,max
Benevento (63,026) 05.VI.1688 6.72 IX
Orvieto (20,909) 11.VI.1695 5.70 VII-VIII
Nicastro* (-) 27.III.1638 7.00 XI
05.II.1783 6.91
Gerace (2,932) VIII
28.XII.1908 7.24
S. Giuliano di Puglia (1,147) 05.XII.1456 6.96 VIII
*
Actually, the town inhabitants are about 50,000, but the area of interest is not populated.
320
All the following analyses were referred to deterministic seismic scenarios, defined by the maximum
historical earthquakes. In each case study, 1D or mostly 2D dynamic analyses were based upon
wide and accurate geotechnical characterization, which is, unfortunately, not the standard practice in
Italy. They will be discussed following the same order as in Table 1, i.e. for decreasing town
population and increasing age of the studies reported hereafter.
THE CASE HISTORY OF BENEVENTO
Site description and subsoil model

Benevento is a city of pre-roman origin covering an urban area of about 10 km2, located in the Sannio
region, about 200 km SE of Rome. It is a typical example of medium-size city representative of the
complex mediterranean reality, due to its high seismicity, its complex geological and geo-
morphological settings, and the presence of valuable monumental heritage, together with many even
recent buildings designed without seismic criteria. Therefore, since the early 90s, it was selected as a
test site for several hazard mitigation projects (e.g. Marcellini et al., 1995). The city rises on a hill
dominating the confluence of the Calore and Sabato rivers. Fig. 2 reports a geological map, including
the location of available geotechnical investigations, and a representative cross-section.
(a)
300 m a.s.l.
(b)
FLF
R Dt
ALG CR
AGA
200
Figure 2. Geological map of Benevento (modified after Improta et al., 2005) (a);
representative NW-SE section (b).
321
The historical centre lies on a hill made of a Pleistocenic (Rissian) conglomeratic slab (CR), overlying
a Pliocenic clay formation (AGA), the top of which lays tens to hundreds of meters below the ground
level. Towards the S-E side of the hill, fine to coarse fluvio-lacustrine deposits (FLF, FLG) overlay the
conglomerate formation. Terraced alluvia (GS), outcrop in the river valleys, and locally are covered by
sand and coarse grained recent alluvia (ALG). Debris/colluvial deposits (Dt) and man-made ground
(R), including large masonry blocks and archaeological ruins, are widespread throughout the whole
urban territory (Improta et al., 2005).
All the available geotechnical data, widespread on the whole urban territory, were georeferenced and
uploaded into a GIS (Penna, 2005). More than 260 boreholes were carefully analyzed; 30 down-hole
profiles were integrated with site-specific correlations between blowcounts (N SPT) and shear wave
velocity (VS). A spatial subsoil model (Fig. 3) describing the variation of VS throughout the whole
town center was obtained by means of a geostatistical approach (Penna, 2005; Penna et al, 2007).
Overall, the 3D geostatistical model shows that the limit between the soft shallower deposits (R, Dt,
ALG, with VS < 500 m/s) and the underlying stiffer formations (GS, CR, AGA, with VS > 500 m/s) is
very variable, and that it is often hard to identify a firm bedrock between the Pleistocenic
conglomerates and the Pliocenic clay formation.
Vs
(m/s)
800
650
500
360
180
Figure 3. 3D shear wave velocity model for the subsoil of Benevento (after Penna et al., 2007).
Table 2 summarizes the soil properties, either measured or assumed on the basis of literature data, of
the geotechnical units. The initial damping, D0, and the linear and volumetric threshold strain levels, l
and v, were used to define the G/G0- and D- curves relevant for the seismic response analyses
(Santucci de Magistris et al., 2004).
Reference input motion

Benevento falls inside the seismogenic zone Z927 Sannio - Irpinia Basilicata (MPS Working Group,
2004), that represents the most hazardous earthquake source for the city. Earthquakes in zone Z927
have a mode focal depth of 10 km and normal faulting mechanism. For a return period of 475 yrs, the
maximum acceleration amax expected in the city on outcropping bedrock ranges between 0.250 and
0.275 g (cf. Fig. 1).
322
Table 2. Geotechnical properties of Benevento subsoil.
VS D0 l v
Soil type
(kN/m3) (m/s) (%) (%) (%)
Man-made ground (R) 17 130 300 5 0.001 0.01
Debris colluvium (Dt) 18-19 250 -330 3 0.0029 0.0371
Recent alluvium (ALG) 18-19 200 500 2 0.001 0.01
Terraced alluvium (GS) 19-21 630 -750 1 0.002 0.02
Fine fluvio lacustrine (FLF) 19 700 3 0.005 0.05
Coarse fluvio lacustrine (FLG) 20 300 500 1 0.002 0.02
Weathered 20 600 800 1 0.0015 0.042
Rissian conglomerate (CR)
Cemented 23 800 900 0.5 0.02 0.2
Shallow 21.3 450 600 3 0.009 0.103
Pliocenic clay (AGA)
Deep 22.5 800 2 0.009 0.106
The maximum historical event that struck the city was the 1688 Sannio earthquake with I max=XI MCS,
moment magnitude MW= 6.72 and epicenter located around 40 km N of Benevento. This earthquake,
featured by a return period Tr 380 years, was adopted as the reference input motion for dynamic
seismic response analyses. The ground motion at rock outcrop was simulated using a hybrid statistical-
deterministic approach, producing three sets of 150 synthetic seismograms, one for each motion
component (NS, EW and UD). The average response spectrum of the NS components of the full
dataset is plotted in Fig. 4, together with the standard deviation (red lines). An appropriate cluster
analysis (Aliperti, 2006) allowed to reduce this numerous dataset to 8 most representative
accelerograms, the average spectrum of which ( ) is also plotted in the same figure (blue lines). Note
that the higher the number of signals, the lower the standard deviation.
2.00
1.75
1.50
1.25 8 signals; avg.

150 signals - average
8 signals - average
1.00
150 signals; avg.
0.75
0.50
0.25
0.00
0 0.5 1 1.5 2
Period, T(sec)
Figure 4. Average response spectra and confidence intervals for the accelerograms simulating
the 1688 Sannio earthquake (after Aliperti, 2006).
Seismic response analysis

The 3D geostatistical subsoil model in Fig. 3 was used for a preliminary seismic zonation of
Benevento, based only on site classification criteria (Grade-2 according to ISSMGE-TC4, 1999). As
suggested by Eurocode 8 (EN 1998-1 2003), each surface point was associated to a site-specific
acceleration spectrum, consistent with the equivalent shear wave velocity VS,30 , this latter computed
along the vertical profiles of the geostatistical model. Indeed, this procedure lead to a quite rough
microzonation map, consisting in an almost homogeneous distribution of site-specific seismic actions
throughout the whole urban area (Santucci de Magistris et al., 2004; Penna, 2005). As of matter of
323
fact, the EC8 ground types do not cover adequately and uniquely all soil conditions which are often
encountered in practice, in terms of both typical soil layering and definition of the classification
parameter, VS,30. Following the suggestions by Bouckovalas et al. (2006) addressing some possible
modifications of the EC8 provisions (Fig. 5a), an upgraded zonation of the Benevento territory
resulted as mapped in Fig. 5b.
a) b)
A1 C
A2 D
B E
Figure 5. Definition of ground types according to Bouckovalas et al (2006) (a) and site
classification map of Benevento (b).
A Grade 3 zonation (ISSMGE-TC4, 1999) was finally carried out, by dynamic response analyses with
reference to the deterministic seismic scenario above defined. 1D linear equivalent analyses were
carried out using EERA code (Bardet et al., 2000) on 28 sites widespread all around the city (see Fig.
6). 16 of them corresponded to the locations of DH VS measurements (red flags in Fig. 6), while in the
remaining sites (green triangles) the VS profile was obtained from the geostatistical model. Site-
specific response spectra were obtained by averaging the results of seismic response analyses with the
clusterized accelerometric dataset as input motion signals. Finally, seven homogeneous zones were
identified by grouping together the sites with similar average spectra.
Zone 1 is characterized by relatively low amplification in the whole range of periods, due to the
limited seismic impedance ratio. In Zone 1 *, higher peak spectral accelerations occur for T = 0.25 0.4
s, since the deformable soil layers are deeper than in zone 1. Moderate spectral amplifications over a
wide range of periods are computed in Zone 2, where deformable soils are presents at higher depth,
and a medium impedance ratio can be evaluated along the shear wave velocity profiles. Zone 3 is
characterized by some significant spectral amplification for periods spanning between 0.2 and 0.6 s,
while the VS profiles indicate large impedance ratios. In Zone 4, where deeper deformable formations
and large impedance ratios are detected, the spectral acceleration reaches even 2g for periods T=
0.2 0.3 s. A quite large spectral amplification, concentrated in a narrow frequency range characterizes
the spectral response of Zone 5, due to the shallow bedrock and large seismic impedance ratio. A
similar situation is finally found in Zone 6, where the highest spectral accelerations (up to 2.5g!) have
been computed for a period of 0.2s.
By comparing Figs. 5 and 6, it appears that some important details in the zonation were not
appreciated even using an improved Grade 2 approach. This could be due to the presence of the stiff
levels somewhere overlying the Pliocenic clay formation assumed as bedrock. Its is worth nothing that
the Grade 3 zonation map resulted quite compatible with the indications of standard spectral ratios of
micro-earthquakes and ambient noise recordings (Improta et al., 2005) and the site intensity
distribution observed after 1688 Sannio earthquakes (Castenetto and Romeo, 1992).
324
3.00
2.50
Zone 1
Ponte
2.00
Galanti
Asse S10
1.50 Asse S3
Viad. SS3
1.00 Ponte 2
Input motion
0.50
0.00
0 0.5 1 1.5 2
Period, T (sec)
3.00
2.50 Zone 1*
2.00
Rufina
1.50 Asse S1
Input motion
1.00
0.50
0.00
0 0.5 1 1.5 2
Period, T (sec)
3.00 3.00
2.50 Zone 2 2.50 Zone 3
2.00 2.00
Avellola AT1 IND_SOND 113
Avellola AT3 IND_SOND 190
1.50 1.50
AvellolaAT2 IND_SOND 260-261
Input motion Input motion
1.00 1.00
0.50 0.50
0.00 0.00
0 0.5 1 1.5 2 0 0.5 1 1.5 2
Period, T (sec) Period, T (sec)
3.00 3.00
2.50 Zone 4 2.50 Zone 5
2.00 Avellola AT6

2.00
Viad. SS11
Avellola AT8
1.50 IND_SOND 243
Viad. SS9 1.50
IND_SOND 253
Avellola AT4
IND_SOND 250
1.00 Input motion
1.00 Input motion
0.50 0.50
0.00 0.00
0 0.5 1 1.5 2 0 0.5 1 1.5 2
Period, T (sec)
3.00 Period, T (sec)
2.50 Zone 6
2.00
Guerrazzi
1.50 IND_SOND 8
IND_SOND 182
IND_SOND 16
1.00
Input motion
0.50
0.00
0 0.5 1 1.5 2
Period, T (sec)
Figure 6. Grade-3 seismic zonation of Benevento: map and site response spectra grouped for
different zones.
325
THE CASE HISTORY OF ORVIETO

The upper part of the Orvieto hill is formed by a slab of lithic tuff and weakly cemented pozzolana
delimited by subvertical cliffs up to 60 m high. The slab is approximately elliptical in shape, with
maximum length and width of 1500 m and 700 m in the WE and NS directions, respectively (Fig. 7).
The lower part rests on a truncated cone carved in an overconsolidated clay formation, with a natural
slope about 100 m high and an average inclination of 18 , overlying the cenozoic bedrock located at a
minimum depth of 200 m. A 5 to 15 m thick succession of weakly cemented silts and dense
sands/gravels (Albornoz formation) is interposed between the slab and the clay deposit. The slow
deformation and yielding of the clay slope produced by the erosion induces severe tensile stresses at
the cliff foot, determining the fracturing of the slab margin which is subsequently involved in
instability phenomena. Major instabilities consist of collapses of pozzolanic spurs and by sliding of
large regularly jointed portions of the cliff. More frequently, block falls are produced by toppling,
sliding and tensile failure along sub-horizontal planes of tuff and pozzolana slices (Manfredini et al.,
1980; Tommasi et al., 1996). The interpretation of historical data and a comparative analysis of
seismic and landslide catalogues reveal that a number of falls can be related to seismic events.
However, damages to structures are the most frequent seismic effects reported in historical sources.
San Benedetto
Stat
e
Um road
b ro
-C
a se
CHP n tin
e se
A'
gully
S7
ne
y
ll
gu
Fosso C avaro
ero
m s.l.m.
A A'
Z
an
S
RU 7
Pozzo d i San Patrizi o RU10 POZZO DI S. PATRIZIO
300
San D omen ico

200
Fo
ss
RU 10 o
de
Pal azzo de i ll a 100
Ca pitan i del Po pol o BR C
i ve
tta
Clay Albornoz Pyroclastic Landslide
lly
Porta Mag gio re

materials
gu
m s .l.m. debris
CATHED RAL
d ia
S ta
Um te R
ba
CHR RU1 RU10

bro oad
Porta
B B'
Ab
300
-C
as
ent
Ro mana
San Fran cesco N RU7
in
A ese
RU 1 CHR S7 CHP
200
100
B 0
0 400 800 1200 1600 2000 m
Figure 7. Plan and geological sections of Orvieto hill (modified after Manfredini et al., 1980).
Geotechnical data on pyroclastic materials were collected at the southern cliff: boreholes were drilled,
static and dynamic properties of materials were determined on the recovered cores and shear wave
velocities were measured by a cross-hole test (Tommasi and Ribacchi, 1998; Rotonda et al., 2002).
The stiff in-situ clay formation is covered by a slide debris blanket of remoulded and oxidized clay. At
the top, the stiff clay is softened and fissured, with closely spaced joints which progressively disappear
with depth. The thickness of the different clay layers varies widely throughout the slope. The
geotechnical parameters of pyroclastic and clayey materials are summarised in Tables 3-4.
Table 3. Geotechnical properties of the Orvieto cliff materials

Red Competent
Material Weak pozzolana
Tuff pozzolana
Bulk dry density (Mg/m3) 1.18 1.15 1.06
Shear wave velocity laboratory 1150 620 540
(m/s) in situ - 550-650 470
Table 4. Geotechnical properties of the Orvieto slope materials

Clay material Stiff Softened Remoulded
Bulk density (Mg/m3) 2.14 2.01 2.0
Natural water content (%) 20 23.5 27
Liquid limit 50 50 53
Plasticity Index 21 20 24
Shear wave velocity (m/s) 560-590 300-450 200-210
326
The geotechnical model along the longitudinal (WE) and transversal (NS) directions is sketched in
Fig. 8. From top to bottom, the slab consists of a 20 m weak pozzolana layer, resting on a 10 m layer
of competent pozzolana, which in turn overlies a 30 m tuff layer. Beneath the slab, a 200 m thick stiff
clayey substratum is superimposed on the bedrock. The different layers were characterised by constant
values of bulk unit weight, , and shear wave velocity, VS, assigned on the basis of the above
mentioned laboratory and cross-hole tests. It is worth noting that the product of Vs and values yields
singular variations of impedance ratios between in-contact materials. More in detail, the clay/tuff
impedance ratio is 1.1; conversely, impedance ratio values for tuff/pozzolana and competent/weak
pozzolana are 1.24 and 1.27, respectively. This implies that the larger impedance contrast is not at the
clay-slab interface, but between the volcanic materials within the slab. The G/G0- and D- curves for
each material were taken from experimental data published in the literature on similar materials
(Muzzi et al., 2001).
1400 m (E-W cross-section)
700 m (N-S cross-section)
300 m 60 m
100 m
100 m
3
Bedrock 25,5 kN/m , VS =2200 m/s
Pozzolana More cemented pozzolana
13,3 kN/m 3 ,VS =530 m/s 13,3 kN/m3 , VS =660 m/s
Tuff Stiff clay
12,5 kN/m3 ,VS =900 m/s 3
21,0 kN/m , VS =600 m/s
Figure 8. Geotechnical model for numerical analyses (modified after Muzzi et al., 2001).

As documented by historical sources, Orvieto suffered the effects of near-field as well as far-field
earthquakes. The strongest near-field event was the 1695 Bagnoregio earthquake (de = 10 km, I 0 = IX
MCS) which also produced the highest site intensity at Orvieto (I S = VIII MCS). Most of the events
were triggered by far-field sources (de = 70-80 km), located in the Umbria Apennines. An analysis of
the historical seismicity of Orvieto and a seismotectonic study of the region were carried out for
evaluating a credible rock outcropping motion (Muzzi et al., 2001). Two controlling near-field (NF)
and far-field (FF) seismogenic zones were identified, which are characterised by de 20 km, Mmax =
5.9 (NF) and d e 70 km, Mmax = 6.7 (FF). Several rock outcropping accelerograms were selected from
the world-wide databases matching the required Mmax and de values. Two records producing the
maximum ground motion at the site in the frequency range of interest were chosen: for the near-field,
the NS horizontal component recorded at Cascia station during the 1979 Valnerina earthquake (Italy),
and the NS horizontal component at Butler Valley station during the 1992 Cape Mendocino
earthquake (USA) for the far-field (Fig. 9).
0.2 0.8
Cascia NS (Valnerina 1979)
0.1 Cascia NS
(near-field)
0
0.6
-0.1
0.15 g near-field
-0.2
0 5 10 15 20 0.4
0.2
Butler Valley NS (Cape Mendocino 1992) Butler V. NS
0.1 0.067g (far-field)
0.2
0
-0.1
fa r-field c)
-0.2 0
0 5 10 15 20 0.01 0.1 1 10
Time (s) Period (s)
Figure 9. Time histories and response spectra of the reference input motions.
327
One- and two-dimensional total stress analyses were respectively performed by means of SHAKE91
(Idriss and Sun, 1992) and QUAD4M (Hudson et al., 1994), along the two sections of Fig. 7. In 2-D
analyses, to minimise the effects of wave reflection from the side boundaries, they were extended
horizontally 900 m away from the slope foot.
As expected, the results obtained on the transversal (NS) cross-section were more significantly
affected by geometrical factors (Fig. 10). In fact, 2D analyses yielded higher peak accelerations with
respect to 1D analyses (Fig. 10a), especially in the slab, with the maximum difference at the ground
surface: peak acceleration raises from 0.12g to 0.21g and from 0.22g to 0.34g in the far and near-field,
respectively. These results, influenced by the relatively small width of the slab in the NS direction,
suggest that 1D analysis do not adequately account for the actual geometry. On the other hand, the
peak shear strain profiles obtained at the hill center from 1D and 2D analyses are in fair agreement
(Fig. 10b). Even though the peak shear strain is higher in the clay than in the slab (Fig. 10b), its
magnitude does not exceed 0.04%. This value is smaller than the volumetric threshold for clays of
similar plasticity (0.05%), thus suggesting that the increase of pore pressure related to shaking is
negligible, and therefore justifying a total stress analysis. The horizontal profile of amax along the hill
surface is plotted in Fig. 10c. It can be noted that the 2D geometry greatly affects the amax distribution,
with maximum values generally at the center and at the edge of the slab. Under the near- field input
motion, amax reaches 0.35g at the center and at the edge of the cliff, while a value of 0.2g is reached at
the slab center in the far-field condition. At the cliff foot, amax abruptly decreases down to about 0.15g
and 0.1g for near- and far-field motions respectively. An estimation of the overall topographic effects
can be expressed by the 2D over 1D ratio of horizontal peak acceleration. The resulting topographic
amplification factor, AT, is about 1.7 for both input motions.
For the longitudinal (WE) section, characterised by a higher slab width, the comparison between 2D
and 1D analyses at the hill center is very satisfactory in terms of amax and max profiles, while the amax
profile computed along surface by 2D analyses show higher values close to the cliff edge (Muzzi et
al., 2001). Estimated topographic amplification factors AT are lower than NS section, being 1.4 for
both near-field and far-field conditions.
amax (g) max (%)

0 0.1 0.2 0.3 0.4 0 0.02 0.04 0.06
0.4
0 NF NS Section
Shake (NF)
FF
Quad4m (NF) 0.3
40 Shake (FF)
Quad4m (FF) 0.2
80
0.1
(c)
120 0
0 100 200 300 400 500 600
160 Distance from the hill center
NS section
200 hill center
(a)
(b)
240
Figure 10. NS section: vertical profiles of peak acceleration (a) and shear strain (b) by 1D and
2D analyses at the hill center, and horizontal amax profiles computed by 2D analyses (c)
(modified after Muzzi et al., 2001)..
328
THE CASE HISTORY OF NICASTRO

Nicastro cliff is located in Calabria (Southern Italy), one of the most active seismic regions of the
Mediterranean area. The cliff is elongated in the NE-SW direction, about 60 m high; the width at the
crest varies between 30 and 60 m moving from NE to SW, while it is about 200 m at the base (Fig.
11). The ridge flanks are quite irregular, with slope angles varying generally between 20 and 50, but
almost vertical in the northern side. Precious remains of an ancient Norman-Swabian castle lie on the
crest. It was built in the period 1100-1500 and severely damaged during two strong-motion
earthquakes in 1638 and 1783. In the last two centuries the castle was completely abandoned, and only
recently has been subjected to consolidation works.
Figure 11. Plan view and sections of the Nicastro cliff.
The cliff is constituted by metabasitic rocks and schists characterised by extremely variable degree of
weathering and jointing, which strongly affect the mechanical behaviour of the cliff. From 1998 to
2003, several geotechnical investigations were carried out (Fig. 11), including 31 boreholes drilled to a
maximum depth of about 40 m, 5 seismic refraction surveys, and 2 down-hole tests (Pagliaroli, 2006).
From the sections 1-1 and 2-2 reported in Fig. 11, it can been seen that the cliff is locally covered by
fill or debris, mainly gravel in a silty matrix, with a maximum thickness of 5 m. Despite the high
degree of heterogeneity, it is possible to recognise some macro-zones from bottom to surface:
namely, moderately weathered and jointed rock, moderately weathered and highly jointed rock, highly
weathered and jointed rock. The physical and mechanical material properties of the above materials
are reported in Table 5. The shear wave velocity VS was taken from down-hole tests results (Pagliaroli
et al., 2006).
Table 5. Geotechnical properties of the Nicastro subsoil.

Material (kN/m3) VS (m/s) D0 (%)
cover 21 0.35 350 1.0
highly weathered and jointed rock 23 0.25 700 1.0
moderately weathered, highly jointed rock 24 0.25 900-1000 0.4
moderately weathered and jointed rock 26 0.25 1200 0.4
bedrock 27 0.25 1500 0.1
329
The historical seismicity of the area was characterised by setting up a site-specific seismic catalogue,
based on the intensities either felt at Nicastro, or estimated through attenuation relationships
(Pagliaroli, 2006). The most destructive event was the earthquake of March 27, 1638 (M=7.1),
producing a site intensity IS = XI MCS.
Two reference seismogenic zones were identified: the Catanzaro strait (near-field condition)
characterised by a minimum source-to-site distance de = 10 km and maximum historically observed
magnitude MS = 7.1; the Reggio Calabria fault system (far-field condition), characterised by de = 100
km and MS = 7.3. The peak ground acceleration resulted 0.35g and 0.04g, using the attenuation
relationships by Ambraseys and Douglas (2003) and Ambraseys et al. (1996), for the near-field and
far-field conditions respectively. The corresponding estimated mean response spectra and their
standard deviations are shown in Fig. 12, together with those relevant to ten accelerograms (five for
near- and five for far-field conditions), selected from European and world-wide records on rock
outcrops, matching the required MS - D values and scaled to the expected amax.
1.75 0.25
near-field average + (attenuation model) a) real accelograms b)
1.5 real accelograms average
0.2
1.25 average far-field
(attenuation model)
1 0.15
0.75 0.1
0.5
0.05
0.25
average - (attenuation model)
0 0
0.01 0.1 1 0.01 0.1 1
Period, T (s) Period, T (s)
Figure 12. Real accelerograms selected for near-field (a) and far-field (b) conditions.
Seismic response analyses

The seismic response of the cliff to both reference earthquakes was simulated by the finite difference
code FLAC (Itasca, 2002). The input motions were applied to the base of the model as vertically
incident SV waves. Due to the high stiffness of the materials, only linear analyses were carried out.
The dynamic response of the cliff was investigated for both sections in Fig. 11, characterised by
different shape ratios H/L (where H is the average cliff height and L is the half-width at the base),
respectively equal to 0.55 and 0.8 for cross-sections 1-1 and 2-2. The steeper section 2-2 is also
characterised by stiffer material with respect to the other.
The representative finite difference meshes shown in Fig. 13 were discretised in order to model
frequencies up to 20 Hz, while radiation damping was simulated by viscous dashpots at the bottom
and free-field boundaries at the both vertical sides. The same figure shows the horizontal profiles of
the peak surface acceleration, normalized to the corresponding value at the outcropping bedrock, with
reference to all the selected input motions. For both sections, the amplification increases from the base
to the crest of the cliff, with maximum values generally between 2 and 2.5. In order to evaluate the
magnitude of the stratigraphic amplification, 1D analyses were carried out for the vertical soil columns
corresponding to two representative nodes at the crest of the cliff (1A and 2A in Fig. 13).
In the time domain, the topographic amplification factor, AT, was then evaluated as the 2D over 1D
ratio of horizontal peak acceleration. The values calculated by averaging the amplification factors
resulting for all the selected near-field and far-field accelerograms were about 1.5 for section 1-1 and
1.9 for the steeper section 2-2. Overall, such factors are 15% higher than those calculated for a simple
homogeneous subsoil model (Pagliaroli et al., 2006).
330
3.5 3.5
3 a) 3 near field b)
far field
2.5 2.5
2 2
1.5 1.5
1 1
0.5 0.5
0 0
0 40 80 120 160 200 240 0 40 80 120 160 200
distance (m) 2A distance (m)
1-1 cross-section 1A 2-2 cross-section
cover (VS = 350 m/s) moderately weathered and jointed rock (VS= 1200 m/s)
highly weathered and jointed rock (VS= 700 m/s) bedrock (V S= 1500 m/s)
moderately weathered, highly jointed rock (V S= 900-1000 m/s)
Figure 13. Normalised horizontal peak accelerations computed at the surface of cross-sections 1-
1 (a) and 2-2 (b) for the near-field and far-field input motions.
In the frequency domain, the topographic amplification was estimated by the so-called TAF function,
expressed in terms of the ratio between 2D over 1D smoothed Fourier spectra at the crest nodes 1A
and 2A (Fig. 14). It can be noted that the input motion does not affect the variation of TAF with
frequency, for all the selected input motions. The maximum peak spectral amplification occurs
between 3-5 Hz for both sections. This peak corresponds to the fundamental frequency of vibration
(f2D) of the cliff, estimated by the Rayleigh method applied to homogeneous triangle-shaped
asymmetric ridges (Paolucci, 2002):
VS
f 2D f SV (1)
2L
where f SV is a factor usually variable between 0.6 and 1.0, according to geometric features of the cliff
and the Poisson ratio. For the case of Nicastro, being 2L=180 m and VS=1200 m/s (representative
value of the average stiffness of the cliff), equation (1) yields f 2D=4-6 Hz. Therefore, topographic
amplification can be interpreted as a two-dimensional resonance phenomenon, as also shown by the
simplified homogeneous model (Pagliaroli, 2006). The amplitude of spectral amplification at crest is
about 1.8 and 2.5 for sections 1-1 and 2-2 respectively, thus increasing with H/L ratio analogously to
amplification factor A T.
4 4
1-1 section near field 2-2 section
3.5 3.5 2A
1A far field
3 mean 3
2.5 2.5
2 2
1.5 1.5
1 1
0.5 0.5
a) b)
0 0
0.1 1 10 0.1 1 10
Frequency, f (Hz) Frequency, f (Hz)
Figure 14. Topographic aggravation factor (TAF) computed at the crest of cross-sections 1-1 (a)
and 2-2 (b) for the near-field and far-field input motions.
331
THE CASE HISTORY OF GERACE
Site description and susbsoil model

Gerace is located at about 480m a.s.l. on a cliff oriented in direction NW-SE (Fig. 15), between the
basins of two rivers, characterised by continuous gully erosion affecting the slope instability of the
valley borders. The upper part of the cliff is formed by a soft rock slab, constituted by calcarenites and
sandstones, overlying a deep formation of clayey marls, which in turn rests on a relatively thinner
layer of interbedded sands and conglomerates. The slab floats on a deep layer varicoloured clay shales,
with an estimated thickness of about 200m. The current geomorphological setting is the result of the
intense erosion processes, producing the progressive removal of the soils overlying the clay formation
(Monteleone, 1993).
Figure 15. Geological map of Gerace (modified after Monteleone, 1993).
The longitudinal and transversal subsoil models for the seismic response analyses (see Figs. 17a,b)
were characterised combining data collected from previous investigations and experimental results of
on-purpose dynamic in situ (Down-Hole, MASW and SASW) and laboratory tests (RC and TS)
(Costanzo, 2007). A synthetic picture of physical and mechanical material properties, used in the
analyses, is reported in Table 6.
Table 6. Geotechnical properties of Gerace subsoil.
VS G K D0 ' c' r v
Material
(kN/m3) (m/s) (MPa) (MPa) (%) () (kPa)
Fractured
16.69 400 0.250 267 445 0.5 35 7
calcarenites Linear
Calcarenites 16.69 750 0.250 939 1565 0.5 40 700
Sandstones 16.94 400 0.322 271 671 0.5 35 54 0.100 0.048
Clayey marls 18.05 800 0.332 1155 3053 1.5 32 34 0.140 0.060
Conglomerates 19.51 750 0.285 1097 2186 0.5 30 1000 0.300 0.160
21.00 658 909 4243 1.5 0.164
Varicolour clay (cliff) 0.400 25 0 0.068
21.00 725 1104 5151 1.5 0.164
Varicolour clay 20.87 454 430 1404 3.0 0.103
0.361 25 0 0.040
(valley) 20.87 643 863 2816 3.0 0.103
Bedrock 22.00 1500 0.300 4950 10725 Linear
332
In less than sixty days of 1783, since February, 5 to March, 28, Southern Calabria was struck by five
strong earthquakes, with magnitudes between 5.9 and 6.9. The sequence caused the destruction of
many towns (Carbone-Grio, 1887), as well as widespread environmental damage (Cotecchia et al.,
1986), including significant ground movements and maybe slope failure in the historical town of
Gerace (Romeo & Delfino, 1997).
From the analysis of the site seismic history (Costanzo, 2007), the first event of the sequence could be
assumed as the maximum historical earthquake. The reference seismic scenario was simulated through
two methods: one based on macroseismic data (estimated epicentre and magnitude) and the other on
seismogenic parameters (position and geometry of the possible sources). The acceleration and
frequency content of the reference motion were evaluated by means of different attenuation
relationships, yielding very close evaluations of the motion amplitude. The estimate of synthetic
ground motion parameters allowed to sort out a set of compatible acceleration time histories
(Costanzo, 2007).
On the basis of the reference parameters, fixing suitably restricted ranges of magnitude and distance,
compatible seismic records were selected as possible input motions from seismic databases. To select
the most reliable accelerogram, the approach suggested by Bommer & Acevedo (2004) was followed,
computing the amplitude scale factor, F sc, and the root mean square parameter, Drms, that supplies a
quantitative evaluation of the similarity between the frequency contents. The most suitable seismic
motion was the record of the Landers earthquake of 1992 (M=7.5, d e=21km) at the Morongo Valley -
Fire Stn 461 (Fig. 16), that showed the better spectral compatibility to the reference response
spectrum, estimated through the relationship by Pugliese & Sabetta (1989).
0.25 1.00 P&S(89) mean
(a) P&S(89) mean (b)

0.15 0.80
LANDERS 1992
0.05 0.60
-0.05 0.40
-0.15 0.20
-0.25 0.00
0 10 20 30 40 50 60 70 0 1 2 3 4
time, t (s) period, T (s)
Figure 16. Landers earthquake of 1992: acceleration time history (a)

and spectral compatibility with the reference motion (b).

The numerical 2D simulations of the seismic response of the cliff to the sequence have been carried
out by FLAC 5.0 (Itasca, 2005), along the two orthogonal sections of the cliff. The bedrock depth was
calibrated on the basis of far-field earthquake records taken by a mobile accelerometric station
installed in the cathedral, at the cross-point between the two sections (Costanzo, 2007).
The plots in Fig. 17a,b report with blue lines the horizontal peak acceleration profiles computed at
surface, while the green lines refer to the values at the top of the varicolour clay formation; the
coloured dots indicate the stratigraphic contacts between different soils.
Along the longitudinal hill section (Fig. 17a), the amplitudes at the top of the clay are comparable or
not significantly lower than the reference input motion (black horizontal lines). For the transversal
section, instead, an apparent amplification is shown at the toes of the cliff (Fig. 17b), which was
shown to affect the clay slope stability along this direction (Costanzo et al., 2007a,b).
333
0.60
surface (a)
0.50
top of varicolour clay
0.40 bedrock
0.30
0.20
0.10
0.00
-1800 -1600 -1400 -1200 -1000 -800 -600 -400 -200 0 200 400 600 800 1000 1200 1400
distance, d (m)
C1 0 0.5 1km
50 0 m
NW S
fractured calcarenites (435ms.l.m.)
calcarenites
sandstones SE
clayey marls
25 0
conglomerates & sands
varicolour clay shales

0
0.60
surface (b)
0.50
top of varicolour clay
0.40 bedrock
0.30
0.20
0.10
0.00
-1600 -1400 -1200 -1000 -800 -600 -400 -200 0 200 400 600 800 1000 1200 1400 1600
distance, d (m)
0 0.5 1 km
C2
500m
(435ms.l.m.)
SW NE
250
0
Figure 17. Horizontal peak acceleration profiles along NW-SE (a) and SW-NE (b) sections.
The surface motion amplitudes (blue lines) along the NW-SE section (Fig. 17a), characterised by a
smoother topographic profile, show that the amplification is visibly affected by the impedance
contrasts, and enhances across the more deformable sandstone layer. Also, the interaction between
incident and diffracted waves induces oscillating amplification factors and out-of-phase motion. In the
narrower NE-SW section, wave focalisation into the hilltop induces a pronounced topographic
amplification, which results into A T of the order of 1.9 (Costanzo, 2007), i.e. a little higher than the
maximum values evaluated for Orvieto (1.7) and equal to those relevant for Nicastro.
As for the case of Nicastro cliff, the net topographic effect was evaluated by the TAF parameter, again
defined as 2D over 1D Fourier spectra ratio (Fig. 18). At the hilltop (sites C1 and C2) and on the
sandstones outcropping (site S), the TAF results sensibly higher than unity in the range between 2 and
8 Hz. At the crest of the longitudinal cliff section (C1), the TAF is about 2.5 around 5 Hz, which falls
in the range of resonance periods of typical Gerace buildings. At the apex of the narrow section (C2),
the TAF shows a peak of 3.8 for frequency of 3.5Hz and an apparent tendency to increase beyond
334
6Hz, due to numerical oscillations of low values of 2D and 1D Fourier amplitudes in the high
frequency range.
5
4 C1
C2
3 S
0
0.1 1 10
frequency, f (Hz)
Figure 18. Topographic aggravation factory (TAF) computed at the crest of both sections
and at the outcrop of sandstones along the section NW-SE.
The vertical ground motion profiles and the plastic strains distribution, not reported here (see Costanzo
et al., 2007a,b), indicate a tendency to slippage of the calcarenite layers along the contact with the
underlying weak sandstone. Irrecoverable strains as high as 0.1% concentrate at the NW and SE toes
of the hill and across the soft rock slab, evidencing an asynchronous motion of independent calcarenite
blocks: this could be responsible for the opening of sub-vertical cracks across the cliff and of a rigid
sliding of the calcarenite above the sand layer at the hilltop (see Costanzo et al., 2007a). Also, the final
deformation pattern of the transversal section showed residual displacement values higher than 1.5 m
in correspondence of the swelled varicolour clay, just beneath the contact with conglomerates: this
deformation pattern seems to highlight the triggering of a rotational slope instability at the base of the
hill.
Therefore, the results of the non-linear 2D numerical analyses indicated not only a frequency-
dependent amplification of the ground motion at the cliff surface, but also a significant ground
deformation under the maximum historical earthquake; both effects have to be accounted for the
protection of Gerace and its cultural heritage from the seismic risk.
THE CASE HISTORY OF SAN GIULIANO DI PUGLIA

San Giuliano di Puglia is a little middle-age town in the Molise region, edified on a narrow hill with
the ridge crest oriented SSE-NNW. The urban centre shows the typical characteristics of the Southern
Apennines villages: buildings are mostly located on outstanding positions, and frequently connected in
heterogeneous structural aggregates. The historical residential houses have typically rubble masonry
walls, while in the newer part, built after the 40s, rubble and brick masonry are mostly present, with
sometimes recent r.c. additions (Mucciarelli et al., 2003). Overall, both historical and newer buildings
of San Giuliano di Puglia are usually 2-4 storeys high, resulting in fundamental periods of 0.1 to 0.5s
(Baranello et al., 2003; Vona, 2007). As a consequence of the historical development of the town, the
older part was edified on a soft rock formation (Faeto flysch) outcropping in the southern part of the
town (Fig. 20a); in the progressive extension towards North, the newer buildings were raised up along
the ridge, on a thick fine-grained soil (Toppo Capuana marly clays).
Until 2002, the town was classified as not hazardous by the Seismic Italian Code. On October 31st,
2002, a moderate earthquake of local magnitude ML=5.6 (CPTI Working Group, 2004) shook the area.
The epicentral distance (D) of San Giuliano di Puglia was only about 5 km (Pacor, 2007) and a degree
of damage observed was VIII-IX MCS, while it never resulted larger than VII in the other towns close
to the epicenter (Stucchi et al., 2007). Detailed surveys on the buildings (Baranello et al., 2003; Dolce
335
et al., 2004) indicated that the damage distribution was strongly non-uniform, with resulting VII-VIII
EMS + in the recent part of the town, and VI in the ancient zone (see Fig. 20b). Such evidence
suggested that differential site amplification significantly affected the seismic response in the town;
the hypothesis was confirmed by the observation of the aftershock sequence recorded by two mobile
accelerometric stations, located on clay and flysch outcrops, shown in Fig. 20c (Puglia et al., 2007).
For the seismic microzonation of the town, the Department of Civil Protection (DPC) of the Italian
Government committed a comprehensive study on the subsoil properties (Baranello et al., 2003). The
field investigation included a large number of boreholes, sampling pits, in situ piezometric
measurements, CPT, cross-hole and down-hole tests. The laboratory experimental programme
involved oedometer, triaxial and cyclic/dynamic torsional shear tests on undisturbed samples of fine-
grained soils, analyzed in detail by Silvestri et al. (2006a). Table 7 describes the geotechnical
properties for the marly clay and flysch formations, as reported by Silvestri et al. (2006b).
Table 7. Geotechnical properties of the San Giuliano di Puglia subsoil

Depth, z [m] [kN/m3] VS [m/s] D0 [%] VP [m/s]
Debris cover 0 - 2.4 19.6 122 3 0.493 1010
Toppo Capuana
Tawny clay 2.4 - 7.3 21.15 250 2.3 0.489 1700

marly clays
Grey clay 1 7.3 - 15 339 0.485 1970

Grey clay 2 15 - 30 364 0.483 2000
Grey clay 3 30 - 60 391 0.481 2050
21.2 2.5
Grey clay 4 60 - 120 421 0.479 2100
Grey clay 5 120 - 240 454 0.477 2160
Grey clay 6 > 240 483 0.475 2210
Faeto flysch - 22 1350 0.5 0.392 3200
Although very accurate and detailed in terms of characterisation of lithological, physical and
mechanical properties of the geomaterials, neither the above investigation nor the numerous surface
geological surveys could assess for certain the actual geometry of the stratigraphic contact between
flysch and marly clay. On purpose, in the framework of the INGV-S3 National Research Project, deep
geophysical surveys were carried out: three geo-electric resistivity tomographies (see Silvestri et al.,
2006b), a gravimetric (Palmieri et al., 2006) and two deep seismic surveys (under processing) seem to
indicate a complex 3D flysch-clay interface, characterised by planar boundaries rather than a basin
geometry, as formulated by the earlier studies (Baranello et al., 2003). To date, the subsoil model most
effective in matching the results of geophysical surveys and reproducing the subsoil response recorded
during the aftershocks (Puglia et al., 2007) is that reported in Fig. 20c, i.e. a wedge-shaped clay
deposit, fully bounded by the soft rock and deep about 300m.

The mainshock of the 2002 Molise seismic sequence was simulated by Franceschina et al. (2006)
through the Hybrid Integral-Composite source model (HIC), considering the fault mechanism
formulated by Basili and Vannoli (2005). Fig. 19a shows the acceleration time history of the
horizontal component on rock outcrop resulting from the HIC simulation, after projection along the
section in Fig. 20. A 7Hz low-pass filter was applied for consistency with the mesh discretisation used
in the numerical analyses described below.
A typical engineering approach was also used to define the seismic input motion, by using the most
significant aftershock (2002.XI.12, ML=5.2) recorded by the mobile station on the rock outcrop. The
horizontal and vertical peak acceleration were scaled to the values predicted by regional attenuation
laws, calibrated on the records of several temporary and permanent networks operating in the area
affected by the seismic sequence (Luzi et al., 2006). Fig, 19b shows the time history of the scaled
aftershock after projection along the section and 7Hz low-pass filtering. The comparison between the
two input motions considered shows that the simulated mainshock has a peak amplitude of about
+
Note that such values in EMS-98 (European Macroseismic Scale) are roughly equivalent to those
corresponding in the MCS scale.
336
0.05g, against 0.03g of the scaled aftershock. On the other hand, the response spectra (Fig. 19c) show
comparable frequency contents, with peak values around 5 Hz (0.2 s), and similar trends throughout
the range of the resonant periods (0.1 to 0.5s) of most of the buildings.
Figure 19. Horizontal time histories of simulated mainshock (a),

scaled aftershock (b) and relevant pseudo-acceleration response spectra (c).
Seismic response analyses

Simulations of the seismic response of San Giuliano di Puglia during the mainshock were carried out
along a representative longitudinal section traced along the hill (Fig. 20a); this direction approximately
follows the alignment of the most severe damage distribution, the locations of the two accelerometer
stations and those of main geotechnical investigations. 2D numerical analyses were executed by
QUAD4M (Hudson et al., 1994) on the wedge-shaped section drawn in Fig. 20c, geometrically
defined with the criteria described in the paper by Puglia et al. (2007).
The horizontal profiles of peak ground acceleration (plotted in Fig. 20c with red lines, right scale)
reproduce the differences in amplitudes between the simulated mainshock and the scaled aftershock;
instead, the amplification factors on the marly clay deposit are similar, varying between 1.5 and 3,
with significant oscillations characterized by a surface wavelength of about 300 m, i.e. equal to the
depth of the clay deposit.
It must be noted that the predicted peak ground acceleration cannot directly be referred to the observed
damage distribution, since they pertain to frequencies higher than those typical of the existing
structures. Therefore, to assess the damage distribution occurred in the actual earthquake, the output of
the analyses were processed in terms of Housner intensity (SI), i.e. the integral of the pseudovelocity
response spectrum, PSV(T), calculated in the range of periods of San Giuliano di Puglia buildings (0.1
to 0.5 s):
0. 5
SI PSV , T dT (2)
0 .1
and assuming a structural damping =5%. The horizontal profile of this integral ground motion
parameter is also plotted in Fig. 20c with black lines (left scale). Note that the ratio between the SI
values pertaining to the two ground motions, as well as their trends with local oscillations, are much
similar to those observed for the peak ground acceleration.
In such a way, the predicted ground motion profiles can be reliably compared to the observed degree
of damage, this latter expressed in terms of standardized damage, as shown in Fig. 20a (Dolce et al.,
2004). This index assumes values between 0 (no damage), and 1 (building collapse), and it does not
depend on the vulnerability class of the building; therefore, it can be viewed as a significant variable
337
to empirically measure the local amplification effects. A closer analysis of the variation of such
parameter at the building size scale appears promising to reveal the likely correlation existing between
the predicted ground motion and the observed damage.
Figure 20. Geological map (from Puglia et al., 2007) and standardized damage of buildings in San
Giuliano di Puglia (from Dolce et al., 2004) (a); average degree of damage in terms of EMS (from
Dolce et al., 2004) (b); horizontal profiles of peak ground acceleration and Housner intensity for
the wedge-shaped subsoil model subjected to two input motions (c).
CONCLUSIONS
In all the above examples, the expected seismic response of urban areas under maximum historical
earthquakes was accurately analysed with comparable methodologies, although with different specific
addresses. In fact, the studies were more or less oriented to prediction or assessment, depending on the
availability of data on the historical damages. For each case study, Table 8 summarizes the main
aspects of input motion and subsoil modelling which significantly influenced the seismic response
analyses.
The overall intensity in terms of EMS (Fig. 20b) was instead obtained through the comparison between
average damage index, as calculated by Damage Probability Matrices (DPM) referred to the vulnerability class
of the buildings, and observed for specific zones (Dolce et al., 2004).
338
Table 8. Main aspects of the analysis of the selected case studies.
Input motion
Case study amax amax Bedrock Soil layering Topography
(hazard map) (this study)
Assumed at Complex Gently
Benevento 0.250 0.275 0.3350.129
variable depth with some impedance inversions sloping hill
0.150 (nf) Assumed at Regular
Orvieto 0.125 0.150 Plateau
0.067 (ff) fixed depth with constant impedance
0.350 (nf) Assumed at Variable
Nicastro 0.250 0.275 Peak
0.040 (ff) fixed depth with weathering at surface
Back-analysed Regular
Gerace 0.200 0.225 0.200 Plateau
(far field records) with some impedance inversions
Back-analysed
S. Giuliano di Continuous variation Gently
0.200 0.225 0.03-0.05 Under deeper
Puglia with depth sloping hill
investigation
In most cases (except for San Giuliano), no representative seismic signals were already available for
the definition of suitable reference rock outcropping motions: therefore, these were defined on the
basis of detailed studies on the historical seismicity and seismotectonic outlines of the regions. Table 8
compares the peak ground accelerations predicted by the official hazard map by MPS Working Group
(2004) to those used in this study. Actually, for Nicastro and Benevento, the ground motion
amplitudes considered herein overestimate the specified values. For San Giuliano, the values of amax
assumed in the analysis significantly underestimate that prescribed by the hazard map, since the recent
earthquake considered is not the maximum historical event. Anyway, in most cases the results
presented herein might be taken as references for seismic microzonation maps for urban planning. In
such perspective, when the available subsoil data are numerous, like in the case of Benevento, the use
of GIS maybe very promising.
The geo-morphological setting and the susbsoil layering were sometimes quite complex; thus, their
study resulted of critical importance for the subsoil modelling. The bedrock depth and geometry were
in most cases unknown; their definition strongly influences the low-frequency response and the
variation of amplification at surface, like verified for San Giuliano (see Puglia et al., 2007). In such
cases, deep geophysical profiling can be of primary utility; also far-field or weak motion records can
help to back-figure the bedrock depth, as shown by Costanzo (2007) and Puglia et al. (2007) for the
two cases of Gerace and San Giuliano, respectively.
The degree of accuracy in the description of soil layering must be adequate to the frequency content of
the input motion, especially when sharp variations of seismic impedance are present. The response to
near-field earthquakes, with significant energy content at high frequencies, can be very sensitive to the
stratigraphic details, such in most of the cases presented. The example of Benevento shows that
unusually excessive spectral accelerations at high frequencies may be induced by the use of synthetic
accelerograms as input motion; as a matter of fact, in the case of San Giuliano, the simulated input
motion was preliminarly low-pass filtered, to be comparable to that of a recorded signal.
The site amplification was seen to be strongly affected by surface topography for the towns lying on
flat (Orvieto, Gerace) to steep (Nicastro) cliffs. It must be noted that, in such cases, an irregular
morphology is often associated with a stratigraphic heterogeneity, because erosion processes that can
affect different materials with variable intensity. Therefore, the interplay between geometry and
layering can be misunderstood, if simple homogeneous linear analysis models are used; only 2D
numerical FEM or FDM simulations, either with linear equivalent or non-linear analysis, are capable
to yield reliable predictions or assessment of dynamic ground motion and deformation.
Summing up, in order to preserve the high value of the cultural heritage from the damage induced by
strong-motion earthquakes, the strongest efforts are needed in terms of definitions of seismic actions,
subsoil characterisation and methods of analysis.
339
ACKNOWLEDGMENTS
The town municipalities of Benevento, Orvieto, Nicastro, Gerace and San Giuliano are kindly
acknowledged for providing the available data on the subsoil, even when not sponsoring the research
projects.
The authors are indebted to the Department of Civil Protection (DPC) of the Italian Government and
the National Institute of Geophysics and Vulcanology (INGV) for funding the research works on
Benevento, Gerace and San Giuliano di Puglia. The coordinators of the INGV-S3 Research Project,
Dr. Francesca Pacor and Prof. M. Mucciarelli, are warmly acknowledged for their valuable scientific
and administrative support.
The oldest Authors are indebted to their former supervisors, Profs. B. DElia, R. Ribacchi and F.
Vinale, for having introduced them to these fascinating research studies.
REFERENCES
(pr)EN 1998-1 (2003). Eurocode 8: Design of structures for earthquake resistance Part 1: General
rules, seismic actions and rules for buildings, CEN European Committee for Standardization,
Bruxelles, Belgium.
Aliperti D. (2006). Un contributo alla zonazione sismica del comune di Benevento, Undergraduate
thesis, Universit di Napoli Federico II (in Italian).
Ambraseys N. N., Simpson K. A., and Bommer J. J. (1996) Prediction of horizontal response spectra
in Europe Earthquake Engineering Structural Dynamics, 25:371-400.
Ambraseys N. N., and Douglas J. (2003) Near-field horizontal and vertical earthquake ground
motions Soil Dynamics and Earthquake Engineering, 23:1-18.
Baranello S., Bernabini M., Dolce M., Pappone G., Rosskopf C., San T., Cara P.L., De Nardis R., Di
Pasquale G., Goretti A., Gorini A., Lembo P., Marcucci S., Marsan P., Martini M.G. and Naso G.
(2003) Rapporto finale sulla Microzonazione Sismica del centro abitato di San Giuliano di Puglia
Dipartimento di Protezione Civile, Roma (in Italian).
Bardet J.P., Ichii K. and Lin C.H. (2000) EERA a Computer Program for Equivalent-linear
Earthquake site Response Analyses of Layered Soil Deposits, Univ. of Southern California, Dep.
of Civil Eng.
Basili R. and Vannoli P. (2005) Source ITGG052 San Giuliano di Puglia. In: DISS Working Group,
Database of Individual Seismogenic Sources (DISS), Version 3.0.1, A compilation of potential
sources for earthquakes larger than M 5.5 in Italy and surrounding areas, Istituto Nazionale di
Geofisica e Vulcanologia (INGV), internet: http://www.ingv.it/DISS/ .
Bommer J. J. and Acevedo A. B. (2004) The use of real earthquake accelerograms as input to
dynamic analysis, Journal of Earthquake Engineering, 8, special issue 1: 43-91.
Bouckovalas G.D., Papadimitriou A.G. and Karamitros D. (2006). Compatibility of EC-8 Ground
Types and Site Effects with 1-D Wave Propagation Theory, ISSMGE ETC-12 workshop, NTUA
Athens, Greece, 53-67, Bouckovalas Ed.
Carbone Grio D. (1884) Terremoti di Calabria e Sicilia nel secolo XVIII, Barbaro editore, Oppido
Mamertina (RC) (in Italian).Castenetto R. and Romeo R. (1992). II terremoto del Sannio del 5
Giugno 1688: Analisi del danneggiamento subito dalla citt di Benevento, Proc. XI C.G.N.G.T.S.,
Roma, 2138 (in italian).
Costanzo A. (2007) Analisi di fenomeni deformativi di pendii e rilievi in condizioni sismiche : il caso
di Gerace, Ph.D. thesis, Universit della Calabria (in Italian).
Costanzo A., DOnofrio A. and Silvestri F. (2007a) Numerical simulations of the ground deformation
recorded in the historical town of Gerace during the seismic events in Calabria (1783), IV
International Conference on Earthquake Geotechnical Engineering, Thessaloniki, Greece, paper no.
1613.
Costanzo A., DOnofrio A., Silvestri F. (2007b) Analisi dei danni registrati nel borgo di Gerace
durante gli eventi sismici della Calabria del 1783, XII National Conference LIngegneria Sismica
in Italia, Pisa (in italian). ANIDIS, Roma.
340
Cotecchia V., Guerricchio A. and Melidoro G. (1986) The Geomorphogenetic Crisis Triggered by the
1783 Earthquake in Calabria (Southern Italy), Symposium: Engineering Geology Problems in
Seismic Areas, 6, Bari.
CPTI Working Group (2004) Catalogo Parametrico dei Terremoti Italiani, versione 2004 (CPTI04),
INGV, Bologna, internet: http://emidius.mi.ingv.it/CPTI04/ (in Italian).
Dolce M., Masi A., Samela C., Santarsiero G., Vona M., Zuccaro G., Cacace F. and Papa F. (2004)
Esame delle caratteristiche tipologiche e del danneggiamento del patrimonio edilizio di San
Giuliano di Puglia, XI National Conference Lingegneria Sismica in Italia, Genova (in Italian).
ANIDIS, Roma.
Fenelli G.B., Pellegrino A. and Picarelli L. (1998) Stability problems of old towns built on relict
plateaux on clay deposits, Proc. Arrigo Croce Memorial Symp. on Geotechnical Engineering for
the Preservation of Monuments and Historic Sites, Napoli, 163-173, Rotterdam: Balkema.
Franceschina G., Pacor F., Cultrera G., Emolo A. and Gallovic F. (2006) Modelling directivity effects
of the October 31, 2002 (MW =5.8), Molise, Southern Italy, earthquake First European Conference
on Earthquake Engineering and Seismology (1st ECEES), Geneva, Switzerland, paper no. 1424.
Hudson M., Idriss I. M. and Beikae M. (1994). QUAD4M - A computer program to evaluate the
seismic response of soil structures using finite element procedures and incorporating a compliant
base, University of California, Davis.
Idriss J. and Sun J.I. (1992). SHAKE91- A computer program for conducting equivalent linear
seismic response analyses of horizontally layered soils deposits, University of California, Davis.
Improta L., Di Giulio G. and Iannaccone G. (2005). Variations of local seismic response in
Benevento (Southern Italy) using earthquakes and ambient noise recordings, Journal of
Seismology, 9:191-210.
ISSMGE-TC4 (1999) Manual for zonation on seismic geotechnical hazards, The Japanese Society
of Soil Mechanics and Foundation Engineering.
Itasca (2002) FLAC Fast Lagrangian Analysis of Continua Version 4.0 Users Guide, Itasca
Consulting Group, Minneapolis, USA.
Itasca Consulting Group (2005) Flac 5.0 USERS Manual, Minneapolis, Minnesota, USA.
Kallou P.V., Gazetas G., and Psarropoulos P.N. (2001) A case history on soil and topographic effects
in the 7th September 1999 Athens earthquake Proc. of 4th Int. Conf. on Recent Advances in
Geotechnical Earthquake Engineering and Soil Dynamics, San Diego, California.
Luzi L., Morasca P., Zolezzi F., Bindi D., Pacor F., Spallarossa D. and Franceschina G. (2006)
Ground motion models for Molise region (Southern Italy) First European Conference on
Earthquake Engineering and Seismology (1st ECEES), Geneva, Switzerland, paper no. 938.
Manfredini M., Martinetti S., Ribacchi R., Sciotti M. (1980) Problemi di stabilit della rupe di
Orvieto XXI Convegno Nazionale di Geotecnica, Firenze, 1, 231-246 (in Italian).
Marcellini A., Bard P.-Y., Iannaccone G., Meneroud J.P., Mouroux P., Romeo R., Silvestri F., Duval
A., Martin C., and Tento A.. (1995) The Benevento seismic risk project: The microzonation
Proc. of the V Int. Conf. on Seismic Zonation, Nice, France, I, pp.810814.
Monteleone S. (1993) Deformazioni gravitative profonde e grandi frane indotte dalle argille
varicolori scagliose a Gerace e liquefazione nei terreni del territorio di Locri, D.Sc. thesis,
Universit della Calabria (in Italian).
MPS Working Group (2004). Redazione della mappa di pericolosit sismica prevista dall'Ordinanza
PCM 3274 del 20 marzo 2003, Final Report for DPC, INGV, Milano-Roma,
http://zonesismiche.mi.ingv.it/ (in Italian).
Mucciarelli M., Masi A., Vona M., Gallipoli M. R., Harabaglia P. and Caputo R. (2003) Quick
survey of the possible causes of damage enhancement observed in San Giuliano after the 2002
Molise, Italy seismic sequences, Journal of Earthquake Engineering, Vol. 7, No. 4, pp. 599-614.
Muzzi M., Lanzo G., Tommasi P. and Ribacchi R. (2001). Analisi numerica della risposta sismica del
colle di Orvieto, X National Conference Lingegneria Sismica in Italia, Potenza (in Italian).
ANIDIS, Roma.
Pacor F. (2007) Personal communication.
Palmieri F., Marello L. and Priolo E. (2006) Rilievo gravimetrico di dettaglio nellarea di San
Giuliano di Puglia (CB), INOGS Research Report, S3 Project, internet: http://esse3.mi.ingv.it/ (in
italian).
341
Pagliaroli A. (2006) Studio numerico e sperimentale dei fenomeni di amplificazione sismica locale di
rilievi isolati Ph.D. Thesis, Universit di Roma La Sapienza, Roma (in italian).
Pagliaroli A., Lanzo G., DElia B. (2006) Numerical study of the topography effects at the Nicastro
(Southern Italy) cliff and comparison with EC8 recommendations, Proc. ETC-12 Workshop,
NTUA, Athens, Greece, January 20-21, paper #9.
Paolucci R., (2002) Amplification of earthquake ground motion by steep topographic irregularities
Earthquake Engineering and Structural Dynamics, Vol. 31, pp. 1831-1853.
Penna A. (2005) Geostatistical application for the local seismic response: seismic zonation of the
town of Benevento. Ph.D Thesis, University of Naples Federico II (in Italian).
Penna A., dOnofrio A. and Santucci de Magistris F. (2007) Geostatistical analysis in the evaluation
of local seismic response Proc. of the XIV European Conf. in Soil Mech. and Geotech. Eng.,
Madrid (in print).
Puglia R., Lanzo G., Pagliaroli A., Sica S. and Silvestri F. (2007) Ground motion amplification in
San Giuliano di Puglia (Southern Italy) during the 2002 Molise earthquake, IV International
Conference on Earthquake Geotechnical Engineering, Thessaloniki, Greece, paper no. 1611.
Pugliese A. and Sabetta F. (1989) Stima di spettri di risposta da registrazioni di forti terremoti
italiani, Ingegneria Sismica, 6:2 (in Italian).
Romeo R. and Delfino L. (1997) CEDIT: Catalogo Nazionale degli effetti deformativi del suolo
indotti da forti terremoti, Technical report SSN/RT/97/04, Department of Civil Protection, Rome
(in Italian)
Rotonda T., Tommasi P. and Ribacchi R. (2002) Physical and mechanical characterization of the soft
pyroclastic rocks forming the Orvieto cliff, Int. Work. on Mechanical Characterization of
Volcanic Rocks, Madeira.
Santucci de Magistris F., dOnofrio A. and Sica S. (2004). A step into the definition of the seismic
risk for the city of Benevento (Italy). Proc. of the Fifth Inter. Conference on Case Histories in
Geotechnical Engineering, New York City, N.Y.
Silvestri F., Vitone C., dOnofrio A., Cotecchia F., Puglia R. and Santucci de Magistris F. (2006a)
The influence of meso-structure on the mechanical behaviour of a marly clay from low to high
strains, Symposium to celebrate Prof. Tatsuoka's 60th birthday, Roma, Italy.
Silvestri F., DOnofrio A., Guerricchio A., Lanzo G., Pagliaroli A., Puglia R., Santucci de Magistris
F., Sica S., Eva C., Ferretti G. and Di Capua G. (2006b) Modelli geotecnici 1D e/o 2D per i
comuni di San Giuliano di Puglia, Bonefro, Ripabottoni, Colletorto e Santa Croce di Magliano,
Deliverable D8, S3 Project, internet: http://esse3.mi.ingv.it/S3_doc.html (in Italian).
Stucchi et al. (2007) DBMI04, il database delle osservazioni macrosismiche dei terremoti italiani
utilizzate per la compilazione del catalogo parametrico CPTI04, Quaderni di Geofisica, INGV,
internet: http://emidius.mi.ingv.it/DBMI04/ (in Italian).
Tommasi P., Ribacchi R., Sciotti M. (1996). Geotechnical aspects in the preservation of the historical
town of Orvieto. Proc. Arrigo Croce Memorial Symposium on Geotechnical Engineering for the
Preservation of Monuments and Historic Sites, Napoli, 849-858. Balkema, Rotterdam.
Tommasi P. and Ribacchi R. (1998). Mechanical behaviour of the Orvieto tuff, Proc. Int. Conf. The
Geotechnics of Hard Soils-Soft Rocks, Napoli, 901-909. Balkema, Rotterdam.
Vona M. (2007) Personal communication.
342
June 25-28, 2007
Paper No. W2-1001
A RESEARCH ON THE PERFORMANCE OF THE CONCRETE STRUCTURES AND THE

REASONS OF THEIR FAILURE IN BAM EARTHQUAKE AND DESIGN SUGGESTIONS
Roozbeh Ettehad , Hamed Jahangiri
ABSTRACT
The Bam earth quake occurred on December 25th at 05: 26: 26 according to the local time (at 01:26:
26 according to the international time GMT) in the historical city of Bam in the south east Iran (south
east Kerman).This earth quake is considered to be the third destroying earth quake of the country after
the Tabas and Manjil earth quakes .Most parts of the city had been destroyed completely and rest parts
had been damaged between 30% to 70%. The earthquake was caused due to the move of the
Bam..which is located between Baravat village and Bam city and has a north south direction and a
length of 15 km.
Investigations on the land showed small scale cracks in the south and north of the .. This
corresponded the rightward movement of it .This .. Were on e of the few of the country which
had not experienced earth quakes at all.
Site inspections from the suspected structures in Bam area show that considering the consumed
material and current performing methods most of them were considerably suseptable against the earth
quake. Concrete structures located in Bam in comparison with the steal constructions had a relatively
acceptable performance because of rigid connections .In the following article along with investigation
on the concrete structures behavior against the earthquakes, the reason of the mentioned failures,
design and site work problems will be discussed as well.
Keywords: concrete structure, Bam earth quake, design problems, site work problems and code
regulations.
INTRODUCTION
Bam area is located in the south east Iran which suffered a severe earth quake .No historical
earthquake had been reported up to the December 25 th 2003 .The mentioned earth quake was
reported in the early morning hours that most of the city residents were asleep and may be that can
be a cause of high civilian death .Most parts of the city like Arg-e-Bam (a famous historical place) was
destroyed completely and other parts the city were susepted between 30% to 70% . After the earth
quake some pats of the city specially city bazzar were on fire ,however on unfortunately the fire
engines were down due to the destruction of the fire station building .The greatness of the earth quake
were 9 in mercalies scale . In the vertical direction to the fault fast reduction of the severe movement
and greatness of the earth quake is noticeable in a way that all the destructions are limited to the urban
parts of the Bam .researches done show that a considerable force is being imposed to the homogenous
buildings which usually is more than the elastic limit .If the studied structure is unable to tolerate
elastic deflections a significant damage will be mad.
B.Sc of Civil Engireening, Young Researchers Club, Islamic Azad University of Bonab. Add: East
Azerbaijan Science and Technology Park, Golgasht Cross, Azady Ave, Tabriz, IRAN P.O.Box:
51745-153. Email: civil_researches@yahoo.com
343
CONCRETE STRUCTURES
Concrete structure in comparison with the other structures have got a higher flexibility .So they can
resist more easily against the earthquakes(they need less resistance).The ability mentioned for concrete
structures would be actual when the site work regulations of this structure system are obeyed correctly
.The most important elements of the system that can resist against the lateral forces are rigid
connections and concrete shear walls .If a frame has rigid connections ,it can bear lateral forces by a
bending reaction and transfer it to the earth.
The concrete structures existing in Bam had rigid connections and their relatively good performance is
an approval of it .The suseption which at times were caused in their structure was due to not obeying
the design and site work regulations according to the code. Some of the problems are as follows:
1) Producing unqualified concrete

2 )un proper performance of concrete
3 )using simple bars instead of lined ones
4)using long bars
5)the high distance between struts
6)not having enough cover on the long bars
7) improper connection between beam and columns
8) performing a weak column against a strong beam
DESIGN PROBLEMS
The most important problems are as follows:
A) Wrong estimation of weather effects:

1) not considering the repetitions of freezing and melting periods
2) not considering the repetitions of getting wet and drying
3) not considering the fluctuations of temperature and damp and their effects on deflections made on
concrete.
B) Wrong estimations of environmental corrosion

1)not considering the unpleasant effects of the soil ,water and atmosphere carrying salts and corrosive
gases
2) selecting improper construction material for example improper cement or bar
3) not predicting the proper cover on the bars in corrosive environment
C) Problem in providing performance plans and documents

1) not defining the type of the consumed material in performance plans and opening the way to further
performance problems
2)planning problems and drawing plans that does not correspond with the calculations
3) insufficient performance documents
4) not having harmony between the different parts of one plan
5) not controlling the plans before performance or inaccurate controlling
D)not considering performance details

1)not having enough internal length different elements of a structure
2)not having enough struts ,connection at the end of the bars
3)not predicting distribution bars in the effect point of the concentrated loads
4)not predicting strut in the bending points of the tensile bars on the inner corners
5)cutting the attachments or big parts of a main bar in one section
6)not considering the curvature radius of thick bars
344
7)not paying attention to the location of bars around the openings in the structure and not having
diagnol bar on the corners of the openings
8)not putting bars on the corners of the slabs that are on the soil and there are walls on them
9)the condition of the bars in the connections between beams and columns which prevents the
concrete flow
10)not continuing the bars of the foundation under the columns in to the columns
11)placing too much bar in one section with low width and therefore not allowing the concrete to flow
properly
12)not connecting and joining the surrounding walls to the frame
13)the connection screws were too close that in case of an earthquake ,theyre dangerous
14)locating the walls and partitions on one beam
15)not providing continius connection of the walls
E)not considering the codes and regulations during preparing the performance plans
1) not putting cheek bar
2) not considering the minimum strut required
3) not continuing the minimum bar required onto the supports
4)not continuing the column struts in connection with beams and producing weak points during an
earthquake
5) not continuing the negative moment bars required from the supports to the center of the beams
6)not defining vertical ,horizontal connectors in the masonary structures particularly in seismic areas
7) using simple bars as main bars and many other problems
PERFORMANCE PROBLEMS
a)problem in framing
1)not considering the section dimensions and distraction from what was predicted
2)the frame wasnt water proof that helped concrete lose water and have a low quality
3) no stability in the frame form specially the slide ones
4)no stability in the permanent supports
b)bar placing problems

1)not considering type, diameter, length, form and the number of the bars
2)not considering the connection length and curvatures
3)not considering the sequence of placing the bars in the joints
4)not considering the performance details in the joints and connections
5)not considering bending the struts end and connecting the in concrete volume
6) not considering the location and length of the waiting bars
7)long strut distance
8) not considering the minimum connection for the column bars
c) problems in pouring the concrete

1) using improper stone material
2) using improper water and cement to prepare concrete
3) not considering the optimum ratios of mixing in the concrete
4) using unknown complements specially using several complements together without considering the
adaptability
5) not considering concrete homogenousity in transportation and using improper transportation
methods
6) using improper methods and equipment to pour and place the concrete, not paying attention to the
corners and around the bars
7) not caring or caring incompletely using ineffective equipment and methods
8) smoothing the concrete surface in a hurry before the main part of the initial shrinkage has settled
345
9) not paying attention to the weather conditions while pouring concrete
10) no homogenousity between the new and the old concrete
d) problem in picking the frames

1)picking the frames before the proper time when the concrete hasnt gained its complete strength
2)picking the frames by pressure and knocking the new concrete elements
3)picking all the frame and reinstalling the safety supports or not installing them in the middle of long
beams
4)leaving wooden parts in the concrete
e)not considering technical principles and regulations:

1)not preparing flow surfaces in the construction openings before restarting to pour the concrete
2)making the bottom part of the column or the wall with a height of 10cm using concrete or mason
3) doing construction work in improper weather conditions
4) not considering the sequence of the performance
5) discribtion and possible changes in plan without the designers permission
6) not considering the stability of the performed section under site work loading
7) not doing obvious things to prevent enviormental conditions effect
8) not considering the levelness of the ceiling concrete
9) not considering the continuity in different concrete parts
10)making defected flow surface (sloped , improper place , improper form)
11) using bars with different strength
12) not connecting the walls upper part in to the under part of the beam
13) not cleaning the frames before pouring the concrete
14)filling the walls made of concrete blocks by cracked blocks
15) not cleaning rusty bars and not cutting their connection with defected concrete
16)improper isolation of the parts related with soil
17) not considering the freezing and leaving openings which reduces the connection strength
f) problem in quality control:

1) not having control or having weak control over the work quality of the workers
2) not having effective and proper internal controle by the builders not outer control by the owner
3) having improper or incorrect quality control methods and misleading the controller about the
quality of the work done
DESIGN ADVICES
Obeying these simple cases has been suggested by Iranian earthquake code (2800):
1)the building plan must be plain and symmetric
2)the columns in different stories must be on eachother
3) the elements resistance against the lateral forces must be in a vertical plain
4) the correspondence between the gravity center and the rigidness center
5)balconies with a length lower than 1.5 m
6) not placing heavy equipment or facility on the balconies
7) reducing the gravity center of the building
8)using high strength structural material and low weighed unstructural material
9)increasing structure ductility
10) designing on the basis of the first plastic joint formation in the beams before columns
CONCLUSION
346
According to the existing codes, instructionsand issues and because most of the destructions were due
to not obeying them in the building it is suggested to the following to improve the quality of the works
done:
a)educating all the project workers
b)quality control for construction material

1)providing technical and professional labs to control the quality of the construction material
2)obligatory standard for all the material
3)proper keeping of the material up to the using time
4)using light and resistant material
5)using modern construction system
c)control in all the constructing places

1)controlling all the phases by resident observer
2)controlling all the phases and preparing required reports
3) performing required test in different phases
4) controlling the builder groups in relation with required profession
5)control and correct use of the proper machinery
d) correct maintenance
1)providing documents for all the buildings
2)professional annual control for all the buildings
3)performing the required repairs on time
4)recording the changes and repairs in the documents provided for the building
REFERENCES
Iranian concrete code (ABA)

Iranian earthquake code 2800
Instructions for national seismic improvement
347
June 25-28, 2007
Paper No. W2-1013
Underground monuments (Catacombs) in Alexandria, Egypt.
Sayed HEMEDA1, Kyriazis PITILAKIS 2, Ioanna PAPAYIANNI 2, Stavros BANDIS 2,

Mohamed GAMAL3
ABSTRACT
The Catacombs in Alexandria, Egypt from the Greek-Roman era represent cultural heritage of
outstanding universal values. They suffer weathering aging as well as multiple geotechnical and
earthquake problems. A pilot study has been carried out for on the Catacombs of Kom El-Shoqafa in
order (a) to define the pathology and the causes (b) to study the seismic performance (c) to assess the
global risk due to several factors and (d) to define the appropriate retrofitting techniques to be applied.
In the paper a general outline of the various tests, surveys and analyses is presented, highlighting the
most important issues related to the static and seismic stability of underground structures like
Catacombs in particularly unfavorable geotechnical and environmental conditions.
Keywords. Catacombs of Kom El-Shoqafa, Alexandria underground monuments, geotechnical

problems, geotechnical investigations, microtremors, creep tests, stability and seismic analysis.
INTRODUCTION
Alexandria is located in eastern part of the Mediterranean Basin (Northern Egypt) and it is a place of
great historical and religious interest. Numerous Catacombs and cemeteries for Greek-roman were
erected in Greek-roman and Christian era have been found. They represent actually a large complex of
an underground necropolis. The aim of the present study is the investigation and documentation of the
existing stability conditions of the geological formations at a site of a catacomb in order to define the
instability problems to interpret the pathology and to propose the best retrofitting procedure
On 28 September 1900 the ground on the Hill of potsherds (Kom El- Shoqafa) spontaneously opened,
and a donkey disappeared into the crevasse. The unfortunate beast had inadvertently discovered one of
Alexandrias most important archaeological sites, the principal hypogeum of a funerary complex
dating from the end of the first century of the Christian era and still in use at the beginning of the
fourth(Empereur,2003). It had been known for some time that this area held antique tombs, since the
hill has being extensively quarried to provide building materials for a fast expanding modern
Alexandria. Much had already been destroyed, though certain archaeologists of the late 19th century
had been able to record other tombs that were subsequently to disappear. These reports have
descriptions and drawings, which show that the complex that can visit today was part of a vast
necropolis, traces of which must still exist under the foundations of the neighboring buildings.
The stability conditions of the historical monuments are of crucial interest, especially in regions like
the Mediterranean Basin and particularly Alexandria, Egypt, where the seismotectonic and weathering
regime is active and the geological structure is complex .Phenomena like settlement and slope
movements as well as earthquakes and tectonic activity contribute to the damage of the historical
buildings. The ground water activity is also an important factor, especially in cases underground
monuments the environmental factor is also necessary to be taken in mind, when different protection
measures are decided to apply.
1
Doctoral Candidate, Department of Civil Engineering, Aristotle University of Thessaloniki, Greece.
E-mail : hemeda@civil.auth.gr
2
Professors, Department of Civil Engineering, Aristotle University of Thessaloniki, Greece.
3
Associate Professor, Geophysics Department, Cairo University, Egypt.
348
PROBLEM DESCRIPTION
The site: The Catacombs of Kom El-Shoqafa lies in the district of Karmouz on the south-west
of Alexandra, not far from the so-called Pompey, s pillar, on the south slope of the hill. The
area was called Kom El- Shoqafa or a pile of shards. This catacomb lies at about 2-2.5
kilometersfrom the seashore, and it is higher in topography than Amod El- Sawari area, Fig
(1).
The structure: The catacombes under study was most likely initially a private tomb
and later converted to a public cemetery. It consists of 3 levels cut into the rock, a staircase, a
rotunda, the triclinium or banquette hall, a vestibule, an antechamber and the burial chamber
with three recesses in, where in each recess there is a sarcophagus. The Catacombs also
contains a large number of Luculi or grooves cut in the bedrock.
Figure 1. Location maps of Catacombs of kom El-Shoqafa.
The
Entrance
Catacombs of kom El-shoqafa site.
349
El-Shatbi tombs site.
Moustafa kamil tombs site.
Amod El-Sawari site (library and temple).
Figure . 2. Underground monuments (Catacombs) in Alexandria, (present condition).
Pathology and causes : The Alexandria Catacombs show some clear indications of yielding
and partial collapse at several locations, Weathering as indicated by in particular honeycomb
,stone surface scaling disintegration of construction material, intense rock meal damp surfaces
in particular for semi-sheltered parts of the excavation, white salt efflorescence and yellowish
brown iron staining can be noted at many parts.
The structural damage is represented by ceiling cracking, stone surface decay and partial
collapse of some parts of the ceilings and walls, rock exfoliation especially noted in the
ceiling of the narrow corridors that are found at the deepest parts and mass wasting from its
ceiling and walls of corridors.
In conclusion the present state of conservation of the great Catacomb of Kom El-Shoqafa, the
best known and most highly prized testimony of Alexandrine funerary architecture culture is
now at the utmost limit of degradation.
350
The majority of structural damages have been caused from one or the combination of the
following factors:
Progressive weakening of rock material due to intrinsic sensitivity to weathering
factors especially the underground water and salt weathering effect.
Earthquake and other man made dynamic loading
Permanent deformation of the rock mass.
Natural wear of material.
Construction history in the area
0 2 4 6 8 10 M
Figure 3.Plan and a typical cross section A-A/ of the Catacombs of kom El-shoqafa.
Figure 4 .Catacombs of kom El-shoqafa.present conditions.
351
GEOTECHNICAL CONDITIONS
The Catacombs are carved in the oolitic sandy limestone (calcareous cemented sand); it is
yellowish white massive, fine to medium grained cross-bedded sandstone cemented with
calcareous cement. Intersected conjugated joints filled with very fine friable sand saturated
with water in the lower parts. This unit is underlined by loose calcareous sandstone. It is
brownish medium to fine grained calcareous limestone over saturated with ground water. It
overlies unconformable the El Hagif formation (Pliocene) or the older Miocene. Surface
quaternary deposits obscure actual contact (Adel ElFouly 2000, Girgis 1956, Said1990).
Bulk structure of rock and construction materials.

Different tests have been made to assess the durability and the weathering effects on the bulk
structure of the rock mass used as building material. The following set of mineralogical
analyses have been performed for the soft rock and the building materials.
-X-ray diffraction
-Thermal analyses DTA&TGA
-X-ray Fluorescence analysis
-Chemical analyses (for plaster and painting layers collected samples)
-Transmitted plane polarized light (for the collected rock samples).
-Scanning electron microscopy (SEM), attached with EDX Microprobe (energy dispersive X-
ray) microanalyses.
-Porous media characterization for weathered and sound rock samples.
-Pore size measurement and Specific surface area by nitrogen BET-TPV.
-Determination of the specific surface area (SSA).
-Grain size characteristics of the weathered (salt contaminated) rock samples.
-Saturation coefficient, S.
-Capillary water uptake measurements.
Based on this extended set of tests a detailed mineralogy description of the rock mass is
achieved together with other construction materials (plaster). A detailed presentation of these
tests may be found in S.Hemeda (2007) (Progress report of the PhD thesis, AUTH). Selected
results are presented herein.
Catacombs of Kom El-Shoqafa are carved in the fossiliferous sandy oolitic limestone
composed of calcite CaCO3 (47 %),quartz SiO2 (31%) ,halite NaCl (12 %),gypsum Ca
SO4.2H2O (10 %).The rock is yellowish and can be characterized as medium grained with
uniform relative grain size, angular to sub angular grain shape with equidimensional form and
rough surface texture. Sound pieces of rock can be characterized of medium compactness and
durability and the weathered pieces characterized of low compactness and durability. It must
be mentioned that weathering attacked strongly the rock materials, started from the surface
and continuing inward thus loosing the mineral fabric.
Diagrams of representative rocks samples are given in figures (5). From this analysis we
noticed that, % W.T of CaCO3 for the collected rock samples is 64, 25 and for the % W.T.
free water is 0, 54 .and for the % W.T Ca (OH) 2 the results are zero.
The elemental arrangement for the sandy oolitic limestone samples collected from Catacombs
of Kom El-shoqafa can be put in a decreasing order according to their
concentration:CaO(51.54),SiO2(14.55),MgO(1.07),Al2O3(2.51),SO3(0.59)Fe2O3(1.05),Na2O(0
.15) ,K2O(0.02),TiO2(0.04),P2O5(0.03)and L.O.I(28.20) total is (99.75).
In the internal structure we can observe the dominant components, which are bioclasts of
gastropods, foraminifera, algae, and shell debris, most of them are with test wall of
neomorphic microspar, while the tests are internally filled with micrite and microspar.
352
Long
contact
Long
cracks
Figure .5. TGA-DTA results,Thin section blue dyed X40, the blue is the porosity.
Figure .6. Photomicrographs and (EDX) micro analysis of the tested sandy Oolitic
limestone samples collected from catacombs of kom El-Shoqafa.
The pore diameter distribution of these rock samples is ,10-20A (4%) ,20-30A(9.17 %) ,30-
50A(11.1%) ,50-100A(15.847%) ,100-200A(17.80%) ,200-1960 A (41.97 %) ,and nm
2.26599 E-05 , and BET(m2/gr) 2.21098, and TPV (ml/gr) 0.00992,and micro porosity % is
1.79327 ,fig(7) . and for the Adsorption / desorption isotherm diagram. But the rock sample
which collected from the sound layers from the same site ( COM2) we noticed that : the pore
diameter distribution of this rock is ,10-20A (9.46%) , 20-30A(12.5%) , 30-50A(14.98%) ,
50-100A(12.72%) , 100-200A(7.93%) , 200-2040A(42.4%) ,and nm 1.5522E-05, BET(m2/gr)
1.51452 ,TPV (ml/gr) 0.00232 ,and micro porosity % is 0.42542 .
catacomb of kom El-shoqafa Catacom b o f kom El-s hoqafa
45 42. 4
45 41.97 40
40 35
35 30
30
25
25 %
% PORES 20 17.8 20
15.847 14. 98
15 15 12. 5 12. 72
11.1
10 9.17 9. 46
10 7. 93
5 4
0.12 5
0
0
<10 10.0- 20-30 30-50 50- 100- 200-
10. 0-20 20-30 30 -50 50-100 100-200 200-
20 100 200 1960 2040
Diameter (A) pore dia m e te r (A)
Weathered layers sound layers

Figure 7. Pore size (diameter A) distribution for collected unweathered rock sample
from catacomb of kom El-shoqafa.
353
Geochemical /hydro chemical conditions.
The geochemical and hydro chemical conditions have been investigated through specific test
as follows:
-Determination of anions by dioxen ion chromatography (for the rock materials).
-Atomic absorption spectrophotometer, determination of soluble cations and anions in the
collected underground water samples.
-Suspension PH and electrical conductivity (EC)
All plaster and rock samples showed much higher salt contents. The sulphate contents were
consistently higher than the chloride content; sulphate and chloride are the dominant salts.
The type of water in the Catacomb of Kom El-Shoqafa is stagnant not sea water, and it can
notice that the dominant anions are the sulphates which are from 845-1156 ppm l, than the
chloride which are 400-390, and this refer that the hypothetical dissolved salts are sulphates
mainly Na SO4 and Na CO3, followed by chlorides mainly Na Cl.
The hydro chemical analysis of the sewage and the stagnant water indicated very close
similarity in salt type between them. Also the dominant salts in construction material of the
excavation (especially at highly populated area of Alexandria city i.e. Amod El-Sawary and
Kom El-shoqafa) are sulphates then chlorides, very similar to the sewage water at study area.
The sewage water can find its way to permeate to deeper parts in this porous rock -as
indicated from the BET and pore size distribution measurements where these excavations
have been constructed.
The PH values for the studied water samples determined using the method described in the
(ASTM D 4972-95a (1999), (see Table1).varies from 7.0 to7.2 while the PH of the sea water
is 7.6.
Geotechnical and geophysical surveys

The geotechnical investigation comprise six sampling boreholes with SPT measurements
carried out at the archaeological site of catacomb of Kom El-Shoqafa Figure (8) show a
representative profile .
Figure 8. Borehole 1. Catacombs of Kom El-Shoqafa.
354
Ambient noise measurements
Array microtremor measurements have been carried out to define the shear velocity profile at
the site. The ReMi method has been used and the estimated Vs profile is given in Figure (9) .
The obtained shear wave seismic velocities for Remi tests Remi-1 show a relatively high
range of shear wave velocities ranging between 260 m/s to 1420 m/s. However it is clear that
the ground conditions were catacombs are excavated cannot be classified as real rock at least
close to the surface.
Site -1 : Supportive Illustration
Dispersion Curve Showing Picks and Fit

1500
1400 Site-1: Vs Model
1300
1200 0 200 400 600 800 1000 1200 1400 1600 1800
1100
1000 0
900
800
700
600
.
500
400 Calculated Dispersion
300 Picked Dispersion
200
100
]
.
0
0 0.05 0.1 0.15 0.2
Period, s -10
p-f Image with Dispersion Modeling Picks

.
-20
-30
. Experimental
Shear-Wave Velocity, m/s
Figure. 9. The dispersive curve, P-f image and corresponding picks for Remi-1 test,
Catacombs of Kom El-Shoqafa site.
Geotechnical properties of the Calcarenitic rock.
In the framework of this project a specific set of laboratory tests has been carried out in order
to define the physical and the mechanical properties of the rock where the catacombs were
excavated. These studies include:
-Very Slow uniaxial creep tests on rock samples.

The objective of these tests is first to quantify the viscous parameters that govern the long
term behaviour of these underground structures and, secondly, to define the necessary
parameters for the numerical modeling. Most specimens, under constant axial stress shows
complete creep phase: transient, steady and tertiary creep phases. Figure (10).
355
0,088 0,5
0,086
0,45
Steady creep stage Steady creep stage
0,084
(II) (II)
0,082 ertiary creep stage

0,4
(III)
0,08
0,35
0,078
Transient creep stage

0,076
(I)
0,3
Transient creep stage 1=1,75 Mpa
0,074 (I) 1 =1,35
Mpa 65% UCS
50% UCS
0,072 0,25
0 50 100 150 200 250 3 0 50 100 150 200 250 300 350 400
Time(days).1/9/2006-1/6/2007 Time (days). 8/5/2006-1/6/2007
Figure 10. Catacomb of kom El-shoqafa, Strain-versus-time in uniaxial creep test Nr.1
and Nr.2.
-Triaxial Creep Tests on rock samples.

The objective of the triaxial creep tests is to determine visco-plastic parameters of the soft
rock specimens under confinement. The time-related parameters are monitored, recorded and
analyzed.
The axial strain-time curves are given in Figures 11. From the results it is proved that there is
an instantaneous transient creep state under constant axial load and confining pressure
followed by a steady creep phase. Post-test deformation increases rapidly at the beginning to
the first few hours of the test and then tends to remain constant. No specimen failed after at
the end of the test duration.Crack propagation (in the brittle field) and pore collapse (at higher
stress conditions) are therefore the dominant deformation mechanisms for the Alexandria
surface rock as it has been also demonstrated by (Aversa,S 991;Maranini and Brignoli, 1998)
4,4
2,8
4,1
Steady creep stage 2,5
(II) Steady creep stage
(3) 1=2,52 MPa
(II)
3,8 2,2
1,9
3,5
yield stress Transient creep stage
1,6 (I)
. (2) 1=2,16 MPa
3,2
Transient creep stage
1,3
(I) 3 = 225 kpa
3= 510 kpa
(1) 1 = 1,45 MPa
1= 1,41 MPa
2,9 1
0 20 40 60 80 100 120 140 160 180 200 0 20 40 60 80 100 120 140 160 180 200
Time ( hours). Time ( hours) .
Figure 11. Catacomb of Kom El-Shoqafa Strain-versus-time during in triiaxial creep

test no.1 and 2.
356
Short term laboratory tests. Mechanical properties, Stress-Strain behaviour.
-Seismic P-wave velocities.
Seismic P-wave velocities were measured in the laboratory (ASTM 597, ASTM D 2845-83)
for the sandy o litic limestone samples; They vary from 1.83 to 2.00 km/s (the mean value
1.96 km/s) with standard deviation 176 m/s for the orientation perpendicular to bedding plane.
P-wave velocities parallel to bedding plane show higher velocities ranging from 1,93 to 2, 44
km/s (with mean velocity of 2.17 km/s).
1300 3,5
1200
Moustafa kamil tombs.
3
1100
Catacomb of kom El-
1000 shoqafa.
2,5
900 EL-shatbi tombs.
Catacom b of kom E l-
800 sho qafa.
2
Mous tafa kam il
700 tom bs.
600 1,5 El-sh atb i tom bs.
500
1
400
1 1,5 2 2,5 3
200 250 300 350 400 450 500
Vertical wave velocity (km/s).
E-static ( MPa).
Figure 12. P-wave velocities and corresponding modulus of elasticity measured in

vertical and horizontal directions.
-Brazilian strength (BTS).

The mean Brazilian tensile strength (BTS) values of the rock is 0.27 MPa .
- uniaxial compression test.

The unconfined uniaxial compressive strength (UCS) is measured between 2.6 and 2.74 MPa.
Uniaxial test (Catacomb of kom el-shoqafa site).
3
Yie ld fric tion
brittle Ductile
2,5
UCS = 2,6-2,7 4 MPa
Initia l fa ilure point
= 74-10 2 MP a
= fa ilure stre ngth
v= 0,24 -0,28
2
Ine la stic
1,5
be dding
Ela stic
0,5
0
0 0,5 1 1,5 2 2,5 3 3,5 4 4,5
Axial strain (%)
Figure 13.Uniaxial unconfined compressive strength.
-Direct shear tests.
357
Figure 14.presents the results of direct shear tests performed on completely deteriorated
samples.
120 120
110
100 100 Cu = 0
90 =30
80 80
70
60 60
50
40 40
n=50 KPa
30
n=100 Kpa
20 20
n=200Kpa
10
0 0
0 50 100 150 200 250
1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 39
Normal stress n (Kpa).
Ho rizo n tal disp lacemen t (mm )
Figure 14. Direct shear test on completely deteriorated rock samples
-Ttriaxial compressive strength.

Figure 15, presents the results of isotropic compression test performed on undistributed
samples. It is found that the sandy oolitic limestone rock samples have the following
mechanical and elastic properties: c = 650 KN/m2, = 36o, E = 270 to 360 MPa, =0.26-
0.29, G=140.6 MPa, K=272.7Mpa.
Mohr circles diagram from quick-undrained triaxial tests,Catacomb of kom
6,00 el-shoqafa.
3000
E=270-360 Mpa C = 650 KN/m2

2500
=0,28 = 36
5,00 G=140,6 Mpa 3 = 0,6 MPa
2000
K=272,7 MPa
1500
3= 0,325
4,00 1000
500
3,00 0
3 = 0.25 MPa 0 1000 2000 3000 4000 5000 6000
2
Principal stresses KN/m
ypical result of triaxial test .
6 Brittle
2,00
= 3 ,5 90 2 3 + 2 ,96 72
5 1c
R2 = 0 ,9 61 1
3 Ductile
1,00
2
0
0 1 2 3 4 5 6
0,00 M ino r P rin cipa l St re s s , 3 ( Pa ) .
0 5 10 15 20 25
Axial Strain (10-3)
Figure 15. Triaxial test at different confining pressures.
358
Table 1. Short term laboratory tests carried out with Calcarenitic rocks under investigation.
Uniaxial compression test Triaxial compression test Direct Brazilian Block Wave
Shear test
test Punch velocity
Depth. for rock
joints (BTS) Index (Vp)
Type of of d MPa (BPI) Km/sec
Site. 3 (GS) ucs Failure E C E {G} {K} C
rock Sample Mg/cm % MPa Strain Youngs [-] [ ] KN/ Youngs [-] Shear Bulk [ ] MPa
(m) f (%) Modulus Poissons m2 Modulus Poissons modul modulus
(MPa) ratio (MPa) ratio us (MPa)
(MPa)
Intact Moustafa M_2 1.62 2.74 36 3.55 68 0.28 31o 400 410 0.28 160 311 0.37 0.7 1.96
Calcarenite kamil
rock ( pure tombs M_3 1.75 2.80 28 3.45 106 0.30 340 0.28 133 258 0.35 0.7 2.08
Limestone). site. 51 0
M_4 1.77 2.79 19 2.30 78 0.24 300 0.30 117 227 0.34 1 2.1
M_5 1.65 2.67 28 3.40 106 0.30 0.34 0.9 2.18
M_7 1.65 2.84 27 3.45 80 0.28 0.39 1.2 2.48
Sandy Catacomb COM_3 1.69 2.66 32 2.6 74 0.24 36o 650 360 0.29 141 273 0.25 0.5 1.93
Oolitic of kom el- Complete
limestone. shoqafa. COM_5 1.70 2.78 31 2.74 102 0.28 345 0.26 135 261 weathered 0.27 0.6 1.96
samples
COM_8 1.68 2.49 21 2.7 74 0.28 270 0.26 105 205 0.28 0.7 1.97
30 0
COM10 1.65 2.80 40 2.6 74 0.24 0.31 1.1 2.06
31 0
COM15 1.68 2.74 31 2.74 80 0.24 0.29 1 2.17
Calcarenitic El-shatbi SH_2 1.80 2.58 23 6.3 147 0.30 53o 700 445 0.28 173 353 0.46 0.8 2.44
rock. bombs
site. SH_3 1.79 2.82 23 4.7 138 0.28 440 0.28 172 333 0.49 1 2.55
SH_4 1.84 2.58 36 5.03 142 0.28 392 0.25 153 297 56 0 0.48 1.2 2.71
SH_6 1.62 2.49 16 6.3 147 0.30 353 0.24 138 267 0.52 1.4 2.83
SH_7 1.82 2.55 24 6.1 140 0.30 233 0.25 91 177 0.60 1.3 3.06
359
SEISMIC ANALYSIS OF A TYPICAL CROSS SECTION.
The catacomb of Kom El-Shogafa has been studied for specific seismic scenarios using the
finite element code [Plaxis V8].A typical cross section is used in this study (see Figure 24).
The material properties (physical, mechanical and dynamic) used through out the following
analysis have been defined in the previous chapters.
Preliminary static analysis
In the initial static analysis the excavation is modeled by assuming non-linear soil / rock
behavior and the Mohr coulomb failure criterion. The following parameters are used = 36o c
=650 kN/m 2, E=2.270E+06 KN/m2 = 0.28, Vs= 714,49 m/sec for the rock material and =
31o c =500 kN/m2E=1.350E+06 KN/m2 , = 0.25, Vs= 557 m/sec for the modern brick
support walls and piers.The results from the preliminary static analysis indicate that the
ground displacements were very small (of the order of few millimeters) and some piers are
under relatively high compression stresses.(see in Figure16 (b) where the calculated peak
effective principal stress is 1.42*103 KN/m 2 ).
(a)
(b)
Figure 16. Static analysis (a) Total displacement = 1.13*10 -3m. (b)Vertical peak effective
stresses (-1.42*103 KN/m2).
360
Seismic analysis
Alexandria during his long history has suffered important seismic damages from near and
distant sources earthquakes. As the under-water archeological remains in Abou Kir bay
strongly support, the city was destroyed by either local or remote earthquakes (El-Sayed,
2004). Seismogenic zones such as the Red Sea, Gulf of Aqaba-Dead Sea Hellenic Arc, Suez-
Cairo-Alexandria, Eastern-Mediterranean-Cairo-Faiyoum and the Egyptian costal area may
all affect the city. However the seismic hazard of the city has not been fully defined.
In the present study we have selected three reference earthquakes. (i) Aqaba, Egypt, 1995 (ii)
Ertzincan,Turkey, 1992 and (iii) Kalamata, Greece, 1986. The time histories (Figure 17) of
these earthquakes representing different seismotectonic settings and frequency content were
scaled to three peak ground acceleration values equal to 0.08g , 0.16g ,and 0.24g respectively
and they are used as input motions at the bedrock. The design acceleration in Alexandria
according to the Egypt code is 0.08g.
0.3
0.1 0.6 y xomponent
Acc(g) EW componen t
0.2
0.4
0.05 0.1
0.2
0 0
0
0 10 20 30 40 50 60 70 0 5 10 15 20 25 0 5 10 15 20
-0.2 -0.1
-0.05 tim e (sec)
tim e (sec) -0.2
-0.4
-0.1 -0.6 -0.3

Time {s}
Ka lam ata
0.12 0.3
Fourier spectrum Fourier spectrum 1.2 Fourier spectrum
0.1 0.25 1
0.08 0.2 0.8

0.15 0.6
0.06
0.1 0.4
0.04
0.2
0.02 0.05
0
0 0
0 1 2 3 4 5 6 7 8
0 2 4 6 8 10 0 1 2 3 4 5 6 7 8 9 10
period {s} Period ( se c)
Pe riod (s ec)
(a) (b) (c)

Fiure. 17. Seismic excitations at the bedrock for the reference ground motions
(acceleration-time history, and Fourier amplitude spectrum). (a) Aqaba, 1995 (b)
Ertzincan 1992,and (c) Kalamata, 1986.
Analysis results
Figures 18 through 22 are summarizing the calculated displacements for the three PGA
scenarios and the three design earthquakes. In case of Aqaba earthquake, it is clear that a great
part of seismic energy is dissipated to the upper parts of catacombs (ground surface) even for
small values of PGA. The Kalamata and Ertzincan input motions give much low displacement
values than for all. The maximum horizontal displacement at the top of catacombs for Aqaba
0.24g earthquake scenario was: ux 15.2 cm, while the peak effective principal stress was 5050
KN/m2. In the case of Ertzincan and Kalamata 0.24g scenarios the respective values were
3760 KN/m 2, 3890 KN/m2.
Given the value of the static strength estimated in the laboratory the seismic analysis of the
Catacombs complex proved that the piers, which are the most vulnerable parts of the whole
complex, are rather safe for PGA values lower than 0.16g in case of the Kalamata and
Erzincan earthquakes and PGA =0.10g for the Aqaba seismic scenario. (Figures 23 and 26b)
For larger earthquake which are most likely to happen in the region, the seismic stability of
the Catacombs is not satisfied and it is necessary to proceed to specific retrofitting works to
upgrade their seismic performance. The maximum differential horizontal displacements of the
top and the base of the piers is of the order of 5mm (Figure 25). Considering that the induced
seismic ground deformations are better correlated with the intensity of damages in
underground structures, the seismic design of the catacombs must be based on these kinematic
forces.
361
(a)
(b)
(c )
Figure. 18. Deformed mesh and peak total displacement at the final step,(a)Kalamata
RQ,(b)Erzincan RQ,(c)Aqaba RQ ,the PGA =0.08g .
362
acc{m/s2}
0.01 9
9.0
Top of Catacombs
Top of Catacombs
6.0 6
0.005
Bottom of Catacombs
Bottom of Catacombs
3.0 3
0
Bottom of mesh
Bottom of mesh 0 0
-0.005
-3.0 3
-0.01 -6.0 016

0 4 8 12 16
0 4 8 12 16
0 4 8 12 16
Time[s]
(a) Time[s]
2
acc{m/s
ation [m/s2] }
0.01
0.8
0.005 0.6
0.4
0
0.2
-0.005
0
-0.01
-0.2
-0.015
-0.4 5
0 5 10 15 20 0 5 10 15 20
(b)
Time [s] Time [s]
acc{m/s2}
ation [m/s2]
0.04
0.8
0.02
0.4
0
0
-0.02
-0.04
-0.4
-0.06
-0.8
0 10 20 30 40 0 10 20 30 40
0 10 20 30 40
time {s} Time{s}
(c)
Figure .19. Computed displacement and acceleration time histories for
(a)Kalamata,(b)Erzincan,(c )Aqaba input motions PGA= 0.08g .
363
(a)
(b)
(c )
Figure . 20. Deformed mesh and peak total displacement at the final step,(a)Kalamata
RQ,(b)Erzincan RQ,(c )Aqaba RQ .the PGA =0.16g .
364
Acc (m/s2)
ation [m/s2]
0.02 1.5
Top of Catacombs Top of Catacombs
1
0.01
0.5 Bottom of Catacombs
Bottom of Catacombs
0
0 Bottom of mesh
Bottom of mesh
-0.5
-0.01
-1
-0.02 -1.5
0 4 8 12 16 0
0 4
4
8 8
12
12
16
16
Time[s] Time{s}
(a)
0.02
tion [m/s2]
Acc (m/s2)
1.5
0.01 1
s
0 0.5
-0.01 0
-0.02 -0.5
-0.03 -1 0 5 10 15 20
0 5 10 15 20 0 5 10 15 20
Time [s] (b)
Time{s}
Acc (m/s2)
tion [m/s2]
0.08 1.5
1
0.04
0.5
0 0
-0.04 -0.5
-1
-0.08
-1.5
-0.12 -22 0 10 20 30 40
0 10 20 30 40
0 10 20 30 40
(c ) Time{s}
time {s}
Figure.21.Displacement and acceleration time histories at the bottom and the top of the
Catacombs (a)Kalamata RQ,(b)Erzincan RQ,(c )Aqaba RQ =0.16g .
365
Acc(m/s2)
0.03 3
3
Top of Catacombs Top of Catacombs
0.02 2 2
Bottom of Catacombs
Bottom of Catacombs
0.01 1
Bottom of mesh
Bottom of mesh
0 0
0
-0.01
-1
-0.02 -2 2
0 4 8 12 16
0 4 8 12 16
0 4 8 12 16
Time [s] (a) Time{s}
0.03
ation [m/s2]
3
Acc(m/s2)
3
0.02
2
2
0.01
0 1
-0.01
0 0
-0.02
-1
-0.03
-0.04 -2 2
0 5 10 15 20
0 5 10 15 20 0 5 10 15 20
(b)
Time [s] tion [m/s2]
Acc(m/s2)
0.08 3
0.04 2
0 1
-0.04
0
-0.08
-1
-0.12
-0.16 -2
-0.2 -3
0 10 20 30 40 0 10 20 }
Time{s 30 40
(c ) time{s }
time {s}
Figure.22. Acceleration and displacement time histories PGA =0.24g.
366
5000.00 400.00
4000.00 300.00
3000.00 200.00
2000.00
100.00
1000.00
0.00
0.00
-100.00
-1000.00
-2000.00 strength of rock -200.00
material
-3000.00 (experimental) -300.00
-4000.00 -400.00
0.00 2.00 4.00 6.00 8.00 10.00 12.00 14.00 16.00 0.00 2.00 4.00 6.00 8.00 10.00 12.00 14.00 16.00
Time [sec] Time (sec)
(a)
4000.00 250.00
3000.00 200.00
2000.00 150.00
100.00
1000.00
50.00
0.00
0.00
-1000.00
-50.00
-2000.00 strength of rock
material -100.00
-3000.00 (experimental) -150.00
-4000.00 -200.00
-5000.00 -250.00
0.00 5.00 10.00 15.00 20.00 0.00 5.00 10.00 15.00 20.00
Time [sec] Time [sec]
(b)
6000.00 1000.00
800.00
4000.00
600.00
2000.00
400.00
0.00 200.00
0.00
-2000.00
strenght of rock -200.00

-4000.00 material
(experimental) -400.00
-6000.00 -600.00
0.00 5.00 10.00 15.00 20.00 25.00 30.00 35.00 40.00 0.00 5.00 10.00 15.00 20.00 25.00 30.00 35.00 40.00
Time [sec] Time [sec]
(c)
Figure. 23. Effective stresses time histories for the most critical Pier 1 ( Figure 24 at the
right side of rotounda) (a)Kalamata RQ,(b)Erzincan RQ,(c )Aqaba RQ. the PGA=0.24g
.
367
Pier 4 Pier 3 Pier 2 Pier 1
Figure. 24. A typical cross section which used in the seismic study.
4 4
3.5 3.5
p ier_1
3 p ier_2 3
2.5 p ier_3 2.5
p ier_4
2 2
pier_1
1.5 1.5 pier_2
1 1 pier_3
0.5 0.5 pier_4
0 0
27 28 29 30 31 32 33 0 0.2 0.4 0.6 0.8 1 1.2
-3 -3
Horizontal displacement {10 m). Vertical displacement {10 m).
(a)
4 4
3.5 3.5
pier_1
3 3 pier_2
2.5 2.5 pier_3
pier_4
2 p ier_1 2
p ier_2
1.5 1.5
p ier_3
1 1
p ier_4
0.5 0.5
0 0
44.5 45 45.5 46 0 0.2 0.4 0.6 0.8
Horizontal displacement {10-3 m). Vertical displacement {10-3m).
(b)
4 4
3.5 3.5
pier_1 pier_1
3 pier_2
3 pier_2
2.5 pier_3 2.5 pier_3
pier_4 pier_4
2 2
1.5 1.5
1 1
0.5 0.5
0 0
146 147 148 149 150 151 152 153 0 0.5 1 1.5 2 2.5 3 3.5 4
Horizontal displacem ent {10-3 m). Vertical displacement {10-3 m).
( c)
Figure. 25 Horizontal and vertical displacements on the top and the base of the most
critical rock pier of the catacomb complex(a)Kalamata at t =23,22 s,(b)Erzincan at t
=7.09 s,(c )Aqaba at t = 4.60 s. PGA = 0.24g .
368
A qaba RQ
6000
0.18
Kalamata RQ
Aqaba RQ 5000
0.15 Erzincan RQ
Erzincan RQ
Kalamata RQ 4000
0.12
strengthof rock
3000 t i l( i
0.09
2000
0.06
1000
0.03
0
0
0 0.08 0.16 0.24 0 0.08 0.16 0.24
PGA Value {g} PGA V alue {g}
(a) (b)
Figure 26 (a) Maximum horizontal displacements for different PGA values (b) Peak
effective principal stresses at the top of the most critical rock pier for different PGA
values
CONCLUSION
A summary of the present state of the Catacombs of Kom El-Shoqafa, in Alexandria Egypt
has been presented. In order to assess the the pathology of the catacombs a detailed laboratory
and field surveys and tests have been carried out. The rock material (oolitic sandy limestone)
has an important intrinsic sensitivity to weathering factors especially the underground water
and salt weathering effect. The infiltration of the water through the porous rock is one of the
main problems of the catacombs. The weathering process is linked to the textural
characteristics, like poor geotechnical properties, chemical carbonated composition, presence
of soluble salts in the porous system, marine climate with characteristic humidity and marine
spray, underground water. The durability of the rock is moderate to low due to its high silica
content. The strength of the sandy limestone where the catacombs are carved is low.
Considering all other affecting factors and the specific geometry of the complex this low rock
strength affects seriously the safety of the catacombs both under static and seismic loading
conditions. The preliminary seismic analysis of the catacombs with three seismic scenarios of
different PGA values proved that the some critical supporting parts of the catacomb structure
(i.e. tock piers) are safe, without any strengthening measures, only for PGA values lower than
0.08-0.10g, which is rather low considering the seismic activity and the past seismic history
of the city.
The present study may be considered as a preliminary pilot study for future conservation
efforts of these historical monuments, in order to assess the vulnerability of these
underground structures to different hazards and to propose appropriate strengthening
retrofitting measures especially to reduce the seismic risk.
REFERENCES
Aversa, S. Mechanical behavior of soft rock: some remarks .proc. of the workshop on
experimental characterization and modeling of soil and soft rocks, Napoli, 191-
223.1991
El-Fouly A. Void investigation at Gabbari tombs, Alexandria, Egypt using ground
penetrating radar technique ICEHM 2000, Cairo University, Egypt, September,84-
90, 2000.
El-Sayed A, and Korrat,I and Hussein, MSeismicity and seismic hazard in
Alexandria(Egypt) and its surrounding pure appl.geophys.161(2004)1003-1019.
Empereur.,J The Catacombs of Kom El-Shoqafa , Alexandria, Egypt.1-3.2003
Girgis., GP Geology of the Pleistocene sediments of the Mediterranean coast west of Abou
QirPh D geol, Fac Sci ,Cairo Univ,23,24, 1956.
Hemeda, S Progress report of PhD thesis, AUTH 2007.
Louie J.N.Faster, Better: Shear-wave velocity to 100 meters depth from refraction
microtremors arrays. BSSA, 91, 347-364. 2001.
369
June 25-28, 2007
Paper No. W2-1016
Effect of Strong Wind to the Central Tower, Bayon,

Angkor Thom, Cambodia
Y. Iwasaki, Geo Research Institute, Japan
Abstract
This paper presents a special mechanism of the failure process of the central
Tower of Bayon temple at Angkor Thom, Angkor, Cambodia. The tower stands as the
symbol of the Bayon with diameter of about 22m at the foundation and a height of
about 42m from the ground. The tower is founded upon the densely compacted sand
mound with a height of about 12m from the ground surface.
It is not likely that the tower fails due to overturn like the Pisa tower. This is because
rather wide base compared to its height. The problem of the Bayon Tower is the fall
off of stones from the Tower, which has resulted in non-symmetry with irregular shape
of the top part of the Tower.
Based upon the measurements of the gap movements and meteorological data, he
found the gaps open and close according to strong wind and further confirmed the
existence of a threshold level of the wind velocity that may trigger some permanent
displacement. The failure process of the central tower at Bayon temple is considered
as has been caused by progressive fall out of stone elements from columns.
The masonry structure as put them together always tends rather to spread out than
to get closer together. The failure process of the central tower at Bayon is Domino
process leading to failure caused by very strong winds.
Photo-1 Bayon temple, Angkor Thom, Angkor

1. Introduction
Japanese Government Team for Safeguarding Angkor (JSA) was established in 1994 after establishing ICC
(International Coordinating Committee for Safeguarding Angkor) during the intergovernmental conference on
370
Safeguarding Angkor by UNESCO in Tokyo. One of the objectives of JSA is preparing a strategic plan how to
restore the Bayon temple. Among various structures of the Bayon temple, the main tower is one of the most important
and a unique masonry monument. JSA has been working on monitoring the behavior of the tower since October 1997.
This paper presents a basic view of the stability problem and failure mechanism of the tower based upon the present
observation and monitored results.
2. The Central Tower at Bayon

Photo-1 and 2 shows general view of the Bayon and the Tower. Fig-1 shows vertical section of the Tower at center by
EFEO. The height of the tower is said as 45m and was measured as 42.2m from the ground level at north of Bayon
temple by total station by Geotechnical Unit, JSA in December 1998.
The central tower stands upon the upper terrace at GL+13m. The diameter of the Tower at the base is about 25meters.
The structure of the main tower consists of central and eight sub towers surrounding the central one with carved face
on the outward side. There is a small circular gallery between the central tower and sub-towers, where one may walk
around the tower at about mid level of 16m from the upper terrace.
In February 1998, JSA installed scaffolding ladder in the inner space of sub-tower at southwestern side to access to
the mid level space of the tower from the upper terrace. Using this scaffolding ladder, Geotechnical unit became to be
able to install various sensors at the upper level of the tower and to maintain them.[1]
Photo-2 General View of the Central Tower, Bayon Fig.1 Vertical Section of the Tower
Photo-2 East Face, Central Tower Photo-2 North Face, Central Tower, Photo-3 A Four Face Tower, Bayon
371
The apparent side view of the central tower is different depending upon the direction to the Tower, which has been
caused by falling out of the stone elements from the side surface of the central part. We need to study the mechanical
process of falling out these stones from the tower body to consider countermeasures to prevent. Since the most stones
are simply piled up from bottom to top without over wrapping horizontally, the stone column has been rather easy to
lean out if some horizontal movement taken place caused by plant rooting or other mechanisms.
There is a gallery within the central tower as shown in Fig.1,
which surrounds the tower at a level of about +25m from the
ground level. We can observe the present situation of the fall
out process of the central tower at the gallery. The Gallery is Central tower
the space between the inner central tower and eight sub-
Concrete bridge
towers as shown in Photo-4. The stones are piled up vertically
without over-wrapping horizontally each other and most
pillars stand independently. There are connecting parts
bridging between the central tower and eight sub towers.
These bridges are found broken and some are missing.
Concrete beams were placed and ion nails were used to
retrofit these bridging elements.
These consolidation works were executed by EFEO since
1930.
Openings are found between many neighboring stone sandstone
columns. These opening gaps have caused large
displacements of the top of the columns that might have
resulted in damages of bridge elements. Photo-5 shows the
inside of stonewalls of the Central Tower around the corner
of northwest. It should be noted that a vertical crack develops
in the side surface of a sand stone at the middle height, which
is enlarged and shown in the left of the Photo-4. The crack on
the stone surface initiates at the upper surface and has sub-tower
developed downwards. It should be also noticed that the (SW)
separation of sandstone above the cracked one from the left
neighbor stone, while no separation of the gap spacing
between the cracked stone and the adjacent one. The Photo-4 Bridge between Central and sub Tower
separation of the spacing above the cracked stone is found
continued upwards. At the same time, no separation is found
at the lower stones than the cracked one.
separation
No separation
No separation
Photo-5 A crack that developed in a sand stone element
372
There are several points at the southwest, west, northwest, north, and northeast corner of the Gallery. These cracks
are the probable points where the fallout shall begin and cause the failure of the Tower.
3. Cracks on the stone surfaces

It is also found that there is opening of the vertical space between adjoining column. The gap increases with its height,
which suggests that the column is inclined outwards from the vertical position.
We measured angles of cracks from horizontal line. Several examples are shown in Table 2. The most angles are
found rather large as large as 70 degrees. The outwards displacement of the upper stone might have produced strong
frictional force against the stone underneath, which has been split due to the tension force.
Table 2 Cracks and inclination angle with position
number Inclination Wall

(in degree) position
1 70 SW corner
2 45 SW corner
3 70 NW corner
4 70 EN corner
5 70 NW corner
6 60 NW corner
Fig.2 Cracks underneath the inclined column
Photo-6 Two cracks at SW corner
The forced displacement is caused by horizontal widening of the spaces between columns at upper level by some
reasons under the condition of being fixed at the bottom level stone to the floor stone. It is expected that an increasing
friction force had been mobilized between the stone surfaces at the bottom due to the horizontal movement of the
column.
Photo-7(left side)
Plant root expanding gap
Photo-8 (right side)

gravels trapped in boundary gap
373
Photo-7 shows plant root grown in the gap of stones. It is obvious the plant root develops with enlarging the size of
the root that makes widening the space between stones.
Photo-8 shows another example of space widening phenomenon. Several gravels are found in a space between stones.
It is not known how these gravels had fallen and stayed, however, these gravels might have acted as non-reversible
ratchet. Once the opening of the space increases by some reason, gravels fall down and keep the space wider than
before.
4 Monitoring of the structural behavior

In 1997, we have installed several gap gages to monitor the changes of gap between the surfaces of the adjacent
stones at the central tower of Bayon. We thought it might be a good idea to develop an initial image on the
mechanism of the decay and possible structural modes of movements. We have installed inclinometers and gap
sensors as well as temperature gages at various points on the central Tower. Inclinometer shows seasonal changes of
the inclination that depends upon temperature and no tendency of the steady change with time.
4.1 Monitoring system

We began to monitor behavior of stone by installing gap sensors between adjoining stone faces in October 1997 and
inclinometers at central tower in February 1999. Since then, though missing data for some period, we have continued
monitoring. At the middle level of the central tower, there are eight sub-towers surrounding the central one. We have
installed six micro crack gages to vertical opening of adjoining stones of side walls of these sub-towers.
Fig.3 Measuring points of gap sensors and temperatures
t=15o C
=0.025mm
374
Fig.4 measurements of stone strain, gap open/close, and temperature Fig.5 surface strain vs. temperature
Photo.9 Installation of gap and temperature gages Fig.6 Seasonal change of Gaps at various points
Fig.3 shows plan view of the middle base level of the tower with location of sensors of temperature and gap. The
analog signals from these sensors are transmitted down to the monitoring box at the base level of the central tower
and are to be digitized with a sampling rate of an hour and stored in a hard memory. Solar battery has been used to
drive the whole system with back up lead battery being charged during day.
4.2 Temperature effect on strain of the stone surface and gap opening
To understand the temperature effects on the gap change more clearly, we have embedded temperature sensor within
stone at 2cm from the surface at SW gap point in September 1999.
Fig.4 shows the relationship between strain of the stone surface, temperature inside the stone, and the gap opening of
space between adjacent stone.
It is clearly found that the gap decreases with temperature increase when the strain increases. The response of the gap
well corresponds to the strain of the stone surface.
Based upon Fig.5, the relationship between strain and temperature is found as
( )= 4 t (oC)
The coefficient of temperature expansion ( ) of the sand stone is obtained as
= 4 x10-6
The average length of the stone is about 50-80cm. The expected gap for 15oC of difference in temperature is expected
as
=15x4 x10-6x(0.5-0.8)(m)
=30-50 m=0.03-0.05mm
The measured gap change is 0.025mm, which well corresponds to the estimated value.
4.3 Daily and seasonal change of gap opening

Photo.8 shows an example of installation of a gap sensor and thermometer at NW sub-tower. The measured data of
changes of gap and temperature from October 1997 to November 1998 are shown in Figs.5. The sensors are named as
NE, NW, W, S, SE, and E according to the directions of sub-towers that are being monitored. Among six sensors, it is
noticed that the trend of gap changes has some common and different characteristics. The daily change in common is
found to close the opening the gap in daytime and to widen in night. The magnitude of the gap closing and widening
is different for each point. Table-4 summarized the characteristics of each monitoring points in terms of daily gap
375
change, daily temperature change, as well as annual gap change. The temperature change at each point differs
according to the monitoring condition. The change is about 5 degrees where it is shaded, but becomes as high as 30
degrees where it is exposed directly to the sun. It should be noted that the daily gap change at each point does not
necessarily correspond to temperature change. Gap at NE and NW are found several times larger gap change among
others.
Table 4 Gap Opening daily and annual change
Gap point NE NW W S SE E
Daily gap change (mm) 0.02 0.05 0.03 0.01 0.01 0.01
Daily temperature change ( oC) 5 5 10 (30)
Annual gap change(mm) 0.5 0.5 0.2-0.5 0.1 0.1 0.1
Fig.7 Change of Gap

from May/1 to Aug/1, 1998
Fig.8 Hourly Max Velocity

from May/1 to Aug/1, 1998
Fig.7 shows the same data of Fig.6 only for three-month period from May 1, 1998, where the more detailed
movement can be observed. Fig.8 shows hourly maximum wind velocity. During the period, it is noticed that larger
movements had taken place for every points at the same time. One is around May 23 and another is on June 25 when
the gaps changed significantly compared to daily movement. The gaps opened except at W point, where the gap
opened first but turned to close immediately. These movements are compared to meteorological condition of the
hourly maximum wind velocity. The strong wind of about 12.5m/sec is found to correspond to the event of May 23.
However, this relationship is not applied to the event of June 25. It also should be noted that the sudden gap change
does not always accumulate in the following period. For example, the gap change at NW point increased and always
takes the reversal direction and returns to the original gap opening after 2-3 days. On the other hand, these sudden
changes at W points stay as shifted values. The rebound response to return to the original position may be due to
mechanical stability of the stone tower.
376
Fig.8 Hourly Gap Change and Hourly max. wind velocity Fig.9 Hourly gap change vs. Hourly max velocity
4.4 Strong wind and change of gap opening

Strong wind causes a small displacement between stone elements, which is basically a result of response of the tower
structure against wind.
Fig.8 shows hourly gap change at NS sub Tower and Hourly max wind velocity with time. Fig.9 shows the relation
between hourly gap change vs. hourly max. wind velocity. It is found that the gap change takes place when the
maximum wind velocity becomes more than 4-5m/sec. The stronger wind shall cause the larger gap change.
5. Conclusion
Several conclusions could be obtained through the observation and monitoring of the central tower as follows,
1. Central Tower lacks its symmetry that might be caused by falling off of sand stones of side face of masonry
structure.
2. Cracks in the foundation stones are found that could trigger domino failure of falling off the above stone column.
3. Gap between stones opens at night and closes in daytime, due to the change of the temperature.
4. Sudden large gap changes have been observed when the strong wind blows with greater than 5m/sec. The wind
action may induce to accelerate the vertical opening between the stone columns.
5. Very strong wind might have resulted to cause developing cracks on stones and become one of major factors to
cause damage of masonry structures in the future.
6. It is well observed that the horizontal sliding between dry masonry stones takes place easily if a few small
particles of sand grain are inserted within the stone surfaces. Botanical root grown between the gaps of the stone
boundaries also causes horizontal displacement resulting in fall out. The mechanical behavior of non-reversible
character of stone masonry depends upon displacement and the resistant force and relative movements among
selected masonry elements may reveal the basic mode of reversible and non-reversible movements.
7. The above-mentioned analysis of response mode of tower against wind is not only useful for estimation of
stability of the tower itself but also useful to select countermeasures against fallout of the stone element from the
tower.[2]
8. The masonry structure as put them together always tends rather to spread out than to get closer together. The
failure process of the central tower at Bayon is Domino process leading to failure caused by very strong winds.
6 References
[1]JSA (1999)Annual Report on the Technical Survey of Angkor Monument 1999, JSA, Tokyo, pp.199-245
[2]JSA (2005)The Master Plan for the Conservation of Restoration of the Bayon Complex, JSA, Tokyo, pp.206-
213
377
June 25-28, 2007
Paper No. W2-1011
DAMAGES OF EARTHEN STRUCTURES AT ARG-E BAM CAUSED

BY THE EARTHQUAKE OF DEC.26, 2003, THE CITIDEL AT BAM,
IRAN
Yoshinori IWASAKI 1, Mahmoud NEJATI2
ABSTRACT
This paper describes characteristics of damage of Bam city and Arg-e bam, the Citadel, one of the
World Cultural Heritage in Iran, during the Bam earthquake of December 26, 2003. The most
buildings in the city were earthen structures of adobe type and the earthquake caused heavy damages.
The Arg-e Bam was the one of the largest citadel consisting from adobe structures. Though the most
adobe structures in the Arg-e Bam were also severely damaged, a few structures have survived that
stands upon the rock foundation. Based upon the comparative analysis of the strong ground motions
recorded in Bam revealed that the critical strong ground motion that caused the damage of adobe
structure in Bam was the acceleration level from 200 to 300 gals with a predominant period of about
1.5second
Keywords: Bam Earthquake, World Cultural Heritage, Arg-e Bam, Earthen Structure
INTRODUCTION
Bam Earthquake occurred in the early morning of 5:26 on Friday, December 26, 2003. Bam locates
about 900km south east of Tehran, the capital city of Islamic Republic of Iran(see Fig.1). The Bam
had a population about 150,000 prior to the earthquake.
The earthquake had a magnitude of Mw=6.5(USGS) and an epicenter location of 29.00N,

58.34E(USGS). Focal depth was estimated about to be about 8-10km.The traditional mud-brick and
clay homes showed little resistance against the violent vibration of the duration of a few tens of
seconds. Walls and roofs crumbled and collapsed. More than 100,000 of victims were trapped beneath
the rubble and around 26,500 lost their lives. Tens of thousands left homeless and up to 6,000 children
were orphaned.
The largest mud brick complex in the world, Arg-e Bam (Bam Citadel), was almost totally collapsed.
THE GEOLOGY AND DAMAGE
Based upon the topological feature in the area, it was identified Bam fault as shown in Fig.2. The Bam
fault is strike slip with NS striking high angle reverse fault. After the earthquake, several organizations
have installed temporal net work of seismic recording stations. The structural damages are reported
concentrated within the city of Bam that is a few km west of the Bam Fault. Damages of Qanat system
that is water supplying underground tunnel concentrated the Bam fault. IIEES(International Institute
of Earthquake Engineering and Seismology) shows the distribution of the after shock is somewhat a
few km away from the surface fault line. The regional distribution of the structural damages in Bam is
1
Managing Director, Geo-Research Institute, Osaka, Japan Email: yoshi-iw@geor.or.jp
2
Recovery Project of Bams Cultural Heritage, Iran, E-mail: m_nejati@argebam.ir
378
shown()ICG Reconnaissance Mission Report(2004) in Fig.3. The geology of the Bam area is also
shown in the lower portion of Fig.3.
Fig.1 Map of Iran Fig.2 Bam Fault and after shock distribution(IIEES, 2004)
Fig.3 and 4 Damage distribution and Geology in Bam(IIEES)
The surface geology in Bam city composes of Quaternary sediment()K.A. Hosseini and et al.(2004). It
is interesting to note that relative resemblance between the heavy damage and geological
379
characteristics. In the Bam city, the difference may depend upon top surface condition. Most heavy
damage was in the center and southern part of the Bam that corresponds to fan silt and argillaceous
sandstones. Very nears the Bam fault the damage was not visible where yellow-tan to orange brown
siltstone and sand stone distribute.
Arg-e Bam - World Cultural Heritage Site

In the north eastern corner of the Bam, a citadel called Arge-Bam, was constructed by adobe brick in
ancient time and completely damaged during the earthquake of December 26, 2003. It was about six
months later when UNESCO has inscribed it as the World Cultural Heritage. Most of the adobe
structures are damaged, a few structure have been survived in the northern part where the base rock is
the foundation and exposed to the surface.
Photo-1 Arg-e Bam before the earthquake Photo-2 Arg-e Bam after the earthquake
THE STRONG GROUND MOTION
A strong ground motion was recorded at the Bam Governors Building (58.353, 29.079) by a net work
installed by BHRC(Building and Housing Research Center, Tehran).
1000 west Max Acc,ew=782gal

500 (sec.)
0
15 20 25 30
-500
-1000 east
1000 up Max Acc,v=968gal
500
(sec.)
0
15 20 25 30
-500
-1000 down
1000 north
Max Acc,ns=623gal
500 (sec.)
0
15 20 25 30
-500
-1000 south
Fig.4 Acceleration Record of Bam Earthquake at Bam (BHRC)
380
100
west Max Vel,ew=1 24kine
50
0
15 20 25 30
-50
(sec.)
-100 east
100
up Max Vel,v=41 kine
50
0
15 20 25 30
-50
(sec.)
-100 down
100 north Max Vel,ns=59kine
50
0
15 20 25 30
-50
(sec.)
-100 south
Fig,.5 Velocity Record of Bam Earthquake, (BHRC)

50 west
Max Dew=34cm
(sec.)
0
15 20 25 30
-50 east
50
up Max Dv=9cm
(sec.)
0
15 20 25 30
down
-50
50 north
Max Dns=24cm (sec.)
0
15 20 25 30
-50 south
Fig,6 Displacement Record of Bam Earthquake (BHRC)
Based upon the strong ground motion of acceleration records, velocity, and displacement waves are
also obtained and shown in Figs.4-6. The maximum values of east-west, vertical, and north and south
directions are 782gal, 968gal, and 623gal for acceleration, 124kine, 41kine, and 59kine for velocity,
and 34cm, 9cm, and 24cm for velocity. It should be noted that EW direction was 124kine in rather
long period of about 1-2 second.
Predominant Direction of Motions

Fig.7 and 8 show the locus of the displacement ground motions. As shown in Fig,6, the wave consists
from two parts. The first one is from 15 to 18sec. The latter is from 18 to 21 sac. The locus clearly
shows the first ground motion sharply directed to westwards that was fault normal direction and lasted
about 2second. The second pulse was followed right after the first one from 15 to 21second. The
direction of motion was more or less fault parallel
Predominant Period of Motions

Fig.9 shows the power spectra of the EW and NS components of the acceleration records. There are
two parts to be discussed. One is short period range of shorter than one second. The other is long
period longer than one second. For short period, there is little difference between the two directions.
381
40 North(cm)
40 North(cm)
30 Locus from 18-21sec
30
20
20
10
10
East
East 0
0 40 30 20 10 0 -10 -20 -30 -40 (cm)
40 30 20 10 0 -10 -20 -30 -40 (cm) 20.0sec
-10 15sec
-10 15sec
-20
-20
-30
-30 Locus from 15-18sec
-40
-40
Fig. Displacement Locus from 15-18sec. Fig.8 Displacement Locus from 18-21sec.
-2
1.0x10
-3
8.0x10
-3
6.0x10
Acc EW direction
-3
4.0x10
Acc NS direction
-3
2.0x10
0.0
0.1 1.0 10.0
Period(sec)
Fig.9 FFT Power Spectra for EW and NS component of the Acceleration records
However, we recognize great difference of the amplitude power spectra. It is found that a large
amplitude is observed in the EW direction compared to NS at the period of about 1.5second.
The damaged adobe structures are reported to have failed to westwards direction, which corresponds
to the displacement characteristics expressed in locus diagram. The fault normal ground motion was
also notified as in Kobe earthquake in 1995. This strong and yet with long period ground motion
might be the killing pulse to the adobe structure. The adobe structures, which have rather short natural
frequency, was considered failed by not structural resonance but being pushed over with horizontal
force.
Critical Values of Acceleration for the Adobe Stability during Bam Earthquake
The direction of failure was predominant in west direction. As already mentioned, this process
corresponds well to the time sequence of displacement shown in Fig.7. The initial movement of about
30cm took place westward and consequently very strong reversal movement to the east followed with
about large displacement of 40cm within one second, when the adobe wall that had been install NS
direction have failed.
382
Comparing to NS direction, walls in EW direction were relatively safe. The accelerations of period of
1.5sec are about 300gals for EW and 200gals for NS direction.
We might conclude the critical acceleration for Adobe stability during Bam earthquake as the
difference between the amplitude of EW and NS components in the acceleration records.
We may conclude tentatively that 200-300gals was the critical acceleration in the Bam earthquake.
Surface Ground Characteristics in the Bam
The geotechnical boring was made at the recording site, however, no shear wave profile was given so
far. However, International Institute of Earthquake Engineering and Seismology carried out seismic
refraction survey in and around the Bam City (IIEES(2004)).
The survey was carried out for ten sites listed in Table-1 where the shear wave in the top surface varies
from 117m/sec to 450m/sec.
Table-1 Refraction survey site and approximate damage density
(m/sec)
600 2000
(m/sec)
1500
400
1000
200
500
0 0
Light Medium Heavy Light Medium Heavy
Damage Intensity Damage Intensity
Fig.10 Damage Intensity vs. Vs, top Fig.11 Damage Intensity vs. Vs, 30m averaged
At site, three or four layered structures are obtained and Table-1 shows the shear wave velocity of the
top surface and the averaged shear wave velocity within top 30m.
Each sites are compared with damage intensity map and approximate intensity of damage at the site
was estimated. Fig. 10 and 11 show the relationship between these velocities and damage intensities.
Fig.10 shows the damage intensity increases with decrease of the velocity. On the other hand, the
averaged velocity within the top 30m shows somewhat reverse relation.
383
(m)
14
12
10
8
6
4
2
0
Light Medium Heavy
Damage Intensity
Fig.12 Depth to Base Layer vs. Damage Intensity
Fig.12 shows the relationship between the intensity of damage and the depth to base layer, which is,
here, assumed a layer with Vs greater than 0.45km/sec. It is shown that the deeper to the base layer
tends to increase the intensity of damage. It is interesting to note that shear velocity in the top surface
is more influence to the damage in Bam city.
Estimated Base Rock motion at Arg-e Bam
If we assume the foundation ground of the strong motion station as 200 to 300m/sec, and the exposed
rock to have Vs=2000m/sec, we could estimate the amplitude on rock as the first order approximation
assuming simple two layered system
Transmitted wave amp=B
Incoming wave Reflected wave

amp=1 amp=A
Fig.13 Incoming, reflected, and transmitted waves

Let the impedance of base layer and surface layer as v1 1 and v2 2, respectively, where v is the
velocity and is the density. The impedance ratio is defined as the ratio a of = v2 2/ v1 1.
If we assume the amplitude of incoming wave as unity(1), then the amplitude of the reflected and the
transmitted waves are A=(1- )/(1+ ) and B=2/(1+ ).
If we put v1=2000m/sec, =2.3g/cm3, v2=300, and =2.0, then =0.13. which results in A=0.77 and
B=1.81. If the amplitude of the ground surface is C, the amplitude of the incoming wave is obtained
as C/B, or 0.4xC. Since the maximum acceleration at the recorded site becomes to decrease almost one
384
half of the amplitude on the rock at Arg-e Bam. The estimated acceleration rock site may be in the
range of 150gals which was smaller than the critical value for the adobe structure at Arg-e Bam.
CONCLUSIONS
The paper showed several intuitive relations in terms of the failure of the adobe structures based upon
the analysis of the strong ground motions recorded in Bam city as follows,
1. The ground motion that caused failure of adobe structure was rather long period of 1.5second.
2. The critical acceleration level as 200-300gals to the damage of the adobe structures in Bam city
with a period of 1.5sec. based upon the comparison of the two horizontal components of NS and EW
direction.
3. The intensity of the damage shows some tendency to increase with the depth of the surface layer
and the lower value of shear wave in the layer rather than Vs,30m of the averaged shear velocity
within the top 30m.
4.At Arg-e Bam, the amplitude in the rock may be the order of about 150gals and the adobe structure
might well be survived.
5. Above conclusions are tentative and should be confirmed by further study of geotechnical
properties at the sites to consider the fundamental measures to protect the Arg-e Bam from earthquake
hazards.
AKNOWLEDGEMENTS
The first author thanks Dr.Nejati who helped me to understand the general conditions of the Arg-e
Bam and various discussions during my stay at the expert camp, the Arg-e Bam. The authors are
grateful for Junko TANIGUCHI, Program Specialist for Cultural UNESCO Tehran Cluster Office,
who supplied us various information on the damage of Bam city.
REFERENCES
ICG Reconnaissance Mission Report(2004), Bam Earthquake of 26 December 2003

IIEES(2004), Preliminary Seismic Microzonation of Bam, Unesco Tehran Cluster Branch
K.A. Hosseini and et al.(2004) Engineering Geology and Geotechnical Aspects of Bam Earthquake
(Preliminary Report),<www.iiees.ac.ir/English/bank/Bam/bam_report_english_geo.html>
385
June 25-28, 2007
Paper No. W2-1002
THE RECONSTRUCTION OF THE TEMPLE OF ZEUS AT NEMEA:

RECENT PROGRESS AND FUTURE PERSPECTIVES
Nicos MAKRIS1, Theodoros PSYCHOGIOS2
ABSTRACT
This paper summarizes the challenges and recent progress on the reconstruction of the Temple of Zeus
at Nemea. Starting with the arguments that support the reconstruction, the paper introduces the criteria
that govern the practice of reconstruction such as structural stability, morphologic continuity and
minimum tolerances at the joints. Examples of technical work associated with the reconstruction of the
crepida, the columns and the frieze of the temple is presented. Finally, the paper concludes with the
future perspectives of this reconstruction.
Keywords: Reconstruction, Classical Temple, Multidrum Columns.
INTRODUCTION
The temple of Zeus at Nemea was built around 330 B.C. and is notable for the unique slender
proportions of its exterior colonnade of Doric style which consists of 32 columns as shown in the plan
view of Figure 1. Among the 32 columns only one has survived (K-31) together with two columns
from the entrance of the Pronaos as shown in the 1977 photograph of Figure 2 (left). The isometric
drawing of Figure 2 (right) depicts the location of the three columns which remain standing from the
ancient times.
For centuries the temple remained in ruins with its building elements partially embedded in the ground
(see Figure 3). This situation is most destructive for the integrity and overall conditions of the ancient
material since large surfaces are exposed to an adverse environment which includes rain, snow, icing,
roots, even microorganisms which have drastically eroded the original material.
Figure 1. Plan view of the temple of Zeus at Nemea. The traces of the three columns standing
from the ancient times are shown with light gray.
1
Professor, Department of Civil Engineering, University of Patras, Greece, Director of the
reconstruction of the Temple of Zeus at Nemea. Email: nmakris@upatras.gr
2
Civil Engineer.
386
K 31
Figure 2. Left: View of the temple of Zeus at Nemea from south-west. Photograph of 1977.
Right: Isometric drawing that shows the location of the three columns standing from the ancient
times.
Figure 3. Aerial photograph of the temple of Zeus at Nemea on July 1977.
PILOT RECONSTRUCTION OF K-25 AND K-26
The preservation of the original building material (see Figure 3) is probably one of the most
convincing arguments for proceeding with the reconstruction of the Temple. The idea of
reconstructing the temple was first put forward in 1978 by Professor Stephen G. Miller from the
University of California Berkeley. Soon after that he initiated a fundraising campaign to finance a
detailed re-synthesis study of the temple. During the period 1980-1982 Professor Frederic A. Cooper
from the University of Minnesota led the effort of measuring and recording approximately 1100
ancient building blocks and subsequently conducted the re-synthesis study for the reconstruction of
the temple. Following the work of Professor Cooper, the Greek Ministry of Culture issued the permit
for the pilot reconstruction of two columns, K-25 and K-26, along the north side of the temple. The
387
reconstruction of the two columns was successfully completed in the summer of 2002 and as a result
of this effort the temple of Zeus today consists of five standing columns as the recent photograph of
Figure 4 (left) shows. The isometric drawing of Figure 4 (right) assists with the understanding of the
morphology of the temple as it stands today. More information of the history of the pilot
reconstruction can be found in the article by Miller (2000).
K 31
K 25
K 26
Figure 4. Left: The temple of Zeus at Nemea from north-west. The two front columns are those
reconstructed by the pilot reconstruction. The three ancient columns are distinguished in the
background. Right: Isometric drawing that shows the present layout with five columns standing.
THE RECONSTRUCTION OF K-27, K-28, K-29 AND K-30
The shape of the temple in the near future

Following the successful reconstruction of K-25 and K-26, the reconstruction team of the temple of
Zeus at Nemea submitted via the American School of Classical Studies a reconstruction study for four
additional columns in order to complete the north-east corner of the temple. The re-synthesis study
conducted by F. A. Cooper was carefully re-evaluated in order to confirm the precise location of the
ancient building components in association with recent discoveries of ancient material. For each
building element (column drums, capitals, architraves and frieze) it was prepared an individual file
which describes its pathology together with the retrofit proposal. Figure 5 shows two among the many
detailed drawings that were submitted to support the reconstruction proposal. The new reconstruction
study of columns K-27, K-28, K-29 and K-30 was approved by the Greek Ministry of Culture in
March 2004 and the new cycle of reconstruction works is already under way. The isometric drawing
of Figure 6 shows schematically the morphology that the temple of Zeus at Nemea will assume once
this second reconstruction phase will be completed. The future morphology of the temple (Figure 6)
will improve appreciably the understanding of the architecture of the temple since its current view
creates some confusion to the visitor (see Figure 4 left). With the reconstruction of the four columns of
the north-east corner (K-27, K-28, K-29 and K-30), there is a closure of the exterior colonnade starting
from the two recently reconstructed columns, K-25 and K-26 to the solely standing column from the
ancient times, K-31. The closure of the north-east corner will also offer to the visitor a sense of the
interior space and magnitude of the monument.
388
Figure 5. Re-synthesis of architectural elements and reconstruction proposal of the north side of
the temple (top) and of the east side (bottom). Drawings No. 18 and No. 22 from Katerina
Skleres architectural studies.
389
K 31
K 30
K 25
K 29 K 26
K 27
K 28
Figure 6. Isometric drawing that shows the future layout of the temple after the end of second
phase of reconstruction which is under way.
The reconstruction of the crepida

Before reconstructing the columns it is necessary to reconstruct the wings of the base of the temple
which consist of four layers of massive rectangular stones which support the gravity loads of the
monument. Figure 7 shows the four layers to be reconstructed at the north-east corner of the temple.
Figure 8 (left) shows the major damage of the stylobate and of the second layer in the vicinity of the
support of column K-31 while Figure 8 (right) shows the temporary support of column K-31 during
the installation of the new filling which is now in place. Figure 9 shows ancient stone blocks from the
base that needed to be pieced together while Figure 10 shows new stone blocks of the crepida which
have been positioned in place and receive final adjustments. In order to complete the section of the
crepida and the area where the four new columns will be erected we need to prepare and install 70 new
stones with dimensions that exceed (2.0 m)x(2.0 m)x(0.4 m) and weight that exceeds 3.5 tons. In
addition to this work 15 ancient blocks need to be pieced together. It should be emphasized that the
reconstruction of the crepida has consumed a substantial fraction of our reconstruction effort mostly
because the rectangular blocks of the crepida were much more attractive to the subsequent occupants
of the nearby area than the cylindrical column drums. The limestone that we use for the preparation of
the new stones originates from nearby area, very close to the ancient quarries. Large masses of stone
=stylobate
=stylobate
=euthenteria =euthenteria
Figure 7. Left: View of the north-east corner of the crepida of the Temple. Right: View of the
same corner where the corner stone of the level A has been retrofitted and replaced.
390
are transported at the periphery of the archeological site and they are sliced with an electrical wire-saw
at the desired dimensions. Figure 11 (left) shows the sawing of a large stone mass while Figure 11
(right) shows new stone blocks that have been prepared for the reconstruction of the stylobate.
Figure 8. Left: The poor condition of stylobate and the 2 nd base layer at the vicinity of the
support of column K-31 which remains standing from the ancient times. Right: Temporary
support of column K-31 during the installation of the new filling which is now in place.
Figure 9. Left: Two pieces of ancient stone of crepida that have been bonded. Right: Two
fragments of an ancient crepida block that have been retrofitted with new stone.
391
Figure 10. Final surface finishing of the new stones of the crepida after they have been
positioned in place.
Figure 11. Left: Wire-sawing of a large formless stone to produce new rectangular stone blocks.
Right: Stack of new stones of stylobate that were extracted via wire-sawing of formless large
stones.
Reconstruction of the columns

Along with the preparation of the crepida, significant progress has been accomplished with the retrofit
of the ancient drums and the production of new in order to reconstruct the columns. The criteria that
govern the state of practice of retrofitting emerge from the need to satisfy structural stability together
with the need to achieve morphologic continuity. Numerical and experimental studies have
demonstrated that the small material loss at the base of the column drums reduce substantially the
dynamic stability of the columns (Psycharis et al. 2000). Further to the issues of the structural
stability, the loss of material and the visible discontinuities between successive drums alter the
392
distinctive elements of classical architecture. These two reasons (structural stability and morphologic
continuity) make imperative the preparation and bonding of a large number of fillings for the column
drums, capitals and epistyles. Figure 12 shows photographs of new stone bonded on the eroded drums
of the columns to be reconstructed. The filings are worked out to the extent that they reach the original
surface of the member (see Figure 13). The edges of the flute are shaped to the extent that the material
can sustain the fine workmanship of the stone carvers. Recent theoretical studies have shown that the
controlled mobility along the interfaces of drums offers to the multidrum column superior seismic
stability than that of the equal size and slenderness monolithic column (Konstantinidis & Makris
2005).
Figure 12. New stone fillings that have been bonded with eroded drums of the columns to be
reconstructed.
Figure 13. Worked out fillings where their final shape reaches the original surface of the
member.
Reconstruction of the epistyles and frieze

The approved reconstruction study includes the reconstruction of the epistyles and the frieze above.
Following a considerable effort we have identified the 10 epistyles (5 exterior and 5 interior) that will
connect the top of the four columns to be reconstructed with the column standing from the ancient
times located to the east; and the two recently reconstructed columns along the north side of the
temple. The placement of the epistyles improves the seismic stability of the solitary columns
(Psycharis et al. 2003). Two among the 10 epistyles have survived nearly intact while the rest have
been broken in two or more pieces as shown in Figure 14, which will be re-bonded and filled. Given
that the epistyles are subjected to combined bending and shear their structural capacity deserves
attention. A comprehensive analysis of the stresses and strains that develop within the retrofitted
epistyles was conducted by Makris and Psychogios (2004). Figure 15 shows schematic sections of a
retrofitted epistyle while Figure 16 shows the distribution of normal stresses of the retrofitted epistyle
under gravity loads.
393
Figure 14. Close view of broken epistyles where the material loss is distinguished.
Figure 15. Ancient stone fragments, new stone fillings and reinforcement with titanium bars of
epistyle A.2 external.
Figure 16. Distribution of normal stresses along axis x of external epistyle A.2.
The shape of the temple today

Figure 17 shows a recent photograph of the north-east view of the temple which shows the partially
reconstructed base and the first drum of column K-30 which has been placed atop the stylobate.
394
Figure 17. Photograph from north-east that shows two recently constructed columns K-30 and
K-29.
The future perspective of the reconstruction

The perspective for the reconstruction of the temple of Zeus at Nemea is the entire reconstruction of
the outer colonnade. This was clearly stated in the master plan of the reconstruction that was
submitted to the Greek Ministry of Culture (Zambas 2004). The re-synthesis of the ancient material
that was conducted by Professor F. Cooper reveals that the large number and the wide distribution of
the existing ancient members suggest for the entire reconstruction of the monument (see Figure 18). In
any event, the degree that the reconstruction will proceed also hinges upon the success of fund raising.
The reconstruction of the north-east corner of the temple which is under way (see Figure 6) is a
segmental reconstruction which paves the way for further reconstruction endeavors until the entire
reconstruction of the outer colonnade (32 columns total). The possibilities for expanding the
reconstruction are illustrated in the isometric drawing of Figure 19. The prioritization is established
based on the needs for preserving the monument in association with the desire for the best possible
end-result.
The first expansion aims to show a portion of the long north side of the temple. The proposed
reconstruction does not reach the north-west corner of the temple since in that area there are
documented foundation settlements of the order of 5 cm. these foundation settlements are associated
with the presence of nearby drain which was excavated in the early Christian period. Nevertheless, the
reconstruction of five additional columns to the west will reveal the majestic character of the temple
that the visitor will enjoy while approaching the monument from the south-west.
The perspective for reconstruction does not stop with Figure 19. It should be emphasized that during
the excavations of the last few years they have been discovered along the west side of the temple
another 11 drums, one epistyle and two stones from the frieze which are not included in the re-
synthesis study by Cooper shown in Figure 18. These findings increase further the challenge for the
entire reconstruction of the temple which we are called to take - each one of us in his capacity.
395
Figure 18. Re-synthesis of the four sides of the exterior colonnade from the scattered architectural elements of the temple which were
identified and recorded by Frederic A. Cooper and his collaborators.
396
K 20
K 21
K 31 K 22
K 23
K 30 K 24
K 25
K 29 K 26
K 27
K 28
Figure 19. A proposition for the next expansion of the reconstruction.
ACKNOWLEDGEMENTS
This reconstruction is happening due to the invaluable contribution of several individuals and
institutions, including the Greek Ministry of Culture, the Ephorate of Corinthia, the American
School of Classical Studies and the Department of Classics of the University of California at Berkeley.
Our reconstruction team consists of 9 stone workers, residents of the nearby area who operate under
the supervision of Architect Mrs. Katerina Sklere who also leads the architectural studies.
The funding to the reconstruction originates primarily from Greek donors including Mr. T.
Papalexopoulos, the National Bank of Greece, the Niarchos Foundation, the Club Hotel Casino
Loutraki, Mr. J. Moschaclaidis and the Samourkas Foundation among others. Part of the funding is
administered by the Opheltis foundation and part is administered by Earthquake Engineering Research
Center of the University of California at Berkeley.
The senor author appreciates the transfer of knowledge and invaluable technical and intellectual
contributions that where offered by Professor Stephen G. Miller former director of Nemea excavations
and reconstruction.
REFERENCES
Miller SG. The temple of Nemean Zeus. A California Landmark, Chronicle of the University of
California, No. 4, Fall 2000 issue, 2000.
Konstantinidis D and Makris N. Seismic response analysis of multidrum classical columns,
Earthquake Engineering and Structural Dynamics, 34:1243-1270, 2005.
Psycharis JN, Papastamatiou DY. and Alexandris AP. Parametric investigation of the stability of
classical columns under harmonic earthquake excitations, Earthquake Engineering & Structural
Dynamics, 29:1093-1109, 2000.
Psycharis JN, Lemos JV, Papastamatiou DY, Zambas C and Papantonopoulos C. Numerical study of
the seismic behavior of a part of the Parthenon Pronaos, Earthquake Engineering & Structural
Dynamics, 32:2063-2084, 2003.
Makris N and Psychogios T. Static and dynamic analysis of columns and part of the entablature of
the NE corner of the temple of Zeus at Nemea, American School of Classical Studies, Athens,
2004.
Zambas K. The reconstruction of temple of Zeus at Nemea: The principles, the objectives and the
perspectives of the intervention, American School of Classical Studies, Athens, 2004.
397
June 25-28, 2007
Paper No:W2-1003
SEISMIC PERFORMANCE OF ROCK BLOCK STRUCTURES WITH

OBSERVATIONS FROM THE OCTOBER 2006 HAWAII
EARTHQUAKE
Edmund MEDLEY 1 , Dimitrios ZEKKOS 2
ABSTRACT
Unreinforced masonry construction using blocks of rock is one of the oldest forms of building, in
which blocks are stacked, sometimes being mortared with various cements. Ancient civilizations used
locally available rocks and cements to construct rock block columns, walls and edifices for residences,
temples, fortifications and infrastructure. Monuments still exist as testaments to the high quality
construction by historic cultures, despite the seismic and other potentially damaging geomechanical
disturbances that threaten them. Conceptual failure modes under seismic conditions of rock block
structures, observed in the field or the laboratory, are presented. A brief review is presented of the
damage suffered by the culturally vital Hawaiian Puukohol and Mailekini Heiaus, rock block
temples damaged by the Mw 6.7 and 6.0 earthquakes that shook the island of Hawaii on October 15,
2006.
Keywords: rock block structures, Hawaii earthquakes, earthquake observations, geomechanical

failures
INTRODUCTION
Construction of unreinforced masonry is common in various earthquake-prone regions, particularly in

developing countries, and rural areas of developed countries. This vulnerable type of construction is
susceptible to often devastating damage, as evident from the effects of the 2001 Bhuj, India
earthquake (Murty et al. 2002), where 1,200,000 masonry buildings built primarily based on local
traditional construction practices, either collapsed or were severely damaged. Buildings constructed
with adobe and unreinforced masonry suffered devastating damage in the Bam, Iran 2003 earthquake
(Nadim et al. 2004). However, in the 2002 Molise, Italy earthquake post-1850 unreinforced masonry
buildings performed worse than medieval and renaissance-age masonry buildings (Decanini et al.
2004) indicating that certain methods and materials of construction used in culturally valuable
archaeological and monumental structures may have properties that have resisted significant
earthquakes. This paper reviews some aspects of the geomechanical performance of structures
assembled as unreinforced masonry using blocks of rock.
SOME COMMON CHARACTERISTICS OF ROCK BLOCK STRUCTURES
A simple grouping of rock block structures can be developed based on the cross-sectional aspect ratio
of structures (such as width and thickness to height). Rock block columns are relatively slender with
height substantially greater than their width and thickness and include towers, pillars, and obelisks.
Rock block walls such as fences, partitions, ramparts, bulwarks, and retaining walls have width of the
order of their height, but smaller than the thickness. Rock block edifices have width and thickness
similar to, or often substantially greater, than their height and include platforms, ramps, terraced
1
Senior Consultant, Geosyntec Consultants, Oakland, USA, Email: emedley@geosyntec.com
2
Professional Engineer, Geosyntec Consultants, Oakland, USA, Email: zekkos@geoengineer.org
398
embankments, dams, mounds and pyramids. Edifices are often of imposing appearance or size and, if
of unique character, may be considered monuments.
The exterior envelopes of rock block edifices or the facing of retaining walls may be composed of
artfully fabricated veneers of high-quality rock blocks. The inner cores or backfill may consist of
stacked courses of blocks, earth or loose rubble. The contacts between facing blocks and core
materials range from dry stacked (no inter-block cement) to partially or fully cemented, although the
culturally significant monuments considered in this paper are mortar-free. Facings and interior cores of
rock block edifices may be penetrated by openings such as windows, doors, corridors, chambers and
stairways, the latter being a feature of monuments such as the pyramids.
SOME SIGNIFICANT CULTURAL ROCK BLOCK STRUCTURES IN SEISMIC REGIONS
A few examples of well-known rock block structures constructed by various civilizations in

earthquake- prone areas, which are still in good to excellent condition, are presented in the following
sections.
Mycenaean civilization
The ancient Greek Mycenaean civilization, best known for the Trojan War, flourished in Greece
between the 17th and 12th Century BC. The civilization abruptly collapsed: military invasion,
economic collapse, social disturbance, diseases and earthquakes have all been suggested as causes.
The Mycenaean-type of construction is mostly known for the fortification size of its cities such as
Mycenae and Tiryns. Zekkos et al. (2005) summarized the currently preserved Mycenaean rock block
infrastructure which includes fortifications, underground shafts (graves), retaining walls,
roads/pavements, bridges, hydraulic works, dams/embankments, and tunnels. For example, the
fortifications in Tiryns consist of building blocks that are typically 2-3 meters long and more than a
meter in thickness and height. The blocks of the walls in all cases are made of locally available
material, e.g. limestone rocks in Tiryns, and mostly conglomerate in Mycenae. In Mycenae, three
different building techniques are evident in the extant fortifications: a) Cyclopean masonry (Figure
1), which is the oldest technique, generally consisting of slightly hewn, large blocks. The stone
dimensions are larger on the two wall faces (interior and exterior), with cores composed of smaller
stones and soil. b) Rectangular masonry that consists of carefully hewn rectangular blocks laid in
regular courses (Figure 2). This technique was first used probably in about 1250 BC. In some areas,
the rectangular technique was used to construct only the outer side of walls while the interior side was
constructed with Cyclopean-type masonry. c) Polygonal masonry is generally composed of polygonal
shaped blocks carefully fitted together (Figure 1).
Figure 1. Polygonal masonry construction (in the Figure 2. Rectangular masonry adjacent to
center) used to repair section of Cyclopean the Lions Gate at Mycenae.
masonry (left and right) at Mycenae.
399
Maya and Inca Civilizations
The classic period of the Maya civilization spanned from the 3rd Century AD to the 10th Century
AD. The reasons for the collapse of the Mayan Civilization are not clear and theories include military
invasions, social revolutions and natural disasters. The preferred fall directions of stelae indicate that a
Magnitude 7.5-7.9 earthquake may have contributed to the collapse of the empire (Kovach, 2007). The
Mayan civilization is known for its impressive temples and pyramids. Locally available limestone
quarried in the vicinity of the building sites was used in the construction of buildings ranging from
pyramids to simple residences. Mortar consisting of pulverized and burnt limestone was sometimes
used to augment poor frictional interlocking between the blocks.
The Inca civilization arose in the early 13th Century AD and had largely collapsed in the 16th Century
AD following the execution of the last Inca Emperor by the Spanish Conquistadors. The Inca
civilization expanded into Peru. The seismotectonics of Peru are dominated by the convergence of the
Nazca plate and South American plates (Dewey and Silva, 2003). An example of the Inca art of
construction is spectacularly evident at Machu Picchu, where the majority of the surviving rock block
structures are retaining walls, fortifications and residential construction. Wright and Zegarra (2000)
identified at Machu Pichu at least 18 different stone wall types and patterns constructed with different
levels of care. The blocks in Machu Picchu are primarily of quarried granite. Mortar, consisting of
clay, soil and small stones was used in many of the walls.
Greek and Roman Civilizations

Many examples of rock block construction are preserved from the classic Greek and the Roman
civilizations (6th Century BC to 3rd Century AD). The Hellenistic empire established by Alexander the
Great was conquered by the Romans and extended from Europe to Asia and Africa. Greek and Roman
infrastructure is still preserved in several earthquake-prone areas. A famous Roman rock block
structure is Hadrians Wall, which for 117 km marked the northern border of the Roman Empire in
Great Britain (not an earthquake-prone region). The size and type of the materials used in the Wall
varied depending on the local geology.
Hawaiian Culture
The islands of Hawaii were populated by seafaring Polynesians who likely emigrated from Tahiti and
the other Society Islands in about 1200 AD. Through a heritage developed by centuries of Pacific
explorations, the Hawaii pioneers had an intimate understanding of how to survive on previously
uninhabited Pacific Islands. Rock (p haku) was vital in traditional Hawaiian culture and was so
important that Hawaiians believed gods could appear as stone (Wong, 2004). Rock was used for tools
and weapons; and the early Hawaiians had well-developed skills to construct robust rock block
structures. Dry-stack stone masonry (uhau humu p haku) is still the traditional Hawaiian craft of
stacking stones to create interlocked rock structures without the benefit of mortar, although it is
nowadays practiced by very few craftsmen (Wong, 2004). The appearance, functionality and stability
of Hawaiian rock block construction depended on available materials, the geometry of the structure,
the skill of the builder and, the purpose of the structure. With the arrival of Europeans in the late
1700s, mortared rock blocks were increasingly used for construction of churches, commercial
buildings and homes.
Traditional Hawaiian rock block structures range from serviceable quality, quickly-erected rock fences
and low walls constructed of irregular, relatively light-weight aa clinker (Figure 3) to imposing, high-
quality and durable structures crafted from intricately conjoined blocks of massive polygonal basalt
(Figure 4).
Large Hawaiian walls typically have large foundation blocks (niho) placed at the bottom, or at corners,
with smaller blocks placed to form a battered face and tapered cross section (Figure 5 and Figure 6).
Wider walls have separated facades with a backfilled core composed of smaller stones (h k h ka)
placed in layers or as loose rubble (Greene, 1993; Wong, 2004) .
400
Figure 3. Rock wall of loosely placed aa clinker Figure 4. Rock wall of intricately interlocked
blocks. massive basalt blocks (Mokuaikaua Church,
Kailua-Kona, Hawaii).
Figure 5. Types of construction for rock block walls. Figure 6. Cross-section through
(Greene, 1993; after Kikuchi, 1973). wall at Puukohola Heiau (Green,
1993; after Kikuchi and Cluff,
1969.
Amongst the largest and most imposing rock block edifices in Hawaii are heiaus, temples composed
of rock block platforms supporting structures used to house priests, sacred items, ritual images, and
altars. Heiaus were sacred places; the Hawaiian caste system of kapu (taboo) allowed access only to
priests and sacred high chiefs: kapu demanded death to common entrants.
The Puukohol Heiau, located south of Kawaihae on the island of Hawaii (Figure 7) is one of the few
surviving relatively intact examples of a finely-constructed rock block edifice and is an important
cultural monument in Hawaii. It is still used for Hawaiian ceremonies and only native Hawaiians may
enter or walk on it. Puukohol is a ritual temple commissioned by the Hawaiian chieftain
Kamehameha between 1790 and 1791 in a tribute to the war God Kukailimoku as a religious adjunct
to Kamehamehas military and diplomatic efforts to conquer all the islands of Hawaii. The heiau was
built on the brow of the hill above the existing Mailekini Heiau and Kamehamehas compound (Figure
8 and Figure 9).
Puukohol Heiau is approximately 225 feet by 100 feet (70m by 30m) in plan dimensions.
Topographically below the Puukohol Heiau is the older Mailekini Heiau (Figure 8). The long axes
of the heiaus are oriented approximately north-south. The walls, terraces and platforms of the heiaus
range from less than 5 feet (1.5 m) to 30 feet (10 m) high (Figure 9). Block sizes range between a few
centimeters to more than one meter with most boulders being in the 30 to 60 cm size range.
401
Figure 7. Mailekini Heiau in middle ground, Figure 8. Aerial view of Puukohol Heiau (top)
and Puukohol Heiau on top of hill. and Mailekini Heiau (bottom); yellow line
shows approximate location of cross section in
Figure 9 (after Greene, 1993).
The Puukohol edifice was constructed

using local rock for core backfill (Greene,
1993) (Figure 6). Water-rounded cobbles
and boulders being prophetically
specified (Yanoviak, 2003), were stacked
outer wall
platform to construct the imposing facades. Although
terrace
ground rough rock is abundant locally, water-
rounded rocks was transported by 20,000
ground men passing blocks hand-to-hand from the
graphic scale (ft)
nearest major source of Pololu Valley
0 50
Figure 9. Cross-section through Puukohola Heiau (Yanoviak, 2003), some 20 miles (30 km)
(after Greene, 1993; Kikuchi and Cluff, 1969 away. Over the years, portions of the heiaus
collapsed prompting stabilizations and
repairs performed in the 1970s (Greene, 1993). The Heiaus and adjacent cultural assets are
administered as a National Historic Site by the United States Federal National Park Service
(www.nps.gov/puhe).
Upon abolition of the kapu system in 1819 by Liholiho, (King Kamehameha II) heiaus were
abandoned as places of worship; many were subsequently exploited to provide building materials for
new structures. Others were destroyed to make way for new developments and commercial
agriculture.
SOME FAILURE MODES OF ROCK BLOCK STRUCTURES
Figure 10 show several modes of failure, which have been observed in the laboratory, on sites of
culturally significant rock block structures or can be expected for rock block structures subjected to
seismic loading. These modes are:
A) Block-on-block sliding of columns: whereas stacked blocks of rocks, that could be for
example part of a column of a temple, move relative to each other as a result of seismic
excitation and exhibit permanent relative seismic displacement.
B) Bending of a (generally mortared) wall or column during seismic loading as a result of the
structural response of the column or the structure that the column supports.
C) Sliding and dislodgement of blocks; that could occur during earthquake loading of the
structure that has been exposed to centuries of weathering and erosion.
402
Figure 10. (A-I) Failure modes of rock block structures.
D) Dislocation of walls due to differential settlement of foundation soils, which could take place
under static or seismic conditions. An example of this mode for an Inca wall in Machu Pichu
is presented by Wright and Zegarra (2000).
E) Loss of strength of loose soil or rubble cores or backfill, and imposition of additional loading
on walls; and (F) separation of block walls due to differing response of fill and exterior walls;
Such modes of failure have been observed in shake table tests performed by Meyer et al.
(2007) on brick and stone walls.
G) Loss of arch key block and deformation of arch legs. Such failure modes have been observed
on Mycenaean underground burial chambers known as treasuries.
H) Increase in earth pressures of core or backfill, possibly by the elevation of water.
I) Wall deformation due to global instability such as failure of the core or backfill; or a failure at
the toe due to inadequate bearing capacity of the underlying soils.
403
J) Raveling, bulging and slumping failures of the outer faces of a rock block edifice, such as
those observed at the Hawaiian heiaus described subsequently. Face failure may be
progressive: the inclination of the bulges increase until they become over-steepened and
collapse, thereby inducing collapse of the face above.
.
Figure 10J. Raveling, bulging and slumping of the face of a rock block edifice.
There are benefits to considering the apparent similarities between rock block edifices and natural
arrangements of rock blocks in rock masses. In-situ blocks in rock masses are bounded by joints,
shears, fractures and other discontinuities which range between open apertures or contain infillings
ranging through soil-like to strongly mineralized. Loose rock debris is analogous to talus fans or
colluvium.
Considered in geomechanical terms, it can be reasonably expected that rock block edifices may behave
under static and dynamic loadings in similar fashion to natural masses of rock and coarse soil. For
structures with tightly-packed or layered block arrangements, roughly circular failures occur where
the failure surfaces negotiate around the boundaries of blocks, with a degree of failure surface
roughness dependent on the size of the blocks. For edifices with internal cores composed of loose
rubble or soil face slumping or bulging is anticipated. Very steep and high wall faces composed of
stacked rocks arranged in columns may topple, much as steeply-jointed slabs of naturally jointed rock
topple. Additional analogies between block rock structures and natural rock masses can be conceived.
In some cases, it may be advantageous to perform preliminary geomechanical analysis of rock block
edifices using conventional rock mass characterization schemes, which require field estimates of water
conditions; intact rock strength: joint orientations, friction angle, spacing, persistence, roughness and
surface conditions. These qualities are estimated by an investigating geological engineer or
engineering geologist, often without the benefit of much exploration data from penetrative borings or
trenches. The need to work with limited observations is also likely a familiar situation for an
investigating archaeologist charged with characterizing a rock block structure without the benefit of
intrusive explorations.
404
PERFORMANCE OF ROCK BLOCK STRUCTURES DURING THE OCTOBER 15 2006
HAWAII EARTHQUAKES
The October 2006 earthquakes

The island of Hawaii has suffered several major earthquakes in the last 150 years; some of which have
killed hundreds of people, generally because of tsunamis. Another major event shook the island at
7:07:48 a.m. (local time) on Sunday October 15, 2006, when a Mw 6.7 earthquake struck off-shore of
the west coast of the island (Figure 11) about 11 km (7 miles) north-northwest of Kalaoa, in Kiholo
Bay, Hawaii, some 99 km from Hilo (the second-largest city in the State) and 250 km from Honolulu
the capital city, on Oahu) A Mw 6.0 event at 07:14 a.m. and depth of 19 km is considered to have been
a separate earthquake. The strongest shaking (MMI VII-VIII) and damage resulting from the two main
earthquakes was concentrated on the northern portion of the Island. The felt effects of the earthquake
extended to Honolulu, where the International Airport was shut down for several hours.
The earthquakes caused significant damage on the island of Hawaii in the form of failed road cut
slopes, landslides, bridges, embankments; rock block retaining walls. Liquefaction and lateral
spreading severely damaged the Port of Kawaihae (Medley, 2006; Robertson et al, 2006; and Chock,
2006). Kalahikiola Congregational Church, located at the northern end of the island, suffered almost
complete collapse (Figure 12). The church was constructed in 1855 of weakly cemented rock blocks.
Although the earthquake damaged many rock block structures, the overall impression was that most
damage was suffered by tall and/or steep structures close to the earthquake epicenters, and where the
contacts between rocks were minimal (Medley, 2006). The cumulative damage from the earthquakes
exceeded $146 million (Chock, 2006).
Figure 11. USGS ShakeMap for the M6.7 Kiholo Bay Figure 12. Kal hikiola Church, Hawaii.
earthquake (after USGS). Weakly cemented rock block walls
failed dramatically (Medley, 2006).
405
Fi gure 13. Mailekini Heiau, pre- earthquake (photo www.pacificworlds.com/kawaihae).
Figure 14. Raveling and slumping of eastern wall of Mailekini Heiau, after earthquake.
406
Figure 15. Pre-earthquake view of Figure 16. Post-earthquake view of collapsed
entrance passageway into Puukohol entrance.
(Photo www.pacificworlds.com/kawaohae).
Figure 17. Pre-earthquake condition of Figure 18. Post-earthquake photo near NE

Puukohol Outer Wall, from near NE corner corner of Puukohol shows bulging of face
(Photo: www.pacificworlds.com/kawaihae). of wall (indicated by yellow line).
Performance of Puukohol and Mailekini Heiaus Rock Block Edifices During the Earthquakes
The Puukohol and Mailekini Heiaus were both severely damaged by the October 15 2006
earthquakes, the epicenters of which were less than 50 km away (Figure 11). At the Mailekini heiau,
much of the interior face of the east wall slumped and raveled (Figure 13 and Figure 14). The worst
damage at the Puukohol Heiau was to high retaining walls bounding the sole access passage (Figure
15 and Figure 16). Many wall bulges were observed at originally steep wall faces (40 to 50 degrees)
resulting in further steepening of the wall faces near the bottom of the edifice (Figure 17 and Figure
18), although it is likely that some bulging had occurred before the earthquakes. The bulging suggests
that there are future vulnerabilities to increased rock face instability and possible eventual collapses
from future earthquakes or creep deformation (Medley, 2006). Some slumps and ravels of the wall
faces occurred where the bulges over-steepened and collapsed (Figure 19) apparently in much the
same fashion as sketched in Figure 10J. The angle of repose for the slumps was about 30 degrees, with
run-outs extending to more than 5 m from the original toe of the walls.
407
The location of the Heiau on the brow of a hill may have focused seismic shaking, although no details
about site response are available. The pre-earthquake geometry of the Puukohola Heiau was
apparently surveyed a few years ago with terrestial LiDAR (Light Detection and Ranging) before the
earthquake (personal communication with National Park Service staff). As shown in Figure 17, there
was apparent wall bulging or steepening of the Heiau walls before the earthquake, but it is unknown
how much of this occurred because of, or since, wall renovations in the 1970s A repeat LiDAR
survey could form the basis for quantifying earthquake-related deformation, back-analysis and
identification of those portions of the edifices vulnerable to further damage in future earthquakes, or
due to creep (Medley, 2006).
Figure 19. Puukohol Heiau. Slump failure due to earthquakes (between dashed lines).
CONCLUSIONS
Construction with unmortared block of rocks is venerable and universal even in seismically active
regions. That so many historic and culturally valuable structures have survived is a testament to
careful engineering craftsmanship. Finely fitted, massive blocks of rock and well-constructed interior
cores and backfill assure the survival of these structures better than poorly constructed unreinforced
masonry of modern communities nearby. Modes of failure of rock block structures under seismic
loading observed on sites or the laboratory are conceptually presented.
The Puukohol and Mailekini Heiaus suffered great damage in the Mw 6.7 and 6.0 earthquakes that
shook the island of Hawaii on October 15, 2006. The causes of the damage have not been adequately
studied, but further analyses should ideally research: ground and structure responses; understanding of
the effects of topographic focusing; consideration of the quality of the interior core; review of the
nature and quality of previous repairs in the last 200 years; and estimation of the occurrence of prior
permanent wall deformations. Since portions of the Heiaus now have potentially hazardous wall
geometries, these should be analyzed to assess future potential performance under static and dynamic
loading.
ACKNOWLEDGMENTS
The authors wish to acknowledge the Geo-Engineering Earthquake Reconnaissance Association

(GEER) and Geosyntec Consultants for supporting the Geological Engineering Reconnaissance after
the Hawaii earthquakes in October 2006 (http://gees.usc.edu/GEER/Hawaii/title.htm). Additional
information and photographs on the geoengineering aspects of historical monuments are available
from the Geoengineer website (http://www.geoengineer.org).
408
REFERENCES
Chock, G, (ed). Preliminary Observations on the Hawaii Earthquakes of October 15, 2006. EERI
Special Earthquake Report, Learning from Earthquakes; EERI, Oakland, CA:
http://www.eeri.org/lfe/pdf/usa_kona_hawaii_eeri_prelim_report.pdf
Decanini L, De Sortis A, Goretti A, Langenbach R, Mollaioli F, and Rasulo A, Performance of
Masonry Buildings During the 2002 Molise, Italy Earthquake. 2002 Molise, Italy, Earthquake
Reconnaissance Report, Vol. 20, Earthquake Spectra, EERI, 2004.
Dewey, J W and Silva, W J, Seismicity and Tectonics, 2001 Southern Peru Earthquake
Reconnaissance Report, Supplement A to Volume 19, January 2003, Earthquake Spectra,
Earthquake Spectra, pp.1-10.
Greene, LW. Cultural History of Three Traditional Hawaiian Sites on the West Coast of Hawai'i
Island, United States Department of the Interior, National Park Service, Denver Service Center,
September 1993: (http://www.cr.nps.gov/history/online_books/kona/history.htm)
Kikuchi, W K. Hawaiian Aquacultural Systems, Ph. D. thesis, University of Arizona, 1973.
Kikuchi, W K and Cluff, D F, An Archaeological Survey of Puu Kohola Heiau and Mailekini Heiau,
Kawaihae, South Kohala, Hawaii Island in Deborah F. Cluff, William K. Kikuchi, Russell A.
Apple, and Yosihiko H. Sinoto, The Archaeological Surface Survey of Puu Kohola Heiau and
Mailekini Heiau, South Kohala, Kawaihae, Hawaii Island, Hawaii State Archaeological Journal 69-
3 (Honolulu: Division of State Parks, Dep. of Land and Natural Resources, 1969), pp. 37-53.
Kovach R L. Personal research website on Ancient Earthquakes of the Americas
http://pangea.stanford.edu/~kov/kovach-home-page.html, Last visited: April 10, 2007)
MacDonald, G A, and Abbott, A T. Volcanoes in the Sea The Geology of Hawaii, University of
Hawaii Press, Honolulu, HI; pp. 441, 1970.
Medley, E W, Geological Engineering Reconnaissance of Damage Resulting from the October 15,
2006 Earthquakes, Island of Hawaii, Hawaii, USA, Report for Geo-Engineering Earthquake
Reconnaissance Association (GEER), Geosyntec Consultants, Oakland, California
(http://gees.usc.edu/GEER/Hawaii/title.htm), 2006.
Meyer, P, Ochsendorf, J, Germaine, J, Kausel, E, The impact of high-frequency low-energy seismic
waves on unreinforced masonry, Earthquake Spectra, Volume 23, No. 1, pp. 77-94, February 2007.
Murty, CVR, Arlekar, JN, Rai, JN, Udasi, HB, and Nayak D. Chapter 11: Masonry structures. 2001
Bhuj, India earthquake Reconaissance Report, EERI, pp. 187-224, July 2002.
Ladd, E J, Ruins Stabilization and Restoration Record, Pu'ukohola Heiau National Historic Site,
Kawaihae, Hawaii. Honolulu: National Park Service, 1986.
Nadim F, Moghtaderi-Zadeh M, Lindholm C, Andresen A, Remseth S, Bolourchi M J, Mokhtari M
and Tvedt E, The Bam Earthquake of 26 December 2003. Bulletin of Earthquake Engineering, 2
(2), Springer, Netherlands, pp. 119-153, 2004.
Robertson, I N, Nicholson, P G and Brandes, H G. Reconnaissance following the October 15th, 2006
Earthquakes on the Island of Hawaii. Research Report UHM/CEE/06-07, University of Hawaii
College of Engineering, Department of Civil and Environmental Engineering; pp. 65, 2006.
Stokes, J F G, "Report 13: Archaeological Features of the Pu'uhonua Area," in Bryan and Emory,
Natural and Cultural History of Honaunau, pp. 163-74; (1915??)
Wong, S K, Cultural Foundations, April 2004 (http://www.oha.org/pdf/kwo04/0404/13.pdf)
Wright K, Zegarra AV, Machu Picchu A Civil Engineering Marvel. ASCE Press, 2000.
Yanoviak, A C, Puukohol Heiau in Cheever, D and Cheever, S., co-ordianting authors; and
Clifford, . J.S. and Haines, F. S., editors, P haku The Art and Architecture of Stonework in
Hawaii, Editions Limited, Honolulu, Hawaii, pp 171, 2003.
Zekkos D, Manousakis J, and Athanasopoulos A, Geotechnical engineering practice in the
MycenaeanCivilization (1600-1100 BC). 2nd International Conference Ancient Greek
Technology, Athens, Greece, 17-21 October 2005 (Also available at: http://www.geoengineer.org).
409
June 25-28, 2007
Paper No. W2-1005
REGIONAL SUBSIDENCE AND EARTHQUAKES AS THREATS TO

ARCHITECTURAL MONUMENTS IN MEXICO CITY
Efran OVANDO-SHELLEY1, Mrcia PINTO DE OLIVEIRA2, Enrique SANTOYO3
ABSTRACT
Much of Mexico City is located over extremely soft lacustrine clays that have been undergoing a
consolidation process due to the exploitation of the aquifers that underlie these soils. Consequently,
the city has been sinking and will continue to do so in the foreseeable future; resulting settlement
differentials have damaged and will continue to damage most of the citys architectural heritage. The
city is also located within a high seismicity zone. Earthquakes have also damaged the citys
architectural treasures and will continue to be threatened by them in the future. In this paper we look at
the way in which these to hazards have affected monuments in the past and at the ways in which they
will combine in the future to pose further threats. We take into account the way in which pore
pressures within the clays will change in the future and the way in which the resulting effective stress
increments will affect soil properties and the seismic response of the soft lacustrine clay deposits.
Keywords: Historic Monuments, Regional Subsidence, Earthquakes.
INTRODUCTION
Modern large cities are frequently exposed to natural and man made risks which affect many aspects
of urban development. Political, social, economic, geographical, morphological and geological
conditions, among others, determine the way in which hazards can be faced in practice and, generally,
authorities will seek means for eliminating or at least mitigating these risks or their effects. More often
than not these conditions combine in a variety of imaginable ways to preclude or limit the
effectiveness of public policies for hazard management.
Adopting the view point of Geotechnical Engineering, we look at the way in which two of these risks
have, in the past, compromised the integrity of architectonic monuments in Mexico City. We consider
earthquakes and regional subsidence and also the way in which they may combine in the coming years
to endanger seriously the permanence of the architectural heritage of the central part of the city.
BACKGROUND
The Aztec city of Mexico-Tenochtitlan was founded in the year 1325 on a marshy islet in the middle
of Lake Texcoco. This lake, the lowest portion of a closed basin surrounded by volcanic mountain
ranges, overlies a series of clays and clayey silts formed by the chemical degradation of pyroclastic
materials driven into it by water flows and by the winds. These clays are notorious for their extremely
high water contents, their very low shear strength and their extraordinarily large compressibility.
1
Research Professor, Instituto de Ingeniera, Universidad Nacional Autnoma de Mxico, Mxico,
Email: EOvandoS@iingen.unam.mx
2
M. Sc. Student, Instituto de Ingeniera, Universidad Nacional Autnoma de Mxico, Mxico, Email:
opmarcia@gmail.com
3
Technical Manager, TGC Geotecnia S. A. de C. V., Mxico, Email: santoyo@tgc.com.mx
410
Aztec constructions concentrated in and around the religious precinct that constituted the core of the
old city. Temples atop extremely heavy earthen pyramidal structures clad with stone and colored
stucco dominated the skyline. Built directly on the very soft lacustrine clays, these pyramids
underwent extremely large settlements and were repaired periodically or, as it was the custom in Mid-
America, substituted by newer pyramids built directly over the older ones.
The Spaniards conquered and razed Mexico-Tenochtitlan in 1521 and a new city was built literally on
top of the former one. Spanish masons adapted and used foundation techniques devised by the Aztecs,
like short stakes with flat stone heads; they also imported European constructive techniques and tried
to adapt them to the difficult subsoil conditions, not always successfully. Nearly all of the Spanish
buildings erected in the 16th century underwent extremely large settlements and have almost
completely disappeared, in many cases swallowed by the ground. In order to limit and mitigate the
problems caused by differential settlements induced by the self weight of structures and in lieu of
other technological resources, the height of most structures was generally limited to two stories, to
avoid their ruin if built taller, as referred by Cervantes de Salazar in 1550 (Cervantes de Salazar, 1978)
who also put forth this limit to avoid earthquake related damage (this is probably the first rule for
building earthquake resistant structures in the Americas). Foreign voyagers that visited the city in
1596, wrote that everything sank or tilted, as quoted by Musset (Musset, 1992).
In the erection of manors, palaces or civic buildings or the larger religious buildings like convents and
churches, engineers used ingenuity, patience and lots of money. Innovative foundation solutions were
often tried and tested, as in the Metropolitan Cathedral, where a masonry platform, actually a
foundation raft of large proportions, approximately 60 by 120 m, was tried for the first time in the city
in 1573. These heavier structures were subjected to recurrent interventions to correct structural
damage caused by differential settlements induced by the consolidation of the underlying clays
produced by overburdens imposed on the soil by the structures themselves. On many occasions
foundations were altered and underpinning of the larger conventual churches and cloisters was not
rare. Overall, the approach for building foundations and foundation systems was empirical and it was
late in the 19th century when attempts to rationalize these techniques were first introduced.
Upon the destruction of the prehispanic system of levees and embankments with which the water in
the lake system was regulated, the city flooded periodically. The most severe flood took place in the
early 17th century, when the city was covered by water for five years (1629-1634) and as a result,
damage in the older structures became even more severe. After it the resolve of the authorities to
eliminate the threat by drying the lake was strengthened. Floods were mitigated with a trench
excavated in the mountains of the northern portion of the basin to divert one of the rivers that reached
the lake. After the completion of the monumental trench the 18th century, no further drainage works
were undertaken until the late 19th and the early 20th centuries. The graph in Fig. 1 illustrates the
evolution of the lake system and, as seen there, drainage and land reclamation for urban expansion
produced the gradual shrinking of the lake, which is now practically nonexistent.
411
Figure 1. Evolution of the lake system.
GEOTECHNICAL CONDITIONS IN CENTRAL MEXICO CITY
Much of the soil under Mexico City consists of low strength, highly compressible lacustrine clays. In
most of the area formerly occupied by the lakes, main soil formations are ordered in a sequence of soft
clay strata interspersed with lenses and layers of harder clayey silts with sands. As seen in the
stratigraphical cross-section of Fig. 2, the main soil strata are: a) archaeological debris and fills, the
uppermost materials; b) a crust of dried low plasticity silty clays; c) the Upper Clay Formation having
the most compressible soils, as seen in Fig. 2; d) the first Hard Layer that appears at an average depth
of about 40 m, formed by sands, gravely sands and thin lenses of softer silty clays; d) the Second Clay
Formation, about 10 m thick, as shown in Fig. 2; e) the so called Deep Deposits very consistent silts
and sandy silts interspersed with hard clays appear at the base of the stratigraphical column (50 m
deep or so).
412
Figure 2. Stratigraphical cross-section.
REGIONAL SUBSIDENCE
Regional subsidence is not an exclusive problem of Mexico City. It affects other cities where water is
extracted from the subsoil. Two of the most alarming cases are Venice and Bangkok where the safe
guard of architectonic monuments has been of paramount importance in devising solutions to the
problem. In the Mexican Republic other cities are now suffering from the effects of regional
subsidence but the capital surpasses in size and complexity any other case.
It can be inferred that subsiding ground probably affected Mexico City even before water began to be
extracted from its subsoil on account of the gradual desiccation of the old lake system, beginning early
in the 17th century. Unfortunately there are no topographical data or records of other observations
before the 1880s to back this assertion. The results of levelings in the city performed in 1860, 1870
and 1888 led to the recognition, early in the 20th century, that the city was actually sinking but
engineers and scientists of the time were unable to relate regional subsidence to extraction of water
from the subsoil.
Initially artesian water was drawn from relatively shallow wells; water was later pumped from deeper
wells as demand increased in the second half of the 19th century. Having Terzaghis Consolidation
Theory as a basis, the relationship between population growth, amount of water extracted from the
aquifer and subsidence rates was established by the pioneering work of Carrillo (Carrillo, 1948), later
continued by Marsal and Mazari (Marsal and Mazari, 1959). Data collected by Tamez (1992) show
that from 1900 to 1920 the settlement rate in downtown Mexico City was 3 cm/year; by the 1940s the
rate was 13 cm/year and in the early 50s it reached 26 cm per year. Wells in downtown Mexico City
were banned and settlement rates decreased to 5 cm/year. In the late 70s and early 80s new wells
were put in operation in the outskirts, mainly near the hills that surround the city in the north and in
the south. Settlement rates increased again and in central Mexico City they now amount 7 to 10
cm/year but at some sites near the newer wells they exceed 30 cm/year (Mazari, 1996). There are
zones in the outskirts, towards the edges of the former lake bed where the ground settles 40 to 60
cm/year (Santoyo et al, 2005). The total subsidence over the last 100 years with respect to a reference
point outside the lake zone now exceeds 10 m in some areas.
413
Surface settlements are an expression of changes that occur within the soil mass due to the
consolidation process triggered by the depletion of pore pressures as a consequence of the exploitation
of the aquifers that underlie the soft clays. Most significantly, effective stress increments within the
clay mass will modify stress and stress path dependent parameters like shear strength, compressibility
and permeability. Consolidation of the lacustrine clays affects the totality of the former lake zone. As
regional consolidation progresses, the clays gain strength and become stiffer; they also lose
compressibility and permeability.
Field and laboratory data collected over the last few decades have been used to document the way in
which mechanical and index properties change. The graph in Fig. 3, for example, shows one
dimensional compression curves obtained from the testing of three samples retrieved from the same
site at three different dates. As seen there, initial voids ratios have reduced gradually over the years
and apparent preconsolidation pressures have increased; the three samples exhibit the same virgin
consolidation line, approximately. Results of CPT tests show increasing point penetration resistance
over a 14 year period, as illustrated in Fig. 4; the CPT profiles in that figure also show changes in the
thickness of the upper clay formation, about 100 cm between 1986 and 2000. Compression of the
upper clay formation contributes with about 50 to 60 % to total settlements measured in the central
part of the city.
Figure 3. One dimensional compression curves (Ovando-Shelley et al, 2007).
414
Figure 4. Results of CPT tests (Ovando-Shelley et al, 2007) .
Estimation of the magnitude and distribution of future settlements in the city is important for a variety
of practical applications including problems related to the conservation of architectural monuments.
Calculation of future settlements should take into account the variability of soil properties, stress
conditions and stress histories in the clay masses as well as their evolution with time. All of these
factors have been studied previously in other investigations (Ovando-Shelley et al, 2003, 2004, 2007).
Evolution of pore pressures in the coming years was estimated assuming that the clayey soils are
elastic-viscous-plastic materials, integrating the following set of differential equations (Yin, 1994;
Yin, 1996):
2
u u 1
c ve g u, z (1)
z2 t mve
z u
mve g u, z (2)
t t
v0 vp v0 z
g u, z exp z z0 (3)
t0 z0
Where, cve is a consolidation coefficient associated to elastic deformations, equal to k/(mve w); k is the
soils permeability and mve is the volumetric compressibility along the elastic portion of the one
dimensional stress-strain curve, equal to ze/ z = ( /vo)/( z); /vo= C /2.3, where C is the slope
along the compressibility curve in log time versus strain; finally /vo is the slope of a reference one
dimensional stress-strain curve.
Equations 1 to 3 were applied at 21 sites in the central part of the city integrating them numerically
with a finite difference algorithm. The Mexico City clay deposits were modeled considering that the
two main clay strata are confined by permeable layers strata undergoing independent consolidation
415
processes (Ovando and Ossa, 2004). Initial conditions were specified from piezometric readings
obtained at nearby locations. Conditions at the boundaries were specified from pore pressure depletion
rates at the permeable layers that confine the clays. Relevant properties for modelling the
consolidation process were estimated approximately from the results of CPT soundings performed in
each site (compressibilities and permeabilities as well as thicknesses of each substratum); during the
integration process, these properties were up-dated after each time interval.
Resulting isochrone curves were used to obtain effective stress changes which were then used to
calculate settlements in each of the sites under study. Deformations (compressions) of the two main
clay strata were added to produce equal settlement contours at different dates. The map in Fig. 5, for
example, provides settlement contours estimated for the year 2075. The area covered in that map
includes the citys main square, the Cathedral, the National Palace and many other important
architectonic monuments. As seen in the map, calculated settlements are not uniform; differential
settlements that will accumulate in the future are quite large and will endanger some of the buildings
of the zone. Two of the most conspicuous buildings shown in the map, the Metropolitan Cathedral and
the National Palace, have been subjected to interventions to avoid or mitigate damage induced by
differential settlements. The former was underpinned partially with point bearing piles and the latter
was first underexcavated to correct excessive differential settlements and afterwards its subsoil was
hardened selectively to attain a more uniform distribution of compressibilities. In regard to these two
monuments, contours shown in that figure represent one of the possible scenarios that could have
affected them, had the interventions in them not taken place.
Figure 5 shows that resulting future settlements will be far from being uniform and as structures built
on these grounds accumulate differential settlements, irreversible structural damage will gradually
accumulate and build up. Differential settlements induce tilts and distortions which may lead to the
appearance of tension zones in masonries that will eventually produce fissures and cracks. This, in
turn, increases structural vulnerability to other detrimental agents including a reduction of their
capacity to tolerate further differential settlements. In earthquake-prone regions however, differential
settlements increase substantially the seismic vulnerability of structures.
TIME VARYING RESPONSE IN CENTRAL MEXICO CITY
In looking at the effects of earthquakes on old structures such as those forming part of Mexico Citys
architectural heritage, one must first look at the detrimental effects of differential settlements, as
explained before. The characteristics and intensity of future seismic events must then be considered.
The characteristics of the source and local site conditions are key factors in the distribution of
damages.
Seismicity in Mxico
Seismicity in Mxico is determined by earthquakes having several distinct mechanisms, as illustrated
in Fig. 6 (Hartmut et al, 1991):
Local earthquakes:
These originate within the basin that contained the old lake system or close to it. Several medium or
low intensity local earthquakes were recorded during the 20th century.
Earthquakes originating in the North American Plate:

The Great Acambay Earthquake of 1912 originated in a graben some 70 km northwest of the city. The
graben still threatens the city and seismologists still ponder whether a similar event can actually occur
in the near future.
416
Figure 5. Settlement contour lines, in cm, estimated for the year 2075.
Subduction earthquakes off the Mexican Pacific Coast:

These originate by the subduction of the Cocos Plate under the North American Plate. By far these are
the events that have caused the greatest damage in the city. The Guerrero earthquakes of 1909, 1957
and 1979 originated there as well as Great Michoacn Earthquakes of 1911 and 1985. By far
subduction earthquakes are the most frequent and are generally the most damaging.
Earthquakes in the subducted Cocos Plate:

These are normal faulting events that originate at intermediate depths, 50 to 100 km. They have
produced damage in inland states east of the city and cannot be overlooked in the assessment of
seismic risks in Mexico City, but normal faulting events, with return periods of a few decades.
417
Figure 6. Main earthquakes mechanisms in Mexico
Historical Background
The Aztecs and other prehispanic peoples in the Basin of Mexico recorded earthquakes and a few
descriptions of the damages they caused have reached modern historians but it is not possible to assess
seismic intensities or degree of damage from these descriptions. Archaeological data have shown that
the pyramids in Tenochtitlan, built from manually compacted clay reinforced with masonry walls
forming cells, suffered local slope failures probably enhanced by settlement differentials (Barrera,
1999). It can be speculated that some of these local failures may have occurred during earthquakes. On
the other hand, it is the authors opinion that the probability of having a massive earthquake induced
slope failure is very small indeed, given the presence of the reinforcing masonry walls within the core
of the pyramids. The houses and palaces for the Aztec royalty were rigid low rise masonry structures
were also exposed to differential settlements making them more vulnerable to suffer damages during
earthquakes on account of their heavy ceilings.
European masons soon learned that in Mexico City heavy and slender structures could not sustain
seismic forces especially if they had tilted or were inclined. Cervantes de Salazar provides the firsts
rule for constructing seismic resistant structures in the Americas which consisted simply in limiting
their height to two stories. Of course, churches, especially their towers and spires did not comply with
this rule. Historians have recorded many accounts of fallen, tilted or collapsed church towers during
earthquakes (Garca Acosta and Surez, 1996). Tilts and distortion brought about by settlement
differentials, for example, prepared the central dome at the church of Santa Teresa la Antigua for its
collapse in the great 1845 earthquake during which the bell tower at San Lorenzo tilted and cracked
diagonally.
418
Still, objective observers that visited the country in the early 19th century considered that earthquake
damage in Mexico City was small, compared to damage and destruction sustained in other cities
(Humboldt, 1822; Ward, 1828). Indeed, most damages concentrated in water and sewage pipes and in
aqueducts. Disasters in the city really began to occur well into the 20th century when newer more
vulnerable structures were built. In the central part of the city where site periods range between 1 to 3
s, approximately, the older more rigid structures having short natural periods (say, 0.3 to 0.5 s) fared
better than some of the newer more flexible constructions with longer natural periods (more than 1 s).
Note that as structures accumulate damage, their natural periods become longer and may even become
dangerously close to the site natural period, especially in the transition zones along the edges of the
old lakes.
An example of how the older colonial buildings behave during an earthquake is provided by looking at
records obtained from a network of seven accelerometers which were installed in the Metropolitan
Cathedra late in 1996. These strong motion instruments have been triggered several times and
information gathered during these events confirmed that seismic motions in the Cathedral's structure
are influenced by a very strong interaction effect with its foundation and the underlying clay strata. An
illustrative example of the soil-structure interaction effects is provided in Fig. 7 which shows three
elastic response spectra (5 % damping) obtained from accelerograms recorded at a site in the western
atrium (free field station), at the basement (crypt level) and at the main dome on 11 January, 1997
during a 7.3 magnitude earthquake whose epicenter was located some 600 km west of Mexico City.
Low period peaks, less than 1 s, which are present in the free field spectrum, are attenuated in the
spectrum calculated at the basement; spectral ordinates show peaks near 1.5 s and these can be
associated to the response of the soil deposits. Spectral peaks at around 2.5 s are due to rocking
motions in the transept, mainly in the N-S direction. Amplification of spectral ordinates is not
significant and this suggests that the transept moved essentially as a rigid body during this earthquake.
Also note that spectral ordinates are rather small; maximum acceleration at the free field station was
only 15 gals. From the analysis of motions recorded on the roof top near the apse and near the western
tower, it was concluded that motions in these two zones of the Cathedral and in its transept are
independent. This fact is explained by the existence of families of fissures separating them.
Figure 7. Elastic response spectra from strong motions recoreded at the Metropolitan Cathedral
on 11 January 1977..
419
Overall, the seismic behavior of the Cathedral has been satisfactory but the situation could change if
additional differential settlements were to accumulate in the future and indeed, this was a major issue
when deciding to underexcavate the cathedral (Meli and Snchez, 1995). Similar conclusions may be
reached when analyzing other colonial structures in the city.
Effects on Dynamic Properties

Local soil conditions, i.e. local site effects, have a major influence on seismic intensity distribution in
the former lake area in Mexico City. In this city, a distinctive feature is the extremely large
amplification of seismic motions associated to periods that coincide with the soft clay deposits natural
periods. Material properties which determine the dynamic behavior of soil deposits, density, shear
modulus and internal damping, being as they are, effective-stress-dependent, are all affected by
regional consolidation induced by deep well pumping. Changes on dynamic soil properties brought
about by pumping-induced regional consolidation have been amply documented from the results of
laboratory and field test (Ovando-Shelley et al, 2007).
The dependency of shear wave velocity on effective stress increments can be stated explicitly from the
following expression for the shear modulus at small strains, Gmax, which was derived from the results
of resonant column and cyclic triaxial tests performed on clay samples retrieved from several sites in
the former lake bed, in Mexico City (Romo, 1995; Romo, 1996):
PI I r 0.82
2 1 p0
Gmax V s 122 pa (4)
PI I r pa
Where is mass density; Vs is shear wave velocity; po is the in situ mean effective stress; pa is the
atmospheric pressure; PI is the plasticity index; Ir is relative consistency. Hence:
PI I r
0.41
pa 1 2 p0 ut
Vs 11 (5)
PI Ir pa
u(t) is a pore pressure depletion function obtained from equations 1 to 3, for example. Index
properties, plasticity index and relative consistency, have been included in expressions 4 and 5, as
noted before by other authors (Dobry, 1997).
The complete shear modulus-strain relationship for the Mexico City clays can be formulated by means
of hyperbolic functions of the type (Gonzalez, 2005):
A
2B
r
G Gmax 1 H Gmax 1 2B (6)
1 r
Where g is the shear strain; A and B are experimentally determined parameters that also depend on soil
plasticity whereas gr is a reference strain (Romo, 1995). Damping ratios can also be expressed by
means of the same hyperbolic function:
max min H min (7)
Where max and min are the maximum and minimum values of damping ratio, experimentally
determined over a convenient range of shear strains.
Effective stress increments brought about by water pumping from the aquifers will also modify
(shorten) the dominant periods of the soft clay deposits undergoing regional consolidation:
420
4 h t 4h t
T0 t (8)
Gavg avg Vavg
Where the sub-indices refer to average values of shear modulus, mass density and shear wave
propagation velocity. Average shear wave velocity is:
4 h t
Vsavg t (9)
Vsi hi t
hi t
As seen in expressions 8 and 9, the shortening of periods is due to the compression of the clay strata
and to the increase of stiffness or shear wave velocity.
Seismic Response
Seismic response analyses were made using a one dimensional wave propagation model in which the
input motion consists of vertically propagating waves. Such simple models have been used previously
to successfully predict the seismic response of Mexico City clay deposits in the largest part of the old
lake zone (Rosenblueth and Ovando-Shelley, 1991). The model used here solves the equations of
motion using Haskell-Thomsons solution for a stratified medium (Haskell, 1953; Thomson, 1950),
and obtains the response of the soil column using random vibration theory. Soil strata are assumed to
be visco elastic materials characterized by frequency independent, non linear, stiffness-strain and
damping-strain curves such as the ones provided by equations 1 to 3. Material non linearities are
introduced with the linear equivalent method. The theoretical bases of the program are explained
elsewhere (Brcena and Romo, 1994).
Average initial values of the relevant properties for analyzing the seismic response of the Metropolitan
Cathedral site in central Mexico City were taken form a high quality geotechnical sounding reported
by Tamez (Tamez et al, 1997). Static and dynamic soil parameters needed for the analysis were
updated from effective stress increments obtained from the integration of equations 1 to 3, at different
dates in the future; thicknesses of the strata were also adjusted. Input motions at the base of the model
were defined from response spectra obtained from accelerograms recorded at a basalt outcrop in the
south of the city i.e. assuming that the strata underlying the soils below 51 m are rigid.
The evolution of the sites seismic response is examined using input motions from the great
Michoacn Earthquake of 19 September, 1985 which originated along the subduction zone off the
Mexican Pacific coast, 300 km from the city (Mc = 8.1) and from the of 15 June, 1999, whose
epicenter was located 186 km south-east of the city (Mc = 7.1). Motions produced by subduction
quakes in the city are generally rich in low frequency components which are amplified by the soft clay
deposits. Normal faulting events, like the Tehuacn Earthquake, generate motions which are richer in
higher frequency components.
Response spectra obtained when the site was excited by motions recorded during the Great Michoacn
Earthquake of 1985 show that the sites dominant period can reduce substantially in the future, Fig. 8.
If the dominant period shortens due to the stiffening of the clays as they become more consolidated,
maximum spectral amplitudes would increase steadily. With the soil properties obtained in 2000, the
maximum spectral acceleration turned out to be about 0.65 g whereas the value estimated for the year
2100 reached 1.15 g. Maximum ground acceleration, on the other hand, can also increase
substantially: from 0.14 g for the year 2000 to 0.3 in the year 2100.
421
1.2 2060
1
2040
0.8
2100
Sa, g
2020
0.6
0.4 Year 2000
0.2
0 0
0.5 1 1.5 2 2.5 3 3.5 4
Period, s
Figure 8. Response spectra at the Cathedral site using time-varying shear moduli. Input motion
obtained from records logged at a basalt outcrop during the Great Michoacn Earthquake of
1985.
Nevertheless, spectral amplification i.e. the quotient between spectral ordinates and maximum
ground acceleration would vary over a relatively narrow range at the sites natural period: between
4.0 and 4.8, as seen in Fig. 9. Interestingly, spectral ordinates associated to the sites second mode
displayed a relatively small amplitude, over the 100 years studied here. Furthermore, spectral
amplification below 0.4 s is definitely insignificant. Another remarkable feature is that, even for the
very large spectral ordinates obtained in the later future, the materials behavior would remain mainly
within the linear range.
6
2020
5
2040
4 2060
max
2000
3 2100
Sa / a
0
0 0.5 1 1.5 2 2.5 3 3.5
Period, s
Figure 9. Spectral amplification obtained from the response spectra shown in the previous
figure.
The sites future spectral response to the Tehuacn Earthquake also shows the effect of the sharp
reduction in its natural period. Accordingly, spectral ordinates at that period increased steadily from
2000 to 2040; between this later year to 2060, they remain basically constant and increase noticeably
again in 2100, Fig. 10. On the other hand, spectral ordinates associated to the sites second mode are
rather more significant than for the 1985 event. Maximum ground accelerations could vary from 0.014
g in the year 2000 to 0.05 g in 2100 and spectral ordinates at the first mode from 0.036 g to 0.128 g in
these same years, respectively. Spectral amplifications at the sites natural period could change from
2.2 to 3.2 and from 2.1 to 2.4 at the second mode, Fig. 11. Nonlinear effects would also be negligible
as could have been expected a priori, given that the motions in the Tehuacn event were considerably
less intense than for the Michoacn Earthquake.
422
0.09
2100 2060
0.08
0.07 2040
0.06
Sa, g 0.05 2020

0.04
0.03
0.02 2000
0.01
0
0 0.5 1 1.5 2 2.5 3 3.5
Period, s
Figure 10. Response spectra at the Cathedral site using time-varying shear moduli. Input motion
obtained from records logged at a basalt outcrop during the 1999 Tehuacn event.
3.5
2060
3 2100
2040
max
2.5 2020
2
Sa / a
2000
1.5
0.5
0
0 0.5 1 1.5 2 2.5 3 3.5
Period, s
Figure 11. Spectral amplification obtained from the spectra shown in the previous figure.
CONCLUSIONS
The combined effects of regional subsidence and earthquakes pose serious threats to architectural
monuments in Mexico City. Low rise stiff structures, as most of the buildings classified as monuments
in the city, display low seismic vulnerability, given their intrinsic dynamic properties and given, also,
the characteristics of seismic movements in the surface of the very soft lacustrine clay deposits(long
period nearly monochromatic displacements). Differential settlements brought about by regional
consolidation induced by deep well pumping to extract water from the aquifers that underlie the
former lake bed in the city have increased the vulnerability of those structures on account of the
damage associated to them: cracks, fissures, tilts, structural distortions and the like.
Regional subsidence increases effective stresses within the soft clay masses. As the consolidation of
the clays advances, their static and dynamic properties will change in the future. Changes will occur in
the totality of the former lake zone and given the present trends in the urban and demographical
evolution of the city, regional subsidence will continue to be present in the foreseeable future.
Expected effective stress increments induced by the exploitation of the aquifers were estimated with a
viscoplastic consolidation model which overcomes some of the limitations of Terzaghis consolidation
theory. The equations of this consolidation model were integrated with a finite difference program in
423
which time dependent boundary conditions were imposed, to specify pore pressure depletion rates
from actual piezometric readings. Future expected settlements calculated with this model show that
monuments in the central part of the city will undergo extremely large differential settlements in the
years to come which will, in turn, produce damage that will add to presently existing damage.
Dynamic soil properties and, consequently, dynamic response in the city will evolve. The future
seismic response at a site near the Metropolitan Cathedral was examined studying its response to two
earthquakes having different source mechanisms: the 19 September, 1985 earthquake, a subduction
event, rich in long period components and the 15 June, 1999, Tehuacn Earthquake where short period
components were important. Other seismogenic regions produce earthquakes that also affect Mexico
City but the two ones considered here represent extreme examples of the motions. The seismic
response analysis presented and discussed here are certainly limited in their scope and aims but cant
be considered by any means exceptional. Several general conclusions regarding the future of listed
buildings in central Mexico City can be drawn from these results, pending further research into this
matter:
Depletion of pore pressures within the clayey soils in Mexico City produces changes in their
dynamic properties that can not be ignored.
Stiffening of the clays modifies (shortens) the dominant period of the soil deposits. Damage
related to the apparition of differential settlements will, on the other hand, lengthen the natural
period of old monumental buildings in the city. This combination of effects is potentially very
dangerous to these structures.
Overall shortening of dominant periods will change the zones in which seismic movements are
more intense in the city.
In the case of subduction earthquakes, the most hazardous zones will slowly migrate towards the
centre of the old lake bed where the thicker clay strata.
The transition zone, located near the edges of the lake, where more rigid clay deposits exist, will
become more vulnerable to high frequency events, i.e. deep normal faulting earthquakes or those
produced by neighboring active faults. The seismic vulnerability of damaged old structures may
increase considerably in these areas.
Monuments located in certain parts of central Mexico City, subjected to high, long standing
external loads and to the effects of regional subsidence, like the site studied here, will also
become more vulnerable to high frequency events.
From the above, it is likely that many of those structures will not survive unless decisive actions are
taken to increase their capacity to sustain further differential settlements in the future and reduce future
increments in seismic vulnerability. The authors have identified nearly 30 structures that require
urgent interventions but there are many others that may need to be intervened in the near future.
REFERENCES
Cervantes de Salazar, F. La vida en Mxico en 1554, 1555. Editorial Porra, Mxico, 1978
Musset, A. El Agua en el Valle de Mxico. Centro de Estudios Mexicanos y Centro-Americanos,
Editorial Prtico, Mxico, 1992
Carrillo, N. Influence of Artesian Wells in the Sinking of Mexico City, Proceedings of the Second
International Conference on Soil Mechanics and Foundation Engineering, Vol. VII, 1948
Marsal, Ral J. and Mazari, Marcos. The Subsoil of Mexico City, School of Engineering, UNAM,
Mexico, 1959
Tamez, E. Hundimientos diferenciales de edificios coloniales en el Centro Histrico de la Ciudad de
Mxico, Undcima Conferencia Nabor Carrillo, Zacatecas, Sociedad Mexicana de Mecnica de
Suelos, Mxico, 1992
Mazari M. La isla de los perros, El Colegio Nacional, Mxico, 1996
424
Santoyo Villa E., Ovando Shelley, E., Mooser, F and Len, E. Sntesis Geotcnica de la Cuenca del
Valle de Mxico, TGC ediciones, Mxico, 2005
Ovando Shelley, E., Romo M. P., Contreras, N. and Giralt, A. Some Implications of the Future
Evolution of Central Mexico Citys Subsoil Properties due to Water Pumping, Geofsica
Internacional, Vol. 42, No.2, 185-204, Mxico, 2003
Ovando S. E. and Ossa L. A. Modelo elastoviscoplstico para la consolidacin de los suelos y su
aplicacin al hundimiento regional de la ciudad de Mxico, Memorias de la XXII Reunin
Nacional de Mecnica de Suelos, Vol. 1, 291-299, Guadalajara, Mxico, 2004
Ovando Shelley E., Romo, M. P. and Ossa, A. The sinking of Mexico city: Its effects on soil
properties and seismic response, Soil Dynamics and Earthquake Engineering, 27, 333-343,
Mxico, 2007
Yin J. H. and Graham J. Equivalent times and one dimensional elastic viscoplastic modelling of time
dependent stress-strain behavior of clays, Canadian Geotechnical Journal, 31(1), 42-52, 1994
Yin J. H. and Graham J. Elastic visco-plastic modelling of one dimensional consolidation,
Geotechnique, 46(3), 515-527, 1996
Hartmut, S, A. Astorga, H. Amann, C. Calvo, W. Kolb, H. Scmidt and J. Winseman. Anatomy of an
evolving island: tectonic and eustatic control in the south Central American fore-arc area.
Sedimentation, Tectonics and Eustasy. Sea-level Changes at Active Margins. David MacDonald,
Editor, Special Publication Number 12, International Association of Sedimentologists, Blackwell
Scientific Publications, London, 1991
Barrera, J. A. El rescate arqueolgico en la Catedral y el Sagrario Metropolitanos de la Ciudad de
Mxico, Excavaciones en la Catedral y el Sagrario Metropolitanos Programa de Arqueologa
Urbana (Coordinador: Eduardo Matos), Instituto Nacional de Antropologa e Historia, Coleccin
Obra Diversa, Mxico, 1999
Garca Acosta, V. and Surez Reynoso, G. Los sismos en la historia de Mxico, Tomo I, UNAM,
CIESAS, Fondo de Cultura Econmica, Mxico, 1996
Humboldt, A. Ensayo poltico sobre el reino de la Nueva Espaa (1822), Editorial Porra, Col.
Sepan Cuntos, Mxico, 1978
Ward, H. G. Mxico en 1827, Fondo de Cultura Econmica, SEP, Lecturas Mexicanas No. 73,
Mxico, 1828
Meli, R. and Snchez, R. Catedral Metropolitana: correccin geomtrica, informe tcnico:
Diagnstico estructural, Asociacin de amigos de la Catedral Metropolitana de Mxico A. C.,
141-188, Mexico City, 1995
Romo M. P. Clay behaviour, ground response and soil-structure interaction studies in Mexico City,
State of the Art Paper, S Prakash Editors, Proc. Third International Conference on Recent
Advances in Geotechnical Engineering and Soil Dynamics, University of Missouri, Rolla, Saint
Louis, Missouri, 1995
Romo M. P. and Ovando E. Modelling the dynamic behaviour of Mexican clays, Proc. XII
International Conference on Earthquake Engineering, Acapulco, Mxico, 1996
Dobry R. and Vucetic M. Dynamic properties and seismic response of soft clay deposits, Mendoza
Mj Editor, Simposio internacional de ingeniera geotcnica de suelos blandos, Sociedad Mexicana
de Mecnica de Suelos, Vol. 2, 49-85, Mxico, 1997
Gonzalez B. M. Modelacin de las Propiedades Dinmicas de Arcillas Marinas, Tesis de Maestra,
Universidad Nacional Autnoma de Mxico, Mxico, 2005
Rosenbleuth, E. and Ovando, E. Geotechnical lessons learned from Mexico and other recent
earthquakes, Second International Conference on Recent Advances in Geotechnical Engineering
and Soil Dynamics, San Luis, Missouri, II, 1799 1819, 1991
Haskell, N. H. The dispersion of surface waves in multilayered media, Bulletin of the Seismological
Society of America, Vol. 43, 17-34, 1953
Thomson, W. T. Transmission of elastic waves through a stratified soil, Journal of Applied Physics,
21, 89-93, 1950
Brcena, A. and M. P. Romo. RADSH: programa de computadora para analizar depsitos de suelo
estratificados horizontalmente sujetos a excitaciones dinmicas aleatorias, Internal Report,
Instituto de Ingeniera, UNAM, Mxico, 1994
425
Tamez, E., Ovando-Shelley, E. and Santoyo, E. Underexcavation of Mexico Citys Metropolitan
Cathedral and Sagrario Church, Special Invited Lecture, Proc., XIVth International Conference on
Soil Mechanics and Foundation Engineering, Hamburg, Vol. 4, 2105-2126, 1997
426
June 25-28, 2007
Paper No. W2-1004
USING CLASSICAL MONUMENTS FOR THE ASSESSMENT OF PAST

EARTHQUAKE SCENARIOS
Ioannis N. PSYCHARIS 1
ABSTRACT
In this paper, the question whether it is possible to use the current state of classical monuments in
order to assess possible earthquake scenarios that have occurred in the area is addressed. Such a
procedure would require a backward analysis with serious uncertainties though, caused by the non-
linearity and the sensitivity of the seismic response and the unknown geometry of the structure during
each era of its life. In spite of these drawbacks, some scenarios can be adopted as reasonable, while
others should be excluded. A case study is presented, in which the current deformation of a column of
Olympieion in Athens, Greece, is used to investigate the seismic history of the area during the last
2000 years that the monument is standing. The results show that the present state of the monument
could be the result of: many medium-size, typical near-field earthquakes with a pgv around 30 cm/s
and with an average return period of about 250 years; or a smaller number of stronger earthquakes
with a pgv around 50 cm/s and a return period of about 500 years; or a single event with a pgv up to
100 cm/s. It seems unlike that earthquakes containing pulses of long periods, greater than 1.2 sec, have
occurred.
Keywords: Monuments, earthquake scenarios, seismic response, rocking, distinct element method.
INTRODUCTION
Classical monuments, like the ones encountered in Greece and South Italy, were built about 2000
years ago, or even more. In many of them, some portions of the original buildings are still standing,
although with damages in most cases. Considering the high seismicity of the regions, in which most of
them are located, it is certain that these structures have been subjected to severe earthquakes during
their long life, proving that they can survive strong seismic motions in spite of the fact that they are
made of individual structural elements (drums in the case of columns), which lie on top of each other
without mortar. This good behaviour has also been shown by analyses and experimental
investigations, which indicate that these structures can sustain strong seismic motions without
collapse, especially if they are of large size.
Although classical monuments do not collapse easily, they usually experience severe deformation
during strong earthquakes, which includes sliding and rocking of the blocks or drums, independently
or in groups. Thus, severe base excitations always produce permanent displacements, which depend on
the characteristics of the base motion. In this sense, we can say that a monument has recorded all the
strong earthquakes that have occurred in the area, where it is built. The question that arises is: can we
cipher out this information in order to assess the seismic history of the region?
The main difficulty of such a backward analysis concerns the non-linearity of the seismic response and
its sensitivity, even to small changes in the geometry of the structure and the base motion
characteristics. The unknown structural condition of the building during long periods of its life and the
1
Associate Professor, School of Civil Engineering, National Technical University of Athens, Greece,
Email: ipsych@central.ntua.gr
427
difficulty in modelling the existing damage are two reasons, related to the geometry, that contribute to
the ambiguities of the analysis. The fact that the present state is the cumulative result of a series of
strong seismic events makes things even worse.
Because of these uncertainties, it is evident that the decoding of the present deformed shape of a
classical monument is not a straightforward procedure and, in general, does not lead to unique and
sound conclusions about past earthquakes. In spite of these ambiguities, however, the quantitative
investigation of the response to a number of possible earthquake scenarios can lead to conclusions,
since some scenarios can be established as reasonable assumptions, while others should be excluded.
Such a procedure is applied in this paper for the case study of the seismic history of Athens, Greece,
based on the current state of a column of Olympieion, which is located at the centre of the city. The
results, however, have to be verified from similar analyses for other monuments in the area, before
they can be adopted.
UNCERTAINTIES OF THE BACKWARD ANALYSIS
Non-linearity and sensitivity of the response

The analysis of the seismic response of classical monuments and, in general, of stacks of rigid bodies,
is a very difficult problem, which can only be treated numerically. Several investigators have
examined this problem, analytically or experimentally (Allen et al 1986, Sinopoli 1989a,b, Psycharis
1990, Manos and Demosthenous 1992, Winkler et al 1995, Psycharis et al 2000, Konstantinidis and
Makris 2005, Papantonopoulos et al 2002, Mouzakis et al 2002, Psycharis et al 2003 among others).
All these studies show that the response is non-linear and very sensitive, characteristics which are
evident even to the simplest case of a rocking rigid block.
Due to the non-linearity of the response, it is possible that a structure collapses for a certain earthquake
and it remains stable for the same earthquake magnified to a larger value of the peak ground
acceleration (pga). This behaviour is shown in Fig. 1 for the case of a rocking rigid block, subjected to
the longitudinal component of the Kalamata, Greece, 1986, earthquake, normalized to several values
of pga. For pga=0.93g, the motion of the block is stable, but the maximum angle of rotation is close to
the slenderness angle. As a consequence, a slight increase in the base excitation to pga=0.94g is
enough to topple the block. In the contrary, a larger increase to pga=1.70g produces large rotations in
the opposite direction during the previous half cycle, stabilizing the response. If we continue to
increase the base excitation, the block overturns in the opposite direction from the one shown in Fig.1,
for pga around 2.00g.
The sensitivity of the response, even to trivial changes of the parameters, has been proved analytically
and experimentally. An example is shown in Fig.2, in which the non-repeatability of experiments is
illustrated for the seismic response of a marble model of a Parthenon column (Papantonopoulos et al
2002). The difference between the identical experiments EQ14 and EQ15 concerned a slight
difference in the base motion, due to the alteration of the input signal by the shaking table and
1.2 Overturning
1.0
0.8
pga=0.94g
0.6
0.4
0.2 pga=0.93g
0.0
-0.2 pga=1.70g
-0.4
0 2 4 6 8 10 12 14
Time (sec)
Figure 1. Rocking of a rigid block for the Kalamata, 1986 earthquake, normalized to different
values of pga.
428
100 100
EQ14 EQ15
Experiment Experiment
Relative Displacement (mm)
Relative Displacement (mm)

3DEC 3DEC
50 50
0 0
-50 -50
-100 -100
4 8 12 16 20 4 8 12 16 20
Time (sec) Time (sec)
(a) (b)
Figure 2. Comparison of the horizontal top displacement between two identical experiments
for a 1:3 model of a column of the Parthenon (reproduced from antonopoulos et al 2002).
unnoticeable alterations in the initial geometry (the drums were repositioned after each experiment).
These trivial differences caused a significant slippage during experiment EQ15, which was not
observed at experiment EQ14. In Fig.2, the corresponding numerical results, obtained using the code
3DEC (Itasca 1998), are also shown. In this case, the only difference between the two runs concerned
only the small alteration in the base motion (the acceleration recorded on the shaking table was used to
reproduce numerically each experiment), which again produced a slight difference in the response.
It should be mentioned that, in most cases, the numerical analysis could predict reasonably well the
real response of the specimen, as shown in Fig. 2(a). The discrepancy between the numerical analysis
and the experimental data depicted on Fig. 2(b) is rather exceptional. In this sense, it can be said that
the numerical analysis can, in general, predict satisfactorily the real response.
In spite of the good accuracy obtained by the numerical analysis, the aforementioned sensitivity of the
behaviour, even to small variations of the geometry, leads to an inherent uncertainty of the results,
since: (a) the exact modeling of the real structure is practically impossible, especially if imperfections
are present; and (b) the actual geometry of the monument during each era of its life is usually
unknown. Surprisingly enough, however, the overall behaviour may not differ as much as one would
expect, even for significantly different geometries. An example is presented in Fig. 3, in which the
comparison of the response of a single column and of set of two columns connected with architrave is
shown for harmonic base excitation: The two responses are quite similar, although the geometry and
the rocking mechanism are quite different.
0.80
15-drum column
Single column
0.40 2 columns with architrav e
0.00
-0.40
-0.80
0 5 10 15
Time (sec)
Figure 3. Comparison of the top displacement between the free-standing (single) column and the
set of two columns connected with architrave, for harmonic base excitation (numerical results).
Parameters used in the numerical model

The modelling of classical monuments as assemblies of rigid blocks requires the definition of several
parameters, the most important of which are the properties at the joints, the coefficient of friction and
the damping. Defining the appropriate values for these parameters is not an easy task in most cases.
Concerning the properties at the joints, it should be noted that most numerical codes cannot handle a
direct contact of rigid blocks, because of numerical instabilities. For this reason, they usually assume
429
elastic contacts at the interface of adjacent blocks. The corresponding coefficients, kn and ks, in the
normal and the tangential directions cannot be directly determined from the material properties.
Experimental data are usually needed, in order to calibrate these coefficients.
The coefficient of friction, , is another parameter that concerns the joint behaviour. The value of this
coefficient is not easily determined, either, and is different at the initiation of the sliding and after it.
Additionally, tests for the determination of usually show a wide dispersion of results. For example,
experiments on marble specimens showed that the angle of friction can vary from 35 to 50
(=0.70 1.20) and that it is affected by the vertical load (Zambas 1989, Papantonopoulos 1995).
However, it seems that the response is not, in general, very sensitive to the value of , except of the
permanent dislocations of the drums, which might be affected in extreme cases. This is shown in Fig.
4, in which the time-histories of the displacements (relative to the base) of three drums of a column of
Olumpieion in Athens are shown for the Kalamata earthquake, normalized to a peak ground velocity
(pgv) of 60 cm/sec. Drum #13 is located at about 4/5 of the height and drum #10 at about 2/3 of the
height. The value of =0.75 (center plot) is typical for marble (Papantonopoulos et al, 2002).
Decreasing this value to =0.55 (left plot) does not affect the residual displacement of the capital, but
increases the displacements of the lower drums. The use of =1.05 (right plot) decreases only slightly
the residual deformations, compared to the case of =0.75.
1.50 1.50 1.50
1.00 1.00 1.00
0.50 0.50 0.50
=0.55 = 0.75 = 1.05

0.00 0.00 0.00
Capital Capital Capital
Drum #13 Drum #13 Drum #13
Drum #10 Drum #10 Drum #10
-0.50 -0.50 -0.50
0 5 10 15 20 25 30 35 40 45 50 0 5 10 15 20 25 30 35 40 45 50 0 5 10 15 20 25 30 35 40 45 50
Time (sec) Time (sec) Time (s ec)
Figure 4. Comparison of the displacement of various drums of a marble column, for different
values of the coefficient of friction (numerical results).
The appropriate value of damping that should be used is not obvious, either. Damping originates from
two mechanisms that produce dissipation of energy: friction sliding of the drums and impact between
adjacent blocks during rocking. The dissipation of energy caused by friction is automatically
considered, if the possibility of sliding is included in the analysis. However, impact of blocks cannot
be numerically simulated if elastic properties are considered at the joints. For this reason, Rayleigh
damping is usually added to the numerical model. Comparison of numerical results with experimental
data (Papantonopoulos et al, 2002) showed that, if rocking is dominant, the addition of even a small
damping during the strong motion might over-damp the response; in the contrary, damping should be
added after the strong motion in order to dissipate the motion (Fig. 5).
200 200
Experiment Experiment
3DEC 3DEC
100 100
0 0
-100 -100
-200 -200
0 5 10 15 20 25 30 0 5 10 15 20 25 30
= Time (sec) Time (sec)
Figure 5. Comparison of experimental and numerical results for the top displacement of a 1:3
model of a column of the Parthenon: (a) zero damping; (b) 0.005 stiffness proportional damping
(reproduced from antonopoulos et al 2002).
430
Identification of seismic damage
The current damaged state of a monument is the result of many destructive events that have hit it
during its long life. In seismic areas, most of the damage is related to earthquakes, but, it is possible
that part of it has been produced by other reasons, as fire, collapse of the roof, vandalism, etc. In most
cases, the isolation of the damage that is attributed to earthquakes is very difficult. Any historical
information on the condition of the monument at various eras of its life, in combination with reports of
strong earthquakes that have occurred in the area, may be precious for this purpose.
Furthermore, even during a strong earthquake, permanent dislocations of drums can occur not only
from the rocking and sliding, but also from other reasons, as for example the collapse of an adjacent
part of the building. The effect of the latter is investigated in Fig. 6, in which time-histories of the
response are presented for: (a) the set of the two columns of Olympieion connected with architrave
and (b) a similar system with a hypothetical third column, which is assumed severely damaged and,
thus, it is collapsing during the seismic motion. It is seen that the collapse of the third column
increases temporarily the displacement of the capital in the N-S direction (normal to the longitudinal
direction of the structure), but it does not affect significantly the residual displacements. Note that
these results are sensitive to the pattern of damage considered for the third column. Therefore,
although it is not obvious from Fig. 6, it is reasonable to assume that some of the present dislocations
of the structural elements have been caused by the collapse of adjacent parts, especially the ones
observed at the capital, which is directly connected to the rest of the structure through the architraves.
2-column structure
3-column structure with collapsing third column
0.30 0.30
Top of capital, displ. in E-W dir. Top of capital, displ. in N-S dir.
0.00 0.00
-0.30 -0.30
0 10 20 30 40 50 0 10 20 30 40 50
Figure 6. Time-histories of the horizontal displacement of the capital for (i) the set of two-
columns and (ii) a hypothetical set of three columns with the third column collapsing
(Olympieion subjected to the Kalamata earthquake normalized to pgv=60 cm/s).
0.60 0.15
Top of capital, displ. in E-W dir. Top of capital, displ. in E-W dir.
0.00 0.00
-0.60 -0.15
0.60 0.15
Top of capital, displ. in N-S dir. Top of capital, displ. in N-S dir.
0.00 0.00
-0.60 -0.15
0.60 0.15
SW piece of drum #14, displ. in N-S dir. SW piece of drum #14, displ. in N-S dir.
0.00 0.00
-0.60 -0.15
0 25 50 75 100 125 150 0 25 50 75 100 125 1 50
(a) (b)
Figure 7. Horizontal displacement at different points of the W column of the two-column set of
Olympieion, caused by a series of 3 Kalamata records normalized to pgv=60 cm/s: (a) without
the building on top; (b) with the building on top.
431
Cumulative effect of many earthquakes
It is reasonable to assume that the present state of a monument is the result of a number of strong
ground motions, which have shaken the structure during its life. Due to the sensitivity to the geometry,
existing dislocations from previous earthquakes might affect the response significantly. An example is
shown in Fig. 7 for the two-column set of Olympieion subjected to three consecutive Kalamata
earthquakes, normalized to pgv=60 cm/s. In most of the cases, the repetition of the seismic excitation
leaded to increasing the residual displacements (e.g. right top plot and bottom-row plots). However,
there were cases in which the repetition of the base motion did not affect significantly the final
displacement of some drums (left top plot) or resulted in a reduction of the displacement caused by the
previous earthquake (middle diagram of Fig. 7(b): the second earthquake practically returned the
capital to its initial position and the third one caused a large residual displacement in the opposite
direction from the one that it had been displaced by the first earthquake). In this model, drum #14 was
assumed broken, according to the actual state of the monument. For the broken drum, the two pieces
had the general tendency to separate from each other after each earthquake (e.g. bottom-row diagrams
of Fig.7), although there were cases, in which this behaviour was not observed.
CASE STUDY: SEISMIC HISTORY OF ATHENS
In spite of the weaknesses of the backward analysis mentioned above, useful conclusions on possible
scenarios of past earthquakes can be drawn from the current state of a monument. A case study is
presented here for a part of Olympieion (temple of Olympios Zeus) in Athens, Greece.
Description of the structure and the numerical model

The part of the monument that was chosen for the analysis is the set of the two columns with
architrave of the SE corner of the remaining part of the building. This set was preferred from a single
column, as more representative of the monuments response, and from the more complicated nearby
sets of five and six columns in a corner layout, for simplicity reasons. Additionally, the upper drum
(drum #14) of the western column, just underneath the capital, is cracked and, in the present condition,
the opening of the crack is more than 100 mm, much larger than the dislocations of the other drums.
The opening of this crack was one of the measures of the effect of each seismic motion to the structure
considered in this analysis.
The height of each column is 16.81m and the distance between column axes is 5.50m. The diameter of
the drums varies from 1.92 m at the base to 1.57 m at the top. It should be noted that the number of
drums is varying from column to column. For the columns under consideration, the number of drums
is 14 and 15 for the western and the eastern column, respectively, without counting the base and the
capital. Also, the height of each drum is not constant, depending on the dimensions of the available
pieces of marble during construction. The architrave connecting the columns consists of three equal
beams placed one by the other. The total architrave width is 1.83 m.
For a long time, a building existed on the top of the architrave of these columns, probably constructed
by stylite monks as a place of isolation. This building, known as the cell of the Athenian stylite, is
shown in many gravures of the monument with a height of two to four stories, depending on the time
of the drawing. The building was demolished in 1886 and it is assumed that its existence lasted for 7 to
8 centuries. For this reason, this structure was also considered in the analysis. Actually, analyses were
performed with and without the building on top.
All the analyses were performed in three dimensions, using the code 3DEC (Itasca 1998), which is
based on the distinct element method. As mentioned earlier, the numerical results produced by 3DEC,
showed satisfactory agreement with experimental data in both the maximum displacements during the
seismic motion and in the residual deformation of the column (Papantonopoulos et al 2002).
432
(a) (b)
Figure 8. Model considered in the numerical analyses: (a) without the building on top; (b) with
the building on top.
The models considered in the numerical analyses (see Fig. 8) were based on the actual geometry (same
number and dimensions of drums). The drums were modeled as pyramidal segments of 22-sided cross-
section (the real columns have 22 flutes). The two pieces of the capital and the base drum were also
modeled as pyramidal segments of 22-sided cross-section. No imperfections were considered in the
model, except of the almost vertical crack at drum #14 of the western column, which split this drum in
two pieces.
The cell of the Athenian stylite was modeled as a rigid block with cross-section equal to the plan
dimensions of the architrave and of weight equal to 1000 kN, in accordance with the dimensions of the
structure given by Bouras 1996. The coefficient of friction between the block representing this
building and the architrave beams was considered large in order to prevent sliding. The model with the
building on top of the architrave can also be thought as a gross representation of the structure in its
original geometry, with the weight of the cell of the Athenian stylite modeling the weight of the roof.
In this sense, the effect of past earthquakes to the original temple, before its collapse, is also
considered approximately.
In the original structure, two steel dowels exist at each joint between drums, oriented randomly in the
directions N-S or E-W. Their cross section varies from 9 to 14 cm2 and their length is 12 to 14 cm. In
the numerical model, the dowels were modeled as non-linear shear connectors with stiffness equal to
5.8 105 kN/m and yield force in shear equal to 220 kN. Horizontal steel connectors initially existed
between the architrave beams too. However, these connectors were removed centuries ago and, for this
reason, they were not considered in the model. In any case, these elements affect mainly the
displacement of the architrave beams and have little effect to the residual displacements of the drums,
which are of the main concern in this analysis.
The joints were considered without cohesion and tensile resistance. For the joint properties, the values
proposed by Papantonopoulos et al 2002 were adopted: kn =5.0 10 6 kPa/m; ks=1.0 106 kPa/m; =0.75.
Zero damping was applied during the first 12 sec of each earthquake motion and 10% of critical, mass-
proportional damping at 0.3 Hz was added after that time. The damping was further increased to 20%
of critical for t>30 sec in order to attenuate the free vibrations and obtain the residual deformation
accurately. Mass-proportional damping was chosen instead of stiffness-proportional one, because the
latter required a smaller time step of integration and longer run-times.
433
0.6 0.6
AIGION (L) AIGION (T)
0.0 0.0
-0.6 -0.6
0.4 0.4
KALAMATA (N-S) KALAMATA (E-W)
0.0 0.0
-0.4 -0.4
Acceleration (g)
0.6 0.6
ERZINCAN (E-W) ERZINCAN (N-S)
0.0 0.0
-0.6 -0.6
0.2 0.2
GRIVA, Edessa (N-S) GRIVA, Edessa (E-W)
0.0 0.0
-0.2 -0.2
0.3 0.3
ATHENS, Syntagma (L) ATHENS, Sy ntag ma (T)
0.0 0.0
-0.3 -0.3
0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20
Figure 9. Time-histories of the horizontal components of the earthquake records considered in

the analyses.
Earthquakes considered
Four Greek and one Turkish earthquake records, with different characteristics, were used in the
numerical analyses. Plots of the acceleration time-histories are presented in Fig. 9. Both horizontal
components of each record were applied as the base excitation.
The Aigion, Greece, 1995 earthquake (M=6.2) contains a strong pulse of amplitude 0.50g and of
period 0.5 sec. It is a typical record of a pulse-type, near-field earthquake. The record of the Kalamata,
Greece, 1986 earthquake (M=6.2) was obtained at relatively stiff soil and is also typical for near-field,
medium-size earthquakes. The strong Erzincan, Turkey, 1992 earthquake (M=6.9) was caused by the
rupture of a strike-slip fault and was recorded close to the fault. The record contains a strong pulse of
long period, greater than 1.2 sec, and of amplitude of about 0.50g; such long-period pulses cause
significant damage to classical monuments. The Griva (Edessa), Greece, 1990 earthquake (M=5.9) was
recorded on soft soil and shows a long duration of the strong motion, about 6.0 sec, characterized by a
series of almost sinusoidal pulses of periods ranging from 0.5 to 0.7 sec and of small amplitude, less
than 0.10g. The last record examined was obtained at a depth of 7.00 m at the Metro station at
Syntagma, during the Athens, Greece, 1999 earthquake (M=5.9). The soil at the site was stiff (schist).
The record shows similar characteristics with the Kalamata one, although with smaller amplitude. It is
examined here because it was recorded during a medium-size earthquake very close to the
Olympieion, in a distance less than 1.0 km.
It should be noted that the monument is built on very stiff soil (schist/limestone), which can be
considered as rock. The S-E part of the structure that is examined here is founded directly on the rock,
which at that place shows a bulge. The stiff soil conditions extend to a long distance in the vicinity of
the temple, which covers all the historical center of Athens. Therefore, the ground motions that can
occur at the base of the structure during an earthquake are not expected to show any soil amplification.
Taking under consideration this soil profile, the seismic motions considered in the analysis, with the
exception of the Edessa record, could, theoretically, occur at the site as a result of near-field
earthquakes. The long-duration, almost sinusoidal Edessa record could only happen as the result of a
strong long-distant earthquake.
All the records were normalized for the numerical analyses, so that they had the same peak ground
velocity (pgv) in both horizontal components. The ground velocity was chosen as the means of
normalization, because it is more representative of the seismic motion than the base acceleration for
the response of spinal structures.
434
Assessment of possible earthquake scenarios
In order to assess possible scenarios of past earthquakes that have occurred in Athens, the following
procedure was followed: The considered earthquake motions, scaled to several values of the ground
velocity, pgv, same in both horizontal directions, were applied to the structure as a single event or in a
series of repeated earthquakes. In each case, the residual displacements produced to the structure were
calculated and compared to the existing displacements of the drums. The comparison was limited to
the deformation of the western column only. Indicative drawings of the final deformation of this
column, with the displacements of the drums magnified 20 times, are presented in Fig. 10 for the
Kalamata record. The measured deformation of the column is presented on the left top plot for
comparison. Similar drawings were derived for the rest of the earthquake motions considered. These
drawings give the effect of each earthquake scenario to the deformation of the structure.
As shown in Fig. 10, the way, in which the column deforms to several scenarios of the Kalamata
earthquake, is quite similar to the real deformation in most of the cases, except of the large
displacements of the two pieces of the capital in the E-W direction, which could not be reproduced by
the numerical analysis. These large dislocations, however, could be caused by the collapse of an
adjacent column. Quite similar results with the ones shown in Fig. 10 were obtained for the Aigion
and the Syntagma earthquakes, although in the latter case, the displacements were, in general, smaller
than the ones for the Kalamata record with the same pgv. This consistency of the results enhances the
conclusions drawn from this analysis.
Fig. 10 shows that, in order to match the magnitude of the real displacements, the structure should be
subjected to a single earthquake with pgv greater than 60 cm/s or to a series of three earthquakes with
pgv around 40 cm/s. A single earthquake with a pgv of about 40 cm/s does not seem capable to
produce the deformation of the real structure. It is interesting to note, however, that very strong single
events, with a peak velocity of up to 100 cm/s, produced displacements which were compatible with
the real deformation of the column, except of the larger deformation of the upper part in the N-S
direction (west view). In the contrary, three seismic events with pgv=60 cm/s produced displacements
which did not match the real situation, especially if the cell of the Athenian stylite was considered on
the top of the model.
Somehow different results were obtained for the Edessa record. In this case, a single event with
pgv=40 cm/s could not cause large enough displacements to the drums, to match the real deformation
of the shaft, but increasing the base motion to pgv=60 cm/s resulted to rather larger displacements than
the real ones. So we could say that a seismic motion with pgv around 50 cm/s could produce the
observed displacements. Alternatively, a series of three earthquakes with pgv=40 cm/s could have a
similar effect. For strong, single ground shakings with pgv=80 or 100 cm/s and for a series of two
events with pgv=60 cm/s, the pattern of the deformation and the magnitude of the displacements of the
upper part of the column did not match the real situation. Collapse happened for a series of three
earthquakes with pgv=60 cm/s.
As mentioned earlier, the stiff soil conditions at the site of the monument (schist) do not favor the
generation of ground motions of the type of the Edessa record from near-field earthquakes. Therefore,
such motions could only be generated by long-distant earthquakes. However, in order the real
deformation of the column to be matched, the ground motion at the site should have a peak velocity in
the order of 50 cm/s and a corresponding peak acceleration of about 0.50g. Such strong ground
shakings in a long distance from the epicenter seem rather unlike to happen and thus, this scenario
could be classified as of small probability that has occurred.
For the Erzincan earthquake, the structure collapses in most cases, due to the long-period pulse (longer
than 1.2 sec) that the record contains. Thus, the structure collapses for pgv=40 cm/s but it remains
stable for single ground motions with pgv=60 and 80 cm/s. If the earthquake with pgv=60 cm/s is
repeated for a second time, the structure collapses again. In the cases that the building does not
collapse, the produced deformation is close to the one observed at the monument. With the building on
top, the structure seems to be more stable and it does not collapse even for a series of three
435
earthquakes with pgv=60 cm/s; the deformation, which is caused to the column in this case, however,
does not match the pattern of the real deformation. However, the possibility that such an earthquake
has hit Athens during the 2000 years of the life of the monument does not seem reasonable, not only
because the structure collapses in most cases, but also, because this earthquake produces
unrealistically large displacements at the base of the column. This is shown in Fig. 11, in which the
time-histories of the base motion for pgv=60 and 80 cm/s are presented for all records. It is seen that
the base displacement for the Erzincan record is much larger than the corresponding displacements for
the others; this concerns both the maximum displacement during the seismic motion and the residual
one. It should be noted that, although it is not known whether dislocations of the columns, as a whole,
have occured, the current position of the columns and the regularity that is observed in the distance
Measured deformation of real 1 earthq. with pgv=40 cm/s 3 earthq.s with pgv=40 cm/s
structure Without the building on top Without the building on top
1 earthq. with pgv=60 cm/s 3 earthq.s with pgv=60 cm/s 1 earthq. with pgv=80 cm/s
Without the building on top Without the building on top Without the building on top
1 earthq. with pgv=100 cm/s 1 earthq. with pgv=60 cm/s 3 earthq.s with pgv=60 cm/s
Without the building on top With the building on top With the building on top
Figure 10. Residual deformation (magnified 20 times) of the western column caused by the
Kalamata record, single or in a series of three earthquakes, with and without the building on
top.
436
0.30 0.30
PGV=6 0 cm/s, Displ. in E-W dir. PGV=60 cm/s, Displ. in N-S d ir.
0.00 0.00
ERZINCAN
AIGION
-0.30 -0.30
KALAMATA
0.30 0.30
PGV=80 cm/s, Displ. in E-W dir. PGV=80 cm/s, Displ. in N-S dir. GRIVA, Edessa
ATHENS, Syntagma
0.00 0.00
-0.30 -0.30
0 5 10 15 20 25 30 0 5 10 15 20 25 30
Figure 11. Time-histories of the horizontal displacement of the base of the western column for all
the records considered in the analysis, normalized to pgv=60 and 80 cm/s (without the building
on top).
between adjacent ones indicate that such large displacements have not occur. Also, if such a large
sliding (in the order of 20 cm) had happened during a strong earthquake in the past, the trace on the
marble base, which supports the columns, should be visible even today. Such traces do not exist. For
these reasons, it is concluded that it is not probable that a strong earthquake of the Erzincan type has
hit Athens.
Another question that should be answered is whether the displacements that are observed at the
monument could be the result of a large number of smaller earthquakes. For this reason, runs were
performed for the Kalamata and the Syntagma records with pgv=20, 30 and 40 cm/s and the ground
motion repeated up to seven times. These analyses showed that a sequence of 7-8 seismic motions
with pgv=30 cm/s could also be a possible scenario, although there is not reference of such frequent,
quite strong earthquakes in Athens. In the contrary, even a larger number of earthquakes with pgv=20
cm/s could not have this effect.
CONCLUSIONS
In this paper, the possibility of using the current state of classical monuments in order to assess past
earthquake scenarios is investigated. First, the uncertainties and difficulties of such a backward
analysis are presented. In spite of these drawbacks, however, qualitive and quantitive conclusions can
be drawn, since some scenarios lead to reasonable results, while others should be excluded.
The procedure was applied to a case study for the investigation of the seismic history of Athens,
Greece, based on the present dislocations of the drums of a column of Olympieion, a monument
located in the centre of the city. The results show that relatively strong seismic motions with the
frequency content of typical near-field earthquakes could have hit Athens more than once during the
last 2000 years. Such earthquakes, with a peak velocity of 40 60 cm/s, are capable to produce the
dislocations of the drums that are observed at the monument, if repeated three to four times, i.e. for a
return period of about 500 years. Weaker seismic motions, with a pgv of about 30 cm/s can also
produce the measured deformation of the structure, if repeated seven to eight times, corresponding to
an average return period of 250 years, approximately. However, reference of such frequent, medium-
size earthquakes in Athens does not exist, making this scenario rather improbable. Smaller-size
earthquakes with a peak velocity of up to 20 cm/s do not seem capable to bring the monument in its
present damaged state, even if repeated many times.
It should be noted that a very strong near-field earthquake, with a peak velocity up to 100 cm/s, is
possible that has occurred. In that case, only one severe event could have happened in the 2000-year
life of the monument and, thus, this scenario is less probable than the previous one. It also seems
improbable that seismic motions, containing strong pulses of long period, greater than 1.2 sec, have hit
437
the city. Such earthquakes cause significant sliding at the base of the columns, which cannot be
verified at the monument.
ACKNOWLEDGEMENTS
A significant part of the work presented in this paper was performed in the framework of a research
program financed by the Greek Earthquake Planning and Protection Organization (OASP), which is
gratefully acknowledged. The author wishes to thank Professor M. Korres of the National Technical
University of Athens for providing the architectural data of the monument and the measurements for
the deformation of the column.
REFERENCES
Allen, R. H., Oppenheim, I. J., Parker A. P. and Bielak, J. On the dynamic response of rigid body
assemblies, Earthquake Engineering and Structural Dynamics, 14: 861-876, 1986.
Bouras, Ch. The so-called cell of the Athenian Stylite, Architectural Studies in Memory of Richard
Krautheimer, C. L. Striker, ed., Verlag Philipp von Zabern, 1996.
Itasca, "3 Dimensional Distinct Element Code (3DEC), Theory and Background", Itasca Consulting
Group, 1998.
Konstantinidis, D. and Makris, N. Seismic response analysis of multidrum classical columns,
Earthquake Engineering and Structural Dynamics, 34: 1243-270, 2005.
Manos, G. C. and Demosthenous, M. Dynamic response of rigid bodies subjected to horizontal base
motion, Proceedings, 10th World Conference on Earthquake Engineering, Madrid, Spain, 2817-
2821, 1992.
Mouzakis, H., Psycharis, I. N., Papastamatiou, D. Y., Carydis, P. G., Papantonopoulos, C. and
Zambas, C., Experimental investigation of the earthquake response of a model of a marble
classical column, Earthquake Engineering and Structural Dynamics, 31: 1681-1698, 2002.
Papantonopoulos, C., Psycharis, I. N., Papastamatiou, D. Y., Lemos, J. V. and Mouzakis, H.
Numerical prediction of the earthquake response of classical columns using the Distinct Element
Method, Earthquake Engineering and Structural Dynamics, 31: 1699-1717, 2002.
Papantonopoulos, C. Temple of Apollo Epicourios: Study of structural restoration, Committee for
the Restoration of the Temple of Apollo Epicourios, 219-221, 1995 (in Greek).
Psycharis, I. N. Dynamic behavior of rocking two-block assemblies, Earthquake Engineering and
Structural Dynamics, 19: 555-575, 1990.
Psycharis, I. N., Papastamatiou, D. Y. and Alexandris, A. Parametric investigation of the stability of
classical columns under harmonic and earthquake excitations, Earthquake Engineering and
Psycharis, I. N., Lemos, J. V., Papastamatiou, D. Y., Zambas, C. and Papantonopoulos, C. Numerical
study of the seismic behavior of a part of the Parthenon Pronaos, Earthquake Engineering and
Sinopoli, A. Dynamic evolution by earthquake excitation of multiblock structures, Proceedings,
Inernational Conference on Structural Conservation of Stone Masonry: Diagnosis, Repair and
Strengthening, Greek Ministry of Culture, Athens, 1989a.
Sinopoli, A. Dynamic analysis of a stone column excited by a sine wave ground motion, Applied
Mechanics Reviews, 44 (11), Part 2: S246-S255, 1989b.
Winkler, T., Meguro, K. and Yamazaki, F. Response of rigid body assemblies to dynamic
excitation, Earthquake Engineering and Structural Dynamics, 24: 1389-1408, 1995.
Zambas, C. Structural problems of the restoration of Parthenon, Committee for the Restoration of
the Monuments of Acropolis of Athens, 2a: 153-180, 1989 (in Greek).
438
June 25-28, 2007
Paper No. W2-1008
SEISMIC PERFORMANCE OF THE 4TH CENTURY A.D., BYZANTINE

LAND WALLS OF THE CITY OF THESSALONIKI, GREECE
Sextos A. G. 1 , Stylianidis, K.C. 2
ABSTRACT
The scope of this paper is to illustrate a comprehensive strategy for assessing the dynamic and seismic
performance of the Byzantine Walls of the city of Thessaloniki, which were constructed at the end of
the 4th century A.D. in the reign of Theodosius the Great. These Walls are of paramount importance
not only due to their historical value but also because they are extending in kilometers within the civil
grid of the modern city. Despite their relatively simple structural system, their seismic behavior as a
3D body has not been thoroughly studied so far, primarily due to the lack of efficient numerical tools
and the high computational cost related, especially towards the study of their dynamic response in the
time domain. Moreover, despite the extensive historical documentation available, the seismic history
of the monument is not known in detail since the historical evidence is often partial or cannot be
directly related to the damage pattern and extent observed. Along these lines, a detailed dynamic
analysis scheme is established and the structural performance of particular parts of the Walls complex
is examined for a number of realistic earthquake scenarios, accounting for the site specific soil
conditions, the spatially variable nature of the incident seismic waves, as well as the overall
geotechnical/geotectonic environment of the area. It is concluded that the advanced simulation of the
dynamic structural response that is feasible nowadays is a useful tool for the assessment of the
anticipated seismic performance of the existing Walls as well as for the understanding of their
historical behavior under earthquake loading through the centuries.
Keywords: Historical Structures, Byzantine Walls, Numerical Analysis, Earthquake Engineering
INTRODUCTION
The impact of earthquakes on monumental heritage is a critical issue that has attracted growing
scientific interest during the last decades. Monuments however, are most often complex structures,
whose preservation and/or seismic strengthening heavily relies on the clear understanding of all
factors affecting their vulnerability as well as on the accurate study of the effects of past earthquakes.
Until recently, the investigation of the seismic performance of important and extended monuments
was restricted by the inherent limitations of numerical analysis thus preventing the engineers from the
study of their dynamic response in the time domain under realistic (recorded or artificial) ground
motion scenarios. This problem was further stressed in the case of (Classical, Roman, Byzantine or
Medieval) city (Sea or Land) Walls whose dimensions were normally significantly large, hence the
complexity of the analytical or numerical procedures to be followed, (as well as the subsequent
computational cost) were considerably high. Along these lines, it was deemed interesting to focus on
the city Walls of Thessaloniki, utilizing state-of-the-art numerical tools and the experience gained
from the seismic study of other historical structures of the Byzantine Era throughout the city, in order
to attempt to shed some light not only on the structural history of the Walls through the centuries, but
1
Lecturer, Department of Civil Engineering, University of Thessaloniki, Greece, Email:
asextos@civil.auth.gr
2
Professor, Department of Civil Engineering, University of Thessaloniki, Greece, Email:
kcstyl@civil.auth.gr
439
if possible, to back analyze and evaluate the reliability of the seismic scenarios developed for the
Metropolitan City of Thessaloniki as a whole.
The Walls, still surrounding partially the old town of Thessaloniki were initially built in 315 B.C. by
the king Kassandros and were completed at the time of Great Theodosius (379-395 A.D.). Nowadays,
the Walls extend in kilometres within the civil grid of the modern city but their continuity has been
disrupted due to partial or complete collapse at numerous locations. Historically, due to their dynamic
nature, in the sense that they could be adapted in accordance with civilian needs, thus being repaired
after sieges and following rules of economy and functionality where the art of war was concerned, the
Walls of Thessaloniki (as those of Constantinople, Nicaea and others), did in fact changed
considerably over the centuries (Bouras, 2002) following the heavy fortification requirements that
arose. From the overall Byzantine Walls complex, a specific part was selected for an in-depth study.
The Walls circuit under study is the northern part of the Byzantine fortification, that forms a statically
independent structural system, extending from the beginning of the West Gate (namely Pyrros Gate)
to the main East Gate along the Eptapygriou street, near the Trigonion Tower, inclusive of the two
twin gates at the East section (widened and named thereafter by Anna Palaiologina) up to the circle
tower (Figures 1-5) constructed later. The particular Walls Section was encircling the Byzantine
acropolis (Velenis, 1998) thus separating it from the Ancient Acropolis and it consists of numerous
rectangular (primarily) and triangle defensive towers. This is essentially a monolithic and straight
complex with its main axis (indicated as x-x in Figure 6) being parallel to the East-West direction.The
visible outer face of the superstructure was constructed purely of bricks, whereas alternate bands of
stones and bricks were used for the invisible reverse side of the masonry (Velenis, 1998). The masonry
displays a mixture of construction methods as can be seen in Figure 4.
The particular Land Walls Section was selected primarily for three main reasons: (a) due to its
structural integrity which has remained almost intact within the centuries thus allowing to assume that
its seismic capacity must have not been exceeded for the last 1600 years, (b) due to its overall length
(reaching approximately 500m) that forms one of the most extended preserved parts of the fortification
and (c) due to the additional opportunity that this length provides, to further investigate (in contrast to
what is commonly performed) the potential effect of the (typically neglected) traveling nature of the
incoming seismic waves to the dynamic response of the system.
The scope of this research effort is therefore, to utilize state-of-the-art knowledge and numerical tools
for the pilot study of the seismic performance of the particular Thessaloniki city Land Walls in order
to:
a) verify the good seismic performance of the Walls (at least at the circuit under study) during
the 20/6/1978 Thessaloniki severe earthquake (Ms=6.5).
b) assess the level of safety during future seismic events through appropriate earthquake
scenarios that account for a realistic representation of the overall seismotectonic and local site
conditions.
c) investigate the potential effect of ignoring spatially variability of earthquake ground motion in
the analysis of such long and monolithic structures.
d) if possible, to compare through back numerical analysis, the level of the highest earthquake
excitation (that could have damaged the particular Wall section but apparently has not
occurred yet) with the Maximum Credible Earthquake (MCE) defined through deterministic
seismic hazard assessment for the same area. It is noted that the definition of upper limits on
earthquake ground motions has been identified as the missing piece for seismic hazard
assessment for both deterministic and probabilistic approaches (Bommer, 2002).
The overview of the structural system, the finite element representation, the definition of the reference
earthquake scenario and analysis results are presented in the following.
440
Figure 1. Contemporary photograph of the 4 th Century A.D. Byzantine Land Walls of
Thessaloniki (North side view).

Thessaloniki (North side view detail).
441
Thessaloniki. East Gate, Trigonion Tower (South side view).
Figure 4. Axonometric view of rectangular wall between along the East-West axis (North side
view). After Velenis, 1998.
442
OVERVIEW OF THE ANALYSIS APPROACH ADOPTED
Finite Element modelling of the Wall Section

Due to the complexity of the problem, it was deemed necessary to primarily establish the overall
strategy that had to be followed for the study of the particular city Walls Section (Figure 7). At first,
the 3-Dimentional Finite Element model was created based on the exact structural geometry as this
was extracted by the satellite image (illustrated in Figure 5), scaled axonometric views (Velenis, 1998)
and on-site measurements. The commercially used Finite Element program ANSYS (ver. 10.0) is used
for the static, modal and transient response of the structure, which is modeled with 4000 (10-node)
solid elements appropriately refined at the locations of the gate openings. The particular FE modelling
was adopted in order to avoid unnecessary stress concentration at locations of ubrupt geometry change
at a reasonable computational cost, especially towards dynamic analysis in the time domain. Although
various approaches have been proposed for the modelling of masonry structures (Casolo, 1998,
Galasco et al., 2004, Manos et al, 2004, Rota et al., 2005), with emphasis on the non-linear response of
the walls, the particular structure was studied herein as purely linear elastic, in order to permit
primarily focus on the macroscopic dynamic response of the system under simultaneous excitation
along the tree principal axes. Subsequently both local material and geometric non-linearities (inclusive
of cracking and sliding at critical locations) were not considered.
The structure was assumed as fully fixed at its base since geotechnical and geophysical studies have
shown that the soil conditions at the particular city location are very stiff (Anastasiadis et al., 2001).
Based on experimental results for monuments of the Byzantine Era (Manos et al., 2004, Stylianidis
and Sextos, 2006) an average uniform compression strength of fmc=2.0MPa was adopted for the
construction materials of the whole system for simplicity, based on the weaker brick masonry. The
corresponding tensile strength was set equal to f mt=0.15MPa. The Youngs modulus of the masonry
was taken equal to E=3.5x10 6 kPa (corresponding to the composite material as a whole) and the self
weight equal to =22 KN/m3 while 6% Rayleigh damping was assumed for the circular frequency
range = 40-200 rad/sec. It is noted that the two twin gates of the Trigonion Tower (i.e. Annis
Palaiologinis Gate) as well as the Pyrros Gate (Velenis, 1998), have not been modeled due to their
relatively smaller dimensions compared to the overall Wall Section length.
Earthquake Scenario for Thessaloniki

Following the development of the Finite Element model of the structure, an effort was made to
identify a reasonable earthquake ground motion scenario for the earthquake under study. Along these
lines, the three components of the seismic motion recorded at the city center (Ambraseys et al., 2000)
during the earthquake that stroke Thessaloniki in 1978 (20/06/1978, s=6.4, PGA=0.15g) were taken
as the reference motion and were appropriately deconvoluted (Figure 8) to the bedrock level at the
location where the records were obtained based on geotechnical data from the Microzoning study of
Thessaloniki (Anastasiadis et al., 2001). As mentioned previously, the local site conditions can bee
assumed to correspond to rock thus classified as soil type A according to Eurocode 8. The outcrop
motion along the axis of the Walls Section under study was then derived. As a result, three scenarios
of earthquake excitation were developed involving the:
a) earthquake ground motion of the 1978 seismic event corresponding to the local site conditions
as illustrated in Figure 8 and described above. (Case A: 1978 excitation)
b) the above seismic motion scaled to the level of PGA=0.22g, that is specified by the
Microzoning study of the city of Thessaloniki, for the particular site. (Case B: Potential
excitation corresponding to an average return period of 475 years)
c) the predicted seismic motion of Case B, applied asynchronously, that is, considering the time
lag that arises from the traveling of the seismic waves in finite velocity through the soil media.
(Case C: Asynchronous excitation corresponding to an average return period of 475 years).
443
Figure 5. Satellite image of the Land Walls under study
Figure 6. 3D Finite Element model of the Land Walls under study
Figure 7. Overall strategy for the investigation of the seismic performance of the Byzantine
Walls of Thessaloniki
444
Figure 8. Generation of earthquake ground motion scenario for the city Walls of Thessaloniki
It is noted that for all the above three cases, the Wall complex is excited simultaneously along the
three principal directions, X,Y,Z. Due to the particular features of Case C, it is further discussed in the
following section.
Spatially variable earthquake ground motion along the Land Walls under study
Development of spatially variable seismic ground motion scenarios for the study of the dynamic
response of extended structures (primarily bridges but also dams and pipelines) is attracting increasing
scientific attention after strong evidence that not only seismic motion may differ substantially in terms
of amplitude and frequency content but also that this difference may have a detrimental effect on the
displacements and stresses of the extended structure under certain circumstances (Sextos et al., 2003).
The sources of this spatial and temporal variations of seismic motion have been well identified (Der
Kiureghian & Keshishian, 1997); i.e.: (a) waves travel at a finite velocity, hence they arrive at
consecutive support points with a time delay; (b) wave coherency loss in terms of gradual reduction of
the waves statistical dependence with distance and frequency, due to multiple reflections, refractions
and superpositioning of the incident seismic waves during propagation and (c) variation in local site
conditions that strongly affects wave amplification and frequency content.
For the particular Wall Section, and since the soil conditions are generally stiff, it can be assumed that
from the above sources of spatial variability, the effect of local site conditions and the loss of wave
coherency is rather small; as a result, the asynchronous excitation is estimated solely on the phase lag
as is can be defined by the distance from the reference point and the apparent velocity of seismic
waves (taken equal to V app=2000m/sec). In order to account for the aforementioned time delay at all
base points (nodes) of the 500m long Wall complex, the Finite Element model was internally
programmed to permit the subsequent excitation of each one of the 5830 supporting points i with a
distinct 3-D ground displacement Ui,x, Ui,y , Ui,z which is a function of space and time t and is equal to:
xi
Ui ,x ( xi , yi , zi , t ) U ref , x ( xref , yref , zref , t ) (1)
Vapp
445
xi
Ui , y ( xi , yi , zi , t ) U ref , y ( xref , yref , zref , t ) (2)
Vapp
xi
Ui ,z ( xi , yi , zi , t ) U ref , y ( xref , yref , zref , t
) (3)
Vapp
where xi, yi, zi are the coordinates of a base point i, Uref,x , Uref,y, Uref,z the three component seismic
motion at the reference point, Vapp is the apparent velocity of seismic waves and xref, yref, zref the
coordinates of the reference point (which does not coincide with the zero coordinates due to the slight
rotation along the z-z- axis of the left part of the Wall as seen in Figure 6).
SEISMIC PERFORMANCE OF THE LAND WALLS UNDER STUDY
Dynamic characteristics of the Walls under study

The dynamic characteristics of the system were initially investigated, and as anticipated despite the
height of the Walls, the overall complex is rather stiff (i.e. its fundamental period is equal to 0.12sec).
It is also notable that due to the presence of the intermediate towers, there is no clear transverse mode;
on the contrary the system vibrates on a large number of similar local modes having similar modal
contribution to the overall response. Clearly, the fundamental mode is also local as it corresponds to
the east main wall which is not only unsupported laterally (i.e. no intermediate defensive towers exist)
but also is on average 2m higher than other parts of the Wall circuit. The most characteristic from
these modes are illustrated in Figure 9 where it is seen that it is only after the 25 th mode that the
structure vibrates (asymmetrically) as a whole.
Seismic response of Land Walls under study in the time domain under synchronous excitation
Having assessed the dynamic characteristics of the system, the Walls were subjected to ground
motions derived from the records obtained at the city center during the 1978 seismic event and were
appropriately deconvoluted to account for the different soil conditions at the between the recordings
and the Land Walls site. For the particular Case A earthquake scenario, the maximum tensile stress
developed in time was approximately equal to 30% of the tensile strength (equal to 0.15MPa) assumed
for the structure. Similarly, the compressive stresses developed where found also rather low (of the
order of 25% of the overall compression strength that was taken equal to 2.0MPa). Apparently, the
Walls remain primarily in compression and it is only few locations around the main gates that develop
tensile forces along particular critical sections but just for a few peaks in time. It can be claimed
therefore, that the very good performance and lack of any damage of the Wall Section under study
during the 1978 earthquake, is generally verified. This can be attributed not only to the massive
dimensions of the walls, but also to the rocky foundation conditions which deamplify ground motions
(compared to sites of softer soil formations as seen in Figure 8). It is interesting to notice that this
reduction of PGA (compared to the record site) appears to be more significant than the fact that
outcrop motion that excites the Walls is of higher frequency content thus closer to the natural period of
the structure (i.e. T=0.12sec). Moreover, similarly good performance was numerically estimated for
another part of the city Walls (Stylianidis and Sextos, 2006) as well, where the results also indicated
that the particular system studied could resist at least twice the earthquake load imposed during the
1978 earthquake without significant damage.
Having assessed the structure for the equivalent level of the 1978 earthquake seismic forces, the linear
elastic dynamic analysis in the time domain is repeated for the same ground motion, this time scaled to
the level of PGA=0.22g, that is specified by the Microzoning study of the city of Thessaloniki, for the
particular site (Case B). As seen in Figure 10, the variation of normal stresses with time for three
characteristic locations of the Walls (i.e. West Gate, East Gate and East main Wall), neither the
compression nor the tensile strength of the structure is exceeded at any point or time.
446
Figure 9. Characteristic eigenmodes of the system
Figure 10. Variation of normal stresses with time for three characteristic locations
447
It is only few (two-three) peaks for each critical location that the stresses developed reach the available
tensile strength of 0.15MPa (being 0.15, 0.12 and 0.13MPa respectively). On the contrary, the
effective tensile stress developed (as taken approximately equal to 2/3 of the maximum) does not
exceed 0.1MPa on average for the most critical locations of the structure. It is noted however that
strictly speaking, the critical checks should be performed at masonry sections instead of local points.
Nevertheless, the overall assessment is that seismic demand developed is clearly inferior to the
estimated capacity of the Walls.
In terms of shear stresses, the most critical location was identified to be the North side of the East Gate
as depicted in Figure 11. Notably, both the shear stresses developed and the shear strength are a
function of time as the latter depend on the compression stresses which also vary in time. For the
assessment of the safety factor against shear, the following equation was used according to the
provisions of Eurocode 6:
f vk ( t ) f vk 0 0.40 z (t ) (4)
where fvk(t) is the shear strength of unreinforced masonry in time, fvk0 is the shear strength of mortar
(taken equal to 0.1MPa) and z(t) is the normal compression stress at the same location. As it is seen in
Figure 11, the safety factor against shear varies between 1.5 and 2.5 in the extreme case, but it can be
estimated to be clearly greater than 2.0 on average. As a result, as was observed for the normal (tensile
and compression) stresses the performance of the Wall structure is very good also for the Case B
earthquake scenario.
This observation is interesting in the sense that, as the Microzoning-induced seismic intensity refers to
a return period of 475 years, it can be claimed that the particular level of seismic forces predicted for
T=475 years is not adequate to produce significant damage to the Wall structure a fact that is
anticipated since the structure is not known to have suffered extensive damage over the entire period
of 16 centuries. As a result, both the satisfactory future seismic performance of the Walls and the
predictions of the seismic hazard assessment for the Thesaloniki area are (if not explicitly confirmed)
at least not counteracted by the refined numerical analysis.
Figure 11. Variation of shear stresses with time. East city Walls gate (North view)
448
Seismic response of Land Walls under study in the time domain under asynchronous excitation
Following the dynamic analysis of the system for Case A and Case B earthquake scenarios, it was also
deemed interesting to investigate the potential effect of ignoring (as it is commonly the case) the
propagating nature of the travelling waves. As already discussed, the Walls where excited with 5830
spatially and temporary variable ground displacement vectors that were enforced to the corresponding
base nodes according to equations (1)-(3). Figure 12 illustrates the variation of normal stresses with
time under asynchronous excitation for the three characteristic and most critical locations identified
(i.e. West Gate, East Gate and East main Wall). By comparing these stresses derived from the
synchronous (Figure 10) and asynchronous excitation, it can be observed that although the response
can be considered as rather similar, the extremes of the stresses are different. In particular, the extreme
tensile stress at the East Gate has increased by 50% (from 0.12MPa to 0.19MPa) while the extreme
tensile stress at the East Main Wall has decreased by 40% (from 0.13MPa to 0.08MPa). Especially for
the middle point (on the East Gate) the maximum stresses occur at a slightly different time (7.15 sec
instead of 7.02 sec). Such distinct response under asynchronous earthquake ground motion has also
been observed in a large number of bridges (Sextos et al., 2005) and is pronounced in the case of long
and stiff (monolithical) structures (Zerva, 1990, Tubino et al., 2003) due to the excitation of higher
(primarily antisymmetric) modes (Sextos et al., 2004, 2006) and the subsequent development of
additional pseudo-static forces. However, it is also noted that the effective tensile and compressive
stresses are again lower than the estimated corresponding strength, hence the comments made
regarding the satisfactory historical and future performance of the Land Walls under study remain
valid.
In can be generally stated though, that the differences observed between the synchronous and the
asynchronous excitation of the particular Wall Section, even if they do not impose a threat for the
structural integrity of the system, are not negligible if one considers that due to the very stiff (rock)
formations at the base of the Wall circuit, the variability of earthquake ground motion was attributed
only to the time delay of the seismic waves and not to equally if not more important coherency loss
and local site conditions variation. As a result, the spatially variable nature of the incoming seismic
waves should not be a-priori neglected when assessing the seismic performance of long and high wall
structures, as there is high probability that ignoring asynchronous excitation, certain aspects of their
dynamic response that are related to their inherent stiffness as a structural system and the potential
excitation of higher modes may be disregarded.
Figure 12. Variation of normal stresses with time under asynchronous excitation for three
characteristic locations
449
CONCLUSIONS
This paper is an attempt to propose a refined approach towards the assessment of the seismic
performance of a part of the Land Byzantine Walls of the city of Thessaloniki. Utilizing advanced
numerical tools and generating various earthquake ground motion scenarios, a 500m long section of
the Walls is studied in the time domain under both synchronous and asynchronous excitation. The
main conclusion drawn can be summarized as follows:
It is feasible to study the dynamic and seismic response of extended systems such as high Walls
through refined finite element models and reliable estimates of the overall geotechnical,
seismological and structural conditions.
For the particular Wall Section under study, its good performance during the severe 1978
Thessaloniki earthquake (20/06/1978, s=6.4, PGA=0.15g) was verified. A similarly good
performance was numerically estimated for another part of the city Walls (Stylianidis and
Sextos, 2006) as well, where the results also indicated that the particular system studied could
resist at least twice the earthquake load imposed during the 1978 earthquake without
significant damage.
It was also numerically predicted that the particular Wall Section should not be expected to
suffer significant damage for earthquakes with a return period of 475 years. As a result, both
the satisfactory future seismic performance of the Walls and the predictions of the seismic
hazard assessment for the Thessaloniki area are (if not explicitly confirmed) at least not
counteracted by the refined numerical analysis.
The overall process adopted and proposed may be potentially implemented in a larger scale as a
means to trace (through back analysis) the seismic history of monuments and possibly to
assess an upper bound of the historical seismicity of a particular region or city.
In the framework of the process proposed, the spatially variable nature of the incoming seismic
waves should not be a-priori neglected, since it was shown that (as for bridges and other stiff
and long structures) the excitation of higher modes may impose significant pseudo-static
forces that did not develop under synchronous excitation.
Further investigation is certainly required with emphasis on the potential effect of material or
geometrical non-linearities (i.e. cracking and/or sliding/rocking at certain interfaces).
Moreover, it is evident that the reliability of the methodology adopted is strongly dependent
on the reliability in estimating the mechanical properties of the superstructure and foundation
soil materials involved (which can be also spatially variable), as well as on the accuracy of the
geometrical data that are used for finite element development.
ACKNOWLEDGEMENTS
The authors would like to thank Prof. G. Velenis for his assistance in determining the historical phases
and construction methods of the city Walls circuit under study.
REFERENCES
Ambraseys N.. P. Smit. R. Berardi. D. Rinaldis. F. Cotton. C. Berge (2000) Dissemination of

European Strong-Motion Data. CD-ROM collection. Council of European Communities.
Environment and Climate Programme. ENV4-CT97-0397. Bruxelles. Belgium
Anastasiadis A., Raptakis D., Pitilakis K. 2001. Thessalonikis Detailed Microzoning: Subsurface
Structure as basis for Site Response Analysis, Pure and Applied Geophysics Pageoph, Vol.
158(12), 2597-2633.
ANSYS Engineering Analysis System 2005. User and Theoretical Manual. ANSYS, Inc., Southpointe,
Canonsburg, Pennsylvania, Version 10.0.
450
Bouras, C. 2002. The Economic History of Byzantium: From the Seventh through the Fifteenth
Century, Laiou, A. (Editor in-Chief), Dumbarton Oaks, Trustees for Harvard University,
Washington,D.C.
Bommer, J. (2002) Deterministic vs. Probabilistic Seismic Hazard Assessment. An exaggerated and
obstructive dichotomy, Journal of Earthquake Engineering, 6 (special issue no.1), 43-73.
Casolo, S. (1998) A three-dimensional model for vulnerability analysis of slender medieval masonry
towers, Journal of Earthquake Engineering, Vol. 2, No. 4, 487-512.
CEN. 2004. Eurocode 8: Design of structures for earthquake resistance. EN 1998-1 General rules,
seismic actions and rules for buildings, Brussels.
CEN. 2004. Eurocode 6. Design of masonry structures. ENV 1996-1 General Rules, Brussels.
Der Kiureghian, A. (1996) A coherency model for spatially varying ground motions, Earthquake
Engineering & Structural Dynamics, 25, 99-111.
Galasco, A., Lagomarsino, S., Penna, A. & Resemini. S. (2004) Non-linear seismic analysis of
masonry structures, 13th World Conference on Earthquake Engineering, Vancouver, Canada, Paper
No. 843.
Manos, G. C., Soulis, V. J. and Diagouma, A. (2004) Preliminary numerical investigation of the
dynamic characteristics of historical monuments, 13th World Conference on Earthquake
Engineering, Vancouver, Canada, Paper No. 93.
Rota, M., Pecker, A., Bolognini, D. and Pinho, R. (2005) A methodology for seismic vulnerability of
masonry bridge walls, Journal of Earthquake Engineering, Vol. 9, Special Issue 2, 331-353.
Sextos, A., Pitilakis, K. and Kappos, A. (2003) A global approach for dealing with spatial variability,
site effects and soil-structure-interaction for non-linear bridges: Part 1: Methodology and analytical
tools, Earthquake Engineering Structural Dynamics, 32, 607-629.
Sextos, A., Kappos, A. and Mergos P. (2004) Effect of Soil-Structure Interaction and Spatial
Variability of Ground Motion on Irregular Bridges: The Case of the Krystallopigi Bridge, 13th
World Conf. on Earthquake Engineering, Vancouver, Canada, Paper No. 2298.
Sextos, A., Kappos, A. and Pitilakis, K. (2005) Recent developments on the effect of asynchronous
earthquake excitation on the dynamic response of soil-foundation-superstructure bridge systems, 1st
Greece Japan Workshop: Seismic Design, Observation and Retrofit of Foundations, 211-228.
Sextos, A., Kappos, A. and Kollias, V. (2006) Computing a 'reasonable' spatially variable earthquake
input for extended bridge structures, 1st European Conference on Earthquake Engineering and
Seismology, Geneva, Paper No. 1601.
Stylianidis, K. and Sextos, A. (2006) Seismic response of Byzantine Walls as a means of assessment
the seismic history of the city of Thessaloniki, 1st Hellenic Conference on rehabilitation works,
Thessaloniki, CD-ROM volume, Paper No. 084 (in Greek).
Tubino, F. Carassale, L. & S. Giovanni (2003) Seismic response of multi-supported structures by
proper orthogonal decomposition, Earthquake Engineering & Structural Dynamics, 32, 1639-1654.
Velenis, G. (1998) The Walls of Thessaloniki: From Kasandrus to Heraklios, University Studio
Press, Thessaloniki, (in Greek).
Zerva, A. (1990) Response of multi-span beams to spatially incoherent seismic ground motions,
Earthquake Engineering and Structural Dynamics, 19(6), 819-832.
451
June 25-28, 2007
Paper No. W2-1010
INFLUENCE OF ENGINEERING AND GEOLOGICAL

ENVIRONMENT ON ARCHITECTURES MONUMENTS
Erbol SHAIMERDENOV 1 , Askar ZHUSUPBEKOV2, and Tursun ZHUNISOV3
ABSTRACT
The article is devoted to the theoretical analysis on: influence of capillary water on substructure
materials and walls because of capillary moistening which in turn brings to the demolition of the
brickwork. An example on reducing the drain effect of moister on the substructures and the walls of
the mausoleum Arystan -Bab is given. The example shows the results of the survey and the research
work which was done on it.
Keywords: monuments, construction, soil, capillary water
INTRODUCTION
Architectural monuments, being under the circumstances of geological environment, also including
engineering and economical activity of human, get two types of impacts. First is demolition of
brickwork of a monument which brings to the deformation of base stock. Second is the demolition of
monuments takes place secretly and relates with changes, states and behavior of foundation soil. In the
term of the second impact the destruction is intensively developed when the change increases because
of economical activity of humans. This is the very factor which becomes more ruinous for
architectural monuments and as times go by its impact will go up. These causes motivated to a
development work of researching the man-caused impact on architectural monuments preservation to
determine a constructions deformation diagnosis in order to chose right treatment course and further
partial and full convalescence. This kind of treatment can be incorrect, and then man-caused factors
impact can not only be reduced but can high up as well. Ignoring such impacts, which are often
noticed during restoration because of lack of knowledge, take place on account of methodical
literatures absence, absence of regulation documents on registration of man-caused factors for
monument safety. As the impact of these factors has invisible character their appearance can be caught
only on a final state of the monuments demolition process.
INFLUENCE OF CAPILLARY WATER ON CONSTRUCTION
Determination of the diagnosis of architectural monuments` deformation reasons requires clear view of
behaviors and structure of that geological environment which consists of the subsystem sphere of
interaction of geological environment with the architectural monument. So the necessity of taking
engineering and geological information is occurred, information content which allows solving the
issue problem of finding the cause and explaining the phenomena of researched object, deformation of
architectural monuments.
1
Director of historical heritage Department of Ministry of Culture and Information of Republic of
Kazakhstan, Astana, Kazakhstan, e-mail: madeni_mura@mail.ru
2
Professor, Geotechnical Institute, Eurasian National University, Astana, Kazakhstan, e-mail: astana-
geostroi@mail.ru
3
Associate Professor, Geotechnical Institute, Eurasian National University, Astana, Kazakhstan
452
The importance of determining the role of hydrological factors of preserving architectural monuments
is considerable and has substantial meaning while diagnosing demolition causes of foundation soil
behavior, foundations, walls of monuments. From the other hand, diversity and uncertainty of impact
conditions, appearances of these factors and their inadequacy investigation are the main causes of
interpreting the right reasons. And these cause this or that hydro geological factors. Such confusion
sometimes brings to the acceptation of an absurd decision which can become the object of huge harm
than to save primary conditions.
Underground water adjacent to the surface and aeration zone are always the object of interest, i.e.
those hydro geological objects that closely interact with monument foundations and determine the
conditions of development of exogenous geological processes in terms of monuments and geological
environment interaction. The latter can be divided into aeration and logged zones. The permeable soil
of the logged zone is waterlogged under hydrostatic pressure. And pore spaces of aeration zone are
not filled by water under hydrostatic pressure but they can be filled by sporadic seasonal water (land
water) which is formed on impermeable soil. If the impermeable soil is between logged and non-
logged soil then it is belonged to an aeration zone.
There is usually one aeration zone and one logged zone, which one of them is under the aeration zone.
However, in some cases, there two or more zones of each belt can be apportioned. If there are two
aeration zones then the lowest zone of aeration is situated between the highest and lowest logged
zones.
Aeration zone according to its water content and its circulation can be divided into three belts: soil
moisture belt, internal belt and last one is capillary cover belt.
The soil moisture belt is adjacent to the surface zone where water evaporates into atmosphere in a big
amount in its result it transforms to plants and into air. Power of this belt is diverse according to its
type and flora. Where there is a grass cover consists 0.1 0.15 m. and where there are gardens and
parks this figure goes up rather higher. Moisture of soil belt depends on its hydroscopic water, of
which its molecular adhesion saves from evaporation and also determines its balance with atmospheric
vaporous water. Soil water is formed at the interaction moment of wet air with dry soil.
Capillary cover determines the belt which is above the logged zone, that contains capillary pores (0.1
mk) and is in permanent contact with the waterlogged zone. Water of the capillary cover is above this
zone and under the influence of capillary drains which rise up along the walls and pro rata to the pore
soil diameter. That is why the height of capillary rise depends on soil behavior and its moisture. For
dry sand soil the height of capillary rise is only few dozens centimeters (in non granulated sands
0.15-0.35 m, small granulated sands 0.35- 1.0 m). In damp condition the height of capillary rise of
sand reaches for 3-4 times higher (3). Clay sand reaches 1.0 1.5 M., Uliginose soils 2.4 m., loamy
soil from 3-to 4 m. Capillary water in clay can raise to 8.0 m., and in losses 4.0 m.
Rising of water in soil through capillary pores can be defined by this formula:
Hk=0,3/d (1)
where: Hk the height of capillary rising, m; ddiameter of capillary, m.
It should be mentioned that in the zone of soil compression foundations under the basis load the height
of capillary rising naturally decreases with the expansion of soil size. Thus, significant compression
of clay soils can lead to the total stoppage of capillary rising of water, due to the compression, ultra
pores, completely filled with fixed water are formed.
The height of capillary rising water in soils spheres of monuments interaction can be magnitude,
which corrects acceptance of these or those engineering decisions, directed to neutralization of
453
negative condition changes of soil foundation, given by its moisture. So, for example, penetration into
capillary water bases causes decrease of its coherence and solidity.
Chemical structure of water also affects on the height of capillary rising; particularly as a result of
chlorous and sulphureous natrium presence in water in moist integument loam raising these saline
through capillaries happens differently. Penetration of these salines brings to that in lower part of
loams sulphates predominate, and in upper one chlorides, greater maintenance of the latters in upper
part of slit is connected also with that chlorides diffuse faster than sulphates. Its especially important
to take that into account now, when salt is used against the icy cover of pavements and roads, which
then migrates up trough seepage of surface water along the capillary.
In this case water first penetrates into foundation soil capillaries, into the construction materials of the
foundation and the walls, consequently it causes the appearance of dampens on not only frame
constructions and interiors, but also brings to the penetration of salty ingredients into whitewash,
natural building materials, layer and their chemical destruction. Capillary water movement in soil goes
not only from down to up, but also tends to impact side of capillary force, which occurs on the
boundary of different components: foundation soil and foundation material. On this basis there is
molecular interaction of water and air with inner surface of ore space of brick work material, in its
wetting and formation of meniscus pores, determining rising force.
Between soil water and capillary cover there is an interim belt of aeration zone. Water occurring onto
this zone with its mass goes down to the logged zone or is kept in capillary pores of transition belt and
is called water of intermediate belt. Up and down boundaries of internal belt is defined according to
local conditions stipulating for the peculiarity of soil dampness regime of aeration zone.
Intervening belt of aeration zone can be missed and then capillary cover closely gets to soil water.
Such phenomenon appears at the time of historical territories impoundment. In this case straight
passage of ground water into atmosphere occurs with a help of evaporation.
Possibility of soil humidity changing aeration zone in the time and space depends on factors
influences which determine distribution of moisture in this zone. Particularly, soil humidity of aeration
zone depends on absorbing ability of soil, and that ability primarily depends on its characteristics. The
surface infrastructure condition which determines evaporation amount impacts on soil moisture. That
is why in usual conditions soil moisture goes down when evaporation takes place. But soil moisture
increases after rain. However if the surface is covered with steam-proof materials like concrete or
asphalt, evaporation dramatically decreases and soil moisture distinctly increases in spite of
atmospheric precipitates seepage exception. In this way natural soil moisture of aeration zone reflects
influence both of soil content and seepage, evaporation conditions.
Architecture monuments have great influence on evaporation way, under which it gets smaller or
entirely excludes. This throws into augmentation of aeration zone, especially in the central part of the
buildings. Peoples technology of buildings maintenance produced the struggle manners with
moisture accumulation under buildings concluded in a facility of cellar which is ventilated in summer
time. Such measure used to provide evaporation of moisture from soil pores and to exclude the
capacity of fully absorbing ability of the grounds. This trick permitted to except moisture
accumulation in soils under the monuments building and its further penetration into bearing
constructions and interiors. As an example can be taken foundations of Arystan-Babs mausoleum
walls.
In the process of researching, geotechnical and restoration measures, escort of process and quality of
geotechnical and engineering works, control of temperature-humidity conditions was organized.
454
Figure 1. Mausoleum of Arystan-Bab
Organization of constant monitoring works and environment has a great importance. Monitoring on
Arystan-Babs mausoleum must include:
- Supervision of UPV changing before and after dewatering,
- Constant control of old splits on the fronts and lurking of new ones appearance,
- Geodetic control of vertical and horizontal deformations of longitudinal and cross-cut walls,
especially buttresses;
- Registration of a ventilation trench and pipes;
- Measurement of TVR parameters to define suitable exploitation parameters;
- Control of measuring the salinization of brick walls and foundations;
- Control on a change of chemical behavior of underground water and its link with agriculture land;
- Gynecological control on ground and underground space.
To provide these tasks it is necessary to be equipped with modern facilities.
CONCLUSION
As it is shown within the world practice, salinization process in favorable geological and hydro
geological conditions takes place on many historical monuments, especially in regions with hot
climate with a notable temperature difference of atmosphere. The process takes place in a capillary
zone in a cyclic way and it can be accompanied by soil grain recovery scoping with its extent
(Massarsh K.P.)(1).This effect can cause a wall crack, particularly on weak joint parts. One of the
starts of solving this problem is decreasing capillary zone below the daytime surface.
Preservation of such monument as Aristan-Bab mouselum is a worldwide problem. It has a link with
environmental and technogeneous effects. In most cases geotechnical and hydro geological conditions
play the most important role. Capillary water rise process in clay soil and pore materials with
455
salinization is observed on most historical monuments, particularly in regions with hot climate with a
notable temperature difference of atmosphere.
In order to work out a special project on a regulation moisture regime of ground and underground soil
damp of the underground parts of mausoleum frontward (Technologicak scheme of SR, cipher 778-
g/04) NIPI PMK has offered to use an evaporation method with the help of air ventilation through
pore material backfilling outside, inside and particularly the lowest parts of the trench, and brick walls
(foundations). Moisture ventilation and evaporation intensity is stimulated with a fixation of vertical
asbestos-cement pipes with a diameter 100 mm.
The project worked out by NIPI PMK SR (6)can not find all decisions connected with geotechnical
problems of the monument. The most complex problem is appeared to fight against salinization.
There are many ways of finding a solution of the problem, but detailed study of the object might be
more useful.
REFERENCES
Massarsh K.P. Preservation of Egyptian monuments of the pharaoh epoch. Reconstruction of cities
and geotechnical construction #7. St.Petersburg, 2003-2004, page 60
Abramov S.K. Underground drainage in industrial and urban development. M.: Stroizdat, 1973, 280
Pashkin E.M. Engineering geology. M., Architecture-C, 2005, 264
Pashkin E.M., Bessonov G.B. Deformation diagnosis of architectural monuments. M., Stroizdat, 1984
Architectural monument restoration methods. Under the editorship of E.V.Michailovskogo M.:
Stroizdat, 1977
A project work documentation on Mausoleum Arystan-Babs. NIPI PMK, Almaty city, 2004 (cipher
778-g/04).
456
June 25-28, 2007
Paper No. W2-1009
THE INVERSE PROBLEM: MODELLING PAST EARTHQUAKES

FROM THEIR EFFECTS N ANCIENT CONSTRUCTIONS-
THE CASE OF THE AD365 EAST MEDITERRANEAN EARTHQUAKE
Stathis C. STIROS1 and Villy A. KONTOGIANNI2
ABSTRACT
Observations of seismic damage and observations of earthquake engineering are used to provide some
constraints to the parameters of the AD365 Eastern Mediterranean earthquake. This is a very particular
and poorly understood event: the Eastern Mediterranean has been shaken in the last 2,000 years by
numerous earthquakes but only this specific event became a legend reported by dozens of ancient
writers for centuries.
Different pieces of evidence (historical, archaeological, geomorphological, marine biological,

radiocarbon datings) provide evidence for a destructive tsunami, impressive coastal uplift and
subsidence and earthquake devastation in various parts of the Eastern Mediterranean, especially the
Nile Delta, Cyprus, Crete, Libya and Sicily in the 4th c. AD. A systematic comparative analysis and
evaluation of these data revealed that all these effects can be assigned to a single seismic sequence in
AD365. This earthquake was associated with a reverse-fault earthquake of minimum magnitude 8.5
offshore SW of Crete, as elastic dislocation analysis of the up to 9m uplift of the coasts of western
Crete reveals.
Still, widespread damage in ancient buildings in Crete, Cyprus, Libya and Sicily during this
earthquake testifies to short period, conspicuously nearly local seismic waves, while the possibility
of local amplification of seismic intensities, or of significant modification of spectral characteristics of
seismic waves in remote areas (for instance Cyprus) is unlikely.
We can therefore conclude that while ancient texts are likely to report a single giant earthquake, the
available evidence indicates a seismic sequence with at least two or three major events which occurred
in a short time interval in summer AD365.
Key Words: Eastern Mediterranean, earthquake, Crete, elastic dislocation, AD365, short-period
structures
1
Ass. Professor, Dept of Civil Engineering, Patras University, Patras 26500, Greece, stiros@upatras.gr
2
Dept of Infrastructure, Tech. University of Larisa, Larisa 41110, Greece, kontogianni@kar.forthnet.gr
457
INTRODUCTION
Historical sources report numerous earthquakes in the Eastern Mediterranean, but among them,
only the July 21 st, AD365 tsunami-associated earthquake became a legend for more than a millennium.
There is precise information that this tsunami flooded coasts of Egypt, as well as other coasts, killing
tens of thousands of people, an event commemorated for more than 200 years as the Day of Horror,
while more than 40 usually vague ancient reports provide information for widespread destruction
during the same more or less period, especially the destruction of 100 towns in Crete (for a summary,
see Guidoboni et al., 1994; Stiros, 2001, Stiros and Papageorgiou, 2001; Stiros and Drakos, 2006).
Based on this evidence, this earthquake is included in various seismological catalogues as an event
with magnitude 8-8.3 (Ambraseys et al., 1994; Papazachos and Papazachou, 1997).
Apart from historical data, there is geological and archaeological evidence for this earthquake: an
up to 9m uplift of western Crete, dated by radiocarbon data in the middle of the 4th century (Pirazzoli
et al., 1982; Kelletat, 1991), as well as evidence of widespread destruction in archaeological sites
mainly in Crete, Libya and Cyprus, precisely dated mostly by coins (accuracy in dating between
months to a few years; Stiros, 2001; Stiros and Papageorgiou, 2001).
Despite all this evidence, it remains a matter of debate whether the AD365 earthquake was a single,
regionally catastrophic earthquake with magnitude above 8, a 'universal', giant earthquake
(terraemotu per totum orbem facto, Migne, PL 27, 694) followed by a tsunami which swept
numerous Mediterranean coasts, a strong seismic sequence, or an amalgamation of at least 10 different
earthquakes which occurred between a long time period AD355- AD450 (see Guidoboni et al., 1994;
Ambraseys et al., 1994; for a critical review see Stiros, 2001; Stiros and Drakos, 2006).
In this article we try to contribute in the solution of this problem from the point of view of earthquake
engineering. In particular, based on the geography of damage can be safely assigned to a very short
time interval defined by the frequent alternation of Emperors during a period of decline of the Roman
Empire we try to investigate whether seismic waves generated by the fault derived from the elastic
dislocation analysis of the coastal uplift in Crete can explain the collapse of the type of structures
(more precisely their spectral characteristics) deduced from archaeological studies.
This is an inverse approach to a basic problem of geotechnical engineering applied to historical and
archaeological sites- an effort at modeling of a seismic sequence based on the reported or inferred
seismic damage, while the common problem is to model the earthquake vulnerability of ancient
remains.
GEOGRAPHY OF DESTRUCTION
Systematic archaeological excavations in Crete, Cyprus, Libya, Southern Italy have revealed clear
signs of extensive seismic destructions (including earthquake victims, with more spectacular those of
Kourion at Paphos, Cyprus, see Sorren, 1988, Sorren and Davis, 1985), which in many cases fall
within a period some months to a few decades long, correlating with literary evidence for a regionally
destructive event (or sequence of events) in AD365 (Table 1; Fig. 1; Stiros, 2001; Stiros and Drakos,
2006). In many cases this dating is based on inscriptions and on numismatic evidence found in
destruction layers defining upper and lower bounds of seismic destructions. Given the political
instability of the period, with emperors and consequently coins figuring emperors, changing
frequently, this dating is accurate even to within months or years (Stiros, 2001; Stiros and
Papageorgiou, 2001).
458
Table 1. Ancient sites destroyed by earthquakes in circa AD365. After Stiros and Drakos (2006).
Crete
Kissamos, West Crete shortly after AD355-361
Eleutherna, West Crete shortly after AD355-361
Gortyn, Central Crete 10-15 years before AD383
Libya
Leptis Magna, Tripolitis, SW coast shortly after AD364-367
Sabratha, Tripolitis, SW coast shortly after AD364-367, before AD378
Balagrae (El Beida), Cyrenaica, NE coast shortly after AD364
Cyrene, Cyrenaica, NE coast shortly after AD361
Ptolemais, Cyrenaica, NE coast shortly before AD364-378
Cyprus
Kourion and Paphos between late AD364-September 365
Sicily
Agrigento shortly after AD364-367
Selinunte, Temple C after AD330
Reggio Calabria (S. mainland Italy) before AD374
Figure 1. Location and summary map. Areas for which there is clear archaeological and
historical evidence for destruction caused by the AD365 earthquake are dotted. An arrow points
to the Nile Delta, affected by the tsunami associated with this earthquake.
FAULT MODELING
Except for the tsunami, for which no clear evidence has still been found, the most important geological
effect of the 365 earthquake is the up to 9m episodic uplift and tilting of western Crete which can
safely be assigned to the 365 earthquake (Fig. 2; Pirazzoli et al., 1982; Kelletat, 1991).
459
Figure 2. Contours of uplift along the coasts of western Crete and of Antikithira Islet, dated to
circa AD365 by 14C (simplified after Pirazzoli et. al., 1982).
Using the theory of elastic dislocation, observations of surface deformation (mostly geodetic data in
the broad or narrow sense of the word, such as displacements relative to the sea-level) and specific
closed functions (Okada, 1985) it is possible to estimate the geometric characteristics (orientation,
geometric characteristics, amount and direction of slip) and subsequently the seismic moment and
magnitude of the causative fault either as though formal inversion or through iterative approximations.
This last approach was adopted in the case of the coastal uplift of Crete and using the RNGCHN
software (Feigl and Dupre, 1999) we computed a least-constraint, best-fitting fault based on the
following reasonable assumptions:
(1) Observed uplift is a very large-scale effect, and hence it can only be related to a major
reverse fault along the Hellenic Trench (e.g. McKenzie, 1978; Papazachos and Papazachou, 1997;
Kontogianni et al., 2002).
(2) The contours of uplift of Fig. 2 are taken to reflect those of the AD365 earthquake, and the
possibility of an amalgamation of various effects, including earthquakes, is rather trivial.
(3) We focused on the most conservative solution, i.e. to fault that corresponds to a minimum
seismic moment . The latter is related to the magnitude w of the corresponding earthquake
according to the formulae
M0 u (1)
Mw 0.6666 log M 0 16 (2)
11
where =310 dynecm is shear modulus of the earths crust taken constant, the (inclined) surface
area of the seismic fault, and u the seismic slip (Papazachos, 1997).
Based on such constraints a large number of possible fault models responsible for the uplift were
calculated. Among them, the most favourable is a model corresponding to a thrust fault cutting
through the crust to the sea bottom offshore Crete, and causing an earthquake of minimum magnitude
8.5. The parameters of our preferred solution are shown in Table 2, while predicted uplift in Figure 3.
The trace of this fault on the seabed is shown in Figure 4, and correlates with a major bathymetric
anomaly. As it can be deduced from Figure 5, the fit between observations and predictions of coastal
uplift are very good, and within the range of accuracy of our data (see above).
460
Table 2. Parameters of the minimum fault responsible for the uplift of Crete in AD365.
seismic corresponding
strike dip depth length width slip
moment magnitude
(o) (o) (km) (km) (km) (m)
(dynecm) (Mw )
292.5 40 70 105 100 16 5.041028 8.5
Figure 3. Contours of uplift predicted for the fault of Table 2 using elastic dislocation analysis.
Parallel dense contours correspond to the causative thrust. Subsidence offshore is not shown.
Figure 4. Bathymetry in the vicinity of West Crete. A dashed line indicates the approximate sea-
bed trace of the AD365 fault (minimum fault model). Fault length may extend to NW without
any effect on the fitting.
461
10
2
observed displacem ents
calculated displacements
0
0 50 100 150 200 250 300
Distance (km)
Figure 5. Observed and calculated displacements due to the modeled fault.
IMPLICATIONS FOR EARTHQUAKE ENGINEERING
The calculated fault model for the AD365 uplift of Crete is reasonable, but it certainly corresponds
to the minimum magnitude earthquake. This earthquake can clearly explain the nearly-total
destruction of Crete, in agreement with historical and archaeological data (Stiros, 2001; Stiros and
Papageorgiou, 2001). The reason is that such an earthquake could have clearly produced the necessary
high-period waves to destroy one or two-storey houses, i.e. high-period structures at a maximum
distance of 200km from its fault (maximum length of Crete approximately 250km).
However, the situation with the other areas affected is radically different. These areas (Cyprus,
Libya, Sicily) are at distance >700km form the inferred causative fault and there is no evidence that
the high-period waves generated by this fault could reach these regions. This is supported by other
large earthquakes of the Hellenic Arc which affected only slender, high-rise buildings in Egypt, for
instance (Ambraseys et al., 1994): only long-period waves could have significant energy for major
destruction at such distances, and they could only affect structures with similar spectral characteristics,
i.e. high-period structures (for their reconstruction, see Sorren, 1988). Furthermore, there is no
evidence that long-period waves could locally generate high-energy short period seismic ground
oscillations to explain observed large scale destructions, for instance in Kourion, Cyprus (see above).
This indicates that faulting offshore Crete should be excluded as a possible cause of destruction of
single or two-storey houses in Cyprus, western Libya, etc. A series of local earthquakes should
therefore be assumed responsible for the observed widespread damage. Hence, we may conclude that
the AD365 seismic sequence consisted of at least 2-3 major events, a main event offshore Crete, a
second, probably smaller event offshore Cyprus and possibly another relatively small event between
Sicily and Libya (Fig. 6). If we also assume that in lack of modern communications techniques this
seismic sequence occurred during a time period short enough to be recorded by the ancients as a single
event, it is well consistent with previous constraints for the timing of destructions. Such a seismic
sequence is clearly consistent with our understanding of other giant earthquakes, for instance the 1960
Chile earthquake (Plafker and Savage, 1970).
462
Figure 6. Suggested focal volumes for the major shocks of the AD365 seismic sequence.
CONCLUSION
The widespread destruction associated with the AD365 earthquake requires short period waves which
could affect near local areas only. This constraint imposed by the theory of earthquake engineering
indicates that the AD365 earthquake could not have been a single giant earthquake, but it was a
seismic sequence consisting of at least two or three major events which occurred in a short time
interval in the summer of AD365 and they gave the impression of a single event to ancient writers.
ACKNOWLEDGEMENTS
Discussions with A. Papageorgiou are acknowledged. This is a contribution to the research project
EPAN-DP28 of the Greek Secretariat for Research and Technology.
REFERENCES
Ambraseys N, Melville C and Adams R. The seismicity of Egypt, Arabia and the Red Sea. Cambridge
University Press, 1994.
Feigl K and Dupre E. RNGCHN: a program to calculate displacement components from dislocations
on an elastic half-space with applications for modeling geodetic measurements of crustal
deformation. Computers and Geosciences. 25, 695-704, 1999.
Guidoboni E, Comastri A and Traina G. Catalogue of ancient earthquakes in the Mediterranean up to
the 10th century. Istituto Nazionale di Geofisica, Roma, 1994.
Kelletat D. The 1550 B.P. tectonic event in the Eastern Mediterranean as a basis for assessing the
intensity of shore processes. Zeitschrift fur Geomorphologie, 81, 181-194, 1991.
Kontogianni V, Tsoulos N and Stiros S. Coastal uplift, earthquakes and active faulting of Rhodes
Island (Aegean Arc): Modeling based on geodetic inversion. Marine Geology, 186, 299-317, 2002.
463
McKenzie D. Active tectonics of the Alpine- Himalayan belt: the Aegean Sea and surrounding
regions. Geophys. J. Royal astron. Soc. 55, 217-254, 1978.
Okada Y. Surface deformation due to shear and tensile faults in a half space. Bull. Seism. Soc.
America. 75, 1135-1154, 1985.
Papazachos B. Introduction to seismology (in Greek). Zitis, Thessaloniki, 1997
Papazachos B and Papazachou C The earthquakes of Greece. Zitis, Thessaloniki, 304pp, 1997.
Pirazzoli P, Thommeret J, Thommeret Y, Laborel J and Montagionni L. Crustal block movements
from Holocene shorelines: Crete and Antikythira (Greece), Tectonophysics, 68, 27-43, 1982.
Pirazzoli P, Laborel J and Stiros S. Earthquake clustering in the Eastern Mediterranean during
historical times. Journal of Geophysical Research, 101, B3, 6083-6097, 1996.
Plafker G and Savage J. Mechanism of the Chilean earthquakes of May 21 and 22, 1960. Geological
Soc. Am. Bull., 81, 1001-1030, 1970.
Sorren D. The day the world ended at Kourion- Reconstructing an ancient earthquake. National
Geographic Magazine, 174, 30-53, 1988.
Sorren D and Davis T. Seismic archaeology at Kourion: the 1984 campaign. Report of the
Department of Antiquities, Cyprus, 293-301& pl. LVI, 1985.
Stiros S The AD 365 Crete earthquake and possible seismic clustering during the 4-6 th centuries AD
in the Eastern Mediterranean: a review of historical and archaeological data. Journal of Structural
Geology, 23, 545-562, 2001
Stiros S and Drakos A, A fault-model for the tsunami-associated, magnitude 8.5 Eastern
Mediterranean, AD365 earthquake. Zeitschrift fur Geomorphologie, 146, 125-137, 2006
Stiros S and Papageorgiou S. Seismicity of Western Crete and the destruction of the town of Kisamos
at AD365: Archaeological evidence. Journal of Seismology, 5, 381-397, 2001.
464
June 25-28, 2007
Paper No. W2-1014
EARTHQUAKE RESPONSE AND VULNERABILITY ASSESSMENT OF

MASONRY STRUCTURES
Costas A. SYRMAKEZIS11, Athanasios K. ANTONOPOULOS2, Olga A. Mavrouli3
ABSTRACT
Historical masonry structures appear to be highly vulnerable to earthquake actions, mainly due to their
low flexural strength, fatigue phenomena, past earthquake damage etc. For their protection, their
seismic rehabilitation is of major importance and it should be based on the thorough understanding of
their dynamic properties. In this paper, the seismic response of some typical examples of existing
outstanding historical structures is presented, in order to investigate the earthquake behaviour of this
type of structures. Uncertainties involving structural properties and ground motion parameters are
taken into consideration, using the finite element analysis and specialized failure and vulnerability
evaluation methodologies. Obtained results are exploited for the classification of historical masonry
structures, aimed at correlating their seismic response with their dynamic properties and applied
earthquake parameters. Furthermore, the behaviour of the same structures is evaluated analytically, for
the same seismic ground motions considering their retrofitting using passive control dampers, where
needed. In this way, the efficiency of the method is demonstrated in terms of displacement, failure
reduction and residual vulnerability.
Keywords: masonry, seismic response, earthquake, vulnerability, dampers, finite element method
INTRODUCTION
A significant part of historical structures and monuments are masonry structures. Load bearing
masonries appear to be highly vulnerable to earthquake actions. This is ought to their low tensile
strength, their geometrical peculiarities, past earthquake damages and cracks, fatigue phenomena,
degradation, lack of proper conservation or even incompatible human interventions. In modern
societies, historical structures and monuments architectural and social value as world heritage
carriers imposes the need for their structural protection against earthquake ground motions.
During the last years, considerable research has been carried out, aiming at the investigation of the
causes of damage and the development of technologies suitable for seismic retrofit and rehabilitation
of existing masonry structures. Research has been either experimentally or analytically orientated,
often focusing on the exploitation of past damage observations, in order to comprehend the pathology
of the structure before proceeding with any intervention measures. For the analytical evaluation of
historical structures seismic response, two main sets of information should be taken into
consideration. The first one involves structural properties (i.e. material strength, dynamic properties)
and the latter concerns the parameters characterizing the seismic event (i.e. earthquake period,
amplitude of vibration, direction). The interaction of these internal and external parameters,
1
Institute of Structural Analysis and Aseismic Research, National Technical University of Athens,
Email: isaasrsyr@central.ntua.gr
2
Email: akadon@central.ntua.gr
3
Email: olgamvr2@central.ntua.gr
465
respectively, results in the final seismic performance. In both cases, many uncertainties are introduced
during the structural and earthquake simulation.
In this paper an overview of the peculiar characteristics of masonry structures concerning their
earthquake response is presented. This is achieved through the analytical investigation of typical
examples of masonry structures, using the finite element method and specialized failure and
vulnerability evaluation methodologies. The obtained results are exploited for an attempt to classify
historical masonry structures, aimed at correlating their seismic response with their dynamic properties
and applied earthquake parameters.
Furthermore, modification of historical masonry structures seismic performance, when passive

control dampers are applied, is assessed. In addition to the advantages offered concerning the
minimization of displacements and the increase of the damping energy, this technique is highly
recommended because it complies with the basic principles of reversibility, compatibility,
minimization of interventions and durability.
EVALUATION OF EARTHQUAKE RESPONSE OF HISTORICAL STRUCTURES
Earthquake response assessment using the finite element method

For the analysis of masonry structures, their peculiarities have to be considered. The distribution of the
structures mass all over its volume and the relative insignificance of masses concentrated at the floor
levels are characteristics that do not allow assumptions similar to the ones made for modern structures.
Lack of monolithic connections is a further masonry peculiarity. As a result, for the adequate
modelling of masonries, proper modification of mathematical models already used for modern
structures is required.
The finite element method is strongly recommended as it can provide a reliable distribution of masses
along the structure and a realistic simulation of inertia forces imposed on it. Additionally, it offers the
advantage of flexibility during simulation as far as it concerns geometry, boundary conditions and
other important parameters.
The most commonly employed finite element type for this purpose is the shell element, activating six
degrees of freedom on each joint. However, in some cases, the widely used discretization using 2D
finite elements (shell, plate or membrane) can be proved insufficient. This may happen when finite
elements thickness (representing the wall width) results to be considerably large in proportion to their
finite element dimensions. Inadequate model formulation using plane finite elements is also observed
in the case of architectural particularities (i.e. walls presenting width variations, interconnection points
of various types of elements etc.) where assumptions required to allow the use of two-dimensional
elements for the analysis, may be proved misleading. In such cases, the use of 3D solid elements is
strongly recommended.
Depending on the required information concernung the structural response, response-spectrum

analysis or time history analysis can be performed, among others. Information concerning ground
motion parameters is different in each case; in the first case the spectrum response should be
introduced, while in the latter case an accelerogram is necessary. These data can be calculated
according to information provided by seismological data for the area of the historical monument.
Otherwise, national and international codes and suggestions may apply for the design values.
Finite element analysis results may provide useful information concerning displacements of the
structure, developed stresses, dynamic properties, structural damping capacity etc. However, given
that the real structure obeys natural and not mathematical laws, analysis mainly aims at the provision
of an indication of the structural performance instead of a unique solution to a given problem.
466
Seismic failure evaluation
An important indicator of historical structures seismic performance is the qualitative and quantitative
calculation of structural failure. Failure analysis, following stress analysis, gives an indication of the
masonry susceptibility to damage, under specific loading assumptions. For this purpose the modified
Von Mises failure criterion is used (Syrmakezis et al, 2001). The modified Von Mises failure surface
is formed by the interaction of four surfaces S1, S 2, S 3, and S 4, as shown in Figure 1 (for the level of
zero shear stress).
Figure 1. The modified Von-Mises failure criterion
The software FAILURE (Syrmakezis et al, 2001) is used for the elaboration of principal stress data
and the generation of graphical and statistical outputs, showing the type, extent and location of failure.
Vulnerability evaluation
Risk assessment of historical structures exposed to threat scenarios comprises the assessment of both
probabilities of failure from one side and consequent effects from the other. Seismic vulnerability is
directly related to the first, under the threat of an earthquake.
The seismic vulnerability evaluation of a structure is subsequent to its failure analysis, incorporating
additionally the uncertainties that are introduced during its analytical simulation. It is associated with
the structures expected performance when subjected to a single seismic event, through the definition
of a correlation function between this action and the probability of exceeding a certain response level,
accounting, at the same time, for random values of a structures property referred to as the observation
parameter. The illustration of this function takes place through a fragility curve diagram, which
associates the cumulative probability of exceeding a certain response level with the seismic intensity
index.
The developed methodology is based on the statistical elaboration of structural failure results that have
been obtained by parametric analyses (Syrmakezis et al, 2005). For this purpose, various values of the
seismic indexes as well as of the observation factor are considered. The most common probability
distribution functions, selected to fit the data, are the normal and the lognormal.
In the case of historical masonry structures, the observation factor is usually selected to be masonrys
tensile strength, since it is closely related to its seismic performance and it usually presents large
dispersion. The seismic intensity index, representing the seismic hazard is usually selected to be the
Peak Ground Acceleration (PGA). Alternatively, other ground motion parameters can be used to
represent seismic intensity such as: Spectral Acceleration (SA), Peak Ground Velocity (PGV),
Spectral Velocity (SV) and Spectrum Intensity (SI). The response index, which represents the seismic
response of the structure, is defined, according to the proposed methodology, to be equal to the ratio of
467
the wall area that has failed divided by the total wall area, as seen in Equation 1. This type of index
can provide an indication of the damage extent and of the repair cost for a wall surface.
A fail
D .I . (1)
Atot
CASE STUDIES
Generally
In this paper, some examples of historical masonry structures are presented, in order to investigate the
seismic response of some typical structural types. These case studies are existing structures and they
involve: a stiff Byzantine church, with a geometrically complex load-bearing system (church of
Kaisariani Monastery, Athens); a neoclassical four story building of non rectangular plan (Pagrati,
Athens); a railway arch bridge of a curve span of 116,00 m (Manari, Peloponnese); a bell tower, 22,00
m tall (Saint John, Symi, Greece) and a slender Minaret of 27,00 m height (Islamic Mosque, Larnaca,
Cyprus).
Wherever needed, the seismic retrofitting of these historical structures is investigated, using passive
control dampers. This retrofitting technique can result in improvement of the earthquake response of
historical structures, through the dissipation of seismic energy by dampers. Its efficiency is
demonstrated in terms of displacements diminuation, failure reduction and residual vulnerability, for
the peak ground accelerations considered.
Moni Kaisarianis
Moni Kaisarianis is part of a Byzantine monastery, built during 11th and 12th century, on the ruins of
a paleochristian Basilica.
Figure 2. Moni Kaisarianis Church
It belongs to the type known as domed cross-in-square churches. The main church has a square ground
plan in which four cross-arms have been inscribed. The dome with its high drum is supported by four
marble columns. A general view of the Church is shown in Figure 2. The bearing system of the church
is a regular shape stone masonry made of porous stone. Bricks have been used as well. The
construction quality is very good.
For the evaluation of the response of the actual state of the structure, the finite element method was
used. The finite element model was discretized using three-dimensional solid elements. The selection
468
of this type of finite elements was made considering the peculiarities of the geometry of the structure
(wall width variations, openings etc.). The finite element model is shown in Figure 3. For the
modelling, SAP2000 v.9 software has been used.
Figure 3. The Finite Element Model of Moni Kaisarianis Church
Masonrys strength in compression has been taken as 1,30 MPa, its strength in tension as 0,24MPa and
its Modulus of Elasticity as 1,30 GPa. These values have been determined according to formulae used
for the estimation of masonry properties.
Response spectrum analysis was performed according to the spectrum definition of the Greek
Aseismic Code, for a PGA equal to 0,16g. For the failure analysis of the structure, a specialized
methodology was appliedThe procedure is based on the adaptation of the 3D problem of determination
of mechanical failure to 2D elasticity assumptions, through substitution of the third principal stress by
its effect on the other two (Syrmakezis et al, 2005). Transformation equations correspond to each one
of the three planes. Equations 2 regard plane (1,2), Equations 3 regard plane (2,3) and Equations 4
regard plane (1,3), respectively.
B
1 = 1 - 3 / 1- 1 = 1 - 3 / 1-
(2)
B
2 = 2 - 1 / 1- 3 = 3 - 1 / 1- (3)
B
1 = 1 - 2 / 1- 3 = 3 - 2 / 1- (4)
After all three equivalent biaxial stress states have been calculated, failure analysis was performed and
superposition of results was made. In Figure 4, the graphical FAILURE output is showed. Dark
shadowed areas represent areas that have failed under the given earthquake and they are indicative of
the location of failure and not of the wall plane on which failure occurs. In a colour print, outputs
distinguish much better between different kinds of failure.
The assessment of the structural vulnerability was made through a fragility curve diagram. The seismic
469
hazard was considered to be the PGA having the discrete values: 0,16g, 0,24g, 0,32g and 0,40g. The
observation factor was determined to be masonrys strength in tension with a range of values from
120-480 KPa and an increment step of 40 KPa. For the statistical elaboration of parametric analyses
failure results, the normal and lognormal distributions have been implemented. The fragility curves
diagram for the normal distribution is presented in Figure 5.
Figure 4. Failure analysis for the front wall of Moni Kaisarianis Church
Figure 5. Fragility curves diagram for Moni Kaisarianis Church
Figure 6. Neoclassical building at Pagrati Figure 7. Failure due to earthquake
470
Neoclassical four-storey building at Pagrati, Athens
The seismic response of a four-storey neoclassical building has been investigated (Figure 6). It is a
masonry building where the first floors have been made by natural stones and the last two floors,
which were added afterwards, have been made by bricks. The actual state of the structure is very bad:
masonry and mortar are of poor quality and cracks are apparent all over the structure.
The finite element model has been developed using shell elements. Masonrys strength in tension
(equal to 0,024 MPa), depending on mortars strength, is very low. As a result, failure outputs revealed
extensive damage for a possible future earthquake event of 0,16g, as shown in Figure 7.
Manari railway bridge at Peloponnese

The Manari Railway Bridge (Figure 8) is a curved masonry arch bridge situated in Peloponnese,
Greece. The Manari bridge is the longest bridge of its type in Greece and it has an optimum
construction quality. The bridges total curve span is equal to 116,00 m, its height is 20,00 m and its
width is 4,25 m. It consists of 8 semicircular arches, of 12,50 m span and of 4,25 m width. Piers are of
pyramid shape and under the train circulation deck, there is a 2-meter infill of stones. The ashlar
masonry is made of stones from a local quarry, of average dimensions 60cmx40cmx40cm. Bed mortar
joints are continuous while vertical joints are interrupted.
Figure 8. A general view of Manari bridge
For the structural identification of the bridge an analytical model has been constructed using three-
dimensional isoparametric solid finite elements, activating three translational degrees of freedom on
each node. The stone infill under the deck of the bridge has been simulated using frame elements of
high stiffness, so as to express efficiently its load-bearing function. Material properties considered for
the bridge are: modulus of elasticity, E=3,00 GPa; strength in compression, f wc=6,00 MPa; strength in
tension, fwt=0,70 MPa. Actions applied on the model include dead and live loads, earthquake loads
(PGA=0,24g), soil pressure loads at the abutments and train circulation actions. In particular, train
circulation actions include a) vertical weight loading b) horizontal centrifugal forces c) horizontal
circulation loads and d) horizontal actions due to start/braking.
In this case, time history analysis was performed, for the evaluation of the bridges seismic response
throughout a seismic event. The imported accelerogram simulated Athens earthquake (1999). Finite
element stress analysis and failure analysis revealed that the bridge, in its actual state, is highly
vulnerable to earthquake when railway loads are applied at the same time.
In an effort to mitigate the involved risk, measures had to be taken for the strengthening of the bridge.
A retrofitting method, introducing passive control devices into the load-bearing system of the bridge
(and showing the dampers effect) was proposed. Dampers were placed under each arch of the bridge,
as seen in the finite element model of Figure 9. The damping coefficient was taken as
c=5000KNsec/m.
The comparison of time-history analysis results, for both, the actual state and the retrofitted models
state, reveals the efficiency of dampers in dissipating seismic energy. Figure 10 shows the energy
471
balance before and after dampers application. Additionally, Figure 11 presents in the same diagram the
kinetic energy of the two models.
Figure 9. Retrofitted model using dampers
Figure 10.Energy balance for the unretrofitted (a) and the retrofitted (b) model
Figure 11.Kinetic energy for the unretrofitted and the retrofitted model
Bell tower of Saint John at the island of Simi

The bell tower of Saint John at Simi consists of seven different parts. Its base is of rectangular shape
of 4,95x4,95 m2 area and its total height is 22,00 m. Its plan dimensions are decreased along its height,
as can be seen in Figure 12. Its main load bearing system is masonry, however some reinforced
concrete and steel members have been added after the initial building of the structure. The bell tower
has suffered an earthquake in 1933 resulting in obvious residual deformation of its upper part.
Furthermore, its location next to the sea has contributed significantly to the degradation of materials as
well as to the corrosion of steel members and of concrete reinforcement (Kazakos, 2003).
For the analysis, a finite element model has been developed. For the assessment of soil-structure
interaction, an elastic semi-space was considered for the foundation of the structure, as presented in
Figure 13. The semi-space was simulated using three-dimensional solid finite elements, with
472
15,5x15,5 m2 area dimensions. Its depth was taken as 12 m, under the base of the structure. Soil
dimensions were determined so as to comprise the soil mass where non-negligible stresses are
developed. On the vertical faces of the semi-space, horizontal restraints were applied, while its base
was pinned.
Figure 12. Bell tower of Saint John at Simi Figure 13. Finite element model
The proposed retrofitting of the structure involves the application of a dampers system. Dampers are
placed as seen in Figure 13. The diminuation of the kinetic energy is seen in Figure 14.
Figure 14.Kinetic energy for the unretrofitted and the retrofitted model
Jami Kepir at Larnaca

The case of a masonry Minaret, part of an Islamic Mosque, situated in Larnaca, Cyprus, has been
investigated. The ratio of its height (19,50 m) to its base diameter (2,00 m), classify it as slender
structure, characterized by high flexibility and cantilever response (Figure 15). The soil-structure
interaction has been taken into consideration, as in the case of the aforementioned bell tower. The
elastic semi-space dimensions are 20x20x30 m3. For the rehabilitation of the structure the use of
473
dampers has been proposed, as seen in the analytical model of Figure 16. The effective stiffness was
taken as c=100 KNsec/m. Energy balance results, before and after retrofitting, are presented in Figure
17.
Correlation of the seismic performance with earthquake parameters

The seismic response of a structure results from the interaction of its dynamic properties with
earthquake characteristics. When the earthquake frequency coincides with the historical structures
natural frequency, the vibration amplitude is maximized and damage is more probable.
Figure 15. Jami Kepir Minaret Figure 16. Finite element model of the
retrofitted Minaret
Figure 17. Energy balance for the unretrofitted (a) and the retrofitted (b) minaret
Byzantine churches are usually stiff structures with low values of natural periods and consequently
they are mainly affected by high frequency ground motions, which usually occur near the epicenter of
the seismic activity. On the contrary, slender structures such as minarets have high eigenperiods and
they are affected by low frequency earthquakes, at a distance from the epicenter. Soil properties are
also crucial parameters affecting the seismic wave propagation, and determining in this way the final
seismic excitation applied on the structure.
474
On this basis, the dynamic properties of the investigated case studies, which have been calculated by
finite element analysis, can be used for the assessment of the vulnerability of each structure, when
subjected to an earthquake event. According to the eigenperiod of the structure, the applied Peak
Spectrum Acceleration (PSA) differs, as seen in Figure 18, for each reference earthquake. Vertical
arrows correspond to eigenperiod values for each one of the investigated structures.
Figure 18.. Seismic spectra for various earthquakes and natural periods
for the investigated case-studies
CONCLUSIONS
In this paper, a methodology is proposed for the analytical evaluation of the response of historical
structures. The methodology comprises the application of specialized software and methodologies,
developed for the failure evaluation and seismic vulnerability assessment of masonry structures.
For the demonstration of the proposed methodology some case studies have been selected, which are
typical examples of historical masonry structures. Their structural analysis was performed using the
finite element method and the elaboration of finite element computer stress output led to failure
analysis results, as well as vulnerability evaluation through a fragility curves diagram. In some cases
the soil-structure interaction was investigated by simulating the soil with three-dimensional solid
elements.
Wherever was needed, the retrofitting of these historical structures has been investigated using passive
control dampers. Their efficiency was proved in terms of energy dissipation and kinetic energy.
According to the dynamic properties of the investigated structures, their seismic vulnerability was
further evaluated for diverse earthquakes. According to the eigenperiod of the structure, the applied
Peak Spectrum Acceleration (PSA) differs, for each reference earthquake. Stiff structures like
Byzantine churches are proved to be affected more by high frequency ground motions, while slender
structures like the Minaret are more vulnerable to low frequency seismic excitations.
475
REFERENCES
Kazakos J. Application of the Finite Element Method for the Evaluation of the Structural
Performance of Historical Structures and Monuments. The case study of the Bell Tower of Saint
Jonn Prodromos of Simi, Master Thesis, Postgraduate course Protection of Historical
Structures: Materials and Interventions, National Technical University of Athens, 2003.
Syrmakezis C.A., Antonopoulos, A.K. & Mavrouli, O.A. Historical structures Vulnerability
Evaluation using Fragility Curves, Proc. of the Tenth International Conference on Civil, Structural
and Environmental Engineering Computing, 29 August-2 September 2005, Rome, Italy.
Syrmakezis C.A. Earthquake Protection of Historical Masonry Structures using Passive Control
Systems Proc. of 4th China-Japan-US Symposium on Structural Control and Monitoring, 16-17
October 2006, China.
Syrmakezis C. A., Asteris P. G, Antonopoulos A. K, Mavrouli O. A Stress-Failure Analysis of

Masonry Structures under Earthquake Loading Proc. of the Symposium on Fracture and Failure of
Natural Building Stones Applications in the Restoration of Ancient Monuments, 16th European
Conference of Fracture: Failure Analysis of Nano and Engineering Materials and Structures, 3-7
July 2006, Alexandroupoli, Greece.
Syrmakezis C.A., Asteris P.G. Masonry Failure Criterion under Biaxial Stress State. Journal of
Materials in Civil Engineering, ASCE, p. 58-64, 2001.
476
June 25-28, 2007
Paper No. W2-1006
GROUTING OF THREE-LEAF MASONRY: EXPERIMENTAL

EVIDENCE ON COMPRESSIVE AND SHEAR STRENGTH
ENHANCEMENT
Elizabeth VINTZILEOU 1
ABSTRACT
The paper summarizes the available experimental results from testing (ungrouted and grouted) three-
leaf masonry wallettes in compression or in diagonal compression.
On the basis of the experimental results, (a) the parameters that affect the behaviour of ungrouted
masonry are detected and commented upon, (b) the behaviour of grouted masonry is studied.
As the experimental results show that the key parameter for the improvement of mechanical properties
of masonry is not the compressive strength of the injected grout, emphasis is given to ternary, as well
as to hydraulic lime based grouts: Those materials are expected to ensure durable interventions, as
they are compatible with the in-situ materials from the physical-chemical point of view; they also lead
to a significant enhancement of mechanical properties of masonry, although they are of lower strength
than the cement based grouts (thanks to the improved bonding properties of ternary and hydraulic lime
based grouts with the in-situ materials).
n the basis of the experimental data on wallettes, as well as based on recent data from tests on
grouted cylinders made of filling material, simple formulae are derived, allowing for the strength of
masonry to be calculated.
Keywords: historic structures, three-leaf masonry, grouting, ternary grout, hydraulic lime based grout,
compressive strength.
INTRODUCTION
Three-leaf (stone or brick) masonry constitutes a construction type that is very common in structures
belonging to the built cultural heritage in Europe: Two external leaves (made of stone or brick
masonry) are constructed with a void of varying thickness between them. The space between the two
external leaves is filled with a loose, low strength material made of small pieces of stones and/or
bricks and mortar. This type of masonry is very vulnerable to various actions. In fact, as the bond
between the external and the interior leaves is deteriorated or lost with time (either due to decay of the
materials or due to vertical and horizontal in- and out-of-plane actions), masonry does not behave as a
whole. The slenderness of the external leaves (that resist the major part of any imposed action) is
increased and the probability of severe damages or even collapse is enhanced.
Grouting is one of the most common intervention techniques applied to improve the mechanical
properties of the interior weak leaf mainly, as well as to re-instate the collaboration between external
and internal leaves in three leaf masonry. This technique was proven to be efficient and it is well
accepted by Conservators, although it is clearly a non-reversible technique. It should be noted that,
although grouting is a technique applied to enhance the mechanical properties of masonry subjected to
1
Associate Professor, Faculty of Civil Engineering, National Technical University of Athens, Greece,
Email: elvintz@central.ntua.gr
477
any kind of actions, the importance of both compressive and shear strength upgrading when historic
structures are subjected to seismic actions cannot be overemphasized.
Due to the importance of the subject, considerable research effort was devoted to this technique during
the last almost two decades, both to design grouts adequate for injection into small voids and cracks
and to measure the enhanced mechanical properties of masonry after grouting.
This paper offers a survey of the literature related to the mechanical properties of three-leaf masonry
in compression or in diagonal compression, before and after grouting.
MECHANICAL PROPERTIES OF MASONRY BEFORE AND AFTER GROUTING
Available experimental data

Despite of the importance of the subject, the experimental data available in the literature are rather
limited in number. It should be noted that only experimental results obtained from testing squat
specimens (thus, representative of the behaviour of masonry-material and not masonry-wall) are
summarised and commented upon.
Vintzileou et al. (1995), Toumbakari (2002), Valluzzi (2000) and Vintzileou et al. (2006) have tested
three-leaf stone or brick masonry wallettes in compression and/or in diagonal compression, before and
after grouting. The grouts cover a wide range of combinations of materials (from cement-based to pure
natural hydraulic lime grouts), as well as a wide range of basic mechanical properties. In Table 1,
geometrical characteristics of the wallettes, as well as mechanical properties of materials used for the
construction of specimens are given, along with the main experimental findings.
Mechanics of three-leaf masonry in compression, before and after grouting

In three-leaf masonry is subjected to compression, the applied load is resisted mainly by the external
leaves. In fact, (Figure 1), the compressive strength (as well as the modulus of elasticity) of the filling
material is normally by an order of magnitude smaller than that of the external leaves. By way of
consequence, as proven experimentally by Egermann (1993), as well as by Binda et al. (2006), the
external leaves carry the larger part of the applied load. Furthermore, due to the incompatibility of
deformations of the leaves, (a) the bond between masonry leaves and filling material may be broken
and (b) the external leaves are subject to horizontal (out-of-plane) deformations due to the larger
lateral dilatancy of the filling material.
N/Aw e e
i /Ee
e /Ei
i
fe,E e, e Fe
fi,Ei , i
Fi
te ti te te ' ti'
'
t =te(1+ e e/Ee )
e
ti '=ti(1+ i i /Ei)
Figure 1. Stresses and deformations in external leaves and filling material in a three-leaf
masonry subjected to compression (out-of-scale)
The latter is, on the contrary, under the beneficial confinement offered by the external leaves, thus
improving somehow its poor mechanical properties. This behaviour leads to the failure mode that was
observed in all specimens subjected to compression (Figure 2): As the load on masonry increases,
cracking is initiated both on the faces of the wallettes and on their sides. Vertical cracks on faces of the
specimens pass through mortar joints or they cross also stones and bricks, depending on the relative
strengths of the materials. On the other hand, transverse (vertical) cracks appear along the interface
between the external and the internal leaves mainly, whereas in a number of cases, cracks open within
478
the filling material as well. The mechanism leading to the failure of masonry can be detected if one
observes the development of the vertical cracks in masonry.
FACE 1 SIDE 1 FACE 2 SIDE 2
Figure 2. Typical failure mode of three-leaf masonry in compression (Vintzileou et al., 2006)
In fact, as shown in Table 1, the opening of transverse cracks is substantial larger than the opening of
those appearing on the faces of masonry. This shows clearly that once transverse cracks are formed,
each leaf behaves almost independently from the others. As the slenderness of each leaf is
substantially larger than that of the whole element, out-of-plane deformation of the external leaves
becomes apparent, as the compression load increases. This phenomenon is accentuated by the
horizontal deformations imposed by the filling material (Figure 1).
When masonry is grouted (using a high injectability grout) all voids and cracks are filled down to a
size of some tenths of millimetre. The major part of the injected material improves the behaviour of
the filling material, as well as that of the interfaces, whereas, secondarily, the properties of the external
leaves are also enhanced. Thus, masonry is more or less homogenized. As a result, (a) the contribution
of the filling material to the resistance against the compression load is increased. Thus, the stresses on
the external leaves are reduced. On the other hand, (b) the difference in the deformability properties of
the leaves decreases. Thus, the additional horizontal deformations on the external leaves, due to the
lateral dilatancy of the filling material are also reduced. Last but not least (c) the grout contributes to
the enhancement of the bond along the interfaces between consecutive leaves. As a result, the opening
of (critical) transverse cracks is delayed and the compressive strength of masonry is enhanced. The
delay in the appearance of transverse cracks is observed in experimental results. In fact, as shown in
Table 1, although before grouting, transverse crack openings (wtrans) of the order of several millimetres
were measured at the moment of attainment of the compressive strength of ungrouted masonry (f wc,0),
for the same value of compressive stress, the opening of transverse cracks in the grouted wallettes
(w trans,fwc0) is practically equal to zero. Nevertheless, when the resistance of (grouted) interfaces is
reached, separation between leaves takes place and the final mechanism of failure is similar to that
occurring in masonry before grouting or in thick single-leaf masonry.
Mechanical properties of ungrouted masonry in compression

On the basis of the experimental summarized in Table 1, the following comments can be made:
(a) The compressive strength of masonry units is not a decisive parameter for the compressive strength
of ungrouted masonry: Compare, for example, the values of compressive strength of masonry
measured by Vintzileou et al. (1995) on wallettes made of stones having compressive strength equal to
100N/mm2, with those obtained by Toumbakari (2002) (for compressive strength of stones equal to 55
N/mm2). On the contrary,
(b) It seems that the compressive strength of mortar and that of the filling material affect more the
compressive strength of masonry. In fact, the wallettes tested by Toumbakari (2002) reached, in
general, higher compressive strengths than the wallettes tested by Vintzileou et al. (1995) thanks to the
higher compressive strength of both filling material and mortar.
479
(c) The values of the compressive strength of masonry, measured on identical wallettes, present a
difference of approximately 30% between the minimum and the maximum value. This fact is taken
into account in Section 4, when the estimation of the compressive strength of masonry is attempted.
(d) The ratio of modulus of elasticity (measured at a stress level approximately equal to 0.3f wc,0) to
compressive strength of ungrouted masonry seems to be very scattered (Figure 3). However, as it is
the case for modern masonries as well, most of the values of this ratio lie between 500 and 1500.
(e) The values of vertical strain corresponding to the compressive strength of masonry before grouting
are very scattered (Figure 4). However, if one excludes the very high ultimate strain values measured
by Valluzzi (2000), it seems that for the types of masonry tested within the examined programs, the
most frequent values of wu,0 are between 0.001 and 0.0025.
Vintzileou et al., 1995

Vintzileou et al., 1995 Toum bakari, 2002
Toumbakari et al., 2002 Valluzzi, 2000
Valluzzi, 2000 Vintzileou et al., 2006
Vintzileou et al., 2006 2.8
2.8
2.6
2.6 2.4
2.2
2.4 2
1.8
2.2
1.6
2 1.4
1.2
1.8 1
0.8
1.6
0.6
0.4
1.4
0.2
1.2 0
0 250 500 750 1000 1250 1500 1750 2000 2250 0 0.002 0.004 0.006 0.008
Ew,0 /fwc,0 wu,0
Figure 3. Modulus of elasticity of ungrouted Figure 4. Relationship between the

masonry, normalized to the compressive compressive strength of ungrouted
strength of masonry vs. compressive strength masonry and the corresponding strain.
of ungrouted masonry
Mechanical properties of grouted masonry in compression

(a) The data of Table 1 prove that, as in the case of the compressive strength of masonry before
grouting, a difference of the order of 30% is observed between the minimum and the maximum
compressive strength after grouting within each individual series of tests. This affects the degree of
accuracy that should be sought for, when estimating the compressive strength of grouted masonry
through any model (be it empirical or analytical).
(b) Figure 5 shows the relationship between the compressive strength of the grout and the respective
compressive strength of grouted masonry. One may observe that there is a clear tendency of the
compressive strength of grouted masonry to increase (almost linearly) with increasing compressive
strength of grout, f gr,c, when the latter does not exceed 10N/mm 2 approximately. For higher, fgr,c
values, no significant further enhancement of the compressive strength of masonry is observed. It
should be noted that grouts with a compressive strength up to approximately 10 N/mm2 are ternary
grouts with reduced cement content (less than 50%wt.) or hydraulic lime based grouts, whereas
cement based grouts exhibit compressive strengths between 15 N/mm2 and 30 N/mm2. This important
observation insinuates that the key property governing the mechanical properties of grouted masonry
is not the compressive strength of the grout.
(c) Toumbakari (2002) made the working hypothesis that the key parameter for the improvement of
the mechanical properties of three-leaf masonry is the bond properties of the interfaces between grout
and in situ materials, as improved bonding properties along the external leaves to filling material
480
interfaces contribute to delayed opening of transverse cracks that lead to failure of masonry. In fact,
shear tests on such interfaces (Toumbakari, 2002) have proven that the maximum bond resistance
obtained by (lower compressive strength) ternary grouts was equal or even higher than the bond
resistance of (higher compressive strength) cement grout to in situ materials interfaces. This result was
fully confirmed by a systematic experimental investigation conducted by Adami et al. (2007).
Toumbakari, 2002 Vintzileou e t al., 1995
Valluzzi, 2000 Toum ba ka ri, 2002
Valluzzi, 2000
Vintzileou e t al., 2006
5.5 DTR R, 2006
30
5
4.5
25
4
3.5 20
3
15
2.5
2 10
1.5
1 5
0.5
0
0 0 0.2 0.4 0.6 0.8
0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 f gr,t/fgr,c
fgr,c(N/mm2)
Figure 5. Relationship between the Figure 6. Ratio of tensile to compressive

compressive strength of the grout and the strength of grout as a function of the
obtained compressive strength of grouted compressive strength of the grout
masonry
Vintzileou et a l., 1995
Vintzileou et al., 1995 Toumbakari, 2002
Toumbakari, 2002 Valluzzi, 2000
Valluzzi, 2000 Vintzileou et a l., 2006
Vintzileou et al., 2006 2.5
5.5
5
2
4.5
4 1.5
3.5
1
3
2.5
0.5
2
1.5 0
0 1 2 3 4 5 0 5 10 15 20 25 30
fgr,t(N/mm 2) fgr,c (N/mm2 )
Figure 7. Relationship between the tensile Figure 8. Relationship between the

strength of grout and the obtained compressive strength of grout and the
compressive strength of grouted masonry. strain at strength of grouted masonry
normalized to the respective strain of
ungrouted masonry.
As bond properties depend on the tensile strength of the binding material, one may expect a better
correlation between the compressive strength of the grouted masonry and the tensile strength of the
grout. A first positive sign for the validity of this assumption is the fact that, as shown in Figure 6, the
ratio between compressive and tensile strength of grout is not constant. On the contrary, there is a
clear tendency of the f gr,t/fgr,c ratio to increase for decreasing compressive strength of the grout. It
should be noted that, as shown in the same Figure 6, for compressive strength of grout varying
481
between 3.0 and 10.0 N/mm2 (corresponding to ternary and to hydraulic lime based grouts), the values
of the f gr,t/f gr,c ratio vary between 0.20 and 0.60. It is not, however, possible to find a relationship
between the compressive strength of the grout and the f gr,t/f gr,c ratio, on the basis of the available test
data.
(d) Furthermore, in Figure 7, the experimental compressive strength values are plotted against the
tensile strength of the respective grout. Taking into account the differences from one series of tests to
the others (in geometry, quality of materials used, construction details, etc.), the quite linear
correlation between fwc,s and fgr,t may be considered as satisfactory. This result (a) explains the fact that
high compressive strength grouts are not as efficient as it might be expected from the mechanical point
of view and, more important, (b) it proves that ternary and hydraulic lime based grouts, that are
expected to satisfy durability requirements, may also enhance the mechanical properties of three-leaf
masonry to a level that complies with the bearing capacity requirements set for historic structures.
(e) The experimental results show that, in general, the stiffness of the grouted masonry (expressed by
means of the modulus of elasticity) is higher than the stiffness of the ungrouted wallettes. Although
the experimental results are quite scattered in this respect, the vast majority of Ewc,s/Ewc,0 values do not
exceed 1.60. Furthermore, the data of Table 2 show that strength enhancement is higher than stiffness
enhancement. Thus, the ratio between the modulus of elasticity and the compressive strength of
masonry after grouting is, as a rule, smaller than the respective ratio before grouting.
(f) Another important finding is illustrated in Figure 8: When a medium to low strength grout is used,
i.e. a grout with reduced cement content or a hydraulic lime based grout, the strain corresponding to
the compressive strength of the grouted masonry is larger than that of the ungrouted masonry. The
opposite occurs in case of cement-based grouts. It seems, therefore, that grouting with cement-based
mixes leads to more brittle behaviour of masonry, whereas ternary and hydraulic lime based grouts
allow masonry to sustain larger compressive strains before its maximum resistance is reached. This is
another sign in favour of the use of low to medium strength grouts.
Mechanical properties of grouted masonry in diagonal tension or in shear

The effect of grouting on the shear strength of three leaf masonry was measured either by testing walls
under vertical load and horizontal force (Tomazevic et al, 1982 and 1993) or by testing wallettes in
diagonal compression. As shown in Table 2, although the available results are rather limited in
number, it is evident that grouting leads to substantial strength enhancement. In this case as well, it
may be observed that the strength increase is definitely not proportional to the compressive strength of
the grout. The significant increase of the stress at failure obtained with medium to low strength grouts
suggests that, in this case too, the decisive parameter is the bond between grout and in situ materials.
Tomazevic et al. (1982) reached similar results through testing of walls in shear (under simultaneous
vertical load). The masonry was constructed in various ways to simulate those of historic structures in
Slovenia. The walls were grouted with a cement grout. The authors found in most cases an increase of
shear resistance varying between 110% and 250%, associated with a rather slight decrease of the
horizontal displacement at maximum shear resistance (between 10% and 30%). It should be noted that
the walls, subjected to cyclic horizontal forces exhibited a quite stable behaviour, without significant
force response degradation for imposed angular distortions up to 1%.
It has to be noted that the available experimental results regarding the shear behaviour of ungrouted
and grouted masonry or walls are rather scarce and they do not allow for reliable prediction of the
efficiency of grouting to be made.
PREDICTION OF COMPRESSIVE STRENGTH OF THREE-LEAF MASONRY BEFORE

AND AFTER GROUTING
482
Analytical modelling of three-leaf masonry walls using finite element methods was attempted by
several researchers (see i.a. Binda et al., 1991, Binda et al., 1994, Binda et al., 2006). In several cases,
the researchers were able to reproduce experimental stress-strain curves quite accurately. Nevertheless,
such valuable works, although they contribute to the understanding of the behaviour of three-leaf
masonry, did not yet yield sound engineering models that could be used in the design of interventions
to historic masonries by means of grouts. Thus, there is still a need for simple physical models that
would allow for the compressive strength of three-leaf masonry before and after grouting to be
predicted. Although in case of major monuments it may be possible or advisable to conduct tests, in
order to determine the mechanical properties of masonry in a reliable way, in the general case of
interventions either to less important historic structures or to buildings belonging to urban nuclei,
neither the time nor the funds are available for such an experimental campaign. Therefore, the
availability of simple (but physically sound) formulae is of great importance.
In what follows, an attempt is made to apply to the available experimental results existing formulae, as
well as to improve them, taking into account the observed behaviour, as described in the previous
sections.
Estimation of the compressive strength of three-leaf masonry before grouting

Simple available formulae allowing for the estimation of the compressive strength of grouted three-
leaf masonry, require the compressive strength of ungrouted masonry to be estimated as well. For this
purpose, Valluzzi (2004) suggests that the compressive strength of the external leaves can be measured
in situ, applying the flat jacks technique. The compressive strength of the filling material can be
measured in the laboratory, on cores taken in situ. Subsequently, an engineering model is needed, in
order to calculate the compressive strength of three-leaf masonry.
The simple model, proposed by Egermann (1993) could be used for this purpose. Nevertheless,
according to this model, two parameters need to be estimated, in order to take into account the
interaction between external leaves and filling material, as described in Figure 1. As stated by
Egermann himself, however, the actual state of knowledge does not allow for accurate estimation of
those two parameters.
Tassios (2004) derived closed formulae for the estimation of the compressive strength of ungrouted
masonry, based on the model by Egermann (1993) and taking into account the available experimental
data. Formulae by Tassios proved to predict rather satisfactorily the compressive strengths measured
by Valluzzi (2000), Toumbakari (2002) and Vintzileou et al. (1995).
Estimation of the compressive strength of three-leaf masonry after grouting

(a) In Vintzileou et al. (1995), a simple formula is developed, allowing for the calculation of the
compressive strength of three-leaf masonry after grouting. The development of the formula is based on
the following assumption: Grouting does not affect significantly the mechanical properties of the
external leaves. In reality, when masonry is cracked, grout fills the cracks, as well as voids in the
mortar and, to some extent, also pores of the materials. Nevertheless, it is on the safe side to assume
that through grouting, the initial compressive strength of the external leaves is re-instated.
On the contrary, grouting contributes to the substantial improvement of mechanical properties of the
infill. The strength enhancement of the infill is taken proportional to the square root of the
compressive strength of the grout (as an indicator of its tensile strength), as failure of masonry is due
to the opening of vertical cracks. This type of failure is governed by the tensile strength of the
materials rather than by their compressive strength. The contribution of the strengthened infill material
to the compressive strength of masonry is proportional to the ratio Vi/Vw (Vi denotes the volume of
infill material in the total volume, Vw , of the wall). Thus, the following formula was derived:
Vi f gr ,c
f wc, i f wc, 0 (1 1.25 ) (1)
Vw f wc,0
483
This formula was applied to the available experimental results presented in Table 1 (Figure 9). It
seems that, in general, the formula overestimates the compressive strength of grouted masonry.
However, if a Rd value equal to 1.35 is applied, the formula can yield safe values for the design of
grouted three-leaf masonry.
(b) Valluzzi (2004) offers a modified version of Equation (1). On the basis of her systematic testing of
cylinders made of filling material and grouted with hydraulic lime based grouts, an empirical formula
was proposed for the prediction of the compressive strength of the grouted infill material, as follows:
f i,s 0.31f gr1.18
,c (2)
Thus, the compressive strength of the compressive strength of masonry is calculated using the
following expression:
Vi fi ,s
f wc,i f wc, 0 (1 ) (3)
Vw f wc, 0
Equation (3) overestimates the compressive strength of grouted three-leaf masonry in most of cases.
Thus, to yield values adequate for the design of grouted masonry, a Rd value equal to 1.80 should be
applied to the predicted values.
(c) The available results from testing cylinders made of filling material before and after grouting, i.e.
results by Miltiadou (1990), Valluzzi (2004), Vintzileou et al., 2006, as well as DTRR (2006), were
evaluated, in order to derive an expression for the compressive strength of grouted filling material.
Taking into account (i) the observations made in previous sections and (ii) the fact that there is a clear
trend to apply only ternary or hydraulic lime based grouts to historic masonry, an effort is made to
relate the compressive strength of the grouted filling material with the tensile strength of the grout
only for those cylinders that were grouted with either ternary or hydraulic lime based grout. Thus, the
following expression was derived:
f i,s 1.60 0.50f gr , t (4)

Toumbakari, 2002
Valluzzi, 2000
4
3
3.5
2.5
Rd~1.35 3
2 2.5
1.5 2
1.5
1
1
0.5
0.5
0 0
0 0.5 1 1.5 2 2.5 3 0 0.5 1 1.5 2 2.5 3 3.5 4
(fw c,i /fwc,0)pred pred.fi,s (N/mm2)
Figure 9. Comparison between predicted Figure 10. Comparison between experimental

and experimental strength enhancement, and predicted values of compressive strength of
based on Equation (1) grouted filling material (equation (4)).
where,
f gr,t denotes the tensile strength of the grout.
The comparison of predicted fi,s (using equ. (2)) and experimental values is shown in Figure 11. It
seems that the predicted values fit quite accurately with the experimental ones.
484
Thus, based on the assumption adopted in Vintzileou et al. (1995), equation (3) can be applied, with fi,s
taken from equation (4). The application of this alternative expression to the available experimental
results is shown in Figure 12. This expression seems to offer a rather conservative estimation of the
compressive strength of grouted masonry that can be used in design.
Toumbakari, 2002
Valluzzi, 2000
2.5
1.5
0.5
0
0 0.5 1 1.5 2 2.5
(fwc,i/fwc,0)pred.
Figure 11. Comparison between predicted and experimental strength enhancement, based on
Equation (4). The compressive strength of grouted filling material is calculated from Equation
(2)
CONCLUSIONS
The survey of literature presented in this paper allows for the following conclusions to be
drawn:
1. Three-leaf masonry subjected to compression exhibited in all cases the same behaviour and the
same failure mode. Although there is a significant inherent scatter of the experimental data, the
effect of parameters, such as mechanical properties of constituent materials, strength of filling
material, etc., on the compressive strength of ungrouted masonry was detected and explained.
2. The positive effect of grouting on the compressive strength of three-leaf masonry, as well as on its
overall behaviour was proven by the totality of experimental results. Furthermore,
3. The preponderance of ternary and hydraulic lime based grouts was demonstrated, as they offer
significant enhancement of compressive strength, associated with substantial increase of the strain
at strength. On the contrary, cement based grouts (of equal injectability and significantly higher
compressive strength) do not contribute to further increase of the compressive strength of masonry,
whereas they lead to a rather brittle behaviour.
4. The advantages of ternary and hydraulic lime based grouts (due to their improved bond properties
with the in situ materials) become more important due to the durability ensured by the use of
materials that are compatible with the existing ones from the physical-chemical point of view.
5. Simple empirical formulae, based, however, on the Mechanics of three-leaf masonry were proven
adequate for the prediction of the compressive strength of grouted masonry. It should be noted,
however, that further research is needed towards the development of physical models able to
describe the behaviour of three-leaf masonry before and after grouting.
REFERENCES
485
Vintzileou, E. and Tassios, T.P. 1995. Three-leaf stone masonry strengthened by injecting cement
grouts. Journal of Structural Engineering, Vol.121, No.5, May, pp. 848-856.
Toumbakari E.-E. 2002. Lime-pozzolan-cement grouts and their structural effects on composite
masonry walls. Doctor Thesis, Katholieke Universiteit Leuven.
Valluzzi M.-R. 2000. Comportamento meccanico di murature storiche consolidate con materiali e
tecniche a base di calce. Doctor Thesis, University of Trieste.
Vintzileou E., Miltiadou-Fezans A. 2006. Mechanical properties of three-leaf stone masonry grouted
with ternary or hydraulic lime based grouts. Submitted to Engineering Structures.
Egermann R. 1993. Investigation on the load bearing behavior of multiple leaf masonry. Proceedings
IABSE Symposium Structural Preservation of the Architectural Heritage, Rome, pp. 305-312.
Binda L., Pina-Henriques J., Anzani A., Fontana A., Lourenco P.B. 2006. A contribution for the
understanding of load-transfer mechanisms in multi-leaf masonry walls : Testing and modelling.
Elsevier, Engineering Structures, 28, pp. 1132-1147
Adami C.-E., Vintzileou E., 2007. Interventions to historic masonries: Investigation of the bond
mechanism between stones or bricks and grouts. Accepted for publication in Materials and
Structures, RILEM.
Tomazevic M., Sheppard P. 1982. The strengthening of stone-masonry buildings for revitalization in
seismic regions. 7th European Conference on Earthquake Engineering, Athens, Vol. 5, pp. 275-282
Tomazevic M., Apih V. 1993. The strengthening of stone-masonry walls by injecting the masonry-
friendly grouts. European Earthquake Engineering, 1, pp. 10-20
Binda L., Fontana A., Anti L. 1991. Load transfer in multiple-leaf masonry walls. Proceedings, 9th
International Brick Block Masonry Conference, pp. 1488-1497.
Binda L., Fontana A., Mirabella Roberti G. 1994. Mechanical behaviour and stress distribution in
multiple-leaf walls. Proceedings, 10th International Brick Block Masonry Conference, pp. 51-59.
Valluzzi M.-R. 2004. Consolidamento di murature in pietra. Iniezioni di calce idraulica naturale.
Collana Scientifica REFICERE, Gruppo Editoriale Faenza Editrice S.p.a., 128pp.
Tassios,T.P. 2004. Rehabilitation of three-leaf masonry. In Evoluzione nella sperimentazione per le
costruzioni, Seminario Internazionale, 26 Sept- 3Oct., Centro Internationale di Aggiornamento
Sperimentale Scientifico (CIAS)
Valluzzi M.-R., da Porto F., Modena C. 2001. Behaviour of multi-leaf stone masonrry walls
strengthened by different intervention techniques. Proceedings 3rd International Seminar on
Structural Analysis of Historical Constructions, Lourenco P.B., Roca P. (ed.), pp. 1023-1032.
486
Table 1. Geometrical and mechanical properties of wallettes and main experimental findings
Speci- Dimensions te/ti/te fc,units fm,c fi,c fgr,c/fgr,t Vgr/Vw fwc,0 fwc,i Ew0 Ews wu,0 wu,s
2 2 2 2 2 2 2 2
men (m) (N/mm ) (N/mm ) (N/mm ) (N/mm ) (%) (N/mm ) (N/mm ) (N/mm ) (N/mm ) () ()
Vintzileou/Tassios (1995)
(1)
1 0.6x0.4x1.2 130/140/130 100 1.70 0.15 30/2.5 43.8 2.10 3.10 7000 6250 0.00104 0.0006
2 0.6x0.4x1.2 130/140/130 100 1.70 0.15 - - 1.30 - 2706 - - -
(1)
3 0.6x0.4x1.2 130/140/130 100 1.70 0.15 30/2.5 67.0 2.40 4.30 5000 5971 0.00141 0.0011
4 0.6x0.4x1.2 130/140/130 100 1.70 0.15 30/2.5 31.75 1.60 - 4442 - 0.0008 -
5 0.6x0.4x1.2 130/140/130 100 1.70 0.15 30/2.5 34.72 1.70 4.20 5670 7778 0.0028 0.0012
6 0.6x0.4x1.2 130/140/130 100 1.70 0.15 13/1.4 29.8 1.35 4.05 5625 8438 0.00058 0.0010
7 0.6x0.4x1.2 130/140/130 100 1.70 0.15 30/2.5 29.8 - 3.70 - 15413 - 0.0009
8 0.6x0.4x1.2 130/140/130 100 1.70 0.15 13/1.4 29.8 - 3.00 - 3333 - 0.0009
Toumbakari (2002)
BC1 0.6x0.4x1.2 90/220/90 25 3.40 1.00 7.3/1.7 31.0 - 5.04 - 2238.2 - 0.00242
BC2 0.6x0.4x1.2 90/220/90 25 3.40 1.00 9.0/1.1 23.0 2.41 3.15 729.6 1564.9 0.00312 0.00254
BC3 0.6x0.4x1.2 90/220/90 25 3.40 1.00 19.5/4.5 19.0 2.09 2.91 1018.3 1404.8 0.00185 0.00155
BC4 0.6x0.4x1.2 90/220/90 25 3.40 1.00 7.3/1.7 25.0 2.18 3.00 1097.6 1040.4 0.00234 0.00294
(1)
BC5 0.6x0.4x1.2 90/220/90 25 3.40 1.00 7.3/1.7 2.28 3.86 1144.9 1170.2 0.00231 0.00314
SC1 0.6x0.4x1.2 130/140/130 55 3.40 1.00 9.0/1.1 50.0 2.02 3.25 720.4 1622.2 0.00142 0.00355
SC2 0.6x0.4x1.2 130/140/130 55 3.40 1.00 19.5/4.5 46.0 2.09 3.36 1138.7 1558.6 0.00165 0.00233
SC3 0.6x0.4x1.2 130/140/130 55 3.40 1.00 7.3/1.7 2.65 3.51 1374.8 1187.6 0.00173 0.00245
SC4 0.6x0.4x1.2 130/140/130 55 3.40 1.00 7.3/1.7 30.0 2.71 3.29 1443.3 1014.5 0.00211 0.00349
Valluzzi, da Porto, Modena (2001)
5I1 0.8x0.5x1.4 180/140/180 164 1.57 5.1/0.8 54.2 1.45 2.49 2390 2273 0.00363 0.00726
6I1 0.8x0.5x1.4 180/140/180 164 1.57 5.1/0.8 54.2 1.95 2.49 2029 3093 0.00457 0.00571
13I1 0.8x0.5x1.4 180/140/180 164 1.57 5.1/0.8 41.5 - 2.54 - 3992 - 0.00991
1I2 0.8x0.5x1.4 180/140/180 164 1.57 3.2/0.4 54.2 1.97 2.57 1450 3449 0.0062 0.00625
8I2 0.8x0.5x1.4 180/140/180 164 1.57 3.2/0.4 47.8 1.91 1.82 1559 2367 0.00622 0.0072
16I2 0.8x0.5x1.4 180/140/180 164 1.57 3.2/0.4 47.8 - 2.48 - 1223 - 0.0107
12I1T 0.8x0.5x1.4 180/140/180 164 1.57 5.1/0.8 54.2 - 2.59 - 1336 - 0.00818
14I1R 0.8x0.5x1.4 180/140/180 164 1.57 5.1/0.8 60.6 - 2.14 - 1617 - 0.00821
17I1RT 0.8x0.5x1.4 180/140/180 164 1.57 5.1/0.8 47.8 - 3.06 - 1772 - 0.00824
Vintzileou, Miltiadou-Fezans (2006)
1 1.0x0.45x1.2 190/120/140 25 4.35 0.15 4.5/2.5 32.8 1.82 3.00 1000 1200 -
2 1.0x0.45x1.2 190/120/140 25 4.35 0.15 8.16/2.2 40 1.74 3.75 1440 1550 0.0016 0.0025
3 1.0x0.45x1.2 190/120/140 25 4.35 0.15 4.5/2.5 36.4 2.26 3.73 1500 1300 0.0025 0.0039
487
Table 1 (contd.)
Wallette wvert wtrans wtrans, wtrans,fwci Ew0/ fwc,0 Ews/ fwc,s fwc,i/ fwc,0 Ews/ Ew0 wu,s/
(mm) (mm) fwc,0 (mm) wu,0
(mm)
Vintzileou/Tassios (1995)
1 1.00 0.231 0.00 0.00 3333 1048 1.48 0.89 0.58
2 0.27 - - 2081
3 3.00 2083 1389 1.80 1.19 0.78
4 2.533 - - 2776
5 1.50 7.59 0.00 1.035 3335 1852 2.47 1.37 0.43
6 1.81 0.00 1.325 4167 2083 3.00 1.50 1.72
7 - - 1.3 4166
8 - - 0.125 1111
Toumbakari (2002)
BC1 1.17 444
BC2 7.07 0.15 5.04 302.7 497 1.31 2.14 0.81
BC3 2.69 0.86 2.59 487.2 483 1.39 1.38 0.84
BC4 2.32 0.38 4.03 503.5 347 1.38 0.95 1.26
BC5 2.00 0.03 1.80 502.1 303 1.69 1.02 1.36
SC1 2.89 0.03 2.19 356.6 499 1.61 2.25 2.50
SC2 6.9 0.01 0.60 544.8 464 1.61 1.37 1.41
SC3 3.63 0.51 2.33 518.8 338 1.32 0.86 1.42
SC4 2.84 1.01 2.98 532.6 308 1.21 0.70 1.65
Valluzzi, da Porto, Modena (2001)
5I1 2.25 0.002 4.59 1648 913 1.72 0.95 2.00
6I1 9.2 0.0015 8.55 1040 1242 1.28 1.52 1.25
13I1 0.06 13.85 1572
1I2 3.97 0.16 3.67 736 1342 1.30 2.37 1.00
8I2 5.9 0.04 4.95 816 1300 0.95 1.52 1.16
16I2 0.11 9.45 493
12I1T 0.02 7.95 516
14I1R 0.01 9.8 756
17I1RT 0.0005 10.95 579
Vintzileou, Miltiadou-Fezans (2006)
1 0.78 8.2 0.0 3.2 550 400 1.65 1.20 -
2 0.26 4.8 0.0 1.3 828 413 2.16 1.08 1.56
3 1.55 4.0 0.0 2.6 655 349 1.65 0.87 1.36
Table 2. Summary of experimental results: Wallettes subjected to diagonal compression

Mechanical properties of Mechanical Mechanical properties of
ungrouted specimens properties of grouted specimens
grout
Speci- Dimensions ft v whor fc ft,fl ft,f v,f whor,ft f
men (m) (MPa) () (mm) (MPa) (MPa) (MPa) () (mm)
Toumbakari (2002)
SDC1 0.8x0.8x0.4 0.47 0. 9 0.85 15.9 3.4 0.50 1.36 0.47 1.1
SDC2 0.8x0.8x0.4 0.34 1.2 0.71 19.5 4.5 0.68 1.17 0.05 2.0
SDC3 0.8x0.8x0.4 0.28 1.5 1.20 7.8 2.5 0.59 2.65 0.02 2.1
BDC1 0.8x0.8x0.4 0.44 0.8 0.50 7.8 2.5 0.60 1.56 0.17 1.4
BDC2 0.8x0.8x0.4 0.34 0.3 1.07 15.9 3.4 0.73 1.07 0.02 2.2
BDC3 0.8x0.8x0.4 0.35 0.6 0.50 19.5 4.5 0.75 1.02 0.02 2.1
Vintzileou et al. 1995
9 (0.15) 30 0.64
10 (0.15) 30 0.64
Vintzileou et al., Miltiadou et al. (2006)
4 1.0x1.0x0.45 0.10 2.1 5.2 4.5 2.52 0.22 1.00 0 2.2
5 1.0x1.0x0.45 0.10 1.1 0.2 8.2 2.29 0.34 0.80 0 3.4
6 1.0x1.0x0.45 0.10 1.0 1.5 4.5 2.52 0.23 1.40 0 2.3
ft and v: tensile strength and respective vertical strain of ungrouted specimen, fc and ft,fl: compressive and
flexural strength of grout, ft,f and v,f: tensile strength and respective vertical strain of grouted specimen,
whor : sum of vertical crack openings on the face of the ungrouted specimen (at ft), whor,ft : sum of vertical
crack openings on the face of the grouted specimen (at ft ), f=ft,f/ft
488
WORKSHOP 4
How can earthquake geotechnical
engineering contribute to safer design of
structures to resist earthquakes?
June 25-28, 2007
Paper No. W4-1001
KEY ISSUES IN THE ANALYSIS OF PILES IN LIQUEFYING SOILS
Misko CUBRINOVSKI 1, Hayden BOWEN1
ABSTRACT
Two methods for analysis of piles in liquefying soils are discussed in this paper: an advanced method
for dynamic analysis based on the effective stress principle and a simplified analysis based on the
pseudo-static approach. The former method aims at an accurate simulation of the complex liquefaction
process and soil-pile interaction while the latter is a design-oriented approach that uses conventional
engineering parameters and modelling for estimation of the maximum pile response. This paper
discusses some key issues in the implementation of these analysis methods with reference to the
assumptions used in modelling the soil-pile interaction in liquefying soils.
Keywords: Effective stress analysis, lateral spreading, liquefaction, pile, pseudo-static analysis
INTRODUCTION
Behaviour of piles in liquefying soils is very complex involving rapid change in stiffness and strength
of soils, large ground deformation and significant inertial loads from the superstructure. A rigorous
analysis simulating this process in detail, such as the seismic effective stress analysis, imposes high
demands on the user in terms of required input data and understanding of the adopted numerical
procedures. On the other hand, when using design-oriented analyses based on the pseudo-static
approach, one encounters great difficulties in the selection of appropriate values for the parameters of
the simplified analysis because nearly all parameters used in the simplified model are subject to large
variation in the course of the pore pressure build-up and eventual liquefaction. Thus, the key issues in
the analysis of piles are quite different for the advanced seismic effective stress analysis and simplified
pseudo-static analysis. This paper addresses some of the key issues in the application of these two
methods of analysis to piles in liquefying soils.
SOIL PILE INTERACTION IN LIQUEFYING SOILS
Soil-pile interaction in liquefying soils is a very intense dynamic process that involves significant
changes in the soil characteristics and loads on pile over a short period of time during and immediately
after the strong ground shaking. Some typical features of the ground response and loads on piles in
liquefying soils are illustrated in Figure 1. During the intense ground shaking in loose saturated sandy
deposits, the excess pore water pressure rapidly builds up until eventually it reaches the level of the
effective overburden stress v' and the soil liquefies. In the example shown in Figure 1a from the 1995
Kobe earthquake, the excess pore pressure reached the maximum level after only 6-7 seconds of
intense shaking, and this was practically the time over which the soil stiffness reduced from its initial
value to nearly zero. The intense reduction in stiffness and strength of the soil was accompanied with
equally rapid increase in the ground deformation, as illustrated with the solid line in Figure 1b where
horizontal ground displacements within the liquefied layer are shown. Note the cyclic nature and
relatively large amplitude of these displacements. The peak displacement of about 40 cm occurred just
1
Department of Civil Engineering, University of Canterbury, New Zealand, Email:
misko.cubrinovski@canterbury.ac.nz
489
before or at the time of development of complete liquefaction in the layer and was accompanied with
high ground accelerations of about 0.4g. During this phase of intense ground shaking and development
of liquefaction, the piles were subjected to kinematic loads due to ground movement and inertial loads
due to vibrations of the superstructure. Both these loads are oscillatory in nature with magnitudes and
spatial distribution dependent on a number of factors including ground motion characteristics, soil
density, presence of non-liquefied crust layer at the ground surface, and predominant periods of the
ground and superstructure, among others.
In sloping ground or backfills behind waterfront structures the liquefaction may result in unilateral
ground displacements due to lateral spreading, as indicated with the dashed line in Figure 1b. Lateral
CYCLIC PHASE SPREADING PHASE

Excess pore water pressure
200
150 v'
(kPa)
100
50
0 (a)
0 5 10 15 20
Horizontal displacement
90 Relative ground displacements

60 between GL-0m and GL-16m
30 Spreading displacement
(cm)
0
-30
Cyclic displacement (b)
-60
0 5 10 15 20
Time (seconds)
(c)
Figure 1. Illustration of ground response in liquefying soils and effects on piles: (a) Excess pore
water pressure; (b) Lateral ground displacement; (c) loads on pile during the cyclic phase and
lateral spreading
490
spreads typically result in large permanent displacements of up to several meters in the down-slope
direction or towards waterways. Provided that driving shear stresses exist in the ground, lateral
spreading may be initiated during the intense pore pressure build up, at the onset of liquefaction or
after the development of complete liquefaction. As compared to the cyclic phase of the response,
ground displacements are approximately one order of magnitude bigger and inertial loads are
relatively smaller during the lateral spreading, as illustrated schematically in Figure 1c. These large
differences in liquefaction characteristics and loads on the pile between the cyclic phase and lateral
spreading phase need to be accounted for in the analysis of piles.
SEISMIC EFFECTIVE STRESS ANALYSIS
The effective stress analysis permits evaluation of seismic soil-pile interaction while considering the
effects of excess pore water pressure and highly nonlinear stress-strain behaviour of soils in a rigorous
dynamic analysis. This method basically aims at very detailed modelling of the complex liquefaction
process through the use of advanced numerical procedures. In principle, the effective stress analysis
permits simulation of the entire process of pore pressure development, onset of liquefaction and post-
liquefaction behaviour including associated ground deformation and loads on piles during both cyclic
phase and lateral spreading phase of the response. For these reasons, the seismic analysis based on the
effective stress principle has been established as one of the primary tools for analysis of liquefaction
problems. Over the past three decades, the application of this analysis gradually has expanded from 1-
D analyses of a level ground to more complex 2-D analyses involving earth structures and soil-
structure interaction problems. Recently, attempts have been made to apply this method to a three-
dimensional analysis of large-deformation problems, as described below.
A comprehensive study on pile foundations in liquefiable soils has been recently carried out in Japan,
with a principal objective to investigate the behaviour of piles in liquefying soils undergoing lateral
spreading, from both experimental and numerical viewpoints. Within this project, a series of shake-
table experiments on piles in liquefying soils undergoing lateral spreading was conducted at the Public
Works Research Institute (PWRI), Tsukuba, Japan (Tanimoto et al., 2003). For all experiments, Class-
B predictions were made using two different constitutive models and numerical procedures for 3-D
effective stress analysis. The principal objective of the numerical simulations was to assess the
accuracy and capability of advanced 3-D effective stress analyses in predicting liquefaction-induced
lateral flow and pile group behaviour. Note that both methods of analysis have been extensively
verified and have shown very good performance in simulations of well-documented case histories,
seismic centrifuge tests and large-scale shake table tests of liquefaction problems (e.g., Cubrinovski et
al., 1999; 2001; Uzuoka et al. 2002; 2007).
Various factors were varied in the aforementioned shake table experiments such as the amplitude and
direction of shaking (transverse, longitudinal and vertical), mass of the superstructure and number and
arrangement of piles. A typical physical model used in these tests is shown in Figure 2 consisting of a
3x3 pile foundation embedded in a liquefiable sand deposit, located in the vicinity of a sheet pile wall.
By and large, the numerical predictions were in good agreement with the observations in the
experiment capturing the rapid pore pressure build-up, development of liquefaction and subsequent
ground flow around the foundation. In fact, the response of the foundation piles were very well
predicted by both methods for all experiments, as indicated in Figure 3a where computed and
measured horizontal displacements at the pile head are shown for different tests. As indicated in
Figure 3b, however, the analyses underestimated the displacement of the sheet pile wall. It was found
that the prediction of the large lateral movement of the sheet pile wall including instability in the
backfills and foundation soils was the most difficult to accurately predict with the advanced seismic
analyses.
A number of issues are important when conducting a seismic effective stress analysis as above, but
probably the most critical one is the performance of the constitutive soil model, both in terms of its
modelling capability and proper implementation by the user. It is essential that the constitutive model
491
a) Model: H14-1, H15-3 & H16-2
SHEET FOOTING
PILE
Coarse sand 40
PILE 90cm
Loose T.S.
Loose (35%)
Toyoura sand
(Dr = 35%) Dense T.S. (90%) 50
150 75 175
Figure 2. Schematic plot of physical model used in the shake table test
(a) Footing (pile head) (b) Sheet-pile wall

16-3
Computed displacement (mm)
Computed displacement (cm)
30 Diana-J 60 Diana-J
Liqca Liqca
16-2
14-1
20 40 (2-D)
16-2 Computed = Measured Computed = Measured
15-3 16-2 15-2 16-1

14-3
10
14-1 20 16-3
14-1 (2-D) 15-1
15-2 15-1 14-1 15-3 16-2

16-1 14-3 14-2
0 10 20 30 00 20 40 60
Measured displacement (mm) Measured displacement (cm)
Figure 3. Comparison between computed and measured displacements of footing and sheet pile
provides reasonably good accuracy in predicting the excess pore pressures and ground deformation in
order to allow proper evaluation of the soil-pile interaction effects.
The initial stress conditions and anticipated ground deformation pattern are equally important for a
correct prediction of pile behaviour. In the above study, for example, detailed initial stress analyses
were conducted in order to identify the initial stress in the soil deposit before the application of
shaking. Specific boundary conditions and soil-pile interfaces were also defined in order to allow
development of large displacements and deformation pattern associated with lateral spreading. In fact,
the latter was found to be one of the major reasons for the underestimation of sheet pile displacements
in the analyses. The aforementioned modelling issues together with inherent limitations of a particular
numerical procedure define the second critical issue in the application of the seismic effective stress
analysis. The complexities associated with current constitutive soil models and numerical procedures
probably explain why this analysis in spite of the unparalleled capability has not found yet an adequate
use in the engineering practice.
492
PSEUDO-STATIC ANALYSIS
Unlike the advanced effective stress analysis which aims at simulating the complex soil-pile
interaction as liquefaction develops through time, the pseudo-static analysis aims at estimating the
maximum pile response by using a relatively simple model and small number of engineering
parameters. The key issue in such analysis is thus the selection of appropriate values for the
parameters that represent the soil conditions and loads at the time when the maximum response of the
pile develops. As discussed in the previous sections, the liquefaction characteristics and lateral loads
on piles vary significantly in the course of the development of liquefaction and are quite different
between the cyclic phase and subsequent lateral spreading phase of the response. For this reason, the
cyclic phase and lateral spreading phase have to be treated separately in the pseudo-static analysis.
A conventional beam-spring model allowing for stiffness and strength reduction in the liquefied layer,
large movement of the liquefied soil, lateral soil pressure from the crust layer and lateral load on the
pile due to inertial effects is shown in Figure 4. While this model generally permits modelling of
multiple layers with different load-deformation spring properties, in essence it distinguishes between
three distinct layers: non-liquefiable crust layer at the ground surface, liquefiable layer and non-
liquefiable base layer. As indicated in Figure 4, degraded stiffness and strength are used for the
liquefied soil, and it is assumed that the crust layer moves as a rigid body on top of the liquefied soil.
The characterization of nonlinear behaviour of the soil and pile shown in Figure 5 is probably the
simplest one that allows adequate treatment of nonlinear behaviour of the soil and pile. Here, the
bilinear p- relationships for the soil are defined by an initial stiffness using the conventional subgrade
reaction approach and by an ultimate lateral soil pressure, pmax. The subgrade reaction coefficients k
can be evaluated using empirical correlations based on the elastic moduli of the soil or SPT blow
count while the ultimate lateral pressure for the non-liquefied layers can be estimated using a factored
Rankine passive pressure, as described in Cubrinovski and Ishihara (2004).
a) Cross section b) Soil-pile model c) Numerical scheme

Upper foundation
UG UP
Lateral load
Non-liquefied H1
crust layer
Ground
displacement
Liquefied Degraded
layer H2 stiffness &
strength
Pile
Deformed pile
Non-liquefied
H3
base layer
Pile
Soil spring
Figure 4. Typical FEM beam-spring model for pseudo-static analysis of piles
493
Hf Footing p
Surface layer
H1 UG2 k1 p1-max
U
M Y
C p
Liquefied 2 k2 p2-max
H2
layer
Pile
Ground
displacement
p
Do
H3 Base layer k 3 p3-max
Figure 5. Characterization of nonlinear behaviour and input parameters of the model
Key parameters in this model are the magnitude of lateral ground displacement (U G2), ultimate lateral
pressure from the crust layer (p 1-max), and stiffness and strength reduction in the liquefied layer as
represented by the stiffness degradation factor 2 and ultimate pressure p2-max respectively. Some
guidance in the selection of the ultimate pressure from the crust layer and variation in the magnitude
of lateral ground displacement can be found in Cubrinovski at al. (2007). Here, effects of the ultimate
pressure form the liquefied soil are examined somewhat in detail.
Cyclic Phase
Assuming that the peak response of the pile during the cyclic phase occurs at or before the onset of
liquefaction, which is a reasonable assumption according to observations from experiments and
analyses, then the following reasoning can be applied to the analysis of piles: (i) The value of 2
commonly varies within a relatively small range of values between 1/20 and 1/10; (ii) The magnitude
of cyclic ground displacements can be estimated reasonably well using simple procedures analogue to
those for evaluation of liquefaction triggering based on empirical SPT / CPT charts, as suggested by
Tokimatsu and Asaka (1998) for example; (iii) The relative displacement between the soil and the pile
is often less than that required for mobilizing the ultimate soil pressure from the crust and liquefied
layers; (iv) In addition to the large kinematic loads due to lateral ground movement, the piles are
subjected to significant inertial loads from the superstructure. The above basically implies that often
the particular values of 2, UG2 and p2-max are not critically important, but rather the key issue in the
analysis of piles during the cyclic phase is how to combine the kinematic loads and inertial loads on
the pile since the peak values of these oscillatory loads do not necessarily occur at the same time.
Clear and simple rules for combining the ground displacements and inertial loads from the
superstructure in the simplified pseudo-static analysis have not been established yet, though some
suggestions may be found in Tamura and Tokimatsu (2005) and Liyanapathirana and Poulos (2005).
Lateral Spreading Phase

In the lateral spreading phase, the potential variation in key parameters is much more significant and
involves: (i) Variation in 2 over relatively wide range of values between 1/1000 and 1/50; (ii) Large
uncertainty in the magnitude of lateral displacement and scatter in the estimates based on empirical
correlations for lateral spreads; (iii) Relative displacements between the soil and pile sufficiently large
to mobilize the ultimate soil pressure from the crust layer and liquefied layer; (iv) Small contribution
of inertial loads relative to the kinematic loads. Thus, in the case of lateral spreading, the values of p1-
max , 2, UG2 and p2-max involve great variation and uncertainty associated with the strength and stiffness
of liquefying soils, post-liquefaction spreading displacements and ultimate pressure from the non-
494
liquefied crust layer at the ground surface. Detailed discussion on the modelling of the crust layer and
selection of p1-max is given in a companion paper presented at this conference whereas here the
combined effects of parameters of the liquefied layer UG2, 2 and p2-max are briefly illustrated through
an application of the analysis to a case study.
STIFFNESS AND STRENGTH OF THE LIQUEFIED SOIL
The pile foundation of twin bridges crossing the Avon River in Christchurch, New Zealand, was
analysed using series of seismic effective stress and pseudo-static analyses. Various combinations of
loads and liquefaction characteristics including pile-group and soil-structure interaction effects were
considered in these analyses. The pseudo-static analyses presented in this paper are used only to
demonstrate the effects of the ultimate pressure from the liquefied layer on the pile response.
Figure 6 shows the soil profile and SPT blow count at the site including the adopted three-layer
interpretation of the deposit. The soils between 2.5 m and 17.5 m were considered liquefiable, with a
dense silty sand base layer below 17.5 m depth. The water table was estimated at 2.5 m depth thus
defining a non-liquefiable crust layer at the ground surface. The 1.2 m diameter reinforced concrete
piles are rigidly connected to a pile cap and extend from 2.5 m to 23 m depth below ground level.
The interaction in the liquefied layer can be treated in a simplified manner by an equivalent linear p-
relationship, i.e. with no limiting soil pressure. Alternatively and more rigorously, a limit can be
placed on the pressure exerted by the liquefied soil. One approach in doing this is to use the residual
strength of the liquefiable soil Su as defined by Seed and Harder (1991) using empirical correlation
with the SPT blow count, as shown in Figure 7. Since the scatter of the data is quite significant for this
correlation, it was adopted to use three different Su values in the pseudo-static analysis corresponding
to an upper bound (S u-ub), best-fit (Su-bf) and a lower bound value (Su-lb) respectively. The purpose of
this parametric study was to examine the effects of the ultimate lateral pressure from the liquefied soil
on the pile response.
Figure 8 comparatively shows the computed bending moments and pile displacements for the analyses
with different Su values. Note that an analysis using an equivalent linear p- approximation was also
Figure 6. Soil profile and SPT blow count at the investigated site
495
60
p Su-ub
50
S S
u-lb u-ub
40 Upper
bound
(kPa)
30 S
u-bf
u
S
Best-fit
20
S
u-lb
10 Lower bound
0
0 5 10 15 20
(N )
1 60cs
Figure 7. Residual strength (Su) of sandy soils back-calculated from case histories (after Seed
and Harder, 1990)
0 0
5 5
Mu Eq. Linear
Depth (m)
Depth (m)
10 10
My p Su-ub
Su-bf
Mc Su-lb
15 15
Su-lb
Su-bf Applied ground
20 Su-ub 20 displacement
Mu Eq. Linear
My
Mc
25 25
-10 -5 0 5 10 0 0.2 0.4 0.6 0.8 1
Bending moment (MN-m) Displacement (m)
Figure 8. Effects of ultimate pressure from the liquefied soil on the pile response
conducted for comparison purpose. By and large, the pile response decreases with decreasing ultimate
pressure from the liquefied layer.
Figure 9 shows the relative displacement between the soil and the pile plotted together with the
reference yield displacement of the soil, for the three analysis cases with different Su values. Here, the
shaded areas denote zones over which yielding occurs in the soil. In other words, throughout these
depths, the ultimate soil pressure is applied to the pile. Note that the resultant load on the pile increases
with increasing ultimate lateral pressure from the soil, with the largest response being observed when
Su-ub was used as a limiting lateral soil pressure. The equivalent linear analysis overestimates the pile
response and is not applicable in this case. The relative contributions of the crust and liquefied layers
to the total pile also change with the value of Su . As the value of Su decreases the role of the liquefied
layer diminishes.
496
0 0 0
Su-lb S S
u-bf u-ub
5 5 5
Depth (m)
Depth (m)
Depth (m)
10 10 10
15 15 15
Relative
displacement,
20 20 20
Yield y y
displacement, y
25 25 25
-0.2 0 0.2 0.4 0.6 0.8 1 -0.2 0 0.2 0.4 0.6 0.8 1 -0.2 0 0.2 0.4 0.6 0.8 1
Displacement (m) Displacement (m) Displacement (m)
Figure 9. Effects of ultimate pressure from the liquefied soil on soil yielding
The reasoning behind placing limits upon the ultimate pressure exerted from the liquefied soil is to
avoid unrealistic loads being imposed in situations with very large ground displacements. Figure 10
shows how the bending moment of the considered pile varies with different levels of ground
displacement, for an assumed value of 2 = 1/50. It can be seen that when limits are placed on the
ultimate pressure, different levels of ground displacement yield virtually the same response. This
clearly demonstrates that for any given set of values for 2 and p2-max there exists a threshold
magnitude of lateral ground displacement above which the pile response is practically unaffected by
the magnitude of ground displacement. Hence, for many analysis cases, a specific determination of the
magnitude of lateral spreading displacement may not be needed.
CONCLUDING REMARKS
The effective stress analysis aims at detailed modelling of the complex soil-pile interaction in
liquefying soils, and hence, this analysis procedure is quite complex and burdened by the large number
of parameters and expertise needed for its execution. Two critical issues in the seismic effective stress
analysis are: (i) the performance of the constitutive soil model, both in terms of its modelling
capability and proper implementation by the user, and (ii) details of numerical modelling including
initial stress state, boundary conditions and possible effects of the adopted numerical procedures.
Bending moment (MN-m)
Ground
15
displacement
0.5m
10 1m
2m
3m
5
Eq. Linear
Su-ub
Su-bf
Su-lb
Figure 10. Effects of ultimate pressure from the liquefied soil on soil yielding
497
The pseudo-static analysis is a simple design-oriented approach that uses conventional engineering
parameters and modelling for estimation of the maximum pile response. Liquefaction characteristics
and lateral loads on piles are quite different between the cyclic phase and subsequent lateral spreading
phase of the response, and therefore, these two phases have to be treated separately in the pseudo-
static analysis.
When evaluating the pile response during the cyclic phase, the key issue in the pseudo-static analysis
is how to combine the kinematic loads due to ground movement and inertial loads from the
superstructure. For the lateral spreading phase on the other hand, the combined effects of parameters
of the crust layer and liquefied layer p1-max, UG2 , 2 and p2-max are important. It is important to note,
however, that for an assumed stiffness and strength of the liquefied soil (set of values for 2 and p2-
max ), there is a threshold lateral ground displacement UG2 above which the pile response is practically
unaffected by the magnitude of ground displacement. This simplifies the use of the pseudo-static
analysis and eliminates the need for an accurate estimate of the magnitude of lateral spreading
displacement.
REFERENCES
Cubrinovski, M., Ishihara, K. and Furukawazono, K. (1999). Analysis of full-scale tests on piles in
deposits subjected to liquefaction, Proc. 2nd International Conference on Earthquake Geotechnical
Engineering, Lisbon, Vol. 2, 567-572.
Cubrinovski, M., Ishihara, K. and Kijima, T. (2001). Effects of liquefaction on seismic response of a
storage tank on pile foundations, Proc. 4th Int. Conf. on Recent Advances in Geotechnical
Earthquake Engineering and Soil Dynamics, San Diego, CD-ROM, 6.15.
Cubrinovski, M. and Ishihara, K. (2004). Simplified method for analysis of piles undergoing lateral
spreading in liquefied soils, Soils and Foundations, Vol. 44, No. 25: 119-133.
Cubrinovski, M., Ishihara, K. and Poulos, H. (2007). Pseudostatic analysis of piles subjected to lateral
spreading, Special Issue Bulletin of NZ Society for Earthquake Engineering (to be published).
Liyanapathirana, D.S. and Poulos, H.G. (2005). Pseudostatic approach for seismic analysis of piles in
liquefying soil. ASCE Journal of Geotechnical and Geoenvironmental Engineering, Vol. 131, No.
12: 1480-1487.
Seed, R.B. and Harder, L.F. (1991). SPT-based analysis of cyclic pore pressure generation and
undrained residual strength, H. Bolton Seed Memorial Symposium Proc., Vol. 2: 351-376.
Tamura, S. and Tokimatsu, K. (2005). Seismic earth pressure acting on embedded footing based on
large-scale shaking table tests, ASCE Geotechnical Special Publication 145: 83-96.
Tanimoto, S., Tamura, K. and Okamura, M. 2003. Shaking table tests on earth pressures on a pile
group due to liquefaction-induced ground flow. Journal of Earthquake Engineering, Vol. 27, Paper
No. 339 (in Japanese).
Tokimatsu, K. and Asaka, Y. (1998). Effects of liquefaction-induced ground displacements on pile
performance in the 1995 Hyogoken-Nambu earthquake, Special Issue of Soils and Foundations,
September 1998: 163-177.
Uzuoka, R., Sento, N., Yashima, A. and Zhang, F. (2002). Three-dimensional effective stress analysis
of a damaged group-pile foundation adjacent to a quay wall. Journal of the Japan Association for
Earthquake Engineering, 2(2): 1-14 (in Japanese).
Uzuoka, R., Sento, Kazama, M., N., Yashima, A., Zhang, F. and Oka, F. (2007). Three-dimensional
numerical simulation of earthquake damage to group-piles in liquefied ground. Soil Dynamics and
Earthquake Engineering, 27(5): 395-413.
498
June 25-28, 2007
Paper No.W4-1004
STATE OF ART KNOWLEDGE VS. STATE OF PRACTICE IN

SEISMIC RISK MITIGATION THE ITALIAN EXPERIENCE AFTER
THE 2002 S. GIULIANO EARTHQUAKE
Mauro DOLCE 1, Giacomo DI PASQUALE2, Agostino GORETTI3
ABSTRACT
Generally speaking the state of the art knowledge needs a long way to go, in order to be translated into
good practice and to achieve the final general goal of seismic risk mitigation. Main steps are:
Transferring SoA knowledge into updated seismic zonation and national norms;
Transferring SoA knowledge to engineers, professionals and operators involved in the design
and construction process and implementing SoA knowledge into reliable software tools;
Re-addressing and funding research activities to get products which can immediately and
robustly exploited in the seismic mitigation process;
Allocating adequate funds for seismic risk mitigation, to make seismic safety assessment and
rehabilitation of existing constructions.
After the 2002 S.Giuliano Earthquake a process of this type has been started by the National Civil
Protection in Italy. In this paper the main steps of this process, the problems encountered, the current
situation and the further needs are described.
Keywords: Seismic risk, vulnerability, seismic code, structural design, retofit
INTRODUCTION
The occurrence of an earthquake is so sudden and the impact so devastating that seismic risk can not
be reduced by relying only on preparedness, response and recovery. Public authorities in seismic-
prone countries are typically responsible for protecting public safety and coordinating emergency
response and reconstruction. Due to the massive financial investments needed, the long term benefits
and the nation-wide social protection aims, public authorities are also responsible for assessing and
mitigating public earthquake risk. However, in absence of any seismic emergency, resources are not
usually allocated to implement policies and strategies for seismic risk mitigation. This is the main
reason why significant innovation and action programs for seismic risk reduction are usually
introduced just after destructive earthquakes, when society is devastated and the risk is highly
perceived.
Though having developed a significant experience in earthquake recovery, dating back several
centuries (Goretti & Di Pasquale, 2002), Italy did not go beyond the above described approach on
seismic risk mitigation. For instance, significant improvements in seismic code and seismic zonation
have been implemented only after destructive earthquakes, while direct activities for the reduction of
seismic vulnerability of buildings have been implemented almost exclusively in the areas affected by
recent strong earthquakes.
1
Director, Seismic Risk and Post-Emergency Office, Department of Civil Protection, Rome, Italy,
Email: mauro.dolce@protezionecivile.it
2
Executive, Seismic Risk and Post-Emergency Office, Department of Civil Protection, Rome, Italy.
3
Official, Seismic Risk and Post-Emergency Office, Department of Civil Protection, Rome, Italy.
499
The 2002 Molise-Puglia earthquake, when the collapse of the primary school of S.Giuliano di Puglia
killed 27 pupils, had a strong impact on public opinion and pushed public authorities to start a new and
more comprehensive action plan. The National Civil Protection was the main promoter and actor of
the changes that were deemed as necessary and urgent, as well as of the medium and long term
actions. In this paper such actions will be described and discussed.
SEISMIC ZONATION AND SEISMIC CODES IN ITALY
In order to understand the innovations implemented after the 2002 Molise-Puglia earthquake, an
overview of the historical developments of seismic zonation and codes is summarized in this
paragraph. Further details are reported in (Decanini et al, 2004) and (Di Pasquale et al., 1999).
Before Italy unification there were several codes enforced for post-earthquake reconstruction. In 1783
a code was issued by the King of two Sicilies after a destructive sequence in Calabria, while in 1859
another one was issued by the Pope after the Norcia earthquake. After the 1908 Messina earthquake
(83,000 victims, Ms=7.3) a seismic zonation was introduced for the first time in the unified Italy,
while a building seismic code was issued, addressing repair of damaged buildings and construction of
new buildings. A unique seismic zone, including Calabria, also stricken by the 1905 earthquake, and
part of Sicily, was considered. Shear coefficient was assumed about 0.1. After the 1915 Avezzano
earthquake (30,000 victims, Ms=7.0) Central Italy was included in the seismic zonation. An additional
zone was established: Southern Italy was classified as zone I, while Central Italy was classified as
zone II. Shear coefficients were respectively assumed equal to 0.1 and 0.07. After 1975, seismic
accelerations were considered to be linearly distributed along the height of the building. Up to that
time they were considered almost uniformly distributed. From the same date the design response
spectrum was introduced to describe the seismic action. After 1976 Friuli earthquake (989 victims,
Ms= 6.5), masonry buildings have been addressed, together with the restoration of cultural heritage,
especially churches. The zone stricken by the earthquake, in Northern Italy, was considered as seismic.
In the subsequent years a wide research project (named Progetto Finalizzato Geodinamica - PFG) was
funded in order to provide a new seismic zonation based on the analysis of all the historical data, as up
to that time only territories stricken by strong earthquakes were inserted in seismic areas. After the
1980 Irpinia earthquake (3000 victims, Ms= 6.9) the new zonation based on the hazard results of PFG
was put in force, with several decrees issued from 1980 to 1984. A moderate seismicity zone III was
also introduced, with shear coefficient 0.04. This zone was limited to the peripheral areas of the
territory affected by the 1980 earthquake (99 municipalities), but many other places all over Italy were
indicated as needing further investigations before deciding their zonation.
In the 1996 seismic code a relative displacement check was introduced, in order to limit the economic
damage associated to moderate magnitude non-destructive earthquakes. This is perhaps the only
significant modification implemented. However the Italian seismic code remained unnecessary
misleading and unclear in its basic scopes. In particular the shear coefficient assumed in the code
presupposed a behavior factor of about 5 for most of the structural types, without enforcing adequate
detailing to obtain this value. The same ductility concept was not explicitly reported in the code and it
was not infrequent that professionals believed that expected peak ground acceleration was equal to the
shear coefficient. Design guidelines to attain section, element and structural ductility were not
enforced and, hence, remained mostly unattended. Few information was given in the code about
assessing, upgrading or retrofitting existing buildings.
On the other hand, several seismic code and seismic zonation modifications were proposed in Italy in
the recent past, but they were not enforced, probably because no destructive earthquake occurred in
the meanwhile. Among them a new seismic code proposal in the 80s and a new seismic zonation in
1998. This latter included all the advances in the field of engineering seismology in the previous 20
years (the ongoing zonation dated back to 1980 Irpinia earthquake). Most of the Italian territory not
yet included in seismic areas was proposed to be inserted in seismic zone II or III, while the most
500
hazardous seismic zone I was proposed to remain almost unchanged. For instance, S. Giuliano di
Puglia would have been included in zone II (PGA=0.25g, shear coefficient 0.07), while it was not in
the 1980 zonation, thus allowing buildings to be designed accounting for only gravity loads.
In this framework of advances in research of seismic code and seismic zonation, but stagnation of
innovation into practice, Molise-Puglia 2002 earthquake occurred. Despite the moderate magnitude,
MW=5.7, the impact was severe with 30 victims, including 27 children and 1 teacher in one school
collapse, and 173 serious injuries. Immediately it was realised that public authorities had not been able
in the past to guarantee appropriate building safety levels, especially in case of strategic buildings or
buildings with relevant consequences in case of collapse, as schools. A process of revision of both
national seismic building code and national seismic zonation started, together with a national plan for
the assessment and retrofit of strategic and relevant buildings, infrastructures and cultural heritage.
In this paper the impact of the 2002 Molise-Puglia earthquake is described. Lessons learnt are then
discussed and the main programs and actions implemented to mitigate seismic risk in Italy are
reported.
IMPACT OF THE MOLISE-PUGLIA 2002 EARTHQUAKE
The 2002 Molise-Puglia earthquake received widespread attention within Italy (Galanti et al, 2004)
and abroad (Foster & Kodama, 2004) because of its devastating impact on the residents of the sparsely
populated regions of Molise and Puglia. The Molise earthquake sequence began at 11:32 on October
31, 2002, with an Mw 5.7 shock. The earthquake affected an area of about 1700 Km2 with a
population of 370,000. Overall, 91 municipalities were stricken by the earthquake, respectively 65 in
the province of Campobasso and 26 in the province of Foggia. Most of the municipalities were
agrarian-based villages and towns: 30% of them have less than 1,000 inhabitants and 60% of them
have less than 2,000 inhabitants (Galanti et al, 2004).
Figure 1. 2002 Molise-Puglia earthquake felt intensities (MCS scale, modified by Galli & Molin,
2004).
The extent of damage varied among the impacted municipalities, with most of the structural damage,
all 30 fatalities, and 61 out of 173 total serious injuries in the small village of San Giuliano di Puglia
(population 1,163). In addition to the 27 children and one teacher died in the collapsed school, two
elderly women were killed in building collapses a few blocks from the school. At 16:08 on November
1, an Mw 5.7 aftershock significantly worsened existing damage, causing additional homeless and
anxiety in the population and in the search and rescue teams. In addition to the main shocks, around
200 smaller shocks (M 2.5 to 3.5) occurred in the first 60 hours following the initial event.
501
S. Giuliano suffered damage corresponding to macroseismic intensity I=VIII-IX MCS. The
towns surrounding San Giuliano also suffered varying degrees of damage, with lower felt intensities
(I=V-VI to VII MCS), no fatalities, and relatively few injuries.
The damage pattern in the area depended mainly on the distance from the epicentre and on the specific
geological, geotechnical and morphological conditions. It was also influenced by the fact that most of
the affected villages were not classified as seismic areas, thus allowing for the construction of
economical, home-made, buildings and the use of seismically inadequate materials and detailing.
Villages dated back several centuries and were located for defensive purposes on hilltops. Many
buildings were constructed on steep slopes, with the down-slope side often more than one story greater
than the up-slope side. The historical centres of most of the affected villages were located on the firm
soils of such hilltops, as in S.Giuliano and Bonefro. More recent buildings were located on degraded
shale or soft clay deposits that amplified seismic waves. This produced different site effects within the
same settlement.
About 80% of the damaged buildings were masonry buildings (Goretti & Di Pasquale, 2004). Main
vulnerabilities were found in the scarce quality of the stone masonry walls, often made of two
unconnected layers, with irregular stonework, little engagement between elements, degraded mortar.
The timber floors and roofs, are often thrusting and not anchored to the masonry walls, and iron ties
are rarely present. The layout of the filed stones varies from regular to irregular. Dimensions of the
stones varied up to 50-60 centimetres. In cases of larger stones, small stones were placed in between.
In large portions of the walls, mortar was often lacking. In the 20th century expansion area, buildings
were originally one story high, and constructed with field or hewn stone masonry. One or two stories
were added in the 60s - 70s. Quite all available materials were used: hollow bricks, perforated bricks
with circular hollows, full bricks. New floors and roofs were usually made of R/C and hollow bricks.
These buildings suffered major damage for the presence of rigid and heavy floors placed on brittle,
little resistant, bearing walls. Peculiar of masonry buildings was also the pre-existent damage of the
oldest buildings, due to scarce maintenance and abandon.
In the most vulnerable buildings, failure modes due to the break up of masonry walls prevailed
( ! ..a). The failure mode originated in the
outer portion of the facade, with the expulsion of stones, especially when walls were made of two
adjacent vertical stone layers. When orthogonal walls were scarcely and ineffectively linked together,
the out-of-plane component of the seismic action caused vertical or almost vertical cracks in the corner
regions and, sometimes, masonry displacement, with separation of perpendicular walls and onset of
overturning mechanisms ( ! ..b). In
buildings where the out-of-plane mechanisms were not prevailing, the observed damage is often
related to inclined or X-shaped cracks, primarily due to in-plane actions ( !
..c). The in plane failure mode has been observed in stone
masonry, hollow bricks masonry, concrete blocks masonry and full bricks masonry. It has been
facilitate by some incorrect details as chimney flue or wide opening, especially at the 1st story. In the
most damaged buildings, diagonal cracks were spread all over the masonry walls with large sliding.
Immediately following the main shock, local technicians conducted initial safety inspections to
determine building usability and to assess damage. Soon after, a long lasting activity of damage and
usability inspection was started. The data collection and the usability judgment was based on the
survey form and manual reported in (Baggio et al. 2002). An estimated 23,000 requests for structural
damage inspections were received. 14% of the total number of public buildings and 23% of private
buildings were determined to be inaccessible immediately following the earthquake. Another 5%
required re-inspection before accessibility could be determined. Inaccessibility was due either to direct
damage or to risk of collapse from an adjacent structure (Goretti & Di Pasquale, 2004).
502
Figure 2. Observed failure modes in masonry buildings: a) wall separation, b) out of plane
wall overturning, c) in plane wall shear failure, d) roof collapse.
People that were unable to return in their home were accommodated in emergency shelters. The
number of shelter seekers increased from an estimated 1,100 on October 31, to more than 3,000 on
November 1 and rose to almost 12,000 by the end of the first week. Emergency shelter operations
continued for more than two months, with the last shelter closing on January 15, 2003 (Galanti et al,
2004). Ultimately, 31 tent champs were established and a surplus of beds was finally available.
Moreover a total of 535 prefabricated houses were provided, 268 of which were constructed in the
temporary village of San Giuliano, and the remainder in numerous villages throughout the damaged
region. The temporary village, including a primary school, town hall, market, pharmacy, and social
centre, was located approximately one kilometer from the damaged town, in a less vulnerable site.
Due to the consequences of the San Giuliano di Puglia school collapse, the schools in the Molise
Region received widespread attention. All the schools were closed following the earthquake for
damage inspections and seismic evaluations, and remained closed for more than two weeks. 80% of
them were found to have little or no damage (Augenti et al, 2004). Requested temporary classrooms
and requested temporary residences for homeless in each municipality are reported in (Galanti et al.,
2005). The most common features determining seismic vulnerability of school buildings are examined
in (Dolce, 2002).
LESSONS LEARNT
Due to its devastating impact, the Molise-Puglia earthquake represented a moment of positive
criticism of all the ongoing aspects of seismic risk reduction: mitigation, preparedness, response and
recovery. Several lesson were learnt. They are briefly reported in this paragraph.
In the affected area earthquakes occurred several centuries ago. In 1456 earthquake damage was
reported in many villages. Most severe effects were reported during the 1627 earthquake. In 1805
seismic intensity was at the damage onset. No destructive earthquake occurred in the area. Lesson
learnt is to beware of areas with rare events. People forget easily the consequences of seismic events,
making it difficult to develop any local seismic culture in the area.
The historical centre of S. Giuliano was built on rock, while the expansion area, where most of the
damage occurred, was built on soft clay soil. A similar event in absence of the expansion area on soft
soil would not have produced the same disaster. Lesson learnt is that site effects can play a critical role
in transforming a non-destructive earthquake in a devastating event and that an incorrect land use
contributed to the disastrous outcome. No destructive earthquake was reported in the past probably
because villages were, originally, built on rock.
503
Since 1998 it was known that the affected area should have been classified as moderate seismicity
area, according to a hazard map with wide scientific consensus. However involved national authorities
did not update the seismic zonation at that time. Moreover the Italian building code was obsolete and
unnecessary misleading and unclear. A behavior factor was implicitly assumed in the code, without
enforcing the necessary detailing. Capacity design was not addressed. It was probably the most
obsolete code among developed, seismic prone, countries. Lesson learnt is that authorities should act
according to the latest available information having reached a wide scientific consensus. There was an
considerable gap between consolidated scientific knowledge and transfer of this knowledge into
practice. Moreover, keeping the norm unchanged for a long time increased the indolence of engineers
in updating their professional skill and the difficulties in their learning and put into practice new
design principles and procedures.
Due to the obsolescence of both seismic code and seismic zonation, buildings constructed in the past
are not safe enough and are not so safe as they could have been. Lesson learnt is that there is an urgent
need of seismic assessment of the buildings and infrastructures having a crucial role in emergency or
with relevant consequences in case of collapse.
Due to the absence of any seismic awareness, local authorities were not ready to manage seismic
emergency. The relief model used by National Civil Protection was paternalistic, telling local
authorities what to do and how to do. Lesson learnt is that local authorities should not be left alone in
emergency preparedness, but central authorities should support and periodically check local authorities
preparedness.
Besides the above listed lessons learnt, other were learnt in several aspects of seismic risk reduction.
In the following, the only technical emergency management, that is building post-earthquake safety
inspection, is examined. During the safety assessment it happened sometimes that survey forms
different from the official ones were used by local authorities. If this sped up the survey, it also posed
difficulties in the harmonization and computerisation of results. In epicentral areas an exhaustive
survey turned out to be very effective. Although it required a slightly greater effort, it considerably
reduced the time needed to identify buildings and to access buildings in case the owners evacuated
their houses. During the inspection management, it turned out that management of the safety
assessment should be preferably at local than centralized level. It requires probably more people
devoted to the management, but it allows an immediate data validation and the establishment of maps
or GIS systems. Inspectors training is essential for a common basis language and hence for a reliable
judgment about building safety and for reliable data collection. The resources required, during the
emergency, to inspect building, manage inspections, computerised forms, train inspectors are
approximately in the range of 1 working day for every 2.3 inspected buildings.
Lessons learnt are that the process of homogenisation of procedures and forms should continue all
over Italy; the feasibility of an exhaustive survey should be explored in case of destructive earthquake
or in densely populated areas; new inspection management solutions, making use also of palm
computers, directly connected to GIS should be explored; during the emergency, a permanent staff
should be devoted to training; procedures have to be revised in order to increase performances.
IMPLEMENTED ACTIONS
Based on the lessons learnt after the Molise-Puglia earthquake, Several actions have been
implemented to reduce the impact of future seismic events. Most of them could be undertaken in a
very short time, by exploiting the rich scientific and technological background developed in Italy in
the seismological and earthquake engineering fields during the 80s and 90s. They can be
summarized in:
Transferring SoA knowledge into updated seismic zonation and national norms;
504
Transferring SoA knowledge to engineers, professionals and operators involved in the design
and construction process and implementing SoA knowledge into reliable software tools;
Re-addressing and funding research activities to get products which can immediately and
robustly exploited in the seismic mitigation process;
Allocating adequate funds for seismic risk mitigation, to make seismic safety assessment and
rehabilitation of existing constructions.
These actions are summarized In the following paragraphs.
Implementing a new seismic zonation

As already discussed, the area stricken by the 2002 earthquake was not considered as seismic and,
hence, constructions were not built according to the seismic code, although it was known, from
available hazard studies, that accelerations of the order of 0.2 g were expected with a return period of
475 years. This unfavorable situation was mainly originated by the existing gap between scientific
knowledge and its translation into risk mitigation tools. This gap was considered unacceptable by the
Government, who established a specific Working Group to obtain scientific advices on the possible
upgrade of both seismic zonation and seismic code (Decree of the Under-Secretary of State n. 4485 of
December 4, 2002). The proposal was prepared in a very short time and on march 20, 2003, the Prime
Minister issued the by-law n. 3274 concerning initial items on the general criteria for seismic
zonation of the national territory and on seismic code.
Usually the seismic code is updated with a decree of the Minister of Infrastructures and of the Minister
of the Interior, after the agreements of the High Council for Public Works, the National Council for
Research and the Unified State-Regions Conference. For the seismic zonation a similar procedure has
been usually adopted in the past. However, since 1998, the State provides the only general criteria for
the zonation, while Regions have to apply these criteria to precisely select the grade of seismicity of
each municipality. As these procedures take normally a long time, the Government decided a quicker
way to put into actions a new code and a new seismic zonation: the by-law procedure. The by-law is
an instrument usually adopted for emergency measures, typical of civil protection actions after
disasters. However, the title of the act clearly stresses the character of a first step to be followed by
further upgrades, planning a progressive return to the standard procedure. In more detail, the by-law n.
3274:
provides a new seismic zonation of the national territory, using the results of a previous
Working Group established in 1998, but not reducing the seismicity with respect to the
previous decrees;
gives general criteria for seismic zonation and requires that Regions, that are responsible for
seismic zonation, enforce the zonation on the basis of the general criteria, modifying the
seismicity by at most 1 grade in each municipality with respect to the new zonation;
After the promulgation of the by-law, all the 8103 Italian municipalities were considered as seismic
with different grades from 1 (highest hazard, PGA = 0.35 g) to 4 (very low hazard, PGA = 0.05g). In
the subsequent months, almost all the Regions adopted or modified the seismic zonation according to
the above items. As any sudden and enforced application of the new seismic zonation would have
caused obvious difficulties in building design, the by-law allowed the owners to freely choose the
new seismic zonation or the old one for a transition period of 18 months.
An important aspect of the seismic shaking, again emphasized by the S.Giuliano Earthquake, is related
to seismic amplification effects and then to microzonation. This also is a concern of the Regions, that
could use totally different criteria, one with respect to the other. In order to harmonize the approaches
and the utilization of microzonation for design and urban and emergency planning, DPC has enforced
a working group involving experts from all the Regions. The working group is drafting guidelines,
whose final version should be available by the end of the year.
505
Implementing a new seismic code
The Italian seismic code promulgated in 1909 after the Messina and Reggio earthquake was one of the
most advanced at that time. It kept up with the times up to about 30 years ago, when the design
response spectrum was introduced. In the subsequent years, modifications were very limited, while in
the rest of the world the concepts of limit states, ductility, and capacity design were introduced. The
obsolescence of the seismic code was not so influential as the lack of seismic provisions caused by the
absence of seismic zonation in the area affected by the earthquake. Nevertheless the Government
decided, again with the by-law 3274, to promote an upgrade also in this field, introducing a
completely new code, set up by the Civil Protection Department, in which:
the seismic action is described in terms of elastic response spectra and in terms of peak ground
acceleration corresponding to defined probability of exceedance;
the effect of soil amplification are considered to modify both the shape and the intensity of the
spectrum;
the expected performances of the structure are clearly stated;
the influence of structure characteristics, like geometry, regularity and constructive rules, on
ductility are clarified;
the influence of ductility on the design action is clarified;
rules to consider fragile and ductile components are given;
the rules of the capacity design are adopted;
new techniques as seismic isolation and energy dissipation are dealt with in the code;
linear and non linear analysis can be used;
proper attention is devoted to existing buildings, for which no assumption can be done on the
overall ductility and on the performance of fragile elements.
The new code complies with the principles of Eurocode 8, although some simplifications are
introduced. A great concern in the new code is given to existing structures, for which considerable
improvements and specific provisions for the Italian typical old buildings are provided, with respect to
the relevant EC8 draft. The operations needed to reach an appropriate knowledge of their features, to
evaluate their seismic safety and to design retrofit interventions are extensively dealt with. Attention
is paid also to the safety evaluation of monumental buildings, which are numerous all over Italy. In
this respect, specific Guidelines for the evaluation and the reduction of the seismic risk of the cultural
heritage have been prepared and approved, and will be promulgated in the near future.
A specific concern is given to seismic isolation, as a protection strategy for buildings and bridges.
Although the Italian technology in this field was one of the most developed in the world, only few
base isolated buildings existed in Italy before the enforcement of the by-law, because of the lack of
specific provisions in the old seismic code. Nowadays most new strategic and relevant buildings and
bridges are being designed with seismic isolation.
Obviously such important changes could not be enforced abruptly, so the new seismic code has been
allowed to be used on a voluntary basis for an 18 months period. During this period, engineers, state
and local officers and software developers had time to apply and familiarize with the novelties.
Furthermore in the same period, a special Committee was instituted to monitor the new code and to
study any possible upgrading. As a matter of fact the Committee received contributions from various
components of the world of constructions: research centers, professionals, software houses,
associations of material producers, associations of construction companies, Regions, offices deputed to
check the design, etc. After an intensive work, an updated version of the code has been issued several
months later with the by-law n. 3431, 2005. In September 2005, a new unified version of all the
technical codes was issued by the Ministry of Infrastructure, with the agreement of the National Civil
Protection. In this new text, which provides the general principles of structural design, the new seismic
code contained in the by-laws 3274 and 3431 is one of the technical rules that can be used to assess
safety requirements. A new 18 months period of voluntary use of these norms was once again
506
established, which has recently been extended to the end of 2007, due to the important modifications
that are being made and the need to check the internal consistency of the various parts of the norms.
Implementing professional training and software tools

The new code has been presented in many congresses and workshops since 2003, and a specific
training program has been actuated in cooperation with the professional associations of engineers and
architects. Obviously, so many important changes, after 20 years in which the code remained
substantially unchanged, caused difficulties in the application of the new code. For this reason some
tens of thousands of professionals, mainly engineers, attended short courses of typically 60 hours. A
unified format of the courses was established by the technical associations along with the Civil
Protection Department. The courses were so conceived as to provide the general concepts of modern
earthquake engineering (from structural dynamic principles to non linear structural behavior, from
limit state design to performance-based design, from structural detailing for high ductility to seismic
isolation and other modern seismic protection technologies) as well as to explain how these concepts
were implemented in the new seismic code. The courses were normally given by university teachers,
involved in earthquake engineering research. Although it cannot be thought that attending a 60-hours
course is sufficient to fully understand all the modern design and verification process implied by
modern seismic codes, a considerable step-forward has been made by many engineers in overcoming
some old design concepts as the unmovable standpoint of their professional activities. Moreover, also
the construction practice is expected to improve, due to the awareness and the better understanding of
the consequences in terms of damage of incorrect structural design and detailing. However greater
efforts should be made to involve in this process also construction companies, master builders and
workers. The interactions and exchanges of opinions with the most skilled engineers was also very
useful to improve the provisions contained in the second version of the new seismic code.
Given the nowadays design practice, it is unconceivable to fully implement new design and
verification procedures without the availability of reliable software tools. This strong requirement was
one of the reason for the postponement of the full enforcement of the new seismic code. However, due
to the supply and demand market law, no special action was needed in order to push the software
houses to implement the new rules and procedures of the seismic code in their structural analysis and
design software. The needed rigor to implement new procedures in computer programs determined
some interactions between the software producers and the seismic code committee, aimed at
eliminating some inconsistencies in the norm. On the other hand, not all the software tools are based
on a correct interpretation of the new criteria and rules, but this is an unavoidable aspect of the free
market. A specific action, which has not been implemented yet, should be based on the preparation of
some benchmark structures to test at least the basic features of the software tools.
Re-addressing research activities

After the Molise earthquake, the attention to seismic problems has raised considerably, and the
research activities in the field have experienced a new vigor. However it was understood that the
research activities had to be re-oriented in order to meet the actual needs for the improvement of the
prevention actions. It was acknowledged that experimental testing on large-scale structures, sub-
assemblages and seismic devices is essential to understand the complex non linear and near-collapse
behavior of real structural systems and then to make well-grounded ultimate limit state verification
criteria and procedures. Moreover, it was recognized that a strong coordination and integration of
activities carried out in several research institutions is required, in order to maximize the use of costly
experimental tests and better finalize research findings to new design criteria and rules, to be include
in seismic codes, manuals, guidelines.
Due to the above reasons, the by-law 3274 has also assigned funds to create a new network of
laboratories of seismic engineering (ReLUIS, Rete dei Laboratori Universitari di Ingegneria Sismica,
http://www.reluis.it) and a new research and training centre (EUCENTRE: EUropean CENtre for
Training and Research in Earthquake engineering, http://www.eucentre.it ) with the aim to support the
progress in knowledge and the transfer of scientific results into best practices, through guidelines, pre-
507
normative documents and other tools for the reduction of seismic risk, as well as the training of
technical operators. The ReLUIS Consortium is made by the 4 Italian largest experimental facilities,
i.e. the structural engineering laboratories of the Unversity of Naples (2 DoF, Dual table system, with
two synchronous or asyncronous 3x3 mq, 20t mass, 1.0 m/s shaking tables), the University of Pavia
EUCENTRE (1 DoF, Large mass shaking table 5.60x7 mq, 140 t mass, 2.0 m/s and a reaction wall), of
the University of Basilicata and University of Trento (reaction walls and oleo-dynamic systems for
pseudo dynamic and dynamic testing of structures and components). The ENEA Laboratory, with a 6
DoF shaking table (4.0x4.0 mq, 20 t mass, 0.5 m/s), is associated to ReLUIS. A great 3-years research
project on Earthquake Engineering has been funded by the National Department of Civil Protection in
2005, to improve knowledge and providing high quality products that can be of immediate use for the
reduction of seismic risk (pre-normative documents, norm proposals, guidelines, databases, etc.). The
DPC-ReLUIS Project involves 40 Italian universities and research institutions and is the biggest one
ever launched in Italy. The large facilities of EUCENTRE have been funded by DPC, to start a large
scale structure experimental program, in order to improve the understanding of the seismic response of
real structure. EUCENTRE has also implemented a machine for seismic device testing, able to test
large seismic isolators in shear-compression tests, applying up to 50000 kN compression force and
2000 kN shear force. A smaller seismic device testing machine has also been upgraded (to reach
15000 kN compression force and 1000 kN shear force) at the laboratory of the University of
Basilicata, with the DPC funds. These new apparatuses will permit not only to make research and set
up new seismic code provisions, but also to meet the increasing demand of full-scale seismic device
testing for the most advanced applications of seismic technologies.
RELUIS and EUCENTRE, along with INGV (National Institute for Geophysics and Volcanology)
have been included in the network of the Capacity Centers of the Italian Civil Protection Department.
With this role, they are part of the National Civil Protection System, and are also involved in the post-
event activities, such as, for instance, the damage and usability investigation of buildings and the
temporary monitoring of strategic structures, under the coordination of DPC.
Implementing actions for seismic risk mitigation

The benefits of the adoption of a new seismic code will develop slowly in the long run because only
new constructions will be safe, while the existing constructions would remain unsafe, unless a specific
action is undertaken. Italy is a nation with a very important cultural heritage, as well as a quite old
building stock. Moreover, the seismic classification has been developed for a long time following
earthquakes, and only in the 80s the zonation was revised on the basis of an hazard study. For these
reasons most of the constructions now in seismic areas were built with no seismic provisions. Today
only one building out of 4 is seismically designed, thus having some lateral force resistance. Among
these buildings there are also strategic buildings (hospitals, municipalities, barracks of fire brigades,
army, etc.) and buildings relevant for the possible consequences of collapse (schools, theatres, stations,
airports). According to the new seismic code, strategic buildings should be immediately operational
after an earthquake and relevant buildings are supposed not to collapse: however, due to the lack of
seismic provisions, most of them can not probably guarantee the aforesaid performances. So, again
with the by-law 3274, the government has promoted the seismic assessment of strategic and relevant
constructions, with priority for those in the highest seismic zones (1 and 2), asking for a complete
evaluation in five years. Strategic and relevant constructions include not only buildings, but also
important infrastructures and industrial installations. It is not possible to evaluate exactly the number
of construction to be assessed, even focusing the attention on buildings only, because also private
buildings with mixed occupancy are included, but a lower bound estimate leads to a number of about
70.000 buildings. This is a very important and demanding operation, considering that the assessment
has to be performed by engineering evaluations fulfilling the new code. Many Regions, State
Organizations and also private owners have started to assess their estate, but activities are slowed
down by the large amount of economical resources needed. So other two by-laws: n. 3362 (for Region
estate), and n. 3376 (for State Organisation estate) have been promulgated, to give financial
contributions for the seismic evaluation and the retrofitting of strategic or important public
constructions. 263 million of euros have been allocated up to now. Due to the quasi-federal
organization of the Italian State, where the Regions assume the co-responsibility for Civil Protection,
508
2/3 of the funds are transferred to them to co-finance the costs for the above said operations. This
virtuous co-financing mechanism has considerably amplified the positive effects of this process,
although it has been considerably slowed down by the long bureaucratic procedures required to make
money available, especially to the beneficiary state administrations.
More than 6000 seismic safety verifications ad more than 200 retrofitting interventions have been
funded. Though significant, these figures represent only a small part of the strategic and relevant
constructions that should be potentially involved in this process. The fund allocations ruled by by-laws
3362/04 and 3376/04 have provided an important chance to be exploited at the best, with the
collaboration of Public Adminitrations, professionals, universities and associations. At the same time
other resources have become available for safety upgrading (electrical, fire, but mainly seismic) of
school buildings. About 40.000 public schools exist in Italy and the retrofitting costs are expected to
be extremely high, in the order of several billions of euros. An initial program has been already funded
with about 300 million euros to upgrade or retrofit the most risky schools (Dolce et al. 2007).
CONCLUSIONS
In this paper lessons learnt and implemented actions after the 2002 Molise earthquake have been
summarized. The earthquake by itself was not strong: in the last 25 years several destructive
earthquakes in Italy caused much more damage and casualties. Nevertheless the death of 27 classmates
and of their teacher, in a zone not considered before as seismic, made clear, also to the wide public
opinion, that the safety conditions of many buildings and schools were inadequate and that a wide gap
existed between scientific knowledge and implemented actions for seismic prevention.
The earthquake was also a further test of the Italian emergency management system (Augustus
method), characterized by a flexible system composed by many bodies (army, regions, municipalities,
volunteers, red cross) under the central coordination of the National Civil Protection. The system
proved to be adequate, mainly because it makes possible to involve a growing number of bodies and
resources, according to the needs of the event. The dynamically allocated functions in the coordination
centres (CCS, COM and COC) and the coordinated cooperation of all the bodies involved proved to be
efficient. Also a contribution to the efficiency came from the preliminary loss estimates made by
means of a specific simulation code in the first hours after the event.
It was evident, as many times in the past, that, after an earthquake, attention to seismic risk raises
sharply. The inadequacy of prevention actions is then perceived by public opinion, making important
changes in codes and regulations possible. In this particular circumstance the seismic code and the
seismic zonation were revised. This produced initially disconcert and difficulties in the application of
the new rules, but also an extraordinary effort at all levels to disseminate new design philosophies:
from professionals to scientific and administrative structures. But, probably, the most important
novelty is that direct risk mitigation countermeasures were undertaken at national levels, by making
the safety verification of strategic and relevant structures compulsory in a limited time framework and
by allocating funds, though limited, to support the start of this action. This effort will certainly results
in the enhancement of seismic awareness and a safer building stock against earthquakes. However the
seismic prevention action at national level can be considered being only at a beginning stage. Actually
measures were taken only after seismic events in the past, but they were limited to the stricken area
and involved only, or mainly, damaged buildings. Some other initiatives were undertaken in limited
areas. In order to become fully effective, actions for a long term perspective are necessary in the
future, with the certainty of the availability of resources. The management of these first fund
allocations can be a precious experience for future developments.
REFERENCES
509
Augenti, N., Cosenza, E., Dolce, M., Manfredi, G., Masi, A., Samela, L., Performance of School
Buildings during the 2002 Molise, Italy, Earthquake, Earthquake Spectra 20, Special Issue 1, 2002
Molise, Italy, Earthquake Reconnaissance Report, edited by P. Bazzurro and J. Maffei, pp 257-270,
2004
Baggio C., Bernardini A., Colozza R., Corazza L., Della Bella M., Di Pasquale G., Dolce M., Goretti
A., Martinelli A., Orsini G., Papa F., Zuccaro G., Manuale per la compilazione della scheda di 1
livello di rilevamento danno, pronto intervento e agibilit per edifici ordinari nellemergenza post-
sismica (AeDES), Dipartimento della Protezione Civile, Roma, August 2002.
Decanini, L., Di Pasquale, G., Galli, P., Mollaioli, F., San, T., Seismic hazard and seismic zonation
of the Region affected by the Molise sequence of October 31 November 1, 2002, Earthquake
Spectra 20, Special Issue 1, 2002 Molise, Italy, Earthquake Reconnaissance Report, edited by P.
Bazzurro and J. Maffei, pp 131-165, 2004.
Decree of the President of the Republic n. 380/2001: Unified codes and regulations concerning the
building construction
Decree of the Prime Minister n.3431, Further modifications to the by-law decree n. 3274/03
concerning initials items on the general criteria for seismic zonation of the national territory and on
seismic code, may, 3, 2005.
Decree of the Under-Secretary of State n. 4485 of December 4, 2002 devoted to obtain an authoritative
technical-scientific contribution for a new seismic code and the general criteria for a new seismic
zonation.
Di Pasquale, G., Fralleone A., Pizza A.G. and Serra C., The Italian seismic code and its history, in
La classificazione e la normative sismica italiana dal 1909 al 1984, Report edited by R. De Marco
and M.G. Martini, Istituto Poligrafico e Zecca dello Stato, Rome, 1999.
Dolce M., Seismic Safety of Schools in Italy, Proc. of the Ad Hoc Experts Group Meeting on
Earthquake Safety in Schools, 9-11 February 2004, Paris, France.DPC, Report to the European
Community on 31 October 2002 Molise-Puglia Earthquake, Technical Report, National Seismic
Survey, Rome, Italy, 2002 (In Italian).
Dolce M., Galanti E., De Sortis A., Di Pasquale G., Goretti A., Ferlito R., Papa F., Papa S., Pizza A. G.,
Sergio S., Severino M., Iniziative nazionali per la valutazione e riduzione del rischio sismico, Proc.
12 Convegno LIngegneria Sismica in Italia, June 2007, Pisa.
Foster, B., Kodama, S., Emergency Management, Recovery and Reconstruction Following the 2002
Molise, Italy, Earthquake, Earthquake Spectra 20, Special Issue 1, 2002 Molise, Italy, Earthquake
Reconnaissance Report, edited by P. Bazzurro and J. Maffei, pp 323-339, 2004.
Galanti, E., Goretti, A., Foster, B., Di Pasquale, G., Managing Civil Protection, in Oliveira C.S.,
Roca A., Goula X. (editors), Assessing and Managing Earthquake Risk, Kluwer ed., The
Netherlands, 2005.
Galanti, E., Albanese, V., De Sortis, A., Papa, F., Sergio, S., Molise-Puglia Earthquake: Emergency
Management, Temporary Shelters and Short Term Countermeasures, Ingegneria Sismica, Anno
XXI, Vol. 1, Patron Editor, Bologna, Italy, pp. 26-33, 2004 (In Italian).
Galli, P., Molin, D., Macroseismic survey of the 2002 Molise, Italy, earthquake and historical
seismicity of San Giuliano di Puglia, Earthquake Spectra, 20, Special Issue 1, 2002 Molise, Italy,
Earthquake Reconnaissance Report, edited by P. Bazzurro and J. Maffei, pp 39-52, 2004.
Goretti, A., Di Pasquale, G., An Overview of Post-Earthquake Damage Assessment in Italy, EERI
Invitational Workshop An Action Plan to Develop Earthquake Damage and Loss Data Protocols,
19-20 September, Pasadena, CA, 2002.
Goretti, A., Di Pasquale, G., Building Inspection and Damage Data of the Molise Earthquake
Sequence of October 31-November 1, 2002, Earthquake Spectra 20, Special Issue 1, 2002 Molise,
Italy, Earthquake Reconnaissance Report, edited by P. Bazzurro and J. Maffei, pp 167-190, 2004.
510
June 25-28, 2007
Paper No. W4-1003
THE CONTRIBUTION OF GEOTECHNICAL ENGINEERING

TO SAFER DESIGN OF EARTHQUAKE RESISTANT
BUILDING FOUNDATION
Michele MAUGERI1, Francesco CASTELLI 2 and Maria Rossella MASSIMINO3
ABSTRACT
The contribution of geotechnical Engineering is a key-point for safer design of earthquake resistant
buildings from many points of view. Preliminarily, the stability of the environment against landslides
and liquefactions has to be checked. Subsequently, local site response has to be checked to evaluate
the site response spectra, on which is based the design spectra, to define the seismic loading to which
the structure must resist. Finally, the role of Geotechnical Earthquake Engineering is to analyse
foundation settlements and stability, including soil-structure interaction. Both the classic inertial
interaction and kinematic interaction must be considered for safer design of old and new structures to
resist earthquake loading coming to the structure from the soil and the foundation. Two case histories
of old structures, damaged by the moderate Sicilian earthquake (Italy) of 13th December 1990 are
presented. The case histories are: the Noto Cathedral, founded on shallow foundation, and some
concrete buildings in Augusta, founded on pile foundations. For both case histories the site response
and the design earthquake were evaluated by determinist approach starting from the source modelling
and the evaluation of synthetic accelerograms. In both cases the retrofitting of the foundations gives a
remarkable contribution to a safer design of the superstructure, and also, it have influenced
significantly the retrofitting of the superstructure.
Keywords: Site characterization, site response analysis, shallow and pile foundations, retrofitting.
INTRODUCTION
The role of Geotechnical Engineering was firstly recognised by Seed et al., (1974), who showed that
the local site response is deeply dependent on geotechnical soil properties. Hudson (1981) pointed out
that Geotechnical Engineering is a key point for safer design of earthquake resistant buildings, as he
wrote: I began to wonder if Geotechnical Earthquake Engineering is in fact any different from
Earthquake Engineering. Even for the detailed problems of steel and concrete structural design, the
importance of soil-structure interaction may be a critical matter. Geotechnical Engineering is indeed
the foundation on which the whole subject is built.
During 12 th World Conference on Earthquake Engineering, held in 2000 in Auckland (New Zealand)
at the debate on the topic: That earthquake professionals are falling behind in reducing earthquake
risk worldwide, that was stated that there is a gap between the new research findings and the
1
Professor, Department of Civil and Environmental Engineering, University of Catania, Italy, Email:
mmaugeri@dica.unict.it
2
Researcher, Department of Civil and Environmental Engineering, University of Catania, Italy, Email:
fcastelli@dica.unict.it
3
Researcher, Department of Civil and Environmental Engineering, University of Catania, Italy, Email:
mmassimi@dica.unict.it
511
applications in common engineering practise. This gap enables the reduction of victims and looses due
to earthquake, which has been increased exponentially in the last twenty years. A new trend to cover
this gap and to go towards applied research in the field of Earthquake Engineering can be the
Geotechnical Engineering.
buildings from many points of view. Preliminarily, the stability of the environment against landslides
and liquefactions has to be checked. Subsequently, local site response has to be checked to evaluate
the structure must resist. Finally, the role of Geotechnical Earthquake Engineering is to analyse
foundation settlements and stability, including soil-structure interaction.
About the stability of the environment, frequently landslides occur during medium-high earthquakes,
as in the case of 1980 Irpinia earthquake (M = 6,5), where large landslide of about 1-2 km occurred in
the towns of Boniventre, Senerchia and Calitri. Particularly, the Calitri landslide (Crespellani et al.,
1996) caused many house collapses. During the 2001 El Salvador earthquake (M = 7,6) a very large
landslide caused the destruction of many houses and the deaths of about five hundred people, about
the same amount of victims caused by the building shaking due to the earthquake. The liquefaction
phenomenon occurred during the 1964 Niigata earthquake (Towhata, 2004), as well as the 1964
Alaska earthquake caused the loose of many buildings, which were severely rotated without suffering
any structural damage. Particularly dangerous for the stability of the environment is the combined
effect of liquefaction in an area prone to landslides. This combined effect, as in the case of 1999
Kocaeli earthquake, leads to lateral flow failures (Biondi et al., 2002) producing severe displacements
to some buildings that move significantly towards the sea.
The stability of the site where a building is founded is outside the aim of the present paper, while the
local site response is taken into account, because it is a fundamental aspect to evaluate the design
spectra to be used for the seismic design of structures. The evaluation of the seismic action that acts on
a building is the most unknown aspect of the whole procedure for seismic design of a building.
Generally the seismic action is prescribed by national regulations on the basis of a probabilistic
approach. Actually, very often in the past, the seismic actions recorded during destructive earthquakes
were higher than those prescribed by national regulations. Alternatively to the probabilistic approach
the deterministic approach, based on the modelling of the source mechanism can be used.
The evaluation of seismic action is a multidisciplinary problem, involving geologists, geophysics and
geotechnical engineers; particularly for the evaluation of the site effects the role of Geotechnical
Earthquake Engineering is fundamental. As the seismic action is a preliminary aspect for the design of
earthquake resistant buildings, the next paragraph will be devoted to the site response assessment, for
the evaluation of the design earthquake, which was used for the two case histories, which will be
illustrated in the following.
SITE RESPONSE ASSESSMENT
The seismic action is generally prescribed by probabilistic approach, considering the probability of
exceeding a certain level during the life of the constructions. In Italy the probabilistic evaluation of the
seismic action is based on the probability of exceeding of 10% in 50 years, considering only the
earthquakes with a return period less than 475 years. This criterion is not a conservative one, because
generally speaking the life of the construction exceeds 50 years; in fact in many Italian medieval cities
the buildings are standing from more than 500 years.
The Code of Federal Regulations of the USA requires new municipal solid waste landfills to be
designed either for a maximum horizontal acceleration taken from a published seismic map for a 10
percent probability of exceedance (90 percent probability of non exceedance) in a 250-year exposure
period or on the basis of a site specific analysis. The related return period for the map-based
acceleration is 2375 years.
512
Increasing the historical knowledge of past earthquakes, the probabilistic approach leads to update the
seismic action, which has been changed in the past in Italy (Figure 1), so, buildings designed to resist
to the given seismic action at the time of construction earthquakes could be not resisting at the updated
seismic action.
(a) (b)
Figure 1. Seismic action: (a) Seismic hazard map of Italy in terms of acceleration proposed by
the GNDT on 1984. (b) Seismic hazard map of Italy in terms of acceleration attached to the
Ordinance O.P.C.M. 3274/2003 according to the Civil Defence of Italy.
Alternatively to the probabilistic approach, the procedure for the evaluation of the design earthquake
takes into consideration the deterministic approach based on the evaluation of source mechanism. For
the two case histories considered, a comparison between the probabilistic seismic action given by the
Italian regulation and the deterministic seismic action based on the modelling of the source mechanism
of the scenario earthquake has been performed.
The scenario earthquake considered is the January 1693 events, as the biggest one, and the December
1990 as a low earthquake (Figure 2). The geological, geophysical and laboratory investigations
performed provide additional constraints to the surface geological setting of the area as well as to the
parameterization of the physical models. Numerical simulations have been applied with the twofold
aim of estimating strong ground motion scenarios for different earthquake hypotheses and evaluating
the effectiveness of 1-D non-linear and 2-D linear methods for the estimation of the local response.
The use of advanced methods capable of generating of synthetic seismograms can give a valuable
insight into the evaluation of a seismic ground motion scenario. The reference event is the catastrophic
earthquake (M7+) that struck Eastern Sicily on January 11, 1693 is linked to the Ibleo-Maltese fault
system, which is the major seismogenic structure of Eastern Sicily.
The deterministic approach (Bottari et al., 2003) is based on the modelling of the Ibleo-Maltese fault.
The fault was modelled by two rectangular sub faults, characterised by the southern segment, 51 km
long and 13 km wide, and the northern one, 22 km long and 13 km wide. The longest segment, with a
346 strike, is facing the coast between Siracusa and Augusta; the shorter segment, with a 352 strike, is
placed in the Gulf of Catania. The two sub faults were sub-divided by a grid of seismic point sources.
For each seismic point source the seismogram was computed by the model summation method
developed by Panza et al., (2000). For the summation of the point-source seismograms each one was
513
delayed according to the rupture propagation times on the fault, scaled to include the distribution of
energy release on the fault (Sara et al., 1998). Three source mechanisms were considered: i) the fault
rupture towards North; ii) the fault rupture towards South; iii) the bilateral fault rupture (Figure 3).
The accelerograms were computed at the city of Noto for the three rupture models and for three
different positions of the receiver, located at the surface and at the depths of 30 m and 70 m,
considering as scenario earthquake the destructive January 11, 1693 earthquake.
The seismicity of the Ibleo-Maltese escarpment is rather anomalous, since it generates a very low
number of low to medium earthquakes. Consequently, there are only few events that can be used to
validate numerical simulations of 1693 earthquake. The December 13, 1990, M = 5,8 earthquake
(Figure 2.b) is actually the only medium size event occurred along the northern segments of the Ibleo-
Maltese system, which was recorded instrumentally. This earthquake is associated to a rupture of the
transcurrent segment of the Ibleo-Maltese fault, and it was recorded by the ENEA-ENEL
accelerometric network.
(a) (b)
Figure 2. The scenario earthquakes: (a) Isoseismal map of the January 11, 1693 scenario
strong earthquake. (b) Isoseismal Map of the December 13, 1990 scenario low
earthquake; the ground acceleration recorded at Catania (2.43 m/s2) has amplitudes
about 2,5 times larger than that recorded at Sortino (1.003m/s2), although both stations
are at about the same epicentral distance (about 30 km).
Noto Noto Noto
Figure 3. Source mechanisms of the Ibleo-Maltese fault: a) unilateral rupture from

South to North; b) unilateral rupture from North to South; c) bilateral rupture (after
Bottari et al., 2003).
514
Another reason of interest for modelling this earthquake is the fact that the seismogram recorded by
the Catania ENEA-ENEL station shows anomalously large ground accelerations. The ground
acceleration recorded at Catania (2,43 m/s2) has amplitudes about 2,5 times larger than that recorded at
Sortino (1,003 m/s2), although both stations are at about the same epicentral distance (about 30 km)
(Figure 2). These anomalies can be attributed mainly to local site effects. The use of a method, the 2-D
Chebyshev spectral element method, which solves the seismic full-wave propagation through a
complex geological structure, is then of maximum interest to verify this kind of hypothesis
(Laurenzano and Priolo, 2005).
Based on the synthetic accelerograms virtually recorded at a depth of 30 m, the site response was
checked by a 1-D non linear model for the evaluation of the design spectra for the two case histories.
The two case histories of old structures, damaged by the moderate Sicilian earthquake (Italy) of 13th
December 1990 are: the Noto Cathedral, founded on shallow foundation, and the concrete buildings in
Augusta, founded on pile foundations.
For both the case histories the retrofitting of the foundations the behaviour of existing foundations was
analysed, which is very often ignored in engineering practice. In both the cases the retrofitting of the
foundations gives a remarkable contribution to a safer design of the super-structure, and also, it have
influenced significantly the retrofitting of the same super-structure.
SEISMIC IMPROVEMENT OF THE SHALLOW FOUNDATION OF NOTO CATHEDRAL
The case history

Due to the December 13, 1990 East Sicilian earthquake the Noto Cathedral, one of the most important
symbols of the Sicilian Baroque, which was declared a World Cultural Heritage by UNESCO, suffered
many and spread cracks. These leaded, on March 1996, to the collapse of the piers on the right side of
the main nave, as well as to the collapse of the roof of the main nave and right naves and to the
collapse of the dome (Figure 4). The huge damage and the historical-monumental importance of the
cathedral leaded to deep and numerous investigations both on the structure (Binda et al, 1999) and on
the foundation soil (Cavallaro et al., 2003), in order to discover the reasons of so huge damage and to
plan carefully the reconstruction of the cathedral. The structural investigations clearly revealed that the
occurred damage was mainly due to the poor workmanship of the piers, the excavation of a hole inside
one of the piers and the reinforced concrete roof, recently made instead of the original wood roof. This
reinforced concrete roof stiffly connected all the piers including the weaker pier with the hole inside it,
so that the failure of the weaker pier leaded to the failure of many other piers on the right side, as well
as to the failure of the naves and of the dome. Certainly other factors contributed to have an
earthquake poorly resistant structure, such as the absence of connections between all the foundations,
whose embedment changed from 0 to 2.40 m without any logical reasons.
Figure 4. View of the collapsed Noto Cathedral after the damage caused
by the 1990 South-East Sicilian earthquake.
515
Site characterization
The geotechnical characterization of the foundation soil of Noto Cathedral involved an area of 3200
m2 (40 m x 80 m) up to a depth of 81 m. Both in-situ tests and laboratory tests were made; in
particular, in situ dynamic tests (SAWS, D-H, C-H) and laboratory dynamic tests (RCT, CLTST,
CLTxT) were performed.
The soil underneath Noto Cathedral consists mainly of: a) talus material up to a depth variable in the
range of 0.5 5.5 m; b) highly weathered, weakly cemented, limestone, sometimes with a large clay
content, up to a depth variable in the range of 12.0 21.5 m; c) sandy clay, which lies below the
limestone and is sometimes mixed with it.
The main results of the geotechnical characterization are reported in Cavallaro et al. (2003) and
Cavallaro and Maugeri (2003). As regards the general characteristics and index properties of Noto
Pliocene soil, which mainly consists on a medium stiff, overconsolidated lighted cemented silty-clay-
sand, it was found: wn = 12-37 %; wL = 37-69 %; wP = 17-22 %; IP = 15-47 %. As regards the strength
parameters it was found: c = 0 85 kPa; = 16 33. In general the site characterization showed a
low degree of homogeneity with depth.
The comparison between SPT tests and CLTST and RCT tests allowed to estimate the shear modulus
(G 0) profiles with the depth. In particular, utilizing the results of the SPT tests, the profiles of the shear
wave velocity V s were obtained through the empirical correlations given by Ohta and Goto (1978) and
Yoshida and Motonori (1988). The value of Vs,30 is in the range of 180 360 m/s, thus, according to
EC8 (EC8-Part 1, 2003), the foundation soil can be classified of a ground type C.
By means of the RCT tests the variations of the shear modulus G and of the damping ratio D with the
shear strain were evaluated. The experimental results were interpolated utilizing the following
expressions by Yokota et al. (1981), in which = 15, = 1.28, = 25.6 and = 1.952 (Figure 5):
G( ) 1
(1)
G0 1 (%)
G
D( )(%) exp
Go (2)
1.2 100 S3C1 - 9.00 m - IP = 15

S1C1 - 13.00 m - IP = 33
1 S7C2 - 15.50 m - IP = 27
S4C2 - 22.20 m - IP = 22
0.8 S1C3 - 51.00 m - IP = 36
Yokota et al. (1981)
S3C1 - 9.00 m - IP = 15
D [%]
G/Go
0.6 S1C1 - 13.00 m - IP = 33 10

S7C2 - 15.50 m - IP = 27
0.4 S4C2 - 22.20 m - IP = 22
S1C3 - 51.00 m - IP = 36
0.2
0 1
0.0001 0.001 0.01 0.1 1 0 0.2 0.4 0.6 0.8 1
[%] G/Go
(a) (b)
Figure 5. Evaluation of soil non-linearity: a) Degradation of normalised shear modulus

with shear strain; b) Increasing of damping ratio with degradation of shear modulus
(after Cavallaro et al., 2003).
516
As regards the soil-structure interaction analyses, discussed in the following, only the first soil layer of
dry clayey limestone was taken into account. According to Richards et al. (1993), for the analysed
case the possible failure surface could be localized in this first layer. This layer is characterized by: =
19.30 kN/m3 , G0 = 92 MPa, = 0.30, c = 15 kPa, = 24.
Site dependent design spectrum

In order to define a design spectrum, which takes in account the site effects, a response analysis of the
foundation soil were performed. For the evaluation of the seismic input, the deterministic criterion was
used. In particular, the synthetic accelerograms, obtained through the source mechanism model
discussed in the paragraph Site response assessment and virtually recorded in Noto at the depth
equal to 30 m, were utilised as seismic input at the bedrock, considering that for the Noto Cathedral
foundation soil, the bedrock (Vs 800 m/s) was at the depth of 4550 m.
The analysis of the dynamic behaviour of the foundation soil was performed by means of the 1-D non
linear GEODIN code (Frenna and Maugeri, 1995), in which the expressions (1) and (2) were
implemented, to take into account soil non-linearity. In particular, for the profiles related to the
boreholes outside the cathedral, the free-field conditions were supposed; instead, for the profiles
related to the boreholes underneath the cathedral, utilising a simplified procedure, a surcharge applied
on the soil surface was considered, in order to simulate the load transferred from the foundation to the
soil (Massimino and Maugeri, 2003a). The obtained elastic response spectra were transformed into
design inelastic spectra, considering the following expressions (Giuffr and Giannini, 1982):
Sd Se / g
(3)
g b
where:
T T
for T T0 b 1 2 q0 1 1 0.5 (4.a)
T0 T0
1 0.87 0. 05T0
with q 0 1
1 0.87 0.05T
for T T0 b 1 (4.b)
where S e is the ordinate of the elastic response spectrum, Sd is the ordinate of the design inelastic
spectrum, according to the terminology of the EC8, T0 is equal to 0.5 sec according to Giuffr and
Giannini (1982) and is the structural ductility, considered equal to 2.0 for the Noto Cathedral.
Starting from all the obtained inelastic spectra, the inelastic medium spectrum was computed.
Subsequently, the last one was compared with that suggested by EC8 (EC8-Part 1, 2003) for a ground
type C, a behaviour factor q = 2.5 and a structure damping ratio = 10% (Figure 6). On the basis of
this comparison the site dependent design spectrum shown in Figure 6 was proposed. Its expression is
given by (Masssimino and Maugeri, 2003a):
T 2.5
for 0 T TB Sd ag S S L 1 1 (5.a)
TB q
2.5
for TB T TC Sd a g S SL (5.b)
q
517
2.5 TC
for TC T TD Sd ag S S L (5.c)
q T
2.5 TCTD
for T TD Sd ag S SL (5.d)
q T2
More precisely, the proposed design spectrum was obtained from the inelastic EC8 spectrum,
changing the values of some parameters and adding the two SL and parameters. In particular, the
values of TB and TC were changed (TB = 0.10 sec and TC = 0.75 sec), in order to enlarge the constant
part characterized by the peak value, in agreement with other seismic codes, such as the NEHRP-97
code (NEHRP, 1997) and different recent earthquake (Bucharest 1977, Mexico City 1985 and Kobe
1995) recordings (Mylonakis and Gazetas, 2000). The new SL (SL = 1.1) parameter represents the ratio
between the mean ordinate of the inelastic medium spectrum in the range TB - TC and the maximum
ordinate of the inelastic EC8 spectrum. So, SL represents the local soil parameter for the Noto
Cathedral foundation soil, to be multiplied to the soil factor S provided by EC8. Finally, the new (
= 1.2) parameter was introduced in order to catch the decreasing part of the inelastic medium
spectrum, maintaining ordinates bigger than the ones of the EC8 spectrum. As it is possible to see
from Figure 6, the maximum design seismic coefficient is equal to 0.22.
0.3
0.25 Computed medium design spectrum
0.2
Sd/g 0.15 Proposed design spectrum
0.1
EC8 design spectrum

0.05
0
0 0.25 0.5 0.75 1 1.25 1.5 1.75 2 2.25 2.5
T (sec)
Figure 6. Comparison between the computed inelastic medium spectrum,

the EC8 spectrum and the proposed design spectrum.
Main aspects of the reconstruction design

Thanks to the geotechnical characterization of the foundation soil and the structural investigations, the
suitable strategies for the seismic improvement of the structure and of its foundations were planned.
These strategies were confirmed and clearly defined by soil-structure interaction analyses, which will
be discussed in the next paragraph Soil-structure interaction.
First of all, it was decided to improve the existing foundations, which significantly contribute to a
good seismic structural response. It was proceeded to: i) make uniform the embedment of the pier
foundations, fixing it equal to 2.50 m; ii) make deeper the foundations of the external walls, by means
an underpinning up to 1.25 m; iii) increase the contact area between soil and foundation, creating
inverted arches along the transversal and longitudinal directions (Cavallaro et al., 2003; Massimino
and Maugeri, 2003b). In Figure 7 the plan view of new foundation of the cathedral is shown. In
particular, it is possible to see the inverted arches and the tie rods. Furthermore, four tie rods to cross
each transversal foundation section (Figure 8.a) and two tie rods to cross each longitudinal transversal
section (Figure 8.b) were considered, in order to reduce significantly the foundation relative
518
displacements, which could compromise the stability of the superstructure. Figure 9 shows the external
wall underpinning.
About the overstructure it was decided not only to rebuilt all that had collapsed, i.e. the piers on the
right side of the main nave, the roof of the naves and the dome, but also to pull down and rebuilt the
piers on the left of the nave, to realize a deeper foundation embedment and, above all, because the
standing piers had significant damage and however they had been made with a poor workmanship,
which was the principal reason of the big collapse of the cathedral (De Benedictis and Tringali, 2000;
Tringali and De Benedictis, 2001a, b). Figure 10.a shows the demolition phase of a pier on the left of
the main nave; Figure 10.b shows the excavation to realize the deeper pier foundation level; Figure
10.c shows the scheme of the new pier with its foundations.
External wall
Inverted arches Tie rods
Figure 7. New foundations of Noto Cathedral.
Tie rods
Foundation underneath the piers
a) TRANSVERSAL SECTION Inverted arches
SEZIONE TRASVERSALE
External walls underpinning
0.0 5.0 10.0
Tie rods
b) LONGITUDINAL SECTION Inveted arches
SEZIONE LONGITUDINALE
Figure 8. Improvement of the Noto Cathedral foundations:

a) transversal section; b) longitudinal section.
519
Furthermore, considering that the previous reinforced-concrete roof connecting stiffly the piers
significantly contributed to increase the damage occurred on March 1996, it was decided to rebuilt the
roof with its original configuration, i.e. to rebuilt a typical wood roof. To reduce the stress level on the
lateral arches, six tie rods per arch were utilized: two below the arch and four inside the arch
(Massimino et al., 2004). At the same time, the soil-structure interaction analysis, reported in the
following, shows that for the central arches, which can be greatly stressed due to foundation
movements, a reinforcement was also needed. In order to guarantee a nice view inside the cathedral,
carbon fibers were utilized instead of tie rods.
TIE RODS
OLD FOUNDATION
NEW FOUNDATION
SAND LAYER
Figure 9. Underpinning of the external walls.
(a) (b) (c)
Figure 10. Improvement of the Noto Cathedral foundations: a) demolition of a damaged pier; b)
excavation to built the new foundation; c) construction scheme of the new foundation and pier.
Soil-structure interaction
The study of soil-structure interaction phenomenon was devoted to investigate on the displacement
and stress distributions on the whole structure, including its foundations. In particular, the Authors
refer here on the typical transversal section not crossing the dome. This transversal section was
520
investigated not including (Figure 11.a) and including (Figure 11.b) the tie rods discussed in the
previous paragraph Main aspects of the reconstruction design. The carbon fibers were not
considered in the numerical analyses, because the decision to use carbon fibers was taken after the
performed analyses.
The study of masonry structures should be carried out by complex 2-D and/or 3-D modeling; however,
simplified 2-D FEM models, as the one shown in Figure 11, can be adopted in the cases in which the
masonry structures can be assimilated to frame structures, i.e. in the case of structures characterized by
slim load-bearing walls. In order to take into account the foundation soil deformability, three springs
per node, for the 1, 11, 33 and 43 foundation nodes, were considered, so vertical and horizontal
foundation displacements and foundation rotations were allowed. These springs characterize the three
degrees-of-freedom of the four soil-foundation macro-elements related to nodes Nos. 1, 11, 33 and 43.
For these springs the elasto-hypoplastic constitutive model (SFCM) reported in di Prisco et al. (2003)
was utilized, while for the superstructure the elastic-linear model was considered.
22 22
21 21 23
19 22 23 23 26 19 22 23 26
20 20 24
21 24 24 21 24
19 26 19 26
20 25 20 25
18 25 18 25
18 27 18 27
17 27 17 27
17 28 17 28
9 10 34 36 9 10 34 36
8 16 29 36 8 16 29 36
7 37 7 37
6
6
7 8 9 15 z 30 35 37 38
39
38 6 7 8 9 15 z 30 35 37 38 38
16 28 6 16 28 39
14 31 40
5 40 51 52 14 31 53 54
5
5 15 29 39 5 15 29 39
55 56 57 58 59 60
4 13 32 41 4 13 41
32
4 14 30 40 4 14 30 40
3 12 33 42 3 12 33 42
3 13 31 41 3 13 31 41
2 11 34 43 2 11 34 43
2 12 32 42 2 42
12 32
1 10 35 44 1 10 35 44
11 33 43 44 45 46 47 48 33 49 50
1 1 11 43
y 45 46 47 48 49 50 51 52 53 54
y
(a) (b)
Figure 11. Plane structural scheme of Noto Cathedral employed for the
numerical analyses: a) without tie rods; b) with tie rods.
The initial values of the horizontal, vertical and rocking stiffnesses Kh , Kv, K for the above-mentioned
springs were determined following the Gazetas (1991) approach for embedded foundations (Table 1).
Two types of equivalent rectangular foundations were considered: the first foundation, under the piers
(nodes 11 and 33 of Figure 11), characterized by a width B = 4.30 m, a length L = 7.0 m and an height
D = 2.50 m; the second foundation, under the external walls (nodes 1 and 43, Figure 11), characterized
by a width B = 5.30 m, a length L = 8.0 m and an height D = 1.25 m.
The SFCM constitutive parameters, which had to add to the previously defined elastic spring
stiffnesses are reported in Table 2. Among them, V m, , , characterize the failure locus, R0, , ,
and characterize the plastic stiffness, 1 and 2, 3 characterize the cyclic mechanical behavior. Some
of these parameters (Vm, , , R0) were directly evaluated for the investigated case history, some other
parameters ( , , ) were assumed to be the same of those previously calibrated for sandy soil (di
Prisco et al., 2006), considering that the last ones have a small influence on the calculated results
(Montrasio and Nova, 1997). For the remaining constitutive parameters 1 and 2 = 3, which must be
determined by calibration tests on the macro-elements, the values obtained by di Prisco et al. (2003)
on small scale and large scale experimental test results in dense sand were used.
521
Table 1. Initial values of the stiffnesses for the foundation-soil systems
at nodes 1, 11, 33 and 43 of the chosen transversal section (di Prisco et al., 2006).
Kh (MN/m) Kv (MN/m) K (MNm)

Foundation of the
external walls 2500 2385 22802
Foundation of the
columns 2847 2373 27144
Table 2. Constitutive parameters of the elasto-hypoplastic SFCM model (di Prisco et al., 2006).
FOUNDATION
CHARACTERISTICS MODEL PARAMETERS
R0
Node B (m) D (m) (kN/m) VM (kN)
1 - 43 8.00 1.25 0.95 2.83 1.71 0.40 0.40 0.80 0.286 4 20 86349 25585
11 - 33 7.00 2.50 0.95 2.83 1.71 0.54 0.46 0.80 0.286 4 20 100393 26028
The frames shown in Figure 11 were firstly investigated by means of the pseudo-static approach, i.e.
applying to each node of the frames the following two forces: a vertical force Wi, which represents the
weight of the sub-structure related to the i th node, evaluated by means of the influence area criterion;
and a horizontal force, equal to Hi = kh Wi, being kh the seismic coefficient fixed equal to 0.22
according to Figure 6. The results of this analysis are shown in Table 3 in terms of horizontal and
vertical foundation displacements not including and including tie rods. Figure 12 shows the deformed
configuration of the frame with tie rods. The tie rods, crossing the lateral arches as well as the
foundations, reduce significantly the relative movements, leading to a less severe stress distribution.
To analyze the behaviour of the cathedral under a moderate and more frequent earthquake, an
additional pseudo-static FEM analysis is performed to the frame of Figure 11.b, utilizing a seismic
coefficient kh* = 0.6 kh, i.e. kh* = 0.143. The horizontal and vertical foundation displacements obtained
from this second analysis are reported in Table 4.
Figure 12. Deformed configuration of the structure, considering tie rods and the elasto-
hypoplastic SFCM soil model - pseudo-static approach (after di Prisco et al., 2006).
522
Table 3. Horizontal (u) and vertical (v) foundation displacements not considering
and considering the tie rods - pseudo-static approach (after di Prisco et al., 2006).
WITHOUT TIE RODS WITH TIE RODS

NODE
u (mm) v (mm) u (mm) V (mm)
1 16.30 36.36 17.76 32.09
11 12.46 43.56 15.57 38.76
33 22.68 39.85 20.84 35.17
43 34.92 42.30 29.74 38.28
Table 4. Horizontal (u) and vertical (v) foundation displacements considering the tie rods
and a seismic coefficient equal to 0.143 - pseudo-static approach (after di Prisco et al., 2006).
NODE u (mm) v (mm)

1 3.73 22.93
11 4.62 25.88
33 7.91 23.58
43 14.29 25.15
The pseudo-static approach represents certainly the most simple and thus utilized approach to study
seismic soil-structure interaction. Nevertheless, it can often lead to erroneous results mainly because
two factors are disregarded: the cyclic change of loading distribution within the structure and the
effect of the input frequency content. Very interesting results were obtained considering only the
former factor and so performing numerical cyclic quasi-static analyses, i. e. applying to each node the
constant vertical force Wi and a horizontal force Hi(t). The seismic coefficient kh* = 0.143 was
considered, due also to the severity of a cyclic input of constant amplitude. In this case it is possible to
observe a decrease with time in the value of the vertical force V for the node No. 1, that leaded to a
very dangerous generalized stress path, because the image point approached the horizontal vs vertical
failure locus in the sliding branch (Figure 13). Table 5 reports, finally, the horizontal and vertical
foundation displacements obtained with the cyclic analysis. Comparing the results of Tables 4 and 5 it
is possible to observe with the cyclic analysis a more uniform distribution of the horizontal foundation
displacements due to the change of loading distribution within the structure, and more severe vertical
foundation displacements, due to the cyclic nature of the applied horizontal forces.
N o de 1 V -H
3000
1500
H (kN)
0
0 1 0000 200 00 30000
-1500
Loading pa th
F ailure locus
-3000
V (kN )
Figure 13. Horizontal (H) versus vertical (V) loading paths and corresponding failure locus
for node No. 1 cyclic quasi-static analysis (after di Prisco et al., 2006).
523
Table 5. Horizontal (u) and vertical (v) foundation displacements considering the tie rods
and a seismic coefficient equal to 0.143 - cyclic analysis (after di Prisco et al., 2006).
NODE u (mm) v (mm)

1 7.13 37.20
11 4.64 29.09
33 4.70 27.90
43 5.62 31.16
SEISMIC RETROFITTING OF THE PILED FOUNDATION OF R-C BUILDINGS
The case history

This section presents the case history of the seismic retrofitting of ten reinforced concrete buildings,
located in Augusta (Italy) (Figure 2.b), damaged by the Sicilian earthquake of 13 th December 1990 in
South-East Sicily (Italy). The extreme severity of the damage could only be justified by the existence
of local amplification effects due to soft soil conditions and the lack of enforcement of seismic
regulations at the time of construction. In fact, a few buildings constructed after the year 1981, when
the area was classified as being of medium seismicity and seismic regulations were enforced, suffered
minor damage or no damage at all. One of the most seriously damaged building was chosen as a case
study for structural upgrading and seismic retrofitting of all buildings (Figure 14).
The severity of the damage of the building in question was evident from the first inspection with
cracks and dislocations at the foot and at the top of some of the ground floor columns, cracks at the
joints and spread cracking of infill walls at the lower storeys. At an early stage of the study it was
recognized that the design of the retrofitting system could be greatly affected by the mechanical
properties of the materials employed in the construction and by the effectiveness of the foundation
system (Oliveto & Decanini, 1998).
To analyze and to empathize the importance of the different phases of a seismic design process, the
approach adopted for the seismic retrofitting of the considered six storey reinforced concrete building,
is described and discussed. The reinforced concrete frame resisting residential building had a first
weak/soft storey just above the ground and five other storeys above this. The plan is rectangular and
the ground floor was used as an open space for social activities while the floors above had four
apartments each, served by two staircases and lifts (Figure 14).
Figure 14. View of the damaged building
524
The site was well investigated by means of laboratory and in-situ tests including seismic tests. Seismic
action has been evaluated by the deterministic approach (Frenna & Maugeri, 1995; Maugeri & Frenna,
1995). A site depending response spectrum has been determined, then a design spectrum more severe
than that adopted by the Italian Seismic Regulation has been derived.
For evaluating the possibility to repair the buildings, an investigation on soil, structures and
foundations was carried out. The foundation system is composed of footings on piles. Tie beams
connect the individual footings in the longitudinal and transverse directions of the building but the
connection is partially absent along the central longitudinal alignment. The nominal diameter of the
piles is 50 cm with a length of about 20 m.
The visual inspection and survey of the foundation was ensured by a perimetral excavation along the
east, south and west sides of the building (Figure 15). The depth of the excavation was about 3 m,
allowing for the inspection of the footings, the tie beams and of the stem of the piles for about 1.80 m.
No signs of damage clearly attributable to the seismic action could be found. Nevertheless, vertical
and horizontal loading tests on two single pile tests were conducted. Also tests on a three-piles group
were carried out. Loading tests showed that the seismic actions have not damaged the effectiveness of
the soil-pile system. By the way considering a seismic action more severe than that adopted by the
Italian Seismic Regulation, a number of piles were added.
Site characterization
The investigated area has plane dimensions of 4100 m 2 (Figure 16). A detailed description of the soil
properties was reported by Castelli & Maugeri (2004).The thickness of the deposits varies from 50 up
to 300 m. The upper part of these deposits mainly consist of alternating layers of grey-silty-clay and
sandy-clay which is locally called Augusta Clay. Layers of sand were found at depths of between 9
and 12 m. The lower deposits mainly consist of a medium stiff, over-consolidated marine clay with
low to medium plasticity index.
The index properties and the mechanical characteristics of the Augusta clay derived from the
laboratory tests are: the clay fraction is prevalently in the range of between 60 70%; the silt fraction
is in the range of about 25 40%; the values of the natural moisture content wn prevalently range
between 30 and 35%. Characteristics values for the Atterberg limits are: wl = 60 65% and wp = 22
26%, with a plasticity index PI = 30 40%.
Finally, an assessment of the values of the undrained shear strength cu determined by dilatometer tests,
cone penetration tests and laboratory tests has been carried out. To express the variation of the
undrained shear strength versus depth, the following relationship was derived: cu (kPa) = 7.5 z, where
z is the depth from the ground surface.
Figure 15. Detail of the pile foundation inspection
525
Cross-hole and down-hole tests indicate an average value of the shear waves velocity equal to Vs = 320
m/s up to -10 m from the ground surface. An average value of the shear waves velocity approximately
equal to Vs = 530 m/s can be estimated up to -30 m from the ground surface.
The small strain shear modulus of soil Go can be considered an essential parameter for estimation of
the soil response, especially for dynamic loading conditions. The modulus Go was determined both by
in-situ than by laboratory tests. In laboratory the shear modulus (G) and the damping ratio (D) was
determined by Resonant Column tests (RCT) and cyclic loading torsional shear tests (CLTST) (Figure
17), even if it was attempted to assess Go also by means of empirical correlations based on in-situ test
results (Cavallaro et al., 1996). By laboratory tests, the values of Go were determined at shear strain
levels less than 0,001 %.
Figure 16. Layout of the site with the location of buildings and in situ tests
It is possible to observe that quite a good agreement exists between the laboratory and in-situ test
results. In particular, the values of Go for the upper silty-clay steadily increase from 20 to 80 MPa with
depth. In the transition zone, where stiff sand layers are present, Go increases up to 110 MPa
(Cavallaro et al., 1996). In the lower blue-grey clay Go values range in the range 80 up to 120 MPa. A
small difference between the values of shear modulus Go determined by this last procedure and the
values derived by resonant column tests can be observed in the first 10 m below the ground surface
(Maugeri & Castelli, 2007).
1.2 100 Test. N. 1

Test. N. 2
Test. N. 3
1 Test. N. 4
Test. N. 5
0.8
D [%]
G/Go
0.6 Test. N. 1 10
Test. N. 2
Test. N. 3
0.4 Test. N. 4
Test. N. 5
0.2
0 1
0.0001 0.001 0.01 0.1 1 0 0.2 0.4 0.6 0.8 1
G/Go
Figure 17. Shear modulus degradation and damping ratio variation
Site dependent design spectrum

Augusta site, situated in the South-Eastern part of Sicily (Italy), is placed on one of the most intense
seismic hazard areas of Italy (Figure 1). The city of Augusta has been destroyed by three disastrous
earthquakes with an MKS intensity ranging from IX up to XI (Postpischl, 1985). Even if 1990
526
earthquake was of moderate magnitude, it caused 19 victims and severe damages to buildings and
infrastructures. The estimation of damages caused by the 13th December 1990 South-Eastern Sicily
earthquake, has emphasized the effects of the phenomenon of local amplification, which caused
serious damages to concrete type of structures.
The response of the soil, schematized as a one-dimensional system with several degrees of freedom
and lumped masses, has been evaluated with a hysteretical-simplified 1-D model, through the
integration of the system of differential equations of the dynamic equilibrium with variable
coefficients, using the Newmarks numerical integration method step-by-step (Maugeri & Frenna,
1987). The model is capable to take into account the decay of the elastic shear modulus with strain.
The excitation at the base of the model were the accelerograms of the 13th December, 1990 earthquake
recorded at Sortino E-W, which is the most representative among those made on rock (Figure 18).
Related to the seismic characteristics of the area, to analyze the possibility that even more intense
earthquake than the one recorded on 13th December, 1990 could occur, the accelerations and the
frequencies of the recordings at Sortino E-W have been scaled up to a maximum acceleration 0,3g and
the predominant period of earthquake T = 0,5 sec.
Figure 18. Accelerogram recorded at Sortino E-W during the 1990 Sicilian earthquake
The model for the evaluation of the seismic soil response (Maugeri & Frenna, 1987) requires the
knowledge of the depth of the conventional bedrock. In this case it has been placed 80 m below the
ground surface. The numerical analysis provides the time-history response of displacements, velocity
and accelerations. Figure 19 shows the profiles of the numerical results versus depth (Maugeri &
Frenna, 1995; Frenna & Maugeri, 1995).
Figure 19. Results of the seismic soil response at Augusta site
527
Based on the 1-D seismic soil response (Maugeri & Frenna, 1987) and on the deterministic modelling
of the source mechanism (Bottari et al., 2003) of the 1693 destructive earthquake, as reported above,
the design spectra, more severe than that prescribed by the Italian Seismic Regulation at the time of
the work, has been evaluated (Figure 20) and used for the retrofitting work described in the following.
1.00 Design Earthquake

0.90 Italian Code II cat. e=1
Italian Code II cat. e=1.3
0.80
0.70
0.60
0.50
0.40
0.30
0.20
0.10
0.00
l T (s)
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0
Figure 20. The design spectra at Augusta site
Tests on piles
The reinforced concrete buildings under consideration are founded on bored piles having a length of
20 meters and a diameter of 50 cm. To check the integrity of the piled foundations after the
earthquake, the piles were inspected and an experimental loading tests program was carried out. Two
single piles of the piled foundation were chosen and their head was disconnected from the cap to
perform vertical (Figure 21.a) and horizontal loading tests (Figure 21.b). A detailed description of the
results of the loading tests carried out on piles is reported by Castelli & Maugeri (2004). At the end of
the loading tests, the connection between the steel bars (Figure 22.a) and the concrete (Figure 22.b) of
the piles to the foundation was restored.
Synthetically, according to the procedure suggested by Chin (1970), the vertical ultimate failure load
derived by the loading tests is ranging between around 1,5 MN and 1,7 MN, in a good agreement with
the value of 1,53 MN computed in undrained conditions by the well-known Vesics (1977)
relationships.
(a) (b)
Figure 21. View of a loading test on pile: a) vertical loading b) horizontal loading
528
(a) (b)
Figure 22. View of the pile-cap reconnection
To evaluate the pile performance due to lateral loads, after conducting the axial loading tests, two
lateral loading tests on single pile were also performed (Figure 23). The ultimate failure lateral load
derived by the load-deflection curves results greater than 20 kN.
The loading tests carried out on a three-piles group (Figure 24.a) subjected to horizontal load (Figure
24.b) furnished the results reported in Figure 25 in terms of load-deflection curve.
Horiz ontal Load [KN] Horizontal Load [KN]

0 2 4 6 8 10 12 14 16 18 0 3 6 9 12 15 18 21
0.0 0.0
0.2 0.5
0.4 1.0
0.6 1.5
0.8 2.0
1.0 2.5
1.2 3.0
1.4 3.5
1.6 4.0
1.8 Test no.1 4.5 Test no.2

2.0 5.0
(a) (b)
Figure 23. Single pile horizontal loading test results: a) Test no.1, b) Test no.2
Foundation retrofitting
A detailed description of the solutions adopted for the seismic retrofitting of the case study building
was reported by Oliveto & Decanini (1998). These solutions can be synthetically reported as follow:
1. substitution of the loosened soil;
529
2. enlargement of the foundation and addition of new piles;
3. better distribution of forces among the piles;
4. strengthening the corners of the building;
5. removal of the soft storey condition;
6. endowement of ductility to cores and corner elements;
7. strengthening of overstressed members of the existing structure;
8. repair of damaged structural members.
The excavation required for the visual inspection of the foundation and the piles loosened the soil
along the perimeter of the building. The excavations for the enlargement of the foundation and for the
connection with the existing one caused further loosening. For this reason all the loosened soil was
removed and substituted by cemented gravel. The soil substitution was extended to the perimeter of
the foundation for a depth of 3 m and a width of 2 m.
The vertical and horizontal load-bearing capacities of the existing pile foundation were not sufficient
for the design seismic actions. A strengthening of the foundation was necessary especially in the areas
adjacent to the stiffening cores. Thirty-six new piles of 50 cm in diameter and 25 m in length, longer
than the existing piles, were disposed as reported in the following.
(a) (b)
Figure 24. a) View of the piles group, b) Horizontal loading test on piles group
Horizontal Load [kN]

0 25 50 75 100 125 150 175 200
0
6
Pile Group
Figure 25. Piles group horizontal loading test results
530
The stiffening cores and corner elements by shear walls (Figure 26.a), involve a concentration of
vertical and horizontal forces in the pile groups around the foot of the shear walls and also for this
reason the pile foundation was retrofitted by the increment of the number of piles and its length
(Figure 26.b).
(a)
(b)
Figure 26. Seismic retrofitting of the Augusta building: a) structure retrofitting
b) foundation retrofitting
Inertial soil-pile group interaction

In the last decades several simplified approaches for the analysis of single piles or pile groups have
been developed that can be used with little computational effort. These methods have given results that
are often in remarkable agreement with the mathematical models (Novak, 1974; Markis & Gazetas,
1992; Abghari & Chai, 1995; Tabesh & Poulos, 2001).
Pseudo-static approaches, as example, for the seismic analysis and design of foundations are attractive
for design engineers because they are simple, when compared to difficult and more complex dynamic
analyses. There are two sources of loading of the pile: inertial loading of the pile head, caused by
the lateral forces imposed on the structure by the earthquake and which are then imposed on the piles,
and kinematic loading along the length of the pile, imposed by the lateral ground movements
developed by the earthquake, as will be illustrated in the following section.
Several approaches are currently available to analyze the soil-pile interaction subjected to lateral load
ranging from complex models, as non linear dynamic analysis by 2D or 3D finite element methods, to
the use of simplified approaches as limit equilibrium (LE) approach and p-y analysis approach.
The p-y approach for analyzing the response of laterally loaded piles is essentially a modification of
the basic Winkler model, where p is the soil pressure per unit length of pile and y is the pile deflection.
The soil is represented by a series of nonlinear p-y curves that vary with depth and soil type. The p-y
method is still widespread in practice and based on literature review it appears to be one of the most
used methods for civil engineers.
531
An improvement of these approaches can be achieved if more reliable and relevant soil properties are
adopted in defining the input parameters even if the design method is empirical or semi-empirical. The
shear modulus Go and for clayely soils, the undrained shear strength cu, can be considered essential
parameters for the estimation of the soil response, especially in the case of dynamic loading.
Researches based on results of field tests on full scale piles, both in cohesive than in cohesionless soil,
suggests to employ non linear p-y relationship (Reese et al., 1974; Reese & Welch, 1975; Reese &
Van Impe, 2001; Juirnarongrit & Ashford, 2006), thus in the proposed approach a hyperbolic p-y
relationship has been adopted:
yp( z )
p( z ) (6)
1 yp( z )
Esi( z ) p lim ( z )
where Esi [FL-2] is the initial modulus of horizontal sub-grade reaction, yp [L] is the pile lateral
deflection, p and plim [FL-2] are the mobilized and the ultimate horizontal soil resistance respectively.
Another term that is sometimes used in place of Esi is the coefficient of horizontal sub-grade reaction
kh, expressed in units of force per unit volume [FL-3]. The relationship between Esi and kh is Esi = kh D
being D the pile diameter. Nevertheless Esi is a more fundamental soil property because it is not
dependent on the pile size
The hyperbolic p-y relationship (6) is defined by the two parameters plim and Esi. By this approach a
situation of layered soil can be easily analyzed assuming for these parameters different values along
the pile length. The numerical model is based on an iterative procedure taking into account the
decrease of the model stiffness with the increasing of the applied horizontal load (Castelli et al., 1995;
Castelli & Maugeri, 1999; Castelli 2002; 2006).
The response of a laterally loaded pile group with relatively closely spaced piles is different from that
of a single pile, because of the interaction between piles through the surrounding soil. Experiments
have been carried out in the last decade with the aim of deriving general rules to adapt p-y curves to
take into account the group effects. Brown et al., (1988), as example, introduced the term shadowing
to mean the phenomenon for which the soil resistance of a pile in a trailing row is reduced because of
the presence of the leading pile ahead of it.
Brown et al., (1988) defined the concept of p-multiplier fm, a multiplier of the limiting values plim
capable of stretching the p-y curve for the single pile to account for the interaction among the piles in a
group. It has been shown that the shadowing effect becomes less significant as the spacing between
piles increases and is relatively unimportant for centre-to-centre spacing greater than about six pile
diameters (Cox et al., 1984; Brown & Shie, 1991; Ng et al., 2001).
Similar to the design of axially loaded piles using a efficiency method, in a pile group subjected to
horizontal load, the p-y relationships for the individual piles are modified to take into account the
group effects by stretching the curve in the direction of deflection.
To extend the method of the p-y curves to the case of a pile group, the p-multiplier concept proposed
by Brown et al., (1988) has been applied. The multiplier has obviously values in the range 0 to 1. As
suggested by Brown et al., (1988), a given row, within the group, could be represented by a single pile
p-y curve multiplied by a load factor to account for row position. With this aim, in the proposed
approach the p-y curve of the piles belonging to each row of the group is obtained by scaling down the
p-y curve of the single pile, multiplying the values of plim by a group reduction factor named fm.
It should be expected that the resulting initial modulus of horizontal sub-grade reaction of a pile group
532
is softer than the value of an isolated pile (Ashour et al., 2004). Thus, to adapt the hyperbolic p-y
relationships defined for a single pile to the case of a pile group, a multiplier should be considered also
for the initial modulus of horizontal sub-grade reaction Esi. Consequently, Esi of an individual pile in a
group will be reduced and we can assume for a pile group:
yp( z )
p( z ) (7)
1 y p( z )
m Esi( z ) f m p lim ( z )
where m is the empirical factor introduced in the proposed method to extend the single pile analysis to
the case of a pile group (Castelli, 2006). The multiplier m has obviously values in the range 0 to 1 and
it can be defined as the ratio between the initial modulus of horizontal sub-grade reaction of the pile in
a group (Esi)group and that of the single pile (Esi) single.
A rather large amount of experiments have been carried out in the last decade for the determination of
the p-multipliers (McVay et al., 1998; Rollins et al., 1998; Zhang et al., 1999; Rollins et al., 2006).
The advantage of the proposed approach is that the load-deflection response of the pile group can be
calculated using solutions for the response of a single pile. Naturally, this approach is based on an
estimation of only the average head lateral deflection of the pile group.
Validation of the proposed approach has been carried out comparing the numerical results with those
measured in loading tests, both on single pile and piles group. Model parameters have been assessed to
obtain numerical results close to those measured. The numerical analysis was carried out for the single
piles and then for the pile groups, considering in this last case the experimental curve in terms of
average lateral load per pile versus average pile head lateral deflection.
Taking into account the profile of the values of the undrained shear strength derived from in situ and
laboratory tests, the numerical analysis has been carried out with values of p lim ranging between 0,024
MN/m2 from the ground surface up to 1,33 MN/m2 at the pile tip (Broms, 1964). The initial modulus
was (Welch & Reese, 1972): Esi = (Esoi + ki z), being z the depth, ki equal to 19,75 N/cm3 the gradient
of the initial modulus of horizontal sub-grade reaction and Esoi the initial modulus at the ground
surface, that in the analysis was assumed equal to 37 MN/m2.
The single piles were in free-head condition while the pile group was in fixed-head condition. As the
horizontal load during the loading test was applied at a distance from the ground surface of 43 cm, the
numerical analysis has been carried out considering the pile loaded with a horizontal force and a
moment at the head. In Figure 27.a is reported the comparison between measured (Test no.1) and
computed single pile head lateral deflection and a good agreement can be observed between
experimental and numerical results.
To reduce the computed load-carrying capacity of the single pile to the pile group rows, according to
the design curve showing p-multiplier values reported by Rollins et al., (1998), a p-multiplier f m equal
to 0,4 was assumed. The values of the initial modulus of horizontal sub-grade reaction of the piles
group were reduced assuming for the empirical factor m the same value of fm. Also in this case a good
agreement between numerical results and experimental evidences can be observed (Figure 27.b).
Kinematic soil-pile group interaction

Generally in the pile design only the effects of inertial loading are considered, even if kinematic
loading is also very important (Gazetas, 1984; Tazoh et al., 1988; Masayuki & Shoichi, 1991;
Kavvadas & Gazetas, 1993). Part 5 of the Eurocode 8, as example, states that piles shall be designed
for the following two loading conditions:
533
Orizzontal Load [kN] Horizontal Load [kN]
0 2 4 6 8 10 12 14 16 0 10 20 30 40 50 60 70
0.0 0
Single Pile Pile Group
0.3 2
0.6 4
Computed
Computed
Measured
Measured
0.9 6
Lenght: L = 20 m Lenght: L = 20 m
1.2 Diameter: D = 0.50 m 8 Diameter: D = 0.50 m
Young Modulus: Reduction Factors:
E = 20 x 103 MPa fm and m = 0.40
1.5 10
(a) (b)
Figure 27. Comparison between experimental and computed load-deflection curve for:
a) single pile and b) pile group
a) inertia forces on the superstructure transmitted on the heads of the piles in the form of axial and
horizontal forces and moment;
b) soil deformations arising from the passage of seismic waves which impose curvatures and thereby
lateral strain on the piles along their whole length.
While there is ample experience of carrying out the equivalent static analysis for the inertial loading
(type (a)), no specific method or procedures are available to predict deformations and bending moment
from the kinematic loading (type (b)).
To take into account these two sources of loading, Abghari & Chai (1995) proposed an analysis in
which the pile was subjected to the free-field soil displacements at each node along its length. These
displacements were obtained from a separate free-field site response analysis. The inertial forces
acting on the pile were obtained from the product of mass and spectral acceleration. These forces were
applied to the pile as static forces. Similarly Tabesh & Poulos (2001) and Poulos (2006) proposed a
simple approximate methodology for estimating the maximum internal forces of piles subjected to
lateral seismic excitation.
The response of the complete system constituted by soil, foundation and superstructure is obtained by
the overlapping of the inertial and kinematics effects, even if, the validity of the principle of the
overlapping of the effects is correct only in the hypothesis of linear behaviour of all components
(Kausel & Roesset, 1974; Gazetas & Milonakis, 1998). Nevertheless, the overlapping of the effects
can be considered acceptable for systems moderately non linear (Milonakis et al., 1997).
The passage of seismic waves through the soil surrounding a pile imposes lateral movements and
curvatures on the pile, generating kinematic bending moments even in absence of a super-structure.
The simplified numerical approach adopted in the paper is based on the idea to couple and to evaluate
the effects due to the applied load at the pile head and lateral movements along the pile length by a
series of independent p-y curves, relating soil reaction and relative soil - pile movements. The
implementation of the method involves imposing a known free-field soil movement profile. When the
expected free-field movement is large enough to cause the ultimate pressure of laterally spreading
soils to be fully mobilized, the ultimate pressure, instead of free-field soil movement, may be used.
534
To take into account that the lateral pile response to static or dynamic loading is non linear, the
following p-y relationship could be adopted (Figure 28):
[ y p ( z ) y s ( z )]
p( z ) (7)
1 | y p ( z ) ys ( z ) |
Esi( z ) p lim ( z )
in which Esi, yp, p and p lim have the same meaning of equation (6) while y p is pile movement and ys is
the soil movement.
plim (z)
Lateral resistance p
single pile
E si (z) p(z)
pile group
p(z) = f m p(z)single
E si (z) = m E si (z) single
Relative pile-soil displacement (yp - y s)

Lateral resistance p
pile group
single pile p lim(z)
Figure 28. p-y curves to model soil-pile interaction due to lateral movements
According to the proposed approach, when an individual pile of a group is taken into consideration, in
equation (7) the empirical reduction factors fm and m should be used:
[ y p ( z ) y s ( z )]
p( z ) (8)
1 | y p ( z ) ys ( z ) |
m E si( z) f m plim ( z )
The approach proposed was employed and implemented in an original computer code. Comparison
with field measurements, shows that this procedure can give lateral deflection, bending moment and
shear force distributions on piles which agree with experimental ones (Castelli & Maugeri, 2007).
To evaluate the influence of the inertial and kinematic effects, the numerical analysis has been carried
out via the pseudo-static approach following the proposed procedure (Castelli & Maugeri, 2007). To
simulate the kinematic effects, the displacement profile of the free-field soil movements (Figure 19)
evaluated by the seismic soil response analysis (Maugeri & Frenna, 1995) has been considered. So that
numerical simulation provides bending moment distribution along the pile length. Two cases have
been considered:
- lateral inertial load but kinematic ground movements not included in the analysis (inertial effects
only);
- lateral inertial load ranging from 12,2 kN to 15 kN and kinematic effects (combined).
The results obtained for the single pile are summarized in Figure 29. Analyzing these results it is
possible to observe that, as expected, if kinematic effects (lateral movements) are ignored and only
535
inertial (lateral load applied at the pile head) effects are considered, the maximum moments along the
pile length can be underestimated. When combined effects are taken into consideration, an increasing
of the maximum bending moment can be observed.
By the way considering a seismic action of Figure 20, more severe than that adopted by the Italian
Seismic Regulation, and also that concentration of load beneath the shear walls and the kinematic soil-
pile interaction a number of piles were added (Figure 26).
Pile Moment (KNm)

-5 0 5 10 15 20 25 30 35 40 45 50
0
Combined Effects (H = 15 kN)

4
Combined Effects (H = 12.2 kN)
6 Inertial Effects (H = 15 kN)
8 Inertial Effects (H = 12.2 kN)
10
12
14
16
18
20
-5 0 5 10 15 20 25 30 35 40 45 50
Figure 29. Numerical results for combined effects on single pile (kinematic and inertial)
CONCLUDING REMARKS
buildings from many points of view. Preliminarily, local site response has to be checked to evaluate
the structure must resist. Two case histories of old structures, damaged by the moderate Sicilian
earthquake (Italy) of 13th December 1990 are discussed: the Noto Cathedral, founded on shallow
foundation, and some concrete buildings in Augusta, founded on pile foundations. For both case
histories a site response and a design earthquake were evaluated by determinist approach starting from
the source modelling and the evaluation of synthetic accelerograms.
The contribution of geotechnical engineering was very important to solve the question to demolish or
the retrofit the two structures. After the foundation retrofitting was cheeked successfully for both cases
the design of the superstructure retrofitting was possible.
The geotechnical design for improvement of the foundation conditioned also the improvement of the
overstructure for the Noto Cathedral. It was decided not only to rebuilt all that had collapsed, i.e. the
piers on the right side of the main nave, the roof of the naves and the dome, but also to pull down and
536
rebuilt the piers on the left of the nave, to realize a deeper foundation embedment and, above all,
because the standing piers had significant damage and however they had been made with a poor
workmanship, which was the principal reason of the big collapse of the cathedral.
Furthermore soil-structure interaction, considering the elasto-hypoplastic behavior of the soil, shows
that to reduce the stress level on the lateral arches, tie rods on the arch were utilized. At the same time,
the soil-structure interaction analysis, shows that for the central arches, which can be greatly stressed
due to foundation movements, a reinforcement was also needed. In order to guarantee a nice view
inside the cathedral, carbon fibers were utilized instead of tie rods.
Also in the case of the seismic retrofitting of a r-c building in Augusta, the geotechnical design for
improvement of the foundation, by additional piles, allows the improvement of the superstructure by
shear walls. In fact, the concentration of loads beneath the shear walls sustained by additional
foundation piles.
The inertial and kinematic soil-pile interaction shows that kinematic interaction, usually ignored in
engineering practice, must be performed, even if with simplified procedure to avoid unsafe design of
piles subjected to earthquake loads.
REFERENCES
Abghari A. and Chai J. Modeling of soil-pile superstructure interaction for bridge foundations.
Performance of deep foundations under seismic loading, J. P. Turner ed., ASCE, New York, 45-59,
1995.
Ashour M., Pilling P. and Norris G. Lateral behavior of pile groups in layered soils. Journal of
Geotechnical and Geo-environmental Engineering, ASCE, 130, (6), 580-592, 2004.
Binda L., Baronio G., Gavarini C., De Benedictis R. and Trincali S. Investigation on materials and
structures for the reconstruction of the partially collapsed Cathedral of Noto (Sicily). Proc. 6th Int.
Conf. Struct. Studies, Repairs and Maintenance of Historical Buildings, STREMAH99, Dresden,
Germany, 323-332, 1999.
Biondi G., Cascone E., and Maugeri M. Flow and deformation failure of sandy slopes. Proc. 10h
International Conference on Soil Dynamics and Earthquake Engineering, Philadelphia, USA, 7-10
October 2001. Soil Dynamics and Earthquake Engineering, 22 (2002), 1014-10, 2002.
Bottari A., Sara A., Teramo A., Termini D. and Carveni P. The earthquake on January 11th 1693 in
South-Eastern Sicily: macroseismic analysis and strong motion modelling in Noto. In:
Geotechnical analysis of seismic vulnerability of monuments and historical sites. Maugeri e Nova
Eds., Ptron Editore, Bologna, 2003.
Broms B.B. Lateral resistance of piles in cohesive soils. Journal of Soil Mechanics and Foundation
Engineering Div., ASCE, XC, (SM-2), 27-63, 1964a.
Brown D.A., Morrison C. and Reese L.C. Lateral load behavior of pile group in sand. Journal of
Brown D.A. and Shie C.F. Modification of P-Y curves to account for group effects on laterally
loaded piles. ASCE Geotechnical Congress 1991, 1, ASCE, Reston, Va., 479-490, 1991.
Castelli F. Discussion on Response of laterally loaded large-diameter bored pile groups by
C.W.W.Ng, L. Zhang & D.C.N. Nip. Journal of Geotechnical and Geo-environmental Engineering,
ASCE, 128, (11), 963-965, 2002.
Castelli F. Non linear evaluation of pile groups lateral deflection. Proceedings International
Conference on Piling and Deep Foundations, Amsterdam, 127-134, 2006.
Castelli F., Maugeri M. and Motta E. Non linear analysis of the deflection of a pile subjected to
horizontal loads. Italian Geotechnical Journal, XXIX, (4), 289-303, 1995 (in Italian).
Castelli F. and Maugeri M. Non linear soil effects in the design of pile group in pseudo-static
conditions. Proceedings IX National Conference Ingegneria Sismica in Italia, Torino (Italy),
1999 (in Italian).
537
Castelli F. and Maugeri M. A pseudo-static analysis for the evaluation of the lateral behavior of pile
groups, Proc. 4th International Conference on Earthquake Geotechnical Eng., Thessaloniki,
Greece, Paper no.1399, 2007.
Cavallaro A.M.F. and Maugeri M. Comportamento Tensionale e Deformativo dell'Argilla di Augusta
Sottoposta a Carichi Ciclici. Ingegneria Sismica, 1, 30-40, 1996.
Cavallaro A., Massimino M. R. and Maugeri M. Noto Cathedral: Soil and Foundation Investigation,
Journal of Construction and Building Materials, (17), 533-541, 2003.
Chin F.K. Estimation of ultimate load of piles not carried to failure, Proc. 2 nd SE Asian Conference
on Soil Engineering, Singapore, 81-92, 1970.
Cox W. R., Dixon D. A. and Murphy B. S. Lateral load tests on 25.4-mm (1-in.) diameter piles in
very soft clay in side-by-side and in-line groups. Laterally loaded deep foundations: analysis and
performance, ASTM STP 835, J. A. Langer, E. T. Mosley, and C. D. Thompson (ed), American
Society for Testing and Materials, 122-139, 1984.
Crespellani T., Madiai C. and Maugeri M. Analisi di stabilit di un pendio in condizioni sismiche e
post-sismiche. Rivista Italiana di Geotecnica, XXX(1), 50-61, 1996.
De Benedictis R. and Trincali R. La ricostruzione della Cattedrale di Noto, L.C.T. edizioni, Ispica,
2000.
Di Prisco C., Nova R. and Sibilia A. Shallow footing under cyclic loading: experimental behaviour
and constitutive modelling. Geotechnical analysis of the seismic vulnerability of historical
monuments, Maugeri M. & Nova R. Editors, Ptron editore, Bologna, 2003.
Di Prisco C., Massimino M.R., Maugeri M., Nicolosi M. and Nova R. Cyclic numerical analyses of
Noto Cathedral: soil-structure interaction modelling. Italian Geotechnical Journal, XL(2), 49-64,
2006.
Dobry R. and Gazetas G. Simple Methods for Dynamic Stiffness and Damping of Floating Pile
Groups. Geotechnique, 38, (4), 557-574, 1988.
EC8-Part 1 Design of structures for earthquake resistance Part 5: General rules, seismic actions and
rules for buildings. European Prestandard, ENV 1998, European Committee for Standardization,
Brussels, Belgium, 2003.
Federal Highway Administration FHWA. Design and construction of driven pile foundations. NHI
Course Nos.13221 and 13222, Publication FHWA-HI-97-013, 1996.
Frenna S.M. and Maugeri M. Il codice di calcolo GEODIN per lanalisi della risposta dinamica dei
terreni. Atti del IX Convegno Italiano di Meccanica Computazionale. Universit di Catania. 20-
22/06/1995, 145-148, 1995.
Garassino A., Jamiolkowski M. and Pasqualini E. Soil modulus for laterally loaded piles in sand and
N.C. clays. Proceedings VI European Conference on Soil Mech. and Found. Engineering, Vienna,
1976.
Gazetas G. Seismic response of end-bearing single piles. Soil Dynamics Earthquake Engineering,
Vol.3, no.2, 82-93, 1984.
Gazetas G. Foundation Vibrations, in Foundation Engineering Handbook, (H.Y. Fang, ed.) Van
Nostrand Reinhold, New York, NY, 553-593, 1991.
Giuffr A. and Giannini R. La risposta non lineare delle strutture in cemento armato. Progettazioni e
particolari costruttivi in zone sismiche. ANCE-AIDIS, 1982.
Hudson D.E. The role of Geotechnical Engineering in earthquake problems. Proc. Int. Conf. On
Recent Advances in Geotechnical Earthquake Engineering and Soil Dynamics. St. Louis, MO,
1981.
Kausel E. and Roesset J.M. Soil Structure Interaction for Nuclear Containment Structures,
Proceedings ASCE, Power Division Specialty Conference, Boulder, Colorado, 1974.
Kavvadas M. and Gazetas G. Kinematic seismic response and bending of free-head piles in layered
soil. Gotechnique, 43, no.2, 207-222, 1993.
Kuwabara F. Settlement behavior of non-linear soil around single piles subjected to vertical loads.
Soils and Foundations, XXXI, (1), 1991.
Juirnarongrit T. and Ashford S.A. Soil-Pile response to blast-induced lateral spreading. II: Analysis
and Assessment of the p-y method. Journal of Geotechnical and Geo-environmental Engineering,
ASCE, 132, (2), 163-172, 2006.
538
Laurenzano G. and Priolo E. Numerical modelling of the December 13, 1990 M = 5.8 eastern Sicily
earthquake. In: Seismic Prevention of Damage: A Case Study in a Mediterranean City. M.
Maugeri Editor. WIT Press, Southampton (UK), 2005.
Liyanapathirana D.S. and Poulos H.G. Pseudostatic approach for seismic analysis of piles in
liquefying soil. Journal of Geotechnical and Geo-environmental Engineering, ASCE, 131, (12),
1480-1487, 2005.
Mandolini A., Russo G. and Viggiani C. Pile foundations: Experimental investigations, analysis and
design. Proceedings of the 16th International Conference on Soil Mechanics and Geotechnical
Engineering (16ICSMGE), Osaka, (1), 177-213, 2005.
Makris N. and Gazetas G. Dynamic pile-soil-pile interaction. Part II : Lateral and seismic response.
Earthquake Engineering and Structural Dynamics, 21, 145-162, 1992.
Masayuki H., and Shoichi N. A study on pile forces of a pile group in layered soil under seismic
loading, Proceedings 2nd International Conference on Recent Advances in Geotechnical
Earthquake Engineering and Soil Dynamics, St. Louis, 1991.
Matlock H. Correlations for design of laterally loaded piles in soft clay. Proceedings II Offshore
Technical Conference, Houston, Texas, (1), 577-594, 1970.
Massimino M.R. and Maugeri M. Design spectrum by synthetic input accelerograms for Noto
Cathedral, Geotechnical analysis of seismic vulnerability of monuments and historical sites,
Maugeri & Nova Eds, Ptron Editore, Bologna, 2003a.
Massimino M.R. and Maugeri M. Retrofitting of the foundation of Noto Cathedral (Italy) according
to Eurocodes, Geotechnical analysis of seismic vulnerability of monuments and historical sites,
Maugeri & Nova Eds, Ptron Editore, Bologna, 2003b
Massimino M.R., Maugeri M. and Perretti A. Analisi non-lineare dellinterazione terreno-
fondazione-sovrastruttura per la Cattedrale di San Nicol a Noto, Proc. XXII Convegno Nazionale
di Geotecnica, Palermo, 22-24 Settembre, 2004, 395-402, 2004.
Maugeri M., and Frenna S.M. Modello isteretico semplificato per la determinazione della risposta dei
terreni in campo non lineare, Proceedings III National Conference Ingegneria Sismica in Italia,
Vol.II, 169-288 (in Italian), 1987.
Mc Vay M., Zhang L., Molnit T. and Lai P. Centrifughe testing of large laterally loaded pile groups
in sands. Journal of Geotechnical and Geo-environmental Engineering, ASCE, 124, (10), 1016-
1026, 1998.
Mylonakis G., Nikolaou A. and Gazetas G. Soil-pile-bridge seismic interaction: kinematic and
inertial effects. Part I: soft soil, Earthquake Engineering Structural Dynamics, 27, (3), 337-359,
1997.
Montrasio L. and Nova R. Settlements of shallow foundations on sand. Gotechnique, 41(2), 243-
256, 1997.
Mylonakis G. and Gazetas G. Seismic soil-structure interaction: beneficial or detrimental?. Journal
of Earth. Eng., Imperial College Press, 4(3), 277-301, 2000.
NAVFAC. Foundations and earth retaining structures design manual. Depart. of Navy, DM 7.2,
Alexandria, Va, 1982.
NEHRP Recommended provisions for seismic regulations for new buildings and other structures,
Parts 1 and 2. Building Seismic Safety Council, Washington D.C, 1997
Ng C.W.W., Zhang L. and Nip D.C.N. Response of laterally loaded large-diameter bored pile
groups. Journal of Geotechnical and Geo-environmental Engineering, ASCE, 127, no.8, 658-669,
2001.
Novak M. Dynamic stiffness and damping of piles. Canadian Geotechnical Journal, Ottawa, 11,
574-598, 1974.
Ohta Y. and Goto N. "Empirical Shear Wave Velocity Equations in Terms of Characteristic Soil
Indexes", Earthquake Engineering and Structural Dynamics, 6, 1978.
Panza G.F., Romanelli F. and Caccari F. Seismic wave propagation in laterally heterogeneous
anelastic media: Theory and applications to seismic zonation. Advances in Geophysics, 43, 1-95,
2000.
Postpischl D. Atlas of isosismal maps of Italian Earthquake, Italian National Research Council
(CNR), Geodynamical Project, Roma, 1985.
539
Poulos H.G. Ground movements A hidden source of loading on deep foundations. John Mitchell
Lecture. Proceedings International Conference on Piling and Deep Foundations, Amsterdam, 2-
19, 2006.
Reese L.C., Cox W.R. and Koop F.D. Analysis of laterally loaded piles in sand. Proc. VI OTC,
Houston, Texas, 473-483, 1974.
Reese L.C. and Welch R.C. Lateral loading of deep foundations in stiff clay. Geotechnical Testing
Journal, GTJODJ, XII, (1), 30-38, 1975.
Reese L.C., Wang, S.T., Isenhower, W.M., Arrllaga, J.A. and Hendrix, J. LPILE Plus Version 4.0
m. Ensoft, Inc. Austin, Texas, 2000.
Reese L.C. and Van Impe W.F. Single piles and pile groups under lateral loading, Balkema,
Rotterdam, The Netherlands, 2001.
Richard R., Elms D.G., Budhu M. Seismic bearing capacity and settlements foundations. Journal of
Geotechnical and Engineering. 119(4), 662-674, 1993.
Robertson P.K., Davies M.P. and Campanella R.G. Design of laterally loaded driven piles using the
flat dilatometer. Geotechnical Testing Journal, GTJODJ, 12, (1), 30-38, 1989.
Rollins K.M., Peterson K.T. and Weaver T.J. Lateral load behavior of full-scale pile group in clay.
Journal of Geotechnical and Geo-environmental Engineering, ASCE, 124, (6), 468-478, 1998.
Rollins K.M., Olsen K.G., Jensen D.H., Garrett B.H., Olsen R.J., and Egbert J.J. Pile Spacing Effects
on Lateral Pile Group Behaviour: Analysis, Journal of Geotechnical and Geo-environmental
Engineering, 132, (10), 1272-1283, 2006.
Sara A, Das S., Suhadolc P. Effect of non-uniform station coverage on the inversion for seismic
movement release istori and distribution for a Haskell-type rupture model. Journal Seism., 2, 1-25,
1998.
Seed H.B., Ugas C., Lysmer J. Site dependent spectra for earthquake resistant design. Report N
EERC 74-12, Earthquake Engineering Research Center, University of California, Berkeley, 1974.
Tabesh A. and Poulos H.G. Pseudo-static approach for seismic analysis of single piles. Journal of
Tazoh T., Wakahara T., Shimizu K., and Matsuzaki M. (1988) Effective motion of group pile
foundations, Proceedings 9th World Conference Earthquake Engineering, Tokyo, Vol.3, 587-592,
1988.
Towhata I. Photographs and motion picture of the Niigata city immediately after the earthquake in
1964. The Japanese Geotechnical Society, 2004.
Tringali S. and De Benedictis R.Il progetto di ricostruzione della cattedrale di Noto: il sistema delle
crociere e gli aspetti di impatto socio-economico sulla citt ed il comprensorio. Proc. Int.
Workshop on Ricostruendo la Cattedrale di Noto e la Frauemkirche di Dresda, Noto (Siracusa),
Italy, 143-149, 2001a.
Tringali S. and De Benedictis R. Il progetto di ricostruzione della cattedrale di Noto: il sistema delle
navate. Proc. Int. Workshop on Ricostruendo la Cattedrale di Noto e la Frauemkirche di Dresda,
Noto (Siracusa), Italy, 151-158, 2001b.
Vesic A.S. Design of pile foundations, Transport Res. Board., Report No.42, 68 p., 1968.
Welch R.C. and Reese L.C. Laterally loaded behavior of drilled shafts. Research Report n.3-5-65-
89, Center for Highway Research, University of Texas, 1972.
Zhang L., Mc Vay M. and Lai P. Numerical analysis of laterally loaded 3x3 to 7x3 pile groups in
sands. Journal of Geotechnical and Geo-environmental Engineering, ASCE, 125, (11), 936-946,
1999.
Yokota K., Imai T. and Konno M. Dynamic deformation characteristics of soils determined by
laboratory tests. OYO Tec. Rep. 3, 13-37, 1981.
Yoshida Y. and Motonori, I. Empirical formulas of SPT blow-counts for gravelly soils. Proc.
ISOPT-1, Orlando (USA), 1988.
540
June 25-28, 2007
Paper No. W4-1002
INFLUENCE OF DYNAMIC LOADS (EARTHQUAKE LOADING AND

AIRPLANE CRASHES) ON THE BEARING CAPACITY OF PILES, A
CASE STUDY
Jost A. Studer 1 , Hansjrg Gysi 2
ABSTRACT
The behavior of a pile subsoil system under dynamic loading is discussed. Based on an actual design
problem, the influence of a dynamic load on the strength and stiffness properties of a granular soil
deposit is investigated. A case study is given by discussing on-site tension tests of a pile, and applying
the corresponding conclusions for the final design.
Keywords: Tension piles, earthquake, dynamic loads, airplane crash
INTRODUCTION
The design actions on an investigated building of high importance comprehended, apart from normal
static loads, also earthquake loads as well as dynamic actions arising from an airplane crash. Due to
operational and geometrical restraints, the structure to be designed was very slender, resulting in high
overturning moments under these extraordinary actions. For this reason, a tension pile foundation (or
even a wall composed of adjacent piles) was foreseen for the design of the building. For preliminary
calculations, admissible dynamic skin friction values were estimated. Subsequently, several full-scale
pull-out tests (of 8m and 12m long piles) have been performed to verify these values in-situ. Gysi
Leoni Mader AG bore the main responsibility for this project, whereas Studer Engineering was in
charge of the influence of dynamic effects on the material properties.
GENERAL SOIL CONDITIONS
The subsoil at the investigated building site consists of lower terrace sandy gravel with pebbles and
boulders, under a man-made filling of clean sandy gravel. Occasionally, sand lenses of alluvial sands
are present, as well as cemented zones. The ground water level lies at a depth of about 7m. Lower
terrace layers generally represent good stable subsoils.
Following the borings performed during the tests, uniform subsoil conditions can be assumed,
predominantly a coarse gravel with USCS-classification "GP". The sand content (coarse to medium
sand) varies between 20 and 35%.
1
Dr. sc. techn., dipl. Bauing. ETH, Studer Engineering, Zrich, Email: studer@studer-engineering.ch
2
dipl. Bauing. ETH, Gysi Leoni Mader AG, Zrich, Email: geotech.gylemad@bluewin.ch
541
MATERIAL PROPERTIES
Compactness
From several performed investigations on this site, such as SPT tests, sheet pile wall construction tests,
drive tests and other values from local subsoil experience, the subsoil can be classified as dense or
very dense. For instance, the SPT values (N30) in the relevant layers show an average of 60 (for 30cm
penetration), which indicates a very compact soil.
Resistance and stiffness parameters

Following the results of triaxial tests on similar lower terrace layers, the friction angle for calculations
was assumed as 'o = 44 2. Concerning the stiffness, an oedometer modulus value (called "ME
modulus" henceforth) of 120 20 MPa (at a depth of 3m) is obtained from different plate tests. The
re-loading oedometer modulus is taken as M'E = 3 ME. Assuming a hyperbolic stress-strain
relationship of the subsoil, the ME modulus can be further calculated for all depths.
Other deformation modules are calculated with the aid of the well-known formulas
(1 2v)
G ME (1)
2(1 v)
(1 2v )(1 v)
E ME (2)
(1 v)
E
G (1) (3)
2(1 v )
where has been taken as = 0.33 0.02 for first loading and ' = 0.20 0.01 for reloading.
Dynamical soil properties
Shear wave velocity profile

Since no specific dynamic material tests were available, the needed dynamical soil properties have
been assessed by different correlations with available field test results. A shear wave profile of the
subsoil (Table 1) had been previously assessed by SASW tests (Spectral Analysis of Surface Waves),
so that the maximum shear modulus Gmax could be calculated and adjusted for the different layers.
Table 1. Shear wave profile of the subsoil

Depth vs
Layer
[m] [m/s]
Sandy gravel 2 200
Lower terrace 5 275
Lower terrace 8 305
Lower terrace 11 400
Bedrock > 25 -
542
Linear equivalent material properties
For the dynamical soil properties, a linear equivalent approach was chosen, so that the maximum shear
modulus Gmax and the minimum material damping Dmin could be attached to standard reduction curves
for linear equivalent behavior. In the present case, the corresponding reduction curves of Ishibashi and
Zhang (1993) have been assumed (Figure 1). This is a conservative choice since a somewhat lower
damping is obtained by these curves in comparison to other possible curves for the present soil
conditions (for example Hardin and Drnevich (1970)).
0.4
1
75 kN/m2 75 kN/m2
0.35
0.9
150 kN/m2 150 kN/m2
0.8 250 kN/m2 0.3 250 kN/m2
0.7
0.25
0.6
0.2
0.5
0.4 0.15
0.3
0.1
0.2
0.05
0.1
0 0
0.0001 0.001 0.01 0.1 1 10 0.0001 0.001 0.01 0.1 1 10
Shear Strain (%) Shear Strain (%)
Figure 1. Linear equivalent material properties
The above definition of the dynamical soil properties was judged to be a conservative approach, since
the energy radiation due to geometric damping (soil structure interaction) was not taken into account.
It has to be mentioned that the geometric damping can be very high for vertical and horizontal loading,
but is generally very small for rotational (tilting) movement.
Dynamical bedding factor

The dynamical bedding factor has been evaluated as about 60 - 80 MN/m3 based on literature values
(Grundbau Taschenbuch (1982)) for the relevant strain interval to be expected.
Admissible foundation soil pressure

Another issue was the assessment of an "admissible dynamical foundation soil pressure". This is a
delicate matter since the "admissible soil pressure", as for the static case, is not a material property
itself, but depends, among other things, on the dynamic resistance and stiffness properties, the loading
velocity and the foundation geometry. For the dynamic case it can be assumed that for well
compacted, dilative granular soils the resistance against a base failure is considerably higher than for
the static case, in the range of about two to three times higher. The following remarks can be given in
this respect for the case of an earthquake action:
The loading velocities for earthquake actions do not strongly influence the resistance and
stiffness properties of granular soils, so that these are only marginally higher than the
corresponding static values.
The dynamic loads do not act permanently, but only for limited periods of time. So even if the
resistance is exceeded during short periods of time, this does not lead to a global failure, but
only to additional deformations of the considered sliding mass.
During the earthquake action, both gravel and clay materials can be considered as undrained.
For moderate earthquake levels with a PGA in the range of 0.15 to 0.25 g and the type of
gravelly materials as in the present case, potential earthquake-induced excessive pore-water
pressures can be neglected for the estimation of an admissible foundation soil pressure.
543
Admissible dynamic skin friction
The admissible dynamic skin friction to be taken into account has been assessed by means of
correlations with static values for the skin friction based on results of SPT tests. For the dynamic case,
the possibility of a pore-water pressure increase has to be investigated, since the skin friction is
reduced in this case. For the present investigations, given the well compacted ground, the earthquake-
induced excess pore-water pressures are expected to be low, however, to remain on the conservative
side, only the static skin friction values have been assumed. The correlation with an SPT value N 30 of
63 leads to an admissible skin friction of about 200kN/m2 for piles with a length between 8 and 14 m.
Possibly, injections are necessary to get this value (although this was not the case for the present
situation).
Dynamic coefficients of earth pressure

Active and passive coefficients of earth pressure could be calculated with the method by Mononobe-
Okabe (1926, 1929); however taking the corresponding static values is a conservative approach in the
present case.
PILE TESTS
Geometry and construction

Both test piles (8 m and 12 m reinforced concrete piles) are situated within the area of the planned
building. A general test situation for one pile is shown in Figure 2. The diameter of both piles is
90 cm. The top edge of the reinforced concrete piles is actually 8 m below ground, while the upper
8 m are filled after the pile completion. The tensile force is applied by means of tension cables at
2.5 m below the top edge of the piles.
The construction of the piles was carried out with the "contractor method", in order to avoid potential
air inclusions. The filling of the upper 8 m consisted of concrete gravel with grain sizes of 0 to 32 mm
enriched with opalite (to minimize the influence of the filling on the test results).
Due to the massive reinforcement, the installations for the application of the tensile forces and the
measuring installations, it was decided to use a fine-grained concrete, with a compression strength of
B45/35. The used concrete is a self compacting concrete (SCC) with a grain size distribution of 0 to
8 mm and a cement content of 450 kg/m3 (CEM I 42.5). The reinforcement cage consists of a spiral
reinforcement with diameter 12 mm and a longitudinal reinforcement of 26 rods with diameter 34 mm.
Performing of the tests

The test execution for each pile comprehended the following steps:
set a new tension load level
metering of dial gauges immediately after the new tension was applied
metering of vibrating wire strain gauges, after about six minutes
conducting the leveling, after about ten minutes
after creep decay (usually after 10 and 30 minutes) apply the next load level
Results
The main results are represented by the leveling measurements and the maximum tension level. For
the interpretation, the influence of the pile head as well as of the filling had to be taken into account
(the influence on the results is in the order of 3-4%). In addition, the influence of the loading
equipment (i.e. the load propagation from the abutment fundaments of the loading equipment) had to
be taken into account, which in average lies in the order of 4-7%.
544
Figure 2. Loading equipment and general situation.
Left: axes 1 and 3 abutments of loading equipment, axis 2 pile location;
Right: section through axis M-M
Test for 8m pile

The test ended at a load level of 7000 kN and 62.5 mm effective uplift. The failure state was not
achieved. The corresponding mean skin friction was assessed as 300 kN/m2.
Test for 12m pile

The test ended at a load level of 7000 kN and 52.6 mm effective uplift. The failure state was not
achieved. The corresponding mean skin friction was assessed as 200 kN/m2.
Skin friction at failure

The average skin friction at failure is calculated as 230kN/m2, according to the formula
Z
m (4)
2r L
where Z = tensile force [kN], r = pile radius [m], L = pile length [m], considering the influence of the
self-weight, the pile head, the filling and the loading equipment. A value of 230kN/m2 has been
obtained, representing the mean value for the 8m and the 12m test, with a corresponding mean depth
of 13m below surface.
Assessment of oedometer modules M E

The oedometer modules ME were assessed with the aid of diagrams by Poulos / Davis (1980), for axial
loading. Actually, no diagrams for tension loading are present, so the corresponding diagrams for pure
friction piles on compression are used. By doing so, the part representing the pile base (included in the
diagrams) had to be taken into account, because this part is not present for tension loading (Figure 3).
Since the mobilized skin friction is determined by the effective horizontal stresses, the resulting ME
modulus is primarily the horizontal value MEh.
545
Figure 3. Percentage of load taken by pile base (Poulos and Davis, 1980)
In the present case, for the load taken by the pile base, a percentage of 11.5% (for the 8m pile)
respectively 8% (for the 12m pile) is calculated.
The calculation of the MEh modulus is then based on the diagram of Figure 4, with which a
displacement influence factor I P is first obtained for L/H = 0. Subsequently the ES modulus is
calculated. The conversion from ES to ME was then done with a Poisson's ratio of 0.33.
Figure 4. Displacement influence factor (Poulos and Davis, 1980)
For the present case, MEh modules of about 50 MN/m2 (8m pile) and 40MN/m2 (12m pile) are
obtained.
The calculation of the MEv modules was based on the measured settlements in the area of the
fundaments for the loading equipment. The settlement s is, under assumption that MEv = constant and
by considering an influence depth of three times the foundation width (3B):
546
3B
1
s v ' ( z ) dz (5)
M Ev 0
By assuming a linear stress reduction from the fundament top ( v ' (0) p o ) till a depth of 3B
( v ' (3 B) 0) , s can be written as
1 po 3 po B
s 3 B (6)
M Ev 2 2 M Ev
For the present case, with the measured settlements of about 8.7mm (8m pile) and 7.7mm (12m pile),
MEv modules of about 140 MN/m2 and 170MN/m 2 are obtained, respectively.
Transfer from single piles to multiple piles

Single piles will influence each other within a certain distance. The minimum distance for which two
piles cease to influence each other can be obtained from the distribution of the effective stresses at
failure. It is assumed that two piles do not influence each other as long as the principal stresses are
more or less not rotated in respect to the vertical direction. For both piles, the corresponding distance
is calculated as 3.5 times the corresponding diameter, measured from the pile axis. So starting from a
clear distance of 6 times the diameter, the piles do no more influence each other.
The horizontal coefficient of active earth pressure can be calculated for a single pile from the effective
vertical stress, the mean pile skin friction and the friction angle along the pile. To get the
corresponding coefficient for a pile group, it is assumed that the horizontal coefficient of active earth
pressure can be linearly interpolated as a function of the pile distance, with the aid of the diagram in
Figure 5.
Figure 5. Influence of the clear pile distance on the horizontal coefficient of active earth pressure
k h; for d = 0: k h according to soil parameters; for d = 6: kh from pile tests
547
CONSIDERATIONS FOR EARTHQUAKE LOADING AND AIRPLANE CRASH ACTIONS
General considerations
Earthquake loads or airplane crashes are dynamic actions, and as such their load velocity influences
the resistance and stiffness parameters of a soil.
An airplane crash action is a short impulse acting on the considered structure. As a consequence, the
structure is first displaced from its initial position and then oscillates with a certain damping. Thereby,
the foundation is primarily subject to compression and tension. In the present case, the tensile forces
are borne by the piles, while the compression forces, due to the construction details, are borne by the
soil. The damping is represented by the system damping as well as by the material damping. While the
system damping is defined by the geometric conditions, the material damping depends on the strain
range in the vicinity of the pile mantle. Higher strains cause a higher material damping than lower
strains. The oscillation frequency of the structure, and consequently the alternation frequency of
tensile forces in the piles, is primarily determined by the eigenfrequency of the structure / subsoil
system.
The frequency range for airplane crash actions is the same as the range expected for earthquake loads.
Resistance and stiffness characteristics of the subsoil

In the frequency range as expected for an earthquake action, the influence of the loading velocity on
the resistance and stiffness properties is only marginal in respect to other influence factors. The
stiffness depends on the soil type, its compactness and its stress state as well as on the material strain.
So both in the case of airplane crash and of earthquake loading, the influence of the loading velocity
on the stiffness properties can be neglected. The resistance parameters too are only secondarily
influenced by the loading velocity, for the considered frequency range. That is, the loading velocity
tends to increase the resistance. So the use of known static parameters is acceptable and lies on the
conservative side.
In loose, granular materials (namely poor graded sands) cyclic loadings can induce a pore-water
pressure increase. For a particular soil with given conditions, this increase depends on the load
amplitude and the number of cycles. The present subsoil is a well compacted sandy gravel. For load
amplitudes as expected for the safety earthquake, only a minor, negligible pore-water pressure increase
has to be expected. So the static resistance is not reduced by the dynamic loads and is therefore an
admissible choice for the cases of airplane crash and earthquake loads.
Influence of cyclic loading on the tension resistance of a pile

There are only sparse investigations concerning the influence of a cyclic load on the tension resistance
of piles to be found in the literature. Figure 6 (Gruber et al., 1985) illustrates model results on axially
cyclic loaded piles. These results show that the behavior of the piles heavily depends on the amplitude
of the alternating load.
548
Figure 6. Load-deformation diagram for cyclic loading (Gruber et al., 1985)
From Figure 6 it is apparent that for a low number of cycles, the deformation increase caused by the
cyclic loading is limited. Only for a high number of cycles the influence is relevant. In the present
case, during the safety earthquake the maximum number of cycles to be expected is very limited,
surely below ten. For the operating basis earthquake, more cycles are possible, but here the
corresponding amplitudes are considerably smaller.
As long as the failure skin friction is not exceeded, in the dynamic case the tension resistance is also
determined by the inertial mass of the pile and of the soil volume activated by the skin friction. This
means that higher dynamic loads can be borne compared to the static case, even if the material
properties were not influenced by the load velocity. So the design based on static material parameters
alone is acceptable and lies on the conservative side. It may be assumed that this is also valid for a pile
group or a continuous wall.
CONCLUSIONS
The planned full-scale pile tests were successfully conducted. The obtained in-situ test values
confirmed and even surpassed the predicted values, even without injections.
As thoroughly discussed, in the present case the dynamic actions do not negatively influence the static
resistance properties; the corresponding admissible values are even higher. So the choice to use the
known static values is acceptable and conservative. The potential risk of earthquake-induced excess
pore-water pressures is not pronounced in the present case, based on the given soil compactness and
the moderate earthquake action.
For the present case, given the advantageous subsoil conditions, it could be shown that using static
instead of dynamic material properties is a conservative approach. However, based on this project
experience, it is found that further research is needed in the field of dynamic loading on piles or pile
groups, particularly for the case of cyclic tensile forces.
549
REFERENCES
Brinkgreve (2002), "Plaxis 2 D-Version 8", Delft University of Technology, Balkema

Oda (1972 a), "Initial Fabrics and Their Relations to Mechanical Properties of Granular Material,
Soils and Foundations, Vol. 12, No 1, pp. 17 - 37
Gruber N, Koreck H-W, Schwarz P (1985), "Beitrge zum Tragverhalten axial zyklisch belasteter
Pfhle, Schriftenreihe TU Mnchen, Heft 5
Grundbau Taschenbuch (1982), Ernst & Sohn, 3, Auflage, Kap. 3.2.1, Tabelle 8
Hardin BO, Drnevich VP (1970), "Shear modulus and damping in soils; I. Measurements and
parameters effects, II. Design equation and curves". technical Reports UKY 27-70-Ce 2 and 3,
College of Engineering, University of Kentucky, Lexington
Ishibashi I, Zhang X (1993), "Unified dynamic shear moduli and damping ratios of sand and clay".
Soils and Foundations, Vol. 3, pp 1129-1143
Lambe, Whitman (1968), "Soil Mechanics, pp. 166
Mononobe N, Matsuo H (1929), "On the determination of earth pressures during earthquakes".
Proceedings of the World Engineering Congress, 9 p
Okabe S (1926), "General theory of earth pressures", Journal of the Japan Society of Civil
Engineering, Vol. 12, No. 1
Poulos, HG, Davis EH (1980), "Pile Foundation Analysis and Design". John Wiley and Sons, New
York.
SIA Norm 267 "Geotechnik", Schweizerischer Ingenieur- und Architektenverein, 2003
SIA Norm 267/1 "Geotechnik Ergnzende Festlegungen", Schweizerischer Ingenieur- und Architek-
tenverein, 2003
SIA Vornorm V 191 "Vorgespannte Boden-und Felsanker", Empfehlung SIA, Schweizerischer Inge-
nieur- und Architektenverein, 1995
SIA Vornorm V 192 "Pfhle", Empfehlung SIA V 192, Schweizerischer Ingenieur- und Architekten-
verein, 1996
Studer JA, Koller MG, "Bodendynamik", Springer Verlag Berlin, 1997
550

4-th ICEGE Workshops - Tessaloniki, Greece PDF

Cargado por

Información del documento

Título original

Derechos de autor

Formatos disponibles

Compartir este documento

Compartir o incrustar documentos

Opciones para compartir

¿Le pareció útil este documento?

¿Este contenido es inapropiado?

Copyright:

Formatos disponibles

4-th ICEGE Workshops - Tessaloniki, Greece PDF

Cargado por

Copyright:

Formatos disponibles

ARISTOTLE UNIVERSITY OF THESSALONIKI

LABORATORY OF SOIL MECHANICS, FOUNDATION & GEOTECHNICAL EARTHQUAKE ENGINEEERING

4th International Conference

NEW ORLEANS LEVEE PERFORMANCE IN HURRICANE KATRINA: R. B. SEED 29

Keynote Lecture Presentation

WORKSHOP 1: LARGE SCALE FACILITIES, GEOTECHNICAL STRONG MOTION ARRAYS

W1-1012 15 YEARS OF EUROSEISTEST Kyriazis PITILAKIS, Dimitris RAPTAKIS, Konstantia 211

W2-1006 GROUTING OF THREE-LEAF MASONRY: EXPERIMENTAL EVIDENCE Elizabeth VINTZILEOU 477

WORKSHOP 4: HOW CAN EARTHQUAKE GEOTECHNICAL ENGINEERING CONTRIBUTE TO

Invited Lectures Presentations

Keynote Lecture Presentation

Inertial damage to superstructure and pile at

Soil depth (m)

Depth (m) 116

We are starting to understand better detailed

PILE RESPONSE TO LATERAL SPREADING: FIELD

Adalier, K. (1996) Mitigation of Earthquake Induced Liquefaction Hazards, Ph.D. Thesis,

Youd, T.L. (1993), Liquefaction-Induced Damage to Bridges, Transportation Research Record,

Zeghal, M., and El Shamy, U. (2004). A Continuum-Discrete Hydromechanical Analysis of Granular

REFLECTIONS ON THE IMPORTANCE OF THE QUALITY OF THE

Jean-Louis CHAZELAS 1, Gopal SP MADABHUSHI2

Keywords: actuators, earthquakes, input motions, geotechnical engineering, centrifuge modelling

LABORATORY SPECIFICATIONS OF A HIGH QUALITY 1D SHAKING TABLE

Table I : Specifications of C-Core and LCPC Earthquake simulators

Figure 1. Actityns EQS - Dynamic equilibrium payload counterweights

Figure 4. Fourier Spectrum of Landers earthquake simulation with spurious accelerations

EARTHQUAKE GEOTECHNICAL ENGINEERING RESEARCH AT CAMBRIDGE

Figure 5. A view of SAM earthquake actuator at Cambridge

Figure 7. Acceleration time-histories

Figure 9. Upward propagation of the suction spike

Figure 10. Schematic view of the 2-D actuator assembly

Note: All parameters above are in prototype scale

Figure 11. Schematic view of the 2-D actuator

Bristol New Oxford

Figure 12. UK-NEES INSORS Facility

The European resources in experimental facilities present an important evolution: Schofield

STUDY OF PILE RESPONSE TO LATERAL SPREADING USING

Ricardo DOBRY1, Sabanayagam THEVANAYAGAM2, Tarek ABDOUN3, Ahmed ELGAMAL4,

(a) (b) (c)

Shape-acceleration-arrays sensor (1g testing)

High speed camera (centrifuge testing)

Visualization tools and System Identification

Finite Element (FEM) Numerical Simulations

TRENDS AND OPPORTUNITIES IN THE FURTHER USE AND

Michel Gradin and Fabio Taucer

VISUALIZATION OF LARGE-SCALE SEISMIC DATA RECORDS

Falko Kuester 1, Tara C. Hutchinson2, Tung-Ju Hsieh 3

Figure 1. Sensor locations throughout the studied building.

Figure 2. The components of the visualization pipeline.

Adaptive 3D Rendering and Terrain Mesh Refinement

(a) (b) (c)

(a) (b) (c)

Rendering Static Seismic Data Sets

(a) (b) (c) (d)

Imposters vs. 3D Proxies

Visualizing the Seismic Response of a Building

NONLINEAR WAVE PROPAGATION AND TRENDS AT A LARGE-

Bruce KUTTER 1 and Dan WILSON2

Keywords: centrifuge, earthquake, wave propagation, instrumentation, equipment, collaboration

RESDAQ-Main: 128 channels of amplifiers connected to a high-speed A/D converter for

Pile strain gauges

Pile strain gauges bundled and

RESEARCH ON NONLINEAR WAVE PROPAGATION IN LIQUEFIABLE SOIL