INTRODUCTION

90 ~

l'-,

110 ~

O0 ~

'

b,

CHINA

BANGLADESH

INDIA

. \'\

,\.\\

.\.\

,,,"

,,.'

20 ~

MYANMAR

"~

::::BENGAL::::::
7

9" " " " FAN[" i" [ 9i--?/

l . . . . . -i

i I

: : : : 2~?::

9: : : : : : :

,\

2r,~.

0'

.....................
. . . . . . . . . . . . .

BORNEO

...........................
/:'.'.'.'.'.'.'.'.'.'.

~:-:-:.:::.:-:-:.:-:.:.:-:.:.:.:..

:::::::
9 .'.'.

:::~i
:':::~:

.J

"Nl

: ;::iiil

.: Z. . :.:.:.:.: : : : : : :
9: : : : : :

: :. :. :.: .: .

\i

: :

-.'1.0ON::': ". :. : - : . :. :. ]

SOUTH
CHINA
SEA

9.........

.
".'.

:::::

FANiiii!iiii!i!iii:....~.~

~i!i!iii!iii!

.
.
.
.
.
.
.
.
.
.
.:.............%....%-..
_ . . . . . . . . . . . .

:.:.:.:.:.:.:.:.:.:.:.:.:.:.:.:.:.:.:.:.:.:.:.
================================================

".

::

: . :. . :. . :. : :

........................

.......

,....-~,-"

. .....-.........-........-...l

............................,~.
. . . . . . . . . . . . .
:::::::::::::::::::::::::
_ _ :::::::::::::::::::

. . . . . .

. . . . . . . . . . . . . .

".,

-10os
Christmas o
Is

INDIAN Keeling
Is~176
i
OCEAN
Post-WWII

90 ~

100 ~

0
I

research

Little geological work was possible during the years immediately

after the end of WWII, but following Indonesian Independence in
1947 the Geological Survey of Indonesia (GSI) was established in
the old Bureau of Mines building in Bandung. From 1969 to 1974
the Mapping Division of (GSI) commenced a systematic programme of mapping in the Padang area of West Sumatra, in collaboration with the United States Geological Survey (USGS), as
part of the First Five Year Development Plan (PELITA I).
Several 1:250 000 Geological Map Sheets were published as a
result of this programme (Silitonga & Kastowo 1975; Rosidi
e t al. 1976; Kastowo & Leo 1973). As part of this collaboration
a senior geologist of the USGS, Warren Hamilton, was commissioned to prepare a series of maps and a memoir reviewing the
geology of the Indonesian region in plate-tectonic terms
(Hamilton 1977, 1979). Hamilton's (1979)Tectonic Map, which
includes Sumatra, shows clearly present views of the tectonic
setting of Sumatra.

500
I

1000km
I

Fig. 1.2. The tectonic setting of Sumatra

with the floor of the Indian Ocean
subducting beneath the southwestern
margin of the SundalandCraton. The
deformation front of the Sumatran
subduction system is indicatedby the
toothed line; spreading centres and
transform faults are shown in the Andaman
Sea (after Curray et al. 1979).

SEATAR Programme
In 1973 a meeting was convened by the United Nations Committee
for the Coordination of Joint Prospecting for Mineral Resources
in Asian Off-shore waters (CCOP) in Bangkok which established
the Studies in East Asian Tectonics and Resources (SEATAR)
Programme. At that time a review of the current understanding
of the tectonics of eastern Asia was prepared by Deryck Laming
on behalf of CCOP-IOC (1974). As a result of the meeting it
was proposed to concentrate research along a series of transects
across the island arc systems of East and SE Asia. Subsequently
A. J. Barber (University of London) and Derk Jongsma (BMR)
were engaged by CCOP as Technical Consultants to prepare a
report on the current state of knowledge along the lines of
these transects (CCOP-IOC 1980). One of the selected transects
ran from the Malay Peninsula across northern Sumatra and the
forearc island of Nias to the Sunda Trench. Although the final
report for this transect was never published, a great deal of
important research was carried out by American researchers

CHAPTER 1

INDIAN OCEAN
SW
01
i
km

'~

Nicobar Fan
.... n

.... m

.... n

....

NIAS
Present
BARISAN MOUNTAINS
Accretionary
Sunda Complex
Sumatran Fault
Trench
Ridge
Forearc Basin
Toba Caldera
~--._1.1.~~ , ~ ~ _.....
"
'.: ,'~'-.~-.__ ~-=~-,'-~_.'~:~
~ _,-~-~-,,-'~S~ ~
_~.-j:l
I I I ~ 1 ~
I l l
I I I I Ifl
y accretionary
- complexes

Z_LLL~L~I
I1\1
I I I IIII
I
Jh,.L-~~-~:.~ Earlier
Z..LLL4J.J~_LLL~-~e]
I I/I I I I I I I I

iiiiiiii!!!!!!!iiiiiiiiiiiiiii::,:~.

:',. . .', .',.',.',. ',. .', .', .: .',. ',. ',. .::. .::. . ',. .: .',.',.:. ',. .', .: .: .:.::. . : : : : : i ....
:: ,,: ,, ,,: ,,,: ,. ,. ,.: ~ ~: ', ', ', ', ', ', ', ', ', ', ', ', ', ', ',:: , ,

~(I

I I I I~TI~I

I III

MALAY
PENINSULA
MALACCA STRAIT
Backarc Basin

v.~.-~,
rI~F-F~UI
ICq
~,l=,l~
i, i~=!i
I I
~UNUALAINU
I I I I I I I I I I I I I I I I I I I
I I I I I I I I I I I I I I I I I I I
-

NE
---~ ~

,~
I I
I I1"1 I I
IIII
I I
I I I I I
I I I I I I
I I I [ I I

I I I I I I I I I "I ' I I I~l

.,.-,-.-,.,=:r~l
km
........................
.......................................
, , , , '
~.~ll,,,,,,,,,,
, ', ', ', ', ', ', ', ', ',: ',,,,,
,~,,
i~
~, i
~ I . N D [ A N :::::::::::::::::::::::::::::::::::::,,'I~'~ :u.~::::::::::::::::::::::::::::::::::::::::::::::
~PLATEiiiiiiI[[[[[[[[[[[[I[[[[[[[[[[[[[[[[[~,~.. --~iilll[[[[[[[[[i[[[[[[[[[[[[[[[[[[[[[[[~i[[
''' :":: ~l:::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::
~,~~~i ~i i I~I~~~i ~~'~~i ~~~~IIIII~IIIIi IIIIIIIIIIlli i IIi ~~EURA~"iA~ ~1~'~ :::::::::::::::::::::::::::::::::::::::::::::::

:~',:::','.:::::![[iiiiilllii[[[[[[[[[[[[[[[[[[i[[[i[[i[[[[[[[[[[[[[[[[::--~- "u.~iiii[illi[[[[[[[[[[[[[[[[[[[[[[[[[[[[

]]][][]]]]]]]][]i]]]]]]]];]]][:IIIlllIllIl::',II:IIllllllllI:II::IlllI',l::,~,~'lllllllll:::::::::::I:::',',',:::-'::[

50 Length of cross-section 800kin

50

F i g . 1.3. D i a g r a m m a t i c section across the S u m a t r a n S u b d u c t i o n System from the floor o f the I n d i a n O c e a n to the M a l a y P e n i n s u l a , d r a w n to scale.

under the auspices of the SEATAR Programme, particularly in Nias

and the surrounding seas (Curray et al. 1982; Karig et al. 1980;
Moore & Karig 1980). Also in conjunction with the SEATAR Programme, Cobbing et al. (1992) made a detailed study, including
isotopic dating, of the granites on the Tin Islands of Bangka and
Billiton, supported by the UK Overseas Development Administration as a contribution to the work of : C O P .
Since the effective termination of the SEATAR Programme,
US research in Sumatra has been concentrated on neotectonics,
an important part of which has been the monitoring of movement
along the Sumatran Fault System, using GPS location systems
(Prawirodirdjo et al. 1997).

Indonesian Petroleum Association

In 1971 the Indonesian Petroleum Association (IPA) was established by petroleum companies operating in Indonesia, in association with the Indonesian national oil company, Pertamina.
Since its inception the IPA has held Annual Conventions which
continue to the present day. At these conventions papers on the
geology of Indonesia are presented and published as the Proceedings of the Indonesian Petroleum Association. The IPA
Proceedings provide an invaluable source of information on the
geology of Indonesia. Most of the papers deal with Tertiary deposits and details of the stratigraphy and structure of the oil and
gas fields of Indonesia, including those of Sumatra, but more
general papers on geology and tectonics have also been published.
The publication of the IPA Proceedings has resolved van Bemmelen's (1949) complaint of the pre-WWII situation, in which large
amounts of geological data, accumulated by the oil companies,
remained unpublished for commercial reasons, and were not available for the compilation of regional geological syntheses.

British and Indonesian Geological Surveys

Major UK involvement in the geology of Sumatra began in
1975 when the Institute of Geological Sciences (IGS, now the
British Geological Survey, BGS), in collaboration with the
Geological Survey of Indonesia (GSI), commenced a five-year
mapping and reconnaissance geochemical survey of northern
Sumatra to the north of the equator (Northern Sumatra Project,
NSP). In 1978 GSI was reorganized into a number of semi-autonomous directorates and the Directorate of Mineral Resources
(DMR) became the designated Indonesian counterpart organisation
in the NSP. The work of IGS in the Northern Sumatra Project,
and subsequent projects by BGS in Sumatra, were funded from
the Technical Assistance and Technical Cooperation budgets of
the U.K. Overseas Development Administration (ODA).

The structural, stratigraphic, geochemical and tectonic results

of the Northern Sumatra Project have been presented in a series
of papers (Page et al. 1978, 1979; Cameron et al. 1980; Rock
et al. 1982; Aldiss & Ghazali 1984) and unpublished reports.
In a continuation of the NSP, geological maps and reports resulting from the project were edited by BGS personnel, and published
by the Indonesian Geological Research and Development
Centre (GRDC), one of the constituent directorates of GSI,
as a series of 18 Geological Map Sheets at 1:250 000 scale,
with accompanying Explanatory Notes. Follow-up studies of
fossil localities, with the view of establishing the stratigraphical
ages of the sedimentary units in Sumatra, were carried out
by Metcalfe (1983, 1986, 1989a, b; Metcalfe et al. 1979) and by
Fontaine and his collaborators, under the auspices of : C O P
(Fontaine & Gafoer 1989). The results of the regional geochemical
stream sediment sampling survey were published in a joint
IGS/DMR Geochemical Atlas (Stephenson et al. 1982) and subsequently DMR published sets of single element proportional
symbol distribution maps at 1:250000, for many of the
quadrangles to the north of the equator. Geochemical anomalies
found during the NSP were followed up by BGS and DMR
in the collaborative North Sumatra Geological and Mineral
Exploration Project (NSGMEP, 1985-1988). The results of a
separate programme of research into the mineralization in north
Sumatra, also funded by UK ODA, have been published by
Bowles et al. (1984, 1985) and Beddoe-Stephens et al. (1987).

University of London Southeast Asian Research Group,

BGS and LEMIGAS
In 1978 members of the University of London Southeast Asian
Research Group which had previously been active in Eastern
Indonesia, commenced a programme of research projects in
Sumatra, in collaboration with BGS, DMR and GRDC. In 1984
a joint University of London/BGS North Sumatra Basins
Study Project, was set up with funding from the UK Overseas
Development Administration, in collaboration with the Indonesian
Research and Development Centre for Oil and Gas Technology
(LEMIGAS) (Kirby et al. 1993). This project built on the major
involvement by LEMIGAS in this productive basin, where
one of the largest exploration blocks is operated directly by
Pertamina. The overall programme was largely concerned with
the stratigraphy, sedimentology and geophysics of the Tertiary
basins in northern Sumatra, with the University contribution
Concentrating on field studies of the relationship of the Tertiary
rocks to the underlying basement, with a view to understanding
the tectonic evolution, of these basins (Turner 1983; Tiltman
1987, 1990; Kallagher 1990). More recently the University of
London contribution, funded by the UK Natural Environment
Research Council (NERC), ODA and a number of oil companies,

INTRODUCTION

96~

98 ~

102 ~

1 O0 ~

104 ~

106 ~
6 ~-

BandaAceh

Active Volcano
Sumatran Fault System

_4~

Deformation Front of the

Sumatran Subduction
Complex

4 o.

~'yY

100

200

300

400

500km

..............

2 ~-

".:.:.:.:.:.:.:.:.

Nias
............

|j

(-:-:.:-:-:-:-:-:.:-:.:[... ...-

:i:~i:i:i:i:i:i:i:i:iQ:i

-0~

0 o-

L i n e of
Section
Fig.l.3.

2 ~-

Q
[ii:iil;i:::i:::i;i;i:~~::iiii~mbang

Bengku[
-4 ~

j~.'.i ::..~::!:iiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiii.
I
::i~'-

"'-'" ""'""'"'"'""'"'-'"'"'-'"'"'-

4 ~-

Quaternary-Recent volcanics

pung

Tertiary Sediments and volcanics

6 ~-

Pre-Tertiary Basement
96~

98 ~

100 ~

102 ~

104 ~

106 ~

Fig. 1.4. Simplified geological map of Sumatra showing the distribution of the main stratigaphic units and the active volcanoes. Toothed line marks the deformation
front of the Sumatran Subduction System. The line of section in Fig. 1.3 is also shown.
became increasingly concentrated in the forearc islands, where
a series of geological mapping and gravity surveys were
completed (Situmorang et al. 1987; Milsom et al. 1990; Harbury
& Kallagher 1991; Samuel & Harbury 1996; Samuel et al.
1997). At the same time LEMIGAS collaborated with the
French CNRS (Centre National pour Recherche Scientifique) in
a number of studies in the forearc region using the Indonesian
Marine Research Vessel Baruna Jaya III (Diament et al. 1992;
Izart et al. 1994). Outside the bounds of the NSP, University of
London Staff and research students with funding from NERC,
ODA and a Consortium of petroleum companies collaborated
with LEMIGAS on studies on the Ombilin interarc basin
in central Sumatra (Lailey 1989; Bartram & Nugrahaningsih
1990; Howells 1997b), the Woyla Group in North Sumatra

(Wajzer et al. 1991; Barber 2000; McCarthy et al. 2001) and a

study of the Sumatran Fault System throughout the island
(McCarthy & Elders 1997).

Southern Sumatra Project

Geological mapping, gravity surveys and geochemical programmes in Sumatra south of the equator were conducted by GRDC
and DMR during PELITA II (1974-79) and in successive five
year development programmes, continuing into the 1980s. In
1988 the Southern Sumatra Geological and Mineral Exploration
Project (SSGMEP) was established, and BGS joined DMR and
GRDC in the completion of these surveys and in research

CHAPTER 1

106<'30
I ,

r./i/o:

[/Banda
Aceh
........
,, ,.
......" / ~" . Lhokseumawe

VZZZEZA
i~".._o4.2o U . / ;

<b
"rebingt!pggi'I

;
/Sidikafang,
" ~'(///"/~

t7

~,/Pematan(
/
siantar"

" Bagansiapiapi
." . / c - ~ / . / ' / / . / '

Padangsidempuan

NE Muarasiberut
., 0913 / ..."
Sungaipenuh'
and
Ketuan

6'-

~
t3

1O0

200

300.................

400

500kin

"

' " t
%'7

"

~ "

-~,.%.f~

Fig. 1.5. Coverage, sheet numbers and names of the 1:250 000 Geological Maps published by the Indonesian Geological Survey, the Geological Research and
Development Centre, Indonesian Ministry of Mines and Energy.

programmes with funding from UK ODA Technical Cooporation

budget. This programme was completed in 1995 with the publication by GRDC of the last of the forty three Geological Map
Sheets at 1:250000 scale, covering the whole of Sumatra
(Fig. 1.5) and 18 1:250000 scale Bouguer gravity anomaly
maps of southern Sumatra, including Bangka and Billiton
islands, but excluding the coastal swamps and the Barisan
Mountains. The collaborative geochemical survey was completed
in 1994 with the publication by DMR of 14 quadrangle boxed sets
of 1:250 000 single element proportional symbol geochemical
maps (up to 15 elements) with accompanying reports on the
geochemistry, geology and mineral occurrences. Subsequently
the Sumatra geochemical data was made available on CD-ROM

(Version 2 in 1999). In 1995 following a one-year 'Sustainability

Phase' of the SSGMEP a Geochemical Atlas of Southern
Sumatra was issued in digital form on CD-ROM (Machali et al.
1995). Publication in book form followed in 1997, with text in
both Bahasa Indonesia and English (Machali et al. 1997). An
evaluation of tectonic models for the Pre-Tertiary history of
Sumatra based on BGS/DMR/GRDC and University of London
research programmes has been published by Barber & Crow
(2003).With the completion of this major phase of UK
involvement in the study of the geology of the Sumatra, the
time is ripe to review the vast increase in our knowledge of
the geology of Sumatra since van Bemmelen's (1949, 1970)
synthesis.

Chapter 2

Seismology and neotectonics

JOHN MILSOM

Sumatra is an active (Andean) continental margin that would be

linked by land to SE Asia if sea level fell by as little as 50 m.
Present-day tectonic processes are controlled by three major
fault systems, the most obvious of which is the subduction thrust
which crops out in the Sunda Trench. The trench curves very
little in the 800 km between Enggano and Nias, i.e. off central
Sumatra (Fig. 2.1), but is markedly convex towards the Indian
Ocean both further north and further south. Water depths of
more than 6000 m are reached in the south but the maximum in
the north may be less than 5000 m. The difference is usually,
and convincingly, attributed to the presence on the Indian Ocean
plate of the Nicobar Fan, consisting of sediments, derived
ultimately from erosion of the Himalayas, which increase steadily
in thickness towards the north (e.g. Hamilton 1979). Continuing
subduction is attested by a Wadati-Benioff Zone (WBZ) that
extends to depths of the order of 200 km (e.g. Newcomb &
McCann 1987) and by volcanic activity in the Barisan mountains,
the peaks of which generally lie within a few tens of kilometres
of the coast. The change, of more than 45 ~ in the trend of the
trench between 96~ and 97~ (the 'Nias Elbow') may have
been initiated by subduction of the 2 km high Investigator Ridge
(Investigator Fracture Zone), which trends approximately northsouth at about 98~
Sieh & Natawidjaja (2000) defined a
'Central Domain' of mainland Sumatra between the Nias Elbow
and the ridge intersection as anomalous in a number of ways
(notably in the differing trends of the Sumatran Fault and the
volcanic line) and as distinct from more regular Northern and
Southern domains on either side (Fig. 2.1).
Inland, the dextral transcurrent Sumatran Fault runs the entire
length of the island, from Banda Aceh to the Sunda Strait
(Fig. 2.1). A variety of names have been used for both the
overall fault system and parts of it, and new nomenclature
developed by Sieh & Natawidjaja (2000) divided it into 19 individual segments. Even this detailed study failed to answer many
fundamental questions, and estimates of total lateral displacement
still vary from several hundred kilometres to as little as twenty
kilometres. The 150km suggested by McCarthy & Elders
(1997) seems to be about the mean of the published values. The
fault trace coincides roughly with the watershed of the Barisans
and with the volcanic line, although most of the volcanoes lie
somewhat to the NE of the fault and only nine of the fifty youngest
centres lie within 2 km of it (Sieh & Natawidjaja 2000). A more
precise correlation is with the subduction thrust, since for most
of its length the distance between the Sumatran Fault and the
trench axis differs by no more than 30 km from the average
value of 290 km. The largest deviations are a narrowing within
the bight of the Nias Elbow and a broadening in the region
further to the NW.
The third and most enigmatic of Sumatra's major fault systems
is the Mentawai Fault, at the outer margin of the forearc basin
(Fig. 2.1). In many publications the name is reserved for the
segment extending from the Sunda Strait to Nias (Samuel &
Harbury 1996) or the Batu Islands (Diament et al. 1992), but the
same disturbance zone continues at least as far as the Andaman
Sea (Malod & Kemal 1996) and possibly to the Andaman and
Nicobar Islands. Movement has been variously interpreted as
normal, strike slip or reverse (Sieh & Natawidjaja 2000). There
are considerable changes in appearance on seismic sections
even within the region from Nias southwards; the structure was

described by Sieh & Natawidjaja (2000) as a homocline and by

Karig et al. (1980) as a 'fault-flexure'.
Magnetic anomalies in the Indian Ocean south of Sumatra
trend east-west and were interpreted by Sclater & Fisher (1974)
as indicating Palaeogene ages for most of the crust adjacent to
the trench, with a possibility of Late Cretaceous crust in the
extreme SE. Transforms such as the Investigator Fracture Zone,
which may offset the anomalies by several hundred kilometres,
run almost precisely north-south. With the trend of the trench
varying from N40~ to N60~ and the direction of the Indian
Ocean-Sumatra convergence vector being about N15~ (Fig. 2.1),
Sumatra has long been recognized as a key area for studies of
the partitioning of strain between thrust and transcurrent faults
during oblique convergence (Fitch 1972; McCaffrey 1992, 1996;
Malod & Kemal 1996). The suggestion, originally made by
Fitch (1972), that the oblique motion is to a first approximation accommodated by orthogonal subduction at the trench and
dextral slip along the Sumatran Fault, is now widely accepted.
To the extent that this is true, the forearc region must be decoupled
from both the Indian Ocean and Eurasia. The commonly used
term 'sliver plate' (e.g. Curray 1989) suggests more strength and
rigidity than could reasonably be expected of such a long and
narrow strip of lithosphere, and any analysis of subduction
beneath Sumatra must take into account the probability of independent movements of forearc fragments (e.g. McCaffrey 1991).
Estimates of the movements of the Indian Ocean relative
to Sumatra are shown in Figures 2.1 and 2.4. Changes in magnitude and direction from NW to SE are dictated by the East
African location of the pole of rotation (Larson et al. 1997). If partitioning of orthogonal and transcurrent strain between, respectively, the trench and the Sumatran Fault were complete (and
movement occurred only along these features), then sites in the
forearc sliver would move parallel to the Sumatran Fault relative
to SE Asia, but at right angles to the trench relative to the
Indian Ocean. Trench-normal relative motion implies that the
forearc sliver 'tracks' across linear features on the Indian Ocean
Plate, such as the Investigator Fracture Zone, which have northsouth trends (Fig. 2.1). If the long term movement between the
forearc and the Indian Ocean has actually been approximately
orthogonal, the intersection point of the Investigator Fracture
Zone with the trench, now near the Batu Islands, would have
been north of Nias less than 10 million years ago. The relief, of
more than 2 km, on the Investigator Fracture Zone might not
only impede such tracking but could be responsible for cyclical
uplift and subsidence in the forearc basin and ridge.
Slip partitioning and subduction of Indian Ocean lithosphere
produce high levels of seismicity in the Barisan Mountains, in
the forearc basin and along the forearc ridge (Fig. 2.2). The potential for extremely destructive earthquakes was most recently
demonstrated by the Magnitude 9 event near Simeulue in December 2004 and by the resulting tsunami, which gave rise to one of
the worst natural disasters in recorded human history. However,
and despite the geological evidence for a long history of subduction (e.g. Page et al. 1979), shocks deeper than 200 km are rare
(Fig. 2.3). Events below 300 km are confined to the extreme SE
and may be associated with north-directed subduction beneath
Java rather than NE-directed subduction beneath Sumatra. The
abrupt change in orientation of the active margin between these
two islands must produce considerable stress in the downgoing

CHAPTER 2

Central Domain
Northern Domain

Southern Domain

LINE --~l
42-43 Batu

1000 ~

pora N.

Elbow'

F i g . 2.1. Sumatra: the neotectonic setting. The figure has been oriented on the main fault direction. The India-SE Asia convergence vector changes significantly in both
direction and magnitude over the length of the island, from 52 mm a-1 directed at N10~ (at 2~ 95~ to 60 mm a - l directed at N 17~ (at 6~ 102~ Convergence
data (and mainland structural domains) are from Sieh & Natawidjaja (2000). Elongated rectangles in the forearc region indicate the locations of the zeros on the seismicity
cross-sections in Figure 2.3. The seismic image along Line 42-43 is shown in Figure 2.7. The white stars mark the epicentres of the Enggano 2000 and Simeulue
2004 Great Earthquakes. Bathymetric contours at 200, 1000, 3000, 5000 and 6000 m are from GEBCO (1997). Shading indicates sea floor deeper than 6000 m. I.F.Z.,
Investigator Fracture Zone. Onshore topography derived from the Global Relief Data CD-ROM distributed by the National Geophysical Data Center, Boulder, Colorado.

slab but this is not obvious in the patterns of shallow seismicity

shown in Figure 2.2 and discussed below.

Shallow seismicity
As in most active continental margins, shallow ( < 6 0 km depth)
earthquakes in Sumatra are distributed over wide areas of the
upper plate and are not restricted to the WBZ (Fig. 2.2).
Maximum shallow earthquake activity occurs within the sliver
defined by the Sumatran Fault in the east and by the subduction
thrust in the west and at depth, and is most intense along the
line of the forearc ridge. There must be considerable forearc
extension (see McCaffrey 1991 ) if the estimates of large variations
in rates of transcurrent slip (more than 400 km of offset in Aceh
but negligible displacements in the Sunda Strait; Curray et al.
1978) are correct (see also Bellier & Sebrier 1995). Although
there have been relatively few shocks of Magnitude 6 or greater
beneath the mainland, some have occurred, most notably in
the vicinity of the 'equatorial bifurcation' in the Sumatran Fault
identified by Prawirodirdjo et al. (2000).
The insets to Figure 2.2 attempt to show separately the distributions of events within the uppermost 40 km of the crust and
at depths of between 40 and 60 km. Because of the uncertainties
inherent in determining the depths of shallow earthquakes (see
discussion in Engdahl et al. 1998), there will be events on one
map that should properly have been plotted on the other, but the
overall differences between the plots are likely to be real. The
4 0 - 6 0 km events are concentrated in a narrow zone centred on
the forearc basin and most are probably directly associated with
the subducted oceanic lithosphere, i.e. with the WBZ. There are,
however, some similarities with the patterns of shallower events,
noticeably in the tendency for epicentres to be concentrated in
short linear zones at right angles to the trench, presumably due
to some form of forearc segmentation. The most obvious examples
can be seen around Enggano and western Simeulue, i.e. close to
the sites of the Great Earthquakes (defined as earthquakes with

MW magnitudes greater than about 7.8) in June 2000 and December 2004 respectively. Interestingly, the Simeulue events cluster
along the crest of a basement ridge that defines the northwestern
boundary of a marine and sedimentary basin (Simeulue Basin)
where maximum water depths exceed 1000 m. The trend of the
linear alignments changes slightly north of the Nias Elbow to
partly match the change in orientation of the trench but, surprisingly, N E - S W alignments of epicentres can be seen east of the
even more dramatic change between Sumatra and Java (Fig. 2.2).
A second feature of the shallow seismicity is the separation
of the shallowest earthquakes (Fig. 2.2; lower inset) into two
divergent zones, one along the forearc ridge (with a bend or
offset where the Investigator Fracture Zone enters the subduction
zone near the Batu islands), the other very approximately along
the west coast of Sumatra. The forearc basin itself is relatively
quiet seismically at these depths. The offset at the Investigator
Fracture Zone is interesting because Newcombe & McCann
(1987) noted that ruptures associated with Great Earthquakes do
not propagate across this region. In 1833 a Magnitude (Mw) 8.7
event faulted the plate margin for about 600 km from Enggano
to the Batu Islands, while the effects of the Mw 8.4 event in
1861 were confined to a 300 km segment between the Batu and
Banyak Islands.

The Wadati-Benioff Zone (WBZ)

In keeping with the continental margin setting, seismicity beneath
Sumatra is more diffuse than beneath a typical intra-oceanic arc.
This is illustrated in Figure 2.3, which shows hypocentre distributions within three typical swathes, each 200 km wide. In the
extreme NE (Fig. 2.3a) the WBZ forms the lower boundary to a
seismogenic zone that extends up to the surface over a distance
of approximately 300 km from the trench. The greatest concentration of events is at about I00 km from the trench and at
depths of about 50 km. In the swathe immediately south of the
equator, near the islands of Siberut and Sipora, there is a much

SEISMOLOGY & NEOTECTONICS

km onJy

+ + + 4 o - 6o

Hypocentres above 40 km o "

200

100

%* **"

200

o qto

%~~

"..,

.
.

~ - :, ~

o+

".-$

~,

00

: o :
o,~,

.:

Oo

Equator

T
100

200

R
0

0,~"

Oo

40ku,o,,~* ]

+ +++:+ +~+

........

+,+++++:+++++++++....
+
:

~ ' ' . +I ~

+ :i~':+i. i

200

, +

/o'|
"ore, +
o o

" <~o
"

+,.
...... o
+

o b o"V
(

104~

~g0

T 100
o

+++
* ,i
o .~.~
o

IO0~

Central
R
O
o

bo

oo

~ o
9 oi~o.Oo
' + . . . . . . : : .o~

%o

:, .... i

100
0

<ll._
,,..
<'*"~lRm~-T~..,,

F200

30g
0 o

~ +,~,.~
,~o

:~ : "

.
~++~_~l[llm~ o

....

Ioo

<>~ ......

: A ~o1 7 6 1 7 6 1 7 6 o

o, oo
.

Fig. 2.2. Shallow selsmicity of Sumatra. Data downloaded from supplementary

material to Engdahl et al. (1998) using only events occurring between 1980 and
1996. Rectangles show locations of swathes cross-sectioned in Figure 2.3. Thick
lines within the rectangles mark the cross-section zeros. Insets show 0-40 km
and 40-60 km hypocentres separately 9

clearer development of a linear W B Z but the scatter is still considerable (Fig. 2.3b). Sieh & Natawidjaja (2000), among others,
have claimed that the depth of the W B Z beneath the volcanic
line is considerably greater in this Central Domain (Fig. 2.1)
than to either the N W or the SE, although the m a x i m u m depth
of the seismic zone is actually smaller. The effect is not,
however, obvious in Figure 2.3.
The most intensely active part of the W B Z is in the extreme
south, near Enggano, where there are two main event clusters, at
about 40 and 70 k m (Fig. 2.3c). The seismogenic zone continues
down to at least 200 km. The two deepest shocks might be
associated with Java subduction but, if associated with Sumatra,
indicate a pronounced steepening of the W B Z between 200 and
300 km.

seismicity

A more comprehensive picture of Sumatra seismicity than is

provided by Figure 2.3 was presented by Hanus e t al. (1996),
who plotted hypocentres within 50 k m wide, N E - S W swathes
that together covered the whole of the island. Arguably their
most interesting plot was A15, which included the northern part
of the forearc island of Nias and much of the Toba caldera
(Fig. 2.1). The W B Z in this region dips at an angle of a little
more than 30 ~ and the deepest shocks occur between 200 and
2 5 0 k m . There is a small but noticeable gap in seismicity
beneath the volcanic line at depths of about 150 to 180 k m and
a corresponding region of shallow seismicity immediately
beneath the volcanoes.
In detail, the picture provided by Hanus e t al. (1996) is suspect
because of the reliance on International Seismological Centre

~ ................

~>o

South
Magnitudes

Toba

I00

,e

...

]++, ;.:,:)+ +i]

/ o

o~

6+ ~

++ +?++++++++++++++++++++;
I +++++++++
+ ........ i oo
' % ~+++++++'++++++ +I. . . . . . . . .
9 ,'+::~:~+;++++++++:~++I
...............

",

+o

'~0s, % :

" e~
o
~o,

~
o,,

. . . .
o

'.++~+

.....................

....

300

,,'-,

/
-

200

II

a)

300 k i n

F
100

oGo o

1200

- i

Simeulue

C O o "' ,

I
1 O0

~ I l~

oF

?o

.e o ' ~ .

~:

+I~

North

,i.

o+o..~

o o

./

300

o 0+~+.,
og~. o,~ ~ o+..++
"~ .o

- . L/Hypocentres a 40-60
km
t e *

oe
N e,++"

100

2OO

c) .:~ 300
3 9

4 o

Fig. 2.3. Cross-sections of seismicity, Sumatra subduction zone 1980 to 1996.

Each cross-section is based on a swathe drawn at right angles to arc. Each swathe
is 300 km across, except for the Simeulue swathe (inset, profile 2.3a), which is
only 100 km across. T, R and F in each case indicate the locations of,
respectively, the trench, the crest of the forearc ridge and the Sumatran Fault.
Locations of cross-section zeros are shown in Figure 2.1 and the swathe areas are
shown in Figure 2.2. For profiles 2.3b and 2.3c these zeros coincide with the crest
of the forearc ridge, but for 2.3a, where the ridge is poorly defined, the zero is in
the centre of the forearc basin. The star on the Simeulue swathe indicates the
location of hypocentre of the December 2004 event, after NEIC (2005). All other
data were downloaded from supplementary material to Engdahl et al. (1998).
Distances in kilometres, no vertical exaggeration.

(ISC, Thatcham, UK) hypocentre locations. These, being

derived from interpretations of teleseismic data based on global
velocity models, are inevitably of fairly low accuracy. The significance of this limitation has been demonstrated by Fauzi e t al.
(1996), who used additional data from a newly established (but
now permanent) network of short-period digital seismometers to
study earthquakes in the vicinity of Toba. The primary aim of
the work reported, which covered the period from October 1990
to April 1993, was to investigate a hypothesized break in the
d o w n g o i n g slab due to subduction of the Investigator Fracture
Zone. Seismic activity was found to be unusually high in the
appropriate area but no discontinuity was detected and a limit
of 20 k m was placed on the magnitude of any possible displacement. There was more success with a subsidiary objective of
defining the shape of the W B Z as it followed the bend in the
offshore trench between Nias and Simeulue. In contrast to both
the ISC and Engdahl e t al. (1998) data, hypocentres derived
from the local study and plotted for narrow cross-strike swathes

10

CHAPTER 2

~.Lake

%
Equator

Batu

IS.

I Indian ~
[Ocean/
SE Asia
convergence

......................

vector;

Sipor

52 mm/yr

k,

pNgrtaih~'~f~ S~

N
0

~l'~gal

300 km

Convergence velocity scale

50mm/yr
% ~

....
[! ~

Contourson theWBZ
in the Tobaregion,after %.
Fauzi
* !~
. . et al.,
. 1996
. .
June 2000 Enggano earthquake
~
Main ~
Aftershocks~
~~"
shocks ~
k.'
~
December 2004 Simeulue &
"~
March 2005 Nias
,IL
earthquakes Mainshocks~
~

Indian.
Ocean/ I
SEAsia [ /
convergence
vector I
58 m m / v r / 7....

BengkulL

Engganc

o
0
..................................... 6"S

were found to be tightly concentrated in very narrow zones that

changed in dip scarcely at all around the bend (Fig. 2.4). Estimated
depths also tended to be smaller than those based only on
teleseismic data, especially beneath the forearc basin.
A more recent seismological study of the Toba area used a temporary network comprising 30 short-period and 10 broad-band
seismographs deployed for four months in the first half of 1995
(Masturyono et al. 2001). Tomographic methods were used
to define velocity variations beneath the caldera. The results
support the hypothesized existence of two distinct eruptive
centres, one in the south-central part of the lake and the other at
its northern end, which erupted at different times (Knight e t al.
1986). Low velocity zones underlying these two centres and
extending down into the mantle are separated by a region with a
more typical crustal velocity structure.

Relative horizontal movements

The information on present-day tectonic processes in Sumatra provided by seismology is now being supplemented by geodetic data
from Global Positioning System (GPS) satellites. Repeated
measurements at fixed pillars provide an essential complement
to earthquake studies, which record only episodic, although sometimes very large, displacements. During seismically quiet periods,
GPS measurements monitor aseismic creep and can indicate

Fig. 2.4. Movements of sites in Sumatra as determined

by GPS observations during the period 1989-1993
(Prawirodirdjo et al. 1997). Vectors show rates of
movement relative to a stable SE Asia. They imply stress
accumulation in parts of the forearc region, some
of which would have been released by the June
2000 earthquake near Enggano, the December
2004 earthquake near Simeulue and the March 2005
earthquake near Nias. The locations and mechanisms of
these earthquakes are indicated by the centres of the
lower hemisphere projection 'beachballs', from
Abercrombie et al. (2003) for Enggano and from NEIC
(2005) for Simeulue and Nias. Locations of aftershocks
of the Enggano earthquake for which fault-plane
solutions were calculated by Abercrombie et al. (2003)
are also shown. Major aftershocks to the Simeulue
earthquake occurred almost entirely NW of the limits of
the map. MS, Muara Siberut. S, Sinabang, PB, Pulau
Babi. The pecked grey lines show the locations of
barriers to propagation of ruptures from Great
Earthquakes inferred by Newcomb & McCann (1987).

regions in which stress is increasing and may be released catastrophically at some time in the future. Because of the short time
intervals over which observations are made (typically 3 to 5
years), GPS measurements must always be considered in the
context provided by estimates of long term relative plate motions.
Most of the GPS site markers in Sumatra were established by
B A K O S U R T A N A L , the Indonesian mapping and geodetic survey
authority, working in collaboration with various US institutes, and
most are located between 2~ and 2~ (Prawirodirdjo et al. 1997;
Genrich e t al. 2000). Additional measurements were made at sites
near Bengkulu and Medan and on Nias and Billiton in the course
of the G E O D Y S S E A study, which covered the whole of SE Asia.
The G E O D Y S S E A results defined a 'Sunda' Block that includes
Borneo, the Malay Peninsula and Indochina and moves east relative to Eurasia at 7 - 1 0 mm a -1 (Chamot-Rooke & Le Pichon
1999; Michel et al. 2001). Billiton Island and Medan are clearly
within this block, as is much of Sumatra east of the Sumatran
Fault, but motions near and to the west of the fault are much
more complex. The main B A K O S U R T A N A L campaign (sites
shown in Fig. 2.4) began in 1989. Detailed analyses of the data
obtained to 1996 in the Central Domain (Fig. 2.1) have been provided by McCaffrey et al. (2000) and by Genrich e t al. (2000). To
supplement these analyses, Prawirodirdjo et al. (2000) also considered the results of conventional triangulation surveys extending
over a period of 100 years in the same area. These generally confirmed the GPS estimates of 2 0 - 3 0 m m a-1 of dextral movement

SEISMOLOGY & NEOTECTONICS

on this portion of the Sumatran Fault, but revealed very considerable differences in detail in both movement magnitudes and
directions.
Figure 2.4 shows the site motions relative to SE Asia as
interpreted by Prawirodirdjo et al. (1997) and (also relative to
SE Asia) the averaged long term Indian Ocean movement
vectors (Demets et al. 1990). Strain partitioning was evidently
only partially achieved, at least over the short time interval
involved, nor were movements confined to the main fault
systems. Sites east of the Sumatran Fault but within 50 km of it
were not stationary with respect to SE Asia but recorded small
but significant displacements to the north and NW. Similar patterns near other major strike-slip features have been interpreted
as recording stress accumulations in wide regions of deformed
rock that are ultimately released by faulting (e.g. Armijo et al.
1999).
Sites in the forearc experienced much larger trench-parallel
displacements, but McCaffrey et al. (2000) argued that only
about two-thirds of the necessary slip was accounted for and
that most of the remainder must have been accommodated oceanward of the crest of the forearc ridge. However, the situation
varied considerably from place to place. On forearc islands in
the Central Domain (between the Batu and Banyak Islands) the
trench-normal components were small, suggesting strong partitioning of convergent and transcurrent movements, but it seems
that the forearc was largely coupled to the downgoing slab everywhere to the south of the Batu Islands. The boundary between
the two regimes occurs in the region where the Investigator
Fracture Zone enters the trench. Prawirodirdjo et al. (1997) tentatively interpreted the northwestwards decrease in coupling as a
consequence of the subduction of thick, water-rich sediments of
the Nicobar Fan, resulting in high pore pressures in the forearc
wedge and weakening of the upper plate by the introduction of
hydrothermal fluids. The change in coupling would thus be due
to the barrier to sediment flow from the NW presented by the
Investigator Fracture Zone, rather than directly to its presence as

Trench-orthogonal

motion

Trench-parallel

11

an asperity on the lower plate. However, the magnitude of the

December 2004 Simeulue earthquake suggests a 'sticky', rather
than well-lubricated, fault zone.
The combination of gradual change in the orientation of the
Indian Ocean/SE Asia convergence vector and the change in
trench orientation at the Nias Elbow implies almost orthogonal
convergence across the trench in the vicinity of Simeulue
and the Banyak Islands. The Sumatran Fault, however, changes
direction much less noticeably, and the differences in curvature
of structures on the mainland and along the forearc ridge
produce a widening and deepening of the forearc basin NW of
Simeulue. Rather surprisingly, the GPS motions of the two sites
in the Banyak Islands were almost perfectly parallel to the trend
of the Sumatran Fault, and so to the trench further south. The
lack of GPS sites on Simeulue means that short-term neotectonic
patterns in this critical area remain, for the moment, undefined.
The data from GPS measurements and triangulation surveys
can be compared with long-term slip estimates based on geologic and topographic offsets at the Sumatran Fault. Slip rates estimated from stream offsets on SPOT imagery vary from
10 mm a -1 at the Sunda Strait to 23 mm a -1 near Lake Toba
(Bellier & Sebrier 1995). Much of this change occurs in the
Central Domain, where the rates estimated by Sieh & Natawidjaja (2000) using geological offsets increase from 11 mm a i in
the SE to 27 mm a-1 in the NW. Slip rates estimated from GPS
observations vary much less, increasing by only 4 mm a -1, from
23 mm a-1 to 27 mm a-1, over the same distance (Genrich et al.
2000). Sieh & Natawidjaja (2000) suggested that the geologically indicated changes in slip rates along the fault must have
developed only during the last 100 ka, because of the absence
of compressional accommodation structures, but left the geological-GPS discrepancy unexplained. They also suggested
that the total slip on the Sumatran Fault might be little more
than the 20 km of the maximum verifiable geological offset,
and that the remainder of the roughly 100 km offset required
by stretching in the Sunda Strait might have been accommo-

motion

Fig. 2.5. GPS vectors and the Great Earthquake of June

2000. The upper diagram shows overall movement
vectors relative to SE Asia and their trench-parallel and
trench-orthogonal resolved components. The lower
diagrams compare these components individually.
Vector 1 is the regional convergence vector, after
Demets et al. (1990). The remaining vectors are GPS
vectors from the 1991-1993 campaign at sites at the
bases of the arrows, after Prawirodirdjo et al. (1997).
'Beachballs' show the locations of the two subevents
proposed by Abercrombie et al. (2002) for the June 2000
earthquake.

12

CHAPTER 2

dated by slip on the Mentawai Fault. Their proposed deformation history (which they emphasized was only one of a multitude of possibilities) involved arc-parallel stretching during the
Pleistocene but provided no role for the segment of the Mentawai Fault north of the Nias Elbow.

Seismic reflection sections from many parts of the basin

favour localized faulting in the forearc basin, since deformation
of Late Neogene sediments is generally confined to the narrow
zone close to the eastern coasts of the forearc islands which was
named the Mentawai Fault by Diament et al. (1992). However,
the now numerous published images of this feature obtained on
crossings reported by Karig et al. (1980), Diament et al. (1992)
(Fig. 2.6a), Malod & Kemal (1996) and Schlfiter et al. (2002)
(Fig. 2.6b) and the excellent multichannel imagery obtained
by the Scripps Institution of Oceanography (SIO) south of
Nias (Fig. 2.7), indicate a very complex and variable structure.
Considerable uncertainties remain as to its true nature. On some
seismic sections (e.g. Diament et al. 1992) it appears to be a
simple faulted anticline, while in other areas the zone of weakness
has been exploited by shale diapirs which conceal fundamental
structures (Milsom et al. 1995). The extreme linearity has been
used as an argument for a fundamentally transcurrent role
(Sieh & Natawidjaja 2000) but subsidence of the forearc basin
and elevation of the forearc ridge imply either normal or thrust
components. Where it emerges on land, in southeastern Nias, the
fault was interpreted by Samuel & Harbury (1996) as an originally
extensional fault that has suffered Pliocene to Recent subductionrelated inversion. Significant transcurrent movement was regarded
as improbable. Interestingly, however, seismic section presented
by Schltiter et al. (2002) (Fig. 2.6b) shows the disturbance
as having moved away from the landward side of the forearc
ridge (which is itself fragmented in this region; see Fig. 3.1) to a

GPS data, the Enggano and Simeulue earthquakes and the

Mentawai Fault
During the period covered by published GPS measurements,
the southern forearc islands (Siberut to Enggano) were moving
NW relative to Sumatra at roughly the same rate as the underlying
Indian Ocean Plate (Fig. 2.4). Enggano, in particular, participated
in virtually all of the motion of the Indian Ocean during the period
of observation, which unfortunately in this particular case
extended only from 1991 to 1993 (Fig. 2.5). Much smaller relative
motions were recorded at two sites on the adjacent coast of the
mainland and therefore only a small part of the trench-parallel
motion required accommodation further inland, in the vicinity
of the Sumatran Fault. More than half the trench-parallel motion
and an even greater proportion of the trench-normal motion
must have been absorbed between Enggano and the coast, either
at one or more discrete faults or by distributed strain over the
width of the forearc basin.

/'~

l"

"

(a)

9 2.5

!
Seismic sections from Diament et al. (1992)

"~

Fig. 2.6. (a) Interpreted single-channel seismic reflection

sections across the Mentawai Fault in the southern part of
the Sumatra forearc basin (after Diament et al. 1992). Line
locations as shown. (b) Multi-channel seismic reflection
section across the Mentawai Fault south of Enggano, after
Schltiter et al. (2002). Location shown on (a). The greater
penetration achieved on the more recent survey suggests a
transcurrent origin for the feature which, in the nearby
southernmost single-channel section, appears to be a
simple faulted anticline.

SEISMOLOGY & NEOTECTONICS

13

Fig. 2.7. SIO Line 42-43, showingthe Mentawai Fault

immediatelysouth of Nias. Section provided by Scripps
Institution of Oceanography.

position within the forearc basin. This fact, and the image itself,
are more compatible with transcurrent than vertical motion.
Indeed, Schltiter et al. (2002) suggested that the transcurrent
function of the Sumatran Fault might be in the process of
shifting to the Mentawai Fault. This is an attractive hypothesis
but difficult to reconcile with the suggestion by Sieh &
Natawidjaja (2000) that the total offset on the Sumatran Fault is
rather small, despite the abundant evidence (including occasional
large earthquakes; Untung et al. 1985) for recent and continuing
offsets along it.
A further complication is introduced by a possible relationship
between the Mentawai Fault and the Batee Fault. The latter is a
dextral splay from the Sumatran Fault that trends offshore
near the Banyak Islands and was interpreted by Karig et al.
(1980) as displacing or terminating the Mentawai Fault near
Nias (Fig. 2.1). The Mentawai Fault is often shown as either
ending near Nias (e.g. Diament et al. 1992) or merging with the
Batee Fault, but a very strong gravity gradient indicates a major
structural discontinuity between the two westernmost islands in
the Banyak group (see Fig. 3.5). This is roughly the position
where a Mentawai Fault continuation would be expected if
the Batee Fault were not present. Moreover, the existence of
Mentawai-type structures still further north has been confirmed
by Izart et al. (1994) and by Malod & Kemal (1996) using
single-channel reflection data.
Additional insights into the role of the Mentawai Fault in
the Enggano area were provided in June 2000 by an Mw 7.9 earthquake followed by a train of strong aftershocks (Fig. 2.5). P and S
wave studies of the primary event suggested that this comprised
two subevents, involving strike-slip within the Indian Ocean
Plate followed by thrust motion on the subduction fault
(Abercrombie et al. 2003). The events were too deep, and in the
wrong plate, to be due to failure on the Mentawai Fault, but
they do provide important data on its relationship to the transition
between the accretionary wedge and the continental margin.
Matson & Moore (1992) suggested that this transition occurs
near the east coast of Nias in the Central Domain and that the
subduction fault originally reached the surface in this area. Its
subsequent migration oceanwards was interpreted as a consequence of the development of the accretionary wedge that now
forms the forearc ridge. This is consistent with the Malod &
Kemal (1996) interpretation of the Mentawai Fault along its
entire length as marking the transition between the wedge and
a rigid backstop of pre-existing basement. On this hypothesis,

the linearity of the fault is a consequence of the linearity of the

original subduction trace, which would, in turn, have been controlled by the linearity of the former passive margin.
The significance of the Enggano composite earthquake to the
backstop concept is that the GPS results shown in Figure 2.5 indicate that in this area, and possibly only for short periods, the accretionary wedge moves with the subducting plate and must therefore
compress against the backstop, resulting in folding and reverse
faulting. Potential energy stored in this folded and faulted zone
can be released in large earthquakes in which the wedge moves
oceanwards and deformation near the backstop is reversed. Presumably such reversals are only partial, so that deformation gradually increases. At no point in this stick-slip cycle would large
earthquakes necessarily occur within the wedge, because accreted
material is usually too weak to sustain large local stress. Large
earthquakes will therefore be associated principally with the
unsticking of the wedge from the backstop or from the downgoing
slab along the main subduction thrust and with relative lateral
movement between locked and unlocked segments of
the forearc. Events of both types appear to have occurred in
June 2000, with the movement between segments of the Indian
Ocean plate increasing the stress and triggering failure along
the subduction thrust (Abercrombie et al. 2003).
The results of future GPS measurements in the Enggano-Bengkulu area (there have, unfortunately, been no measurements on
Enggano since the earthquake) are thus likely to be very different
from those obtained between 1991 and 1993. Amongst other
things, they can be expected to provide insights into the highly
controversial question of the extent to which trench-parallel
motion is accommodated by the Mentawai Fault. It seems probable that the new vectors will resemble the vectors shown in
Figure 2.4 for the islands north of Siberut, i.e. they will show
almost entirely trench-parallel motion, implying a primarily transcurrent long-term function. The characteristics of both the main
earthquake and the extensive aftershock sequence suggest that
effects of the Enggano Great Earthquake are unlikely to be seen
in the forearc north of Bengkulu (Abercrombie et al. 2003), and
in fact no such effects have been observed in post-earthquake
GPS studies in the Central Domain (Bock et al. 2003). If this is
the case, then dangerous levels of stress must be accumulating
in the region from South Pagai to Siberut.
The June 2000 Enggano earthquake was completely overshadowed by the December 2004 Simeulue event, information on
which was posted on the National Earthquake Information

14

CHAPTER 2

Center website within a few days (NEIC 2005). The suggested

maximum displacement was 15 m, in a region where convergence
is more nearly orthogonal to the trench than it is further south (see
Figs 2.1 and 2.4). Bizarrely, in view of this latter fact, the results
from the only GPS site NW of the change of strike, on Pulau Babi
(PB on Fig. 2.4), suggest that during the 1989-1993 period the
forearc moved slightly further in a direction parallel to the
trench than did the Indian Ocean, the supposed driver of the
forearc motion. It also seems that about half of the Indian Ocean
trench-normal motion was accommodated between Pulau Babi
and Sumatra, which is less than at Enggano, but much more
than predicted by simple sliver-plate models. The motion of
Simeulue, a few tens of kilometres to the NW, might, of course,
have been different but there is no bathymetric or other evidence
for the placement by NEIC (2005) of an extensional (or any other)
boundary to a 'Burma Plate' immediately east of Pulau Babi.
Fault-plane solutions for the Simeulue earthquake are consistent with either SW-directed thrusting dipping at about 10 ~ to the
NE or NE-directed reverse faulting dipping at about 80 ~ (NEIC
2005). The first of these is much the more likely, but thrusting
on a surface so nearly horizontal, when the Benioff Zone dips at
about 30 ~ in the vicinity of the hypocentre, raises some questions.
The Harvard Centroid Moment Tensor solution, however, places
the centroid west of the forearc ridge and beneath the eastern
wall of the trench (at 3.09~ 94.26~ cf. the NEIC epicentre at
3.30~ 95.96E~ Since, subject to errors introduced by faulty velocity models, hypocentres correspond to points of rupture intiation
whereas centroids represent weighted average locations of
moment release (Meredith Nettles pers. comm. 2005), the results
can be interpreted as describing an event initiated in the vicinity
of the Mentawai Fault and propagating oceanwards and also
NW along the forearc. The complexity of stress patterns in the epicentral area is indicated by the multiplicity of previous smaller
shocks, some of which had strike-slip solutions and others solutions similar to that of the December 2004 event (see
Newcomb & McCann 1987, Fig. 2). The fact that the region
around the Mentawai Fault appears to respond to stress in different
ways at different places and at different times is consistent with the
fault itself being the expression of a fundamental geological discontinuity rather than a simple break through an essentially homogeneous rock mass.
The Simeulue event also spectacularly confirmed the extreme
segmentation of the forearc. Aftershocks occurred along 1200
km of the arc, from the site of the main shock as far as the northern
tip of the Andamans, but there was virtually no activity to the SE
(NEIC 2005). The bathymetric high northwest of Simeulue where
the epicentre was located may therefore be the surface expression
of a discontinuity similar to those associated with the Banyak and
Batu highs further south. The extents of Great Earthquake ruptures
are strongly correlated with the extents of deep marine basins
between Sumatra and the forearc ridge and, given that the NW
limit of the rupture zone of the 1861 event was not at Simeulue
but at the Banyak Islands (Newcomb & McCann 1987), it seems
possible that stress is still building up in a 'Simeulue Basin'
segment, to be catastrophically released at some time in the not
too distant future.

Vertical movements
It is more difficult to monitor vertical movements with GPS than
horizontal movements, both because of the generally smaller
displacements and because the accuracy is inherently lower.
At present, more reliable estimates of rates of vertical motion
are being obtained by observing short-term changes in relative
sea level. Natawidjaja et al. (2000) studied the submergence and
emergence of corals and deduced a pattern of progressive
landward tilting of the forearc ridge, with uplift within about

115 km of the trench axis and subsidence at all greater distances.

Instantaneous vertical movements of tens of centimetres associated with large earthquakes were superimposed on this pattern.
Individual islands in the northern part of the forearc often record
similar tilting. Islets shown on Dutch colonial maps as protecting
Sinabang harbour, at the eastern end of the north coast of Simeulue
(S in Fig. 2.4), are now permanently submerged, and palm trees
are dying along much of the coast as salt water invades the soil
around their roots. Muara Siberut, the main town on Siberut
(MS on Fig. 2.4), is regularly flooded at high tide and some
nearby offshore 'islands' consist entirely of mangroves with
their roots submerged even at low tide.
On Nias the situation is more complicated, since the west coast
can be divided into two very different sectors. In the north the
coastal region is flat and swampy and the beach is broad and
gently sloping, but in the south there are cliffs 5 0 - 1 0 0 m high
and the sea floor shelves steeply. This section of the coastline is
concave seawards and appears to be a scarp created by failure of
an unstable slope (see Fig. 2.1). The relatively low gravity field
along the coast and offshore (see Fig. 3.5) suggests loss of mass
from this region and also supports the concept of failure of a
slope that has been uplifted to unsustainable elevations. On
the opposite (eastern) side of the island, rivers have been incised
in narrow valleys to depths of 5 - 1 0 m within a broad coastal
plain east of the Mentawai Fault, suggesting recent and rapid
uplift, but further north there is evidence of both uplift and
subsidence.
The uplift of the coastal plain on Nias could have been associated with great earthquakes. Zachariasen et al. (1999) interpreted
the results of a detailed study of coral heads exposed around
the Mentawai Islands of Sipora and North and South Pagai,
south of Siberut, as recording aseismic subsidence followed by
co-seismic uplift related to the great earthquake of 1833. In this
area, and in contrast to areas further north, both aseismic and
co-seismic movements appear to have involved tilting towards
the trench. Deducing long-term regional displacement patterns
from measurements of movements over a few years, or even
over tens of years, is clearly never going to be a simple exercise.

Note added in proof

The earthquake activity in the central Sumatra forearc between 26
December 2004 and the end of April 2005 is summarized in Figure
2.8. The first four plots show how the seismicity associated with
the 26 December event gradually died away during the succeeding
three months. It is clear that even as late as March 2005, the
majority of events were part of the aftershock sequence.
However, on 28 March 2005 there was a further Great Earthquake,
with an epicentre just west of the Banyak Islands and an estimated
magnitude of 8.6. The distribution of aftershocks to this event indicated that rupture extended throughout the whole of the region
between the Banyaks and the December 26 epicentre. It was, in
fact, being quite widely predicted in the first few months of
2005 that this would be where the next break would occur.
However, and unexpectedly, the zone of aftershocks also extended
south as far as the Batu Islands (Fig. 2.8e). It seems therefore that
not only had the last remaining segment that had no historic record
of Great Earthquakes failed, but that the segment that ruptured in
1861 moved with it.
Fault plane solutions by both the NEIC and the Harvard group
indicated a shallow thrust, at an even smaller angle of dip than
had been the case the previous December. Once again, movement
seems to have been initiated close to where the Mentawai Fault
(assumed to be near vertical) would reach the subduction fault at
depth, and once again there was a significant displacement
between the calculated positions of the epicentre and the centroid.
In this case, however, the centroid lay south rather than west of the

SEISMOLOGY & NEOTECTONICS

6~

~lc

oo-,\

oO~O-

....

: December 26 - Decemb'er 31, 2004 I

%.,

,o~

(~)

. o

0o ::;:

.....................................................

\ ,

...........

...............~ .........................
\

.............

February, 2005

................... ~

........... ...............

............ .....
%
:

o o% ~

a,,,i,,,,~.,20osl

.....

o -w=.

"-,

~ ......~I

F2"

,o.

., .

~'

. . . . . .

..................................................

.................-..~ , . ,
+

............... ..................~ .

.~

..........

M a r c h ! - March 27. 2005

,.~ o
O ~

...........

,'4"N

"
.

..................?....... ~

'

15

"l

f ...........................
"2

.C)
.....

~. . . . .

..............April 10 - April;30, 2005

d
. . . . . . . . . . . . . . .

'

I ~)

' ....

o'o
......
~

epicentre, and was still a considerable distance from the trench.

Also, and as might have been expected in view of the smaller magnitude of the shock, and hence the probable smaller width of the
slip zone, the displacement between centroid and hypocentre
was considerably less than in December. The greater distance of
the centroid from the trench, together with the smaller magnitude,
may be sufficient explanation for the much smaller associated
tsunami, which was only about 3 metres high on exposed coasts
of Nias and Simeulue and decreased rapidly in amplitude at
more remote locations. It is also possible that submarine slides,
which may have contributed to the destructive power the December wave, did not occur in March because of the absence of any
remaining potentially unstable slopes. The aftershock sequence
(Figures 2.8e and f) was notable for being much more tightly constrained to the region immediately beneath the forearc ridge than
had been the case following the December event.

.........................
~,....;

.............,,~
.......

,~

~:

............
;.: ........ i o ~

..................................
2~

Fig. 2.8. Central Sumatra seismicity,

December 26 2004 to April 30 2005.
Epicentres plotted from catalogues available
on the Internet from http://
www.ngdc.noaa.gov. Note that the periods
covered (shown in the top fight hand corner
of each diagram) are not of uniform length,
being dictated in part by the dates of
initiation of significant earthquake swarms.
The circles corresponding to the NEIC
epicentres of the two Great Earthquakes (in
plots a and e) are shaded and the locations of
the centroids of their Harvard CMT
solutions are indicated by fault-plane
solution 'beachballs'.

A new train of events began still further south and just seaward
of Muara Siberut in the following weeks. There were a few relatively weak shocks in this area in the period immediately after
March 28 (Figure 28e), but the first major event (Mw=6.7) took
place on April 10, and was followed three quarters of an hour
later by another strong (Mw=6.5) shock. Once again, the Mentawai Fault appears to have controlled the location at which
failure was initiated. Both events were compressional but, in contrast to the two Great Earthquakes, the slip planes were much
steeper (from 30 ~ to 60~ There followed numerous weaker
events in the same area but, again in contrast to the pattern associated with the Great Earthquakes, there was no significant rupture
propagation (Fig. 28f). It is to be hoped that the earthquakes in
this isolated cluster will prove to be the last major events in the
current phase of southward-propagating unzipping of subduction
west of Sumatra.

Chapter 3

The gravity field

JOHN M I L S O M & A D R I A N W A L K E R

Data sources
The gravity field of Sumatra and the surrounding marine areas is
shown in Figure 3.1. Contours in the onshore area of Bouguer
gravity, but offshore are of free-air gravity. Terrain corrections
have not been applied. Although marine gravity measurements
have been made in the forearc basin and elsewhere on a number
of research cruises (e.g. Kieckhefer et al. 1981), the data from
these generally widely spaced lines have not been used in preparing the maps because free-air gravity values obtained from
inversion of satellite radar altimetry provide more systematic
coverage and can resolve anomalies with widths of as little as
7 km (Sandwell & Smith 1997). The onshore and satellitederived offshore data were matched at coastlines without undue
difficulty, as should be the case because both free-air and
Bouguer corrections are zero at sea level. However, gradients
tend to be steep at the coasts in the forearc region, partly
because of the change from free-air gravity, which is strongly
correlated with local bathymetry, to Bouguer gravity, which is corrected for local topography.
Figure 3.2 shows the locations of the onshore stations used in
preparing Figure 3.1, but not of the offshore estimates, distributed
on a regular 2 minute grid. Onshore data were obtained from a
variety of sources, but unfortunately the results of the many
detailed gravity surveys carried out by oil companies remain
confidential. The largest single available data set was assembled
as part of the collaboration between the British Geological
Survey (BGS) and the Geological Research and Development
Centre (GRDC) during the period 1988-1995. Almost all of
Sumatra south of the equator was covered at a reconnaissance
level, although there are significant gaps in a few areas where
access would have been especially difficult. In addition to the
Sumatra mainland, measurements were made on Bangka and
Billiton islands in the northeast and the Mentawai islands in the
west (Fig. 3.2). GRDC have published numerous Bouguer
maps at 1:250 000 scale showing contours, generally at 2 mGal
intervals, and station locations. There are also two summary
maps at 1 000 000 scale (Padang and Palembang sheets), contoured at 5 mGal intervals and without station positions. Terrain
corrections, of up to 12 mGal, were applied in preparing the
summary maps but were not used for any of the 1:250 000 detailed
maps. The two versions of Bouguer gravity are therefore slightly
different in the mountainous areas close to the Sumatran Fault
but gradients in these areas are in any case steep, and overall
patterns are very similar.
Coverage north of the equator, principally by GRDC and
LEMIGAS (the Indonesian Petroleum Research Institute), is less
complete than in the south but is progressing rapidly. Moreover,
Japanese universities working between 1977 and 1979 obtained
data along many of the more important roads in the Lake Toba
area (Fig. 3.2). In the northern forearc LEMIGAS collaborated
with the University of London in surveys of all of the major
islands (Milsom et al. 1991). Stations were mainly along the
coasts, except on Nias. L E M I G A S / U o f L stations on Siberut
were restricted to the southeastern corner, but the island was
subsequently covered at a reconnaissance level by GRDC.
In 1991 and 1992, stations were established along major roads
throughout Sumatra by BAKOSURTANAL, the Indonesian

geodetic survey authority. A map showing the locations of the

BAKOSURTANAL stations and Bouguer gravity contours after
the application of a severe high-cut filter has been circulated
on a very limited basis, but these stations are not included in
Figure 3.2. An unfiltered but very small scale version of the
BAKOSURTANAL Bouguer map was published by Kadir et al.
(1996), and the data may also have been used by GRDC in preparing the 1:10000000 Bouguer anomaly map of Indonesia
(Sobari et at. 1993). BAKOSURTANAL Bouguer values around
the Toba caldera are generally 1 0 - 2 0 m G a l higher than those
reported by the Japanese groups, a difference probably due to
the lack of terrain corrections in the Japanese work.
The onshore contours in Figure 3.1 are based on actual point
gravity data where available, supplemented where necessary by
values estimated at known BAKOSURTANAL station positions
using the contours of Kadir et al. (1996). Accuracy is inevitably
low where this has been done, and even so some significant gaps
remain. The problem of making full use of good regional coverage
where this exists and at the same time displaying in an acceptable
way the results of interpolation across larger gaps has been
addressed by overlaying the map based on a relatively fine
(0.1 ~ grid, which is blank in areas of inadequate coverage, on a
map produced using a much coarser grid and a greater degree of
interpolation. This is obviously unsatisfactory as a quantitative
method, but Figure 3.1 is intended to be used only qualitatively
and the general patterns can be considered sufficiently well established to support regional interpretation. It is just possible on
Figure 3.1 to identify discontinuities in the colour patterns at the
edges of areas where the coarse grid has been used.
Extending Figure 3.1 to include Billiton has brought western
Java within the boundaries of the map. The data used were
obtained in 1970 by the BGS, working in conjunction with the
Geological Survey of Indonesia. The results of recent more
detailed work on Java by GRDC are not shown but are generally
compatible with the BGS survey.

Regional gravity patterns

The most prominent features in Figure 3.1 are offshore. Gravity
highs with north-south or N N E - S S W trends are associated
with fracture zones and seamount chains on the Indian Ocean
Plate and these control the positions of individual culminations
on the broad flexural high at the outer margin of the Sumatra
Trench. Two deep NW-SE-trending free-air lows, associated
respectively with the trench and the forearc basin, intervene
between this oceanic domain and the Sumatran mainland and
are separated from each other by a high along the forearc ridge.
The low over the trench exists because the mass deficit of the
water column is not in local isostatic equilibrium but is balanced
elastically by the offset mass of the subducting slab.
Although the available gravity coverage is much less complete
north of the equator than in the south, there can be no doubting
the existence of fundamental differences between SE and NW
Sumatra. In the south the Barisan mountains are associated with
a narrow, discontinuous and rather weak Bouguer low that,
where it exists, coincides quite precisely with the axis of the
mountain range, but in the north the low deepens and expands to

GRAVITY FIELD

17

Fig. 3.1. The gravity field of Sumatra and the surrounding seas, based on data from sources discussed in the text. Contours are of free-air gravity offshore and Bouguer
gravity onshore. The Bouguer reduction density is 2.67 Mg m -3. Faint white contours are bathymetry, at 200 m and at intervals of 500 m thereafter, from the GEBCO digital
atlas prepared by the British Oceanographic Data Centre. The continuous black line running the length of Sumatra marks the approximate surface trace of the Sumatran Fault.
The yellow line crossing the forearc basin near the equator marks the location of the interpreted profile of Figure 3.6. The black outlines enclosing the letters 'O' and 'B'
indicate the locations of the gravity surveys of the Ombilin and Bengkulu basins shown in Figures 3.3 and 3.4 respectively. The letter B also indicates the approximate
position of the town of Bengkulu. TS and T indicate, respectively, Lake Toba (including Samosir Island) and Lake Tawar. The letters 'IFZ' at about 97 ~ 30'E mark the central
trough of the Investigator Fracture Zone. The inset shows the GEM-T3 long wavelength gravity field in the Sumatra region (see Lerch et al. 1994).