Está en la página 1de 9

Glenn J. Rix,1 Carlo G.

Lai,2 and Sebastiano Foti 3

Simultaneous Measurement of Surface Wave

Dispersion and Attenuation Curves

REFERENCE: Rix, G. J., Lai, C. G., and Foti, S., Simultaneous Nazarian 1984; Snchez-Salinero 1987; Rix 1988; Stokoe et al.
Measurement of Surface Wave Dispersion and Attenuation 1989). Traditionally, differences in phase between pairs of re-
Curves, Geotechnical Testing Journal, GTJODJ, Vol. 24, No. 4, ceivers have been used to construct a dispersion curve showing the
December 2001, pp. 350358.
variation in surface wave velocity with frequency. This dispersion
curve is inverted to obtain a shear wave velocity profile. More re-
ABSTRACT: In existing surface wave test procedures, experi-
mental dispersion and attenuation curves are determined separately cently, Rix et al. (2000) used surface wave measurements to deter-
(i.e., uncoupled) using different source-receiver configurations and mine the material damping ratio profile of a layered soil deposit. In
different interpretation methods. A new procedure based on dis- a manner similar to velocity measurements, an attenuation curve is
placement transfer functions is proposed in which dispersion and at- constructed from the observed spatial attenuation of Rayleigh wave
tenuation data are derived simultaneously (i.e., coupled) from a sin-
gle set of measurements using the same source-receiver array. The amplitudes and then inverted to obtain the shear damping ratio
new approach is motivated by the recognition that in dissipative me- profile. In these studies, however, the measurement and analysis
dia, Rayleigh phase velocity and attenuation are not independent as procedures used to obtain the shear wave velocity and material
a result of material dispersion. Therefore, a coupled analysis of dis- damping ratio profiles at a site were performed separately (i.e., un-
persion and attenuation is a more robust, fundamentally correct ap- coupled). This practice ignores the close relationship between ve-
proach. The new approach is also more consistent with coupled in-
version techniques to obtain the shear wave velocity and shear locity and attenuation that exists in a linear viscoelastic material
damping ratio profiles. The proposed approach is illustrated using (Lai 1998).
data measured at a site in Atlanta, Georgia. Lai (1998) and Rix and Lai (1998) developed a method to si-
multaneously invert experimental dispersion and attenuation
KEYWORDS: dispersion, attenuation, surface wave, Rayleigh curves to obtain the shear wave velocity and shear damping ratio
wave, SASW, shear wave velocity, material damping ratio profiles at a site. The technique is based on the application of a
newly developed algorithm for the solution of the strongly coupled
Rayleigh wave eigenproblem in linear viscoelastic media. How-
Engineers, geophysicists, and seismologists often use surface
ever, the dispersion and attenuation curves were still measured sep-
waves for material characterization across a broad range of scales
arately. In a consistent, coupled approach to surface wave testing
from the near surface to the whole of the earths crust. Surface
the dispersion and attenuation curves should be both measured and
waves include both Rayleigh and Love waves; however, the meth-
inverted simultaneously. This paper presents a new experimental
ods used for near-surface site characterization utilize Rayleigh
procedure suited to this purpose. The technique is based on the use
waves almost exclusively. Surface wave tests are becoming more
of the concept of displacement transfer function between the source
widely used in geotechnical engineering, primarily because they
and a linear array of receivers. When combined with the above-
are noninvasive and, hence, do not require the use of boreholes or
mentioned coupled inversion technique, the new procedure consti-
probes. This aspect of the test is essential for certain types of site
tutes a robust tool for engineering site characterization.
investigations (Stokoe et al. 1988; Rix et al. 1998) and allows the
The paper is organized in three sections: First, the techniques
determination of soil parameters of fundamental interest for static
currently used in conventional surface wave testing for both dis-
(Burland 1989) and dynamic (Dobry 1991) engineering applica-
persion and attenuation measurements are reviewed. Then the new
tions. Surface wave measurement techniques consist of both active
approach in which the experimental dispersion and attenuation
and passive methods (Tokimatsu 1995). This study focuses exclu-
curves are measured simultaneously via the displacement transfer
sively on the interpretation of active surface wave methods.
function concept is described and illustrated. Finally, the new pro-
In the engineering literature, surface wave tests are often called
cedure is applied at a site located in Atlanta, Georgia and the results
Spectral-Analysis-of-Surface-Waves (SASW) tests (Nazarian
are compared with those obtained with the conventional interpreta-
1984; Stokoe et al. 1989; Tokimatsu 1995). The SASW test was de-
veloped during the 1980s to determine the shear wave velocity pro-
file of pavement systems and soil deposits (Nazarian et al. 1983;
Site Description
Associate professor, School of Civil and Environmental Engineering, The testing site is located on the campus of the Georgia Institute
Georgia Institute of Technology, Atlanta, GA. e-mail: of Technology in Atlanta, Georgia and was used for a field demon-
Project engineer, Studio Geotecnico Italiano, Via Ripamonti 89, Milan,
20139 Italy. e-mail: stration during the International Conference on Site Characteriza-
Research engineer, Dipartimento di Strutturale, Politecnico di Torino, tion (ISC 98) held in April 1998. The upper 3 m is a controlled fill
1 0129 Italy. e-mail: composed of silty sand. The residual soils below 3 m are mostly
2001 by the American Society for Testing and Materials

FIG. 1Soil profile at ISC 98 test site.

silty sands with some clayey silts and partially weathered schistose
rock (Finke 1998). The observed water table was more than 10 m
below the surface. The log of a boring performed at the site is
shown in Fig. 1.

Conventional Measurement Techniques

The equipment used in SASW testing consists of a transient or
harmonic source (for active measurements), one or more re-
ceivers, and a recording device. Harmonic sources are more re-
peatable and capable of higher signal-to-noise ratios than tran-
sient sources (Rix 1988). Moreover, the use of harmonic sources FIG. 2Source-receiver configuration for SASW phase velocity mea-
greatly simplifies the interpretation of surface wave data since the surements.
analysis of Rayleigh wave propagation in vertically heteroge-
neous media is more difficult for transient than for harmonic
waves. As with other types of sources, the maximum depth of in-
vestigation achieved using a harmonic source is controlled by the The receivers are typically spaced with a ratio r2/r1  2 (Snchez-
frequency and energy content of the output. The receivers used in Salinero 1987).
surface wave tests for near-surface site characterization are usu- For both the conventional and new procedures described subse-
ally vertical velocity transducers with natural frequencies ranging quently, special attention should be given to reducing near-field ef-
from 1 to 4.5 Hz or seismic accelerometers. The recording device fects. The near field is defined as the area in the vicinity of the
is usually a Fast Fourier Transform (FFT) dynamic signal ana- source where the body-wave field is of the same order of magni-
lyzer. tude as the surface-wave field. Numerical studies simulating wave
propagation in vertically heterogeneous media (Holzlohner 1980;
Rayleigh Phase Velocity Measurements Tokimatsu 1995) have shown that in normally dispersive media the
near-field effects are important up to a distance from the source
Figure 2 shows a typical source-receiver configuration used in equal to approximately one-half of the wavelength. However, in
conventional phase velocity measurements. The configuration is media where the material properties vary irregularly with depth,
referred to as the two-station method (Aki and Richards 1980). In the near-field effects may extend up to distances from the source
this arrangement, a wave field is generated by a harmonic source equal to two times the wavelength. In the far field, it is assumed
oscillating at a circular frequency , while two receivers located at that Rayleigh waves dominate the particle motion at the ground
distances r1 and r2 from the source are used to measure the vertical surface and that dispersion curves derived from measurements at
particle velocity spectrum Uz(r,). The latter is defined as the dis- the ground surface reflect the dispersive characteristics of Rayleigh
crete Fourier transform of the particle velocity time history uz(r,t). waves.

Efforts should be made to eliminate or minimize near-field ef- A

fects, unless they are explicitly accounted for during the inversion
process (Rosset et al. 1991; Ganji et al. 1998). Ganji et al. (1998)
summarize receiver spacing criteria used by a variety of investiga-
tors to minimize near-field effects. In this study, the ISC 98 site is
normally dispersive and the near field was assumed to extend to a
distance of one-half the wavelength from the source.
Spatial aliasing is another important concern in surface wave
measurements. It is the spatial equivalent of undersampling a sig-
nal in the time domain. According to the Nyquists criterion, spa-
tial aliasing is avoided if the receiver spacing is chosen to be
smaller than one-half the wavelength being sampled. Near-field
and spatial aliasing effects control the selection of receiver offsets
and frequency ranges in both conventional and new surface wave
measurement techniques.
From the particle velocity spectrum Uz(r,) the signal analyzer
calculates the auto-power spectrum Grr() at each receiver and the B
cross-power spectrum Gr1r2() between the two receivers:
Grr()  Uz(r,)U
Gr1r 2()  Uz(r1,)U
The symbol () used in Eq 1 denotes complex conjugation. The
time delay between receivers as a function of the circular frequency
is given by arg[Gr1r 2()]/. Thus, the phase velocity VR() of the
far-field Rayleigh wave can be computed from:
(r2  r1)
VR()   (2)
arg[Gr1r 2()]
Equation 2 yields the experimental dispersion curve associated
with the pair of receivers located at r1 and r2. This procedure is then
repeated for different receiver spacings to obtain data at additional
frequencies, and the individual dispersion curves are combined to-
gether to form the composite dispersion curve of the site (Nazarian FIG. 3(a) Cross-power spectrum phase for r1  5 m and r2  10 m
1984). and (b) composite dispersion curve at the ISC 98 site from conventional
Figure 3a shows a plot of the cross power spectrum phase mea- two-station method.
sured at the ISC 98 site for r1  5 m and r2  10 m. Additional
measurements were performed at r1  10 m and r2  20 m and at
r1  20 m and r2  40 m. Figure 3a displays the phase in a wrapped
(modulo 2) format. To apply Eq 2 the phase must be unwrapped. eight or more receiver offsets are a reasonable choice. Measure-
For data like that shown in Fig. 3a, this process is straightforward ments may be performed at all of the receiver offsets simultane-
and may be automated easily. For more complex data, manual in- ously if multiple transducers are available, or measurements
terpretation is often required. Figure 3b shows the composite dis- involving fewer transducers may be combined to form a synthetic
persion curve resulting from the application of Eq 2 to all three array if a repeatable source is used (Zywicki 1999). For attenuation
pairs of receivers. measurements, which depend on accurate measurements of particle
It is important to note that the phase velocity computed with Eq motion amplitude, it is essential that the verticality and physical
2 is an apparent Rayleigh phase velocity (Tokimatsu 1995; Lai coupling of each receiver be checked carefully.
1998) that may reflect the contribution of several modes of The relevant spectral quantity is the particle velocity auto-power
Rayleigh-wave propagation. In normally dispersive media such as spectrum Grr(), which is calculated at each receiver location. The
the ISC 98 site, the apparent phase velocity usually corresponds to experimental particle displacement spectrum can be readily com-
the fundamental mode (Gucunski and Woods 1992; Tokimatsu puted by:
1995). It can be shown that the apparent phase velocity is also a lo- | Uz(r,) | Grr
cal quantity, i.e., its value depends on the location where it is mea- | Uz(r,) |     (3)
sured, and thus, is a function of r1 and r2 (Lai 1998). C1() C1()
where C1() is a frequency-dependent calibration factor that con-
Rayleigh Attenuation Measurements verts the electrical output of the velocity transducer (volts) into
kinematical units (e.g., m/s). Once the particle displacement am-
The source-receiver configuration used in Rayleigh wave atten- plitudes | Uz(r,) | have been determined, the Rayleigh attenuation
uation measurements is shown in Fig. 4. Unlike conventional phase coefficients R() can be computed from a nonlinear regression
velocity measurements where two receivers are used, attenuation based on:
measurements involve a larger number of receiver offsets arranged
in a linear array (multistation method). In the authors experience, | Uz(r,) |  FzG (r,)eR()r (4)

FIG. 4Source-receivers configuration for SASW attenuation measurements.

FIG. 5(a) Particle displacement spectra from uncoupled multistation method at the ISC 98 site.

where G (r,) is the geometrical spreading function describing the phase velocity and attenuation measurements. By introducing the
geometric attenuation of multimode Rayleigh waves in vertically concept of displacement transfer function, both the experimental
heterogeneous media, and Fz is the magnitude of the harmonic dispersion and attenuation curves are determined using a multista-
source applied at the free surface. The geometrical spreading func- tion method. Furthermore, in the new procedure, dispersion and at-
tion is calculated from the solution of the Rayleigh forward problem tenuation curves are determined simultaneously using a single set
using the shear wave velocity profile determined from the inversion of measurements. The new procedure is an important part of a ro-
of experimental dispersion curve (Lai 1998; Rix et al. 2000). bust approach to surface wave testing in which the dispersion and
Particle displacement spectra measured at the ISC 98 site are attenuation curves are both measured and inverted simultaneously
shown in Fig. 5a for r  5, 10, 20, and 30 m. Figure 5b shows the (Lai 1998).
results of the nonlinear regression to determine the attenuation co-
efficients R() at a frequency of 51.55 Hz. By repeating this cal- Displacement Transfer Function
culation at other frequencies, the Rayleigh attenuation curve shown
in Fig. 5c was determined. As for the phase velocity measurements, In a linear system, which in this case is a linear viscoelastic
the experimental attenuation coefficients R() determined by Eq medium, the ratio between an output and an input signal in the fre-
4 are apparent attenuation coefficients that may reflect the combi- quency domain is called the frequency response function, or trans-
nation of several modes of Rayleigh wave propagation (Lai 1998). fer function, of the system (Oppenheim and Willsky 1997). In the
In summary, the conventional procedures used to obtain VR() test configurations shown in Fig. 4, the input signal is the harmonic
and R() are inconsistent. A two-station method is used for dis- force Fzeit applied by the vertically oscillating shaker, while the
persion measurements, while a multistation method is used for at- output signal is the vertical displacement Uz(r,) measured at a dis-
tenuation measurements. Moreover, because the dispersion and at- tance r from the source. For far-field measurements, the vertical
tenuation curves are determined separately, the relationship displacement Uz(r,) induced in a linear viscoelastic vertically het-
between them is neglected. erogeneous medium by a harmonic source Fzeit located at the
ground surface can be written as (Lai 1998):
New Measurement Techniques
Uz(r,)  FzG (r,)e i[t (r,)] (5)
This section introduces a new approach to surface wave disper-
sion and attenuation measurements. An essential aspect of the new where (r,) is a complex-valued phase angle. Hence, the dis-
procedure is that it overcomes the inconsistencies of conventional placement transfer function T(r,) between the harmonic source

B In the implementation of the algorithm, there is a difficulty as-

sociated with the fact that the geometric spreading function G (r,)
in Eq 7 is not known a priori because it depends on the still unde-
termined shear wave velocity profile. This difficulty can be over-
come by means of an iterative procedure that combines the use of
Eq 7 with the solution of both the Rayleigh forward and inverse
problems in viscoelastic layered media (Lai 1998). Initially, the ex-
perimental dispersion and attenuation curves are obtained from Eq
7 with a geometric spreading function G (r,) assumed proportional
to 1/r, which is the geometric spreading relation of Rayleigh
waves in homogeneous media. These curves are inverted to obtain
approximate profiles of shear wave velocity and material damping
ratio that are then used to calculate an improved estimate of G (r,).
The subsequent iteration uses the updated G (r,) to determine new
dispersion and attenuation curves from Eq 7. This procedure is re-
peated until convergence. In most cases G (r,) does not differ sig-
nificantly from 1/r, and the iterative scheme converges rapidly.

Displacement Transfer Function Measurements

The equipment configuration used for displacement transfer
function measurements is the same as that used for conventional at-
tenuation measurements (i.e., the multistation method). Surface
waves were generated by a vertically oscillating, electrodynamic
shaker operating in swept-sine mode. The frequency range used in
the field tests was 5 to 100 Hz. Rayleigh waves were recorded by
vertical transducers with a natural frequency of 1 Hz at various off-
sets from the source. The geophones were spaced at 1-m intervals
up to an offset of 6 m, at 2-m intervals from 6 to 10 m, at 5-m in-
tervals from 10 to 30 m, and at 10-m intervals from 30 m to the
maximum offset of 60 m. At each receiver location, the experi-
mental transfer functions were obtained from an average, in the fre-
quency domain, of 10 measurements to reduce the variance of the
FIG. 5(b) uncoupled particle displacement regression and measured spectral quantities. The acceleration of the armature
(c) Rayleigh attenuation curve from uncoupled multistation method at the mass was measured with a piezoelectric accelerometer.
ISC 98 site.
The spectral quantity recorded by the dynamic signal analyzer at
each receiver location was the mobility M(r,), which is defined as
the ratio of the particle velocity measured at the receiver to the in-
and the receiver is given by: put force measured at the source. From the mobility M(r,), the ex-
perimental displacement transfer function can be readily computed
T(r,)    G (r,)ei (r,) (6) by:

Assuming (r,)  K()r with K()  
VR()  T(r,)  

eliminates the implicit dependence of the complex-valued phase where C1() and C2() are the frequency-dependent calibration
angle on the source-to-receiver distance and Eq 6 becomes: factors of the velocity transducer and accelerometer, respectively.
Uz(r,) The mass of the armature is included in C2(). The dynamic signal
T(r,)    G (r,)eiK()r (7) analyzer calculates the transfer function in such a way that uncor-
Fze it
related output noise is eliminated (Bendat and Piersol 1986).
The assumption (r,)  K()r is equivalent to considering the Figure 6 shows the phase and amplitude of the displacement
phase angle (r,) to be the result of a single mode of propagation. transfer function spectra measured at the ISC 98 site for r  5, 10,
Because the dynamic signal analyzer allows direct measurement of 20, and 30 m. The phase shown in Fig. 6a is unwrapped to facili-
T(r,), Eq 7 can be used as a basis of a nonlinear regression analy- tate comparisons between different receiver offsets.
sis for determining the complex-valued wavenumber K(). It is im-
portant to note that the method determines the apparent Rayleigh Nonlinear, Complex Inversion of Transfer Function
phase velocities and attenuation coefficients from the experimen- Measurements
tally measured displacement transfer functions T(r,) over the
same linear array of receivers. Thus, it provides consistency be- Once the experimental transfer functions have been measured,
tween phase velocity and attenuation measurements. It should also Eq 7 is used to determine the frequency-dependent Rayleigh
be noted that in this procedure VR() and R() are determined si- wavenumbers K(). As mentioned earlier, this requires an iterative
multaneously from the complex-valued transfer functions via the scheme, since the geometric spreading function G (r,) appearing
complex-valued wavenumber. in Eq 7 is not known a priori. With the initial assumption G (r,)

A curves in an uncoupled analysis may individually fit the experi-

mental data better than the corresponding curves obtained with a
coupled analysis. This is not surprising since an uncoupled anal-
ysis separately minimizes the misfit between the theoretical
phase and amplitude curves and the corresponding experimental
measurements. In other words, in an uncoupled analysis the
phase and amplitude of the experimental transfer functions are
regarded as independent quantities, and as such they are fit inde-
pendently using different theoretical models. In a coupled anal-
ysis, a single theoretical model is used to fit both phase and am-
plitude data.

Results Obtained at the ISC 98 Site

Figure 8 shows the experimental dispersion and attenuation
curves obtained from displacement transfer function measurements,
B together with the results of conventional measurements. The raw
transfer function data were selected to minimize near-field and spa-
tial aliasing effects, and only frequencies ranging from 12.6 to 100
Hz were used in the analysis. Furthermore, experimental transfer
functions measured at distances from the source greater than 30 m
were rejected due to a poor signal-to-noise ratio. Both the dispersion
and attenuation curves vary smoothly with frequency.
In Fig. 8a, the dispersion obtained with the conventional two-
station method is more irregular. This is expected, since the mul-
tistation method determines an average Rayleigh phase velocity
over the entire array of receivers, instead of between only two re-
ceivers. Note also that the largest differences between the two
curves occur at higher frequencies. This also is expected, since
the receiver spacings used for high frequencies in the two-station
method are small compared to the total length of the receiver ar-
ray. As such, lateral variability of soil properties and/or changes
FIG. 6Displacement transfer function (a) phase and (b) amplitude in the local phase velocity (Lai 1998) will have a more pro-
measured at the ISC 98 site.
nounced influence on the dispersion curve derived using the two-
station method. Figure 8b compares the attenuation curves ob-
tained with the conventional (uncoupled, multistation) and new
1/r, one must determine the value of K() that best matches the (coupled, multistation) measurement techniques. The conven-
experimental transfer functions T(r,). The value of K() can be tional method yields attenuation coefficients that are less than the
obtained with a classical least-squares technique; however, since proposed method. The differences between the two attenuation
K(), G (r,), and T(r,) are complex-valued quantities, the actual curves are attributed to two phenomena: (1) the use of transfer
implementation of the algorithm requires the definition of norm functions instead of displacements and (2) the use of a coupled in-
valid in a Pre-Hilbert space (Parker 1994). Mathematically, K() is stead of an uncoupled inversion.
obtained by minimizing the misfit between the experimental and The coupled, multistation technique yields dispersion data over
the predicted displacement transfer functions at a given frequency: the same range of frequencies as the conventional technique. For
both tests, this range is dictated by the energy of the source and the

{[T(r,)  G (r,)eiK()r ]
ambient noise levels at the site. As such, the maximum depth of in-
vestigation is approximately the same for both methods.
conj [T(r,)  G (r,)e ]}  min
where N is the total number of data and conj(. . .) denotes complex Conclusions
conjugation. The algorithm used for minimizing Eq 9 is the Leven-
In existing surface wave test procedures, experimental disper-
berg-Marquardt method (Press et al. 1992). The particularly simple
sion and attenuation curves are determined separately (i.e., uncou-
form of the partial derivative of Eq 7 with respect to K() makes
pled) using different source-receiver configurations and different
the implementation of this algorithm very efficient.
interpretation methods. A new procedure based on displacement
The analysis yields a series of complex wavenumbers K() 
 transfer functions has been proposed in which dispersion and at-
iR()]. This is, in essence, equivalent to a simultaneous tenuation data are derived simultaneously (i.e., coupled) from a
determination of the experimental dispersion and attenuation single set of measurements using the same source-receiver array.
curves at a site. Figure 7 shows the results of the nonlinear regres- The new approach is motivated by the recognition that, in dissipa-
sion to determine the complex-valued wavenumber K(), and tive media, Rayleigh phase velocity and attenuation are not inde-
hence VR(), and R(), at the ISC 98 testing site for 34.81 Hz. pendent as a result of material dispersion (Aki and Richards 1980).
The analysis is repeated at other frequencies to define the complete Therefore, a coupled analysis of dispersion and attenuation is a
dispersion and attenuation curves. more fundamentally correct approach. Dispersion estimates using
It is important to note that theoretical phase and amplitude an array-based, coupled analysis are more robust than conventional

FIG. 7Coupled transfer function regression for (a) phase velocity and (b) attenuation coefficient at 34.81 Hz at the ISC 98 site.

A Small Strains, Canadian Geotechnical Journal, Vol. 26, pp.

Dobry, R., 1991, Soil Properties and Earthquake Ground Re-
sponse, Guest Lecture, Proceedings, X European Conference
on Soil Mechanics and Foundation Engineering, Florence, Italy.
Finke, K. A., 1998, Piezocone Penetration Testing in Piedmont
Residual Soils, M.S. thesis, Georgia Institute of Technology.
Ganji, V., Gucunski, N., and Nazarian, S., 1998, Automated In-
version Procedure for Spectral Analysis of Surface Waves,
Journal of Geotechnical and Geoenvironmental Engineering,
Vol. 124, No. 8, pp. 757770.
Holzlohner, U., 1980, Vibrations of the Elastic Half-Space Due to
Vertical Surface Loads, Earthquake Engineering and Struc-
tural Dynamics, Vol. 8, pp. 405414.
B Gucunski, N. and Woods, R. D., 1992, Numerical Simulation of
the SASW Test, Soil Dynamics and Earthquake Engineering,
Vol. 11, pp. 213227.
Lai, C. G., 1998, Simultaneous Inversion of Rayleigh Phase Ve-
locity and Attenuation for Near-Surface Site Characterization,
Ph.D. dissertation, Georgia Institute of Technology.
Nazarian, S., Stokoe, K. H., II, and Hudson, W. R., 1983, Use of
Spectral Analysis of Surface Waves Method for Determination
of Moduli and Thicknesses of Pavement Systems, Transporta-
tion Research Record 930, Transportation Research Board,
Washington, D.C., pp. 3845.
Nazarian, S., 1984, In Situ Determination of Elastic Moduli of Soil
Deposits and Pavement Systems by Spectral Analysis of Surface
Waves Method, Ph.D. dissertation, The University of Texas at
FIG. 8Rayleigh (a) dispersion and (b) attenuation curves obtained Austin.
with conventional and new measurement techniques at the ISC 98 site. Oppenheim, A. and Willsky, A., 1997, Signals and Systems, Pren-
tice-Hall, Englewood Cliffs, New Jersey.
Parker, R. L., 1994, Geophysical Inverse Theory, Princeton Uni-
versity Press, New Jersey.
two-station measurements because they are based on a larger quan-
Press, W. H., Teukolsky, S. A., Vetterling, W. T., and Flannery, B.
tity of data. Finally, the new approach is also more consistent with
P., 1992, Numerical Recipes in FortranThe Art of Scientific
coupled inversion techniques to obtain the shear wave velocity and Computing, 2nd ed., Cambridge University Press, Cambridge,
shear damping ratio profiles. United Kingdom.
The use of displacement transfer functions has several limita- Rix, G. J., 1988, Experimental Study of Factors Affecting the
tions as well. Like conventional methods, the new method cannot Spectral Analysis of Surface Waves Method, Ph.D. disserta-
resolve multiple modes of surface wave propagation. This may or tion, The University of Texas at Austin.
may not be a serious drawback depending on site conditions (i.e., Rix, G. J., Lai, C. G., and Spang, A. W., Jr., 2000, In-Situ Mea-
normally dispersive versus inversely dispersive profiles (Toki- surement of Damping Ratio Using Surface Waves, Journal of
matsu 1995)) and on the type of inversion program employed by Geotechnical and Geoenvironmental Engineering, Vol. 126,
the user. The user should consider more advanced methods of dis- No. 5, pp. 472480.
persion analysis using, for example, frequency-wavenumber anal- Rix, G. J. and Lai, C. G., 1998, Simultaneous Inversion of Surface
ysis (Zywicki 1999) if the ability to resolve multiple modes is im- Wave Velocity and Attenuation. Geotechnical Site Character-
portant. Second, a coupled analysis of dispersion and attenuation ization, P. K. Robertson and P. W. Mayne, Eds., Balkema, Rot-
requires an iterative technique because the geometric spreading terdam, pp. 503508.
function depends on the shear wave velocity profile. Third, the Rix, G. J., Lai, C. G., Foti, S., and Zywicki, D., 1998, Surface
computational requirements of the coupled analysis exceed those Wave Tests in Landfills and Embankments, Geotechnical
of a conventional, two-station dispersion analysis. Earthquake Engineering and Soil Dynamics III, ASCE Geotech-
nical Special Publication No. 75, P. Dakoulas, M. Yegian, and R.
Acknowledgments D. Holtz, Eds., pp. 10081019.
Rosset, J. M., Chang, D. W., Stokoe, K. H., II, 1991, Compari-
This material is based upon work supported by the National Sci-
son of 2-D and 3-D Models for Analysis of Surface Wave Tests,
ence Foundation under Grant No. CMS-9616013.
5th International Conference on Soil Dynamics and Earthquake
Engineering, Karlsruhe, Germany, pp. 111126.
Snchez-Salinero, I., 1987, Analytical Investigation of Seismic
Aki, K. and Richards, P. G., 1980, Quantitative Seismology: The- Methods Used for Engineering Applications, Ph.D. disserta-
ory and Methods, W. H. Freeman and Company, San Francisco. tion, The University of Texas at Austin.
Bendat, J. and Piersol, A., 1986, Random Data - Analysis and Mea- Stokoe, K. H., II, Nazarian, S., Rix, G. J., Snchez-Salinero, I.,
surement Procedures, 2nd ed., John Wiley & Sons, New York. Sheu, J. C., and Mok, Y. J., 1988, In Situ Seismic Testing of
Burland, J., 1989, Small is BeautifulThe Stiffness of Soils at Hard-to-Sample Soils by Surface Wave Method, Earthquake

Engineering and Soil Dynamics IIRecent Advances in Ground Tokimatsu, K., 1995, Geotechnical Site Characterization Using
Motion Evaluation, ASCE Geotechnical Special Publication No. Surface Waves, First International Conference on Earthquake
20, J. L. Von Thun, Ed., pp. 264278. Geotechnical Engineering, K. Ishihara, Ed., Balkema, Rotter-
Stokoe, K. H., II, Rix, G. J., and Nazarian, S., 1989, In Situ Seis- dam, pp. 13331368.
mic Testing with Surface Waves, Proceedings, XIIth Interna- Zywicki, D. J., 1999, Advanced Signal Processing Methods Ap-
tional Conference on Soil Mechanics and Foundation Engineer- plied to Engineering Analysis of Seismic Surface Waves, Ph.D.
ing, Balkema, Rotterdam, pp. 331334. dissertation, Georgia Institute of Technology.