ARTICLE IN PRESS

Soil Dynamics and Earthquake Engineering 27 (2007) 234241

www.elsevier.com/locate/soildyn

Earthquake ground motion simulation with frequency-dependent

soil properties
Jiewu Meng
WPC Inc., 1017 Chuck Dawley Boulevard, Mount Pleasant, SC 29464, USA
Accepted 4 July 2006

Abstract
A one-dimensional earthquake ground motion simulation with frequency-dependent dynamic soil parameters, which include shear
wave velocity and quality factor, is performed within the frequency band between 10
2 and 30 Hz. For the simulation, the general
frequency-dependent dynamic soil properties are obtained by using a new non-resonance technique on various soils. Comparisons
between transfer functions involving frequency-dependent and -independent dynamic soil parameters indicate that soil amplication
throughout the considered frequency range is sensitive to the values of the dynamic soil parameters, in particular to the quality factor.
Simplication of frequency-dependent dynamic soil parameters into constants, which prevail as a convenient practice, incurs large
amplitude distortion. This study therefore demonstrates that the dynamic properties of soils play an important role in the near surface
earthquake ground motion analyses and need to be used in the same way as how they were measured.
r 2006 Elsevier Ltd. All rights reserved.
Keywords: Earthquake ground motion; Frequency dependence; Dynamic soil properties

1. Introduction
The inuence of soil deposits on seismic ground motion
is enormous in terms of site amplication and thus
structural damage and ground failures. Ground motion
amplication is determined by two competing factors:
impedance and attenuation of the soil deposits encountered
along the seismic wave traveling path. In current North
America building codes (e.g., International Building Code,
Edition 2003 [1]), the upper 30 m surface soil deposits
overlying the higher impedance earth crust is regarded as
most relevant and important in characterizing the seismic
behavior of a site [2,3]. A large amount of literature
studying site amplication is available on topics involving
linear vs. non-linear, homogeneous vs. layered, and onedimensional vs. multi-dimensional in modeling [4]. Assessment of the site amplication is primarily performed by
using dynamic soil properties, which include stiffness (the
shear modulus or the shear wave velocity, V s f ) and
attenuation (the damping ratio or the quality factor, Q(f )).
Tel.: +1 843 884 1234; fax: +1 843 884 9234.

E-mail address: jmeng@wpceng.com.

0267-7261/$ - see front matter r 2006 Elsevier Ltd. All rights reserved.
doi:10.1016/j.soildyn.2006.07.002

Despite the broadband nature of seismic events over a

range of frequencies below 30 Hz, the effect of frequency
on dynamic soil properties, in particular attenuation, is
generally not considered in seismic analyses such as by
Seed and Idriss [5]. For rocks where the effect of frequency
is slight, this practice is likely acceptable. However, for
natural soils that exhibit greater frequency dependence, this
practice may become undesirable. This is particularly true
for attenuation, which is a known to be a function of
frequency in the earthquake frequency band. For natural
soils with nes content of only 5% or more, as the loading
frequency increases, the shear modulus increases at an
amount of approximately 6% per decade of loading
frequency [6,7] and attenuation in the damping ratio is a
function of the loading frequency with an obvious damping
ratio trough (or, a quality factor hump) existing between
about 10
110 Hz (e.g., [8]). The damping ratios fall into a
range of approximately 15%, or equivalently, the quality
factors range from 10 to 50. In other words, oversimplifying the frequency dependency of the damping ratio into
constants will articially underestimate the capacity of soil
deposits to dissipate energy in some frequency bands and
overestimate it in others. This practice will directly and

ARTICLE IN PRESS
J. Meng / Soil Dynamics and Earthquake Engineering 27 (2007) 234241

adversely result in a distorted site amplication estimate.

Therefore, it is essential to investigate the importance of
the frequency-dependent dynamic soil properties used in
seismic ground motion analyses.
In terms of characterization of the dynamic soil properties with respect to the earthquake frequency, they can be
effectively explored with various geotechnical laboratory
experiments. Amongst the methods, resonant column and
torsional shear tests has been the primary tests throughout
the past half-century [9]. However, these methods are
single-point or narrow band measurements which are
limited by their testing mechanisms and are thus not well
equipped with spectral measurement capacity in the seismic
frequency range under consideration. To overcome sparseness of the measurements, excessive interpolation based on
one to two measurements per frequency decade is typically
exercised. This situation is pointed out by Kibblewhite [10],
who considers it fraught with traps and difculties. A
non-resonance method [11,12] has been introduced recently
to address this technical limitation. The method permits
spectral measurements in a virtually continuous manner
(i.e. very dense) in the range of frequencies from 10
2 to
30 Hz, in typical. It is the unique advantage of this study to
exhibit the signicance of the frequency effect on seismic
ground motions by using the measured frequency-dependent dynamic soil properties.
The essential aim of this paper is to demonstrate the
importance of the frequency-dependent dynamic soil
properties used in earthquake ground motion amplication
analyses through the following steps:

(1) Observe the general dynamic properties of representative natural soils in the earthquake frequency range by
using a newly developed experimental technique
[11,12]. The measurements will be obtained in a
spectrum manner and provide sufcient details to
overcome the undesirable need of excessive interpolation.
(2) Establish frequency-dependent dynamic soil properties
formulae, which consist of shear wave velocity and
quality factor per frequency, by using the measurements from reconstituted soil specimens. In addition,
the known depth-dependent dynamic soil properties
will also be demonstrated and addressed in the
formulae.
(3) Utilize a one-dimensional synthetic model consisting of
a single soil layer overlying half-spaced bedrock subject
to vertically incident SH shear waves to investigate the
importance of using the frequency-dependent dynamic
soil properties. Transfer functions between the surface
of the soil layer and the top of the underlying bedrock
are calculated by using a discrete-time wave propagation technique based on the method of S- afak [13].
While this model is certainly an oversimplication of an
actual case, it provides an appropriate basis for
demonstration purposes.

235

By nature of the dynamic soil properties under

consideration and the transfer function used in the
synthetic model simulation, the procedures performed in
this study are more suitable for weak-motion events for
which the site response is presumably linear. In addition,
the amount of the surface soils of 30 m represents less than
1% of the distance from the source and spatial attenuation
is likely limited when compared with intrinsic attenuation
[14]. It is therefore considered an appropriate substitute of
attenuation by intrinsic attenuation in this study. Overall,
this study is fundamental and essential because it provides
insight and a basis for more advanced studies such as nonlinearity and multi-dimensionality.
2. The non-resonance method
The main concept of this experiment was originally
established in mechanical engineering studies of polymers
and biomaterials, which typically exhibit signicant linear
viscoelasticity (i.e., frequency-dependent behavior) in
common testing conditions. The ratio between loading
stress and resulting strain is used to deduce a complex
elastic modulus, from which the shear wave velocity V s f
and the quality factor Q(f ) can be subsequently extracted.
For a forced harmonic torsional oscillation of a viscoelastic
cylinder at small amplitudes as illustrated in Fig. 1, a
transfer function between the excitation, T 0 ei2pf t , and
response, f0; t, of a specimen is obtained by

4

T 0 ei2pf t
pR
cot On
2

rl


J
(1)
0 2pf ,
n
f0; t
2
O
q
where O f r2pf 2 l 2 =G  f , f is loading frequency,
R, l, and r are radius, length, and density of the test
specimen, respectively, and J0 is the mass polar moment of
the electromagnetic motor. By denition, the complex
shear modulus is G n f G 1 f i G 2 f , where G1(f) and
G2(f) are storage and loss modulus, respectively . The shear

T
Rigid
mass

R
Specimen

Fig. 1. Conceptual conguration of the non-resonance experimental

method.

ARTICLE IN PRESS
J. Meng / Soil Dynamics and Earthquake Engineering 27 (2007) 234241

236

wave velocity V s f and the quality factor Q(f) can be

calculated using the following denitions, respectively:
s
 G n f 


(2a)
V s f 
,

r 
Q
1 f

G2 f
.
G1 f

(2b)

Amongst its advantages, the non-resonance method can be

congured to yield spectral measurements in the seismic
frequency range of 10
230 Hz with reasonable effort. For
example, experiments can be performed with 150200
measurements per decade within their linear strain
amplitude of approximately 10
5.

3. Tested soils
To explore the dynamic soil properties per loading
frequency on a general basis, measurements were performed on three representative natural soil specimens
including sand, silt, and clay:
1. Clayey sand (SC), which was classied per the Unied
Soil Classication System [15]. It was formed during the
Late Tertiary to Quaternary periods and sampled at a
depth of 14 m below the existing ground surface in
Charleston, South Carolina.
2. Sandy elastic silt (MH) and sandy silty clay (CL-ML).
They are of the Piedmont geologic province and were
sampled at depths of 12 and 23 m, respectively, below
the existing ground surface in Opelika, Alabama. The
samples have clayey components greater than 5%.
In addition to the frequency effect on the dynamic soil
properties, embedment depth (or, conning stress) is
another important factor. For modeling purposes, it is
complicated to determine the embedment depth effect
regarding the measurements from the natural soils that
involve uncertainty from the geological origin, embedment
depth, and soil classication. However, this problem can be
tackled by using uniform soils remolded at certain levels of
conning stresses to mimic the embedment depths between
0 and 30 m. Kaolin is one of the most available modeling
minerals that have been broadly used in the study of static
and dynamic behavior of soils. In particular, remolded
kaolin demonstrates a similar attenuation behavior to that
of various natural soils [8] and was therefore selected for
this study. The kaolin used in this study was mined from
middle Georgia and was of the Cretaceous and Tertiary
geologic regions. The specimens were prepared at selected
conning stresses of 173, 242, 380, and 483 kPa, which
correspond to vertical stress levels at depths of approximately 10, 15, 20, and 30 m, respectively.

4. Observations of the frequency-dependent dynamic soil

properties
Fig. 2 shows the general dynamic soil properties of the
natural soil specimens per frequency. The spectra in the top
panel indicate that shear wave velocity moderately
increases with frequency at a rate of approximately 2%
per frequency decade. On the other hand, the quality factor
spectra in the bottom panel uniformly show an apparent
trend of frequency dependency of attenuation behavior and
echo the previous limited ndings by Shibuya [8]. However,
the measurements in this study provide a much higher level
of details per frequency and avoid the need of excessive
interpolation between measurements in the most demanding frequency range for the subsequent seismic motion
analyses. Fluctuations in the quality factor spectra were
mainly from the effect of ambient noise and are obvious in
the sense of the strain amplitude (i.e., o10
5) of the
measurements. In spite of the effect of ambient noise, the
hump-trend is still distinguishable during visual survey
with the advantage of dense measurements. In addition, it
is apparent that the shear wave velocity is proportional to
the conning stress levels (or, the embedment depth)
whereas it is less apparent in the quality factors. Due to
the convoluted difference in the ne contents, geological
origins, and sample depths between the specimens, it is
difcult to single out the depth dependence in a denite
manner with these natural soil specimens.
Figs. 3 and 4 show the spectral measurements on the
remolded kaolin specimens. In brevity, the measurements
consistently agree well with the general frequency-dependent dynamic soil properties from the natural soil
measurements. The four shear wave velocity spectra are
monotonically increasing with respect to the frequency.
The amount of increment with frequency (i.e., dispersion)
is modest at an approximate rate of 2.5% per decade. It is
noted that the frequency dependency in quality factor
spectra is much more obvious than that illustrated in the
corresponding dispersion spectra. In addition, the depth
dependence can be observed with limited effort: shear wave
velocity is proportional to the conning stress levels;
whereas the quality factor spectra do not appear to be
sensitive to the stress levels under consideration. This depth
dependence nding agrees well with the common understandings (e.g., [16]).
5. Formulation of empirical frequency-dependent spectra
A characteristic quality factor spectrum of the remolded
soils is deduced from a simple quadratic polynomial
regression as shown in Fig. 4. Their frequency-dependent
quality factor formula can thus be described as
Qf
9:9 lg2 f
5:8 lg f 49:8,

(3)

where f is frequency in Hz. The frequency dependency and

pressure dependency of shear wave velocity formula can be

ARTICLE IN PRESS
J. Meng / Soil Dynamics and Earthquake Engineering 27 (2007) 234241

237

Fig. 2. Frequency-dependent shear wave velocity spectra and quality factor spectra of natural soils tested between 10
2 and 30 Hz. The samples were
obtained from the upper 30 m depths below their individual existing ground surfaces.

Fig. 3. Frequency-dependent shear wave velocity spectra of remolded kaolin tested between 10
2 and 30 Hz. The selected conning pressures of 173, 242,
380, and 483 kPa correspond to embedment depths of approximately 10, 15, 20, and 30 m.

modeled as
V s f ; s F 1 f F 2 s,

(4)

where F 1 f , F 2 s, and s are the frequency-dependent

component, the pressure-dependent component, and the

conning pressure, respectively. A characteristic shear

wave velocity spectrum, as shown in Fig. 5, is established
from the four normalized spectra:
F 1 f 0:002 lg3 f 0:002 lg2 f 0:013 lg f 1:032.

(4a)

ARTICLE IN PRESS
238

J. Meng / Soil Dynamics and Earthquake Engineering 27 (2007) 234241

Fig. 4. Frequency-dependent quality factor spectra of remolded kaolin tested between 10
2 and 30 Hz. The selected conning pressures of 173, 242, 380,
and 483 kPa correspond to embedment depths of approximately 10, 15, 20, and 30 m, respectively.

Fig. 5. The normalized shear wave velocity spectra and the characteristic spectra obtained from polynomial regression based on measurements in Fig. 3.

F 2 s is established with arbitrarily selected reference

values at 10
2 Hz and the typical exponential format of
other pressure-dependent formulae [17] is similarly determined
F 2 s 42:5 s0:3 ,
where F 2 s and s are in m/s and kPa, respectively.

(4b)

6. One-dimensional synthetic modeling

By using a synthetic model as schematically shown in
Fig. 6, the frequency response spectra by using the
frequency-dependent and -independent parameters were
compared. A discrete-time transfer function as proposed by
S- afak [13], which models SH shear waves passing through a

ARTICLE IN PRESS
J. Meng / Soil Dynamics and Earthquake Engineering 27 (2007) 234241

239

and maximum of the frequency-dependent shear wave

velocities, V s f , are 228, 235, and 243 m/s, respectively,
and the minimum, mean, and maximum of the frequencydependent quality factor, Q(f), are 20, 35, and 50,
respectively. The parameters are grouped in the following
as the frequency-independent parameters:

Fig. 6. The one-dimensional single soil layer over Bedrock model for
shear wave propagation.

single-soil layer over bedrock was selected for the

comparison. The transfer function of the synthetic model
can be established as
n

Y f 1 rn e
i2pf t

Hf
,
X f
1 rn e
i4pf tn

(5)

where X(f ) and Y(f ) are the Fourier transforms of input

x(t) and output y(t) time history, r and tn are complex
reection coefcient and complex traveling time, respectively. It can be shown that rn 4Qr
i1
r=4Q
i1
r and t 2Q=2Q it; and r rr vr
rs vs =rr vr
rs vs and t h=vs ; where r and t are reection coefcient
and traveling time, respectively; rs , vs and rr , vr are the
mass density and SH shear wave velocity in the bedrock
and soil layer, respectively; and h is the thickness of the soil
layer. A complex shear wave velocity vns vs1 i vs2 was
used to derive the linear viscoelastic model from a linear
elastic model. In this study, the reection coefcient is unit.
In this seismic simulation, the single soil layer is assumed
uniform throughout its thickness of 30 m. Soils located in
the middle depth of 15 m of the single soil layer is assumed
representative for the layer and therefore used to determine
the conning stress s. Afterward, Eqs. (3) and (4) are used
to determine the frequency-dependent shear wave velocities, V s f , and quality factors, Q(f ). The conning stress
of the soils located in the middle depth of 15 m is estimated
approximately 273 kPa and the corresponding shear wave
velocity is 228 m/s from Eq. (4b). From Eqs. (4a) and (4),
the frequency-dependent shear wave velocity, V s f ,
increases monotonically from 228 m/s at 10
2 Hz to
243 m/s at 30 Hz. On the other hand, the frequencydependent quality factor, Q(f ), is determined from Eq. (3)
to be approximately 22 and 20 at 10
2 and 30 Hz,
respectively, and 50 between 10
1 and 2 Hz, which is
illustrated also in Fig. 4.
In practice, measurements from torsional shear and/or
resonant column tests are often generalized as constants
throughout the entire frequency range and assumed valid
under most circumstance. To reect the practice of
simplication and for comparison purposes, values of the
maximum, average, and minimum of their frequencydependent parameters are used as the frequency-independent counterparts for comparisons. The minimum, mean,

1. Constant shear wave velocity, Vs, of 228 m/s and

varying quality factors, Q, of 20, 35, and 50;
2. Constant shear wave velocity, Vs, of 235 m/s and
varying quality factors, Q, of 20, 35, and 50; and
3. Constant shear wave velocity, Vs, of 243 m/s and
varying quality factors, Q, of 20, 35, and 50.

7. Results and discussions

The results are illustrated in Fig. 7 and Tables 1 and 2.
The solid lines in each plot represent the transfer function
calculated with the frequency-dependent dynamic soil
parameters, or the frequency-dependent transfer function
(FDTF); and the dashed lines in each plot show the
variation of the transfer function with constant dynamic
soil properties, or the frequency-independent transfer
function (FITF). In general, the shear wave velocity
determines the location (or, frequency) of the peaks,
whereas the quality factor affects the amplitudes of the
peaks and their decay with frequency.
To quantify the difference between the corresponding
peaks in each comparison group, Table 1 summarizes the
location difference between the FDTF and FITF at
amplication peaks ranked from 1st to 7th. For example,
the frequencies of the 4th peak from the FDTF and the
FITF with Vs equal to 228 m/s are 14.02 and 13.34 Hz,
respectively. Their difference is approximately
0.68
( 13.34
14.02) Hz, which is shown in the row 1 of the
column 4 in Table 1. In general, the frequency difference
between the corresponding peaks is slight with a maximum
of approximately
1.53 Hz. Amongst the constant shear
wave velocities of 228, 235, and 243 m/s, which corresponds
to the minimum, the mean, and the maximum of the
frequency-dependent shear wave velocities, V s f , the
difference between the peak frequencies is the slightest
when the maximum constant shear wave velocity, Vs, of
243 m/s was used as shown in the bottom panel in Fig. 7.
On the other hand, to quantify the difference between
the amplitudes of the corresponding peaks in each
comparison group, Table 2 summarizes the ratio (in dB)
between the FDTF and the FITF at peaks ranked from 1st
to 7th. For example, the amplitudes of the 7th peaks from
the FDTF and the FITF with Vs of 243 m/s and Q of 50 are
2.03 and 5.67, respectively. Their ratio is approximately 8.9
(
20 lg(2.03/5.67)), which is shown in the row 9 of the
column 7 in Table 2. In general, the ratios between the
corresponding peaks are highly dependent on the quality
factors and range from
7.7 to 9.2 dB. It is therefore
obvious that the amplitudes of the peaks are sensitive to

ARTICLE IN PRESS
J. Meng / Soil Dynamics and Earthquake Engineering 27 (2007) 234241

240

Fig. 7. The seismic simulation results comparison between the frequency-dependent transfer function (FDTF) and frequency-independent transfer
function (FITF). The solid lines represent the transfer functions calculated with the frequency-dependent dynamic soil properties.

Table 1
Peak frequency difference between the peaks of frequency-independent transfer function (FITF) and frequency-dependent transfer function (FDTF)
Dynamic soil parameters

Difference between corresponding peaks (Hz)

Vs 228 m/s and varying Qa

Vs 235 m/s and varying Qa
Vs 243 m/s and varying Qa
a

1st

2nd

3rd

4th

5th

6th

7th

0.08

0.02
0.06

0.29

0.12
0.06

0.49

0.20
0.10

0.68

0.28
0.14

1.06

0.54
0.00

1.29

0.66
0.00

1.53

0.78
0.00

Constant Q 20, 35, or 50 per frequency.

Table 2
Ratio (in dB) between the peak amplitudes of frequency-independent transfer function (FITF) and frequency-dependent transfer function (FDTF)
Dynamic soil parameters

Ratio (dB) between the corresponding peak amplitudes

1st

2nd

3rd

4th

5th

6th

7th

Vs 228 m/s

Q 20
Q 35
Q 50

7.7

2.8
0.0

5.8

0.8
2.3

4.8
0.2
3.3

3.7

2.8

1.9

2.6
4.8
7.9

1.8
5.2
8.4

0.9
6.0
9.2

Vs 235 m/s

Q 20
Q 35
Q 50

7.7

2.9

0.2

5.9

0.9
2.2

4.8
0.2
3.1

3.7
1.4
4.6

2.6
2.6
5.8

1.8
3.4
6.6

0.9
4.4
7.5

Vs 243 m/s

Q 20
Q 35
Q 50

7.7

2.8
0.0

5.9

1.1
1.8

4.9

4.0

3.7

3.7
3.9
6.7

2.6
4.8
8.0

1.9
5.1
8.2

0.8
5.8
8.9

ARTICLE IN PRESS
J. Meng / Soil Dynamics and Earthquake Engineering 27 (2007) 234241

the quality factor specied at each frequency as shown in

Fig. 7.
In practice, shear wave velocity and quality factor are
mainly obtained from torsional shear tests between 0.01
and 2 Hz and/or from resonant column tests between 30
and 200 Hz, which depends on soil stiffness and individual
equipment conguration. The parameters obtained this
way are often used as constants throughout the entire
frequency range and assumed valid under many circumstances. However, it is indicated in the preceding comparisons that the largely discounted frequency effect on the
dynamic soil parameters should be restored to avoid soil
amplication distortion. It therefore casts doubt on the
validity of the conventional practice of dynamic soil
parameters determination used for ground motion analyses, for example.
8. Conclusions
This paper presents a fundamental study of frequencydependent dynamic properties of various natural and
remolded soils with the aid of a newly developed nonresonance experimental technique. More importantly, the
role that these frequency-dependent dynamic soil properties play in earthquake ground motion analyses was
demonstrated explicitly by using a one-dimensional seismic
ground motion simulation model. A series of comparisons
involving the frequency-dependent and -independent dynamic soil properties concluded that soil amplication is
sensitive to the values of quality factor in the entire
earthquake frequencies under the consideration. Simplication of the frequency-dependent dynamic soil parameters
into constants incurred substantial amplitude distortion in
the simulation. The dynamic properties of soils play an
important role in earthquake ground motion analyses
when soils are involved and should be referenced in the
same way as how they were measured. The practice of
parameters selection and simplication needs to be
exercised with caution.
Acknowledgments
The author appreciates the anonymous reviewers
comments for improving the quality of this article. The

241

author thanks Messrs Bryan Shiver and Matt Silveston of

WPC, Inc. for proofreading the manuscript.
References
[1] International Building Code Council, International Building Code.
Falls Church, VA, 2003.
[2] Martin GR, Dobry R. Earthquake site response and seismic code
provisions. NCEER Bull 1994;8:16.
[3] Borcherdt RD. Estimates of site-dependent response spectra for
design (methodology and justication). Earthquake Spectra EERI
1994;10:61753.
[4] Joyner WB, Boore DM. Measurement, characterization, and prediction of strong ground motion. In: Soil dynamics and earthquake
engineering II, Proceedings of the ASCE Geotec Eng Div specialty
conference, Park City, UT, 2730 June 1988. p. 43102.
[5] Seed HB, Idriss IM. Soil moduli and damping factors for dynamic
response analyses. Berkeley, CA: Earthquake Research Center,
University of California; 1969 (EERC 70-10).
[6] Leroueil S, Marques MES. State of the art: importance of strain rate
and temperature effects in geotechnical engineering, measuring and
modeling time dependent soil behavior. In: ASCE Geotechnical
special publication no. 61, New York, 1994. p.160.
[7] Stokoe KHI, Darendeli MB, Andrus RD, Brown LT. Dynamic soil
properties: laboratory, eld and correlation studies, In: The second
international conference on earthquake geotechnical engineering,
Lisboa, Potugal, 1999. p. 81145.
[8] Shibuya S, Mitachi T. Strain-rate effects on shear modulus and
damping of normally consolidated clay. Geotech Test J 1995;18(3):
36575.
[9] Woods RD. Laboratory measurement of dynamic soil properties. In:
Dynamic geotechnical testing II, Philadelphia. 1994. p. 165201.
[10] Kibblewhite AC. Attenuation of sound in marine sediments: a review
with emphasis on new low-frequency data. J Acoust Soc Am
1989;86(2):71638.
[11] Meng J. The inuence of loading frequency on dynamic soil
properties. PhD. thesis, Georgia Institute of Technology, 2003,
178pp.
[12] Rix GJ, Meng J. A non-resonance method for measuring dynamic
soil properties. Geotech Test J 2005;28(1):18.
[13] S- afak E. Discrete-time analysis of seismic site amplication. J Eng
Mech 1995;121(7):8019.
[14] Anderson JG, Lee Y, Zeng Y, Day S. Control of strong motion by
the upper 30 m. Bull Seismol Soc Am 1996;86(6):174959.
[15] ASTM D2487. Standard practice for classication of soils for
engineering purposes (unied soil classication system). ASTM
International; 2000.
[16] Dobry R, Vucetic M. Dynamic properties and seismic response of
soft clay deposits. In: International symposium on geotechnical
engineering of soft soils, vol. 2, Mexico City, 1987. p.5187.
[17] Ishihara K. Soil behaviour in earthquake geotechnics. Oxford, UK:
Clarendon Press; 1996.