You are on page 1of 10

Soil Dynamics and Earthquake Engineering 34 (2012) 89–98

Contents lists available at SciVerse ScienceDirect

Soil Dynamics and Earthquake Engineering


journal homepage: www.elsevier.com/locate/soildyn

Modeling shear rigidity of stratified bedrock in site response analysis


Hing-Ho Tsang a,n, M. Neaz Sheikh b, Nelson T.K. Lam c
a
Department of Civil Engineering, The University of Hong Kong, Hong Kong
b
School of Civil, Mining and Environmental Engineering, University of Wollongong, Australia
c
Department of Infrastructure Engineering, The University of Melbourne, Australia

a r t i c l e i n f o abstract

Article history: Where a distinct soil–rock interface exists, the bedrock medium is commonly treated as elastic half-
Received 8 April 2011 space and the bedrock surface as the lower boundary of the soil-column model for site response
Received in revised form analyses (or the lower boundary of the finite element model for soil-structure interaction analyses).
4 August 2011
While shear wave velocity in bedrock varies with depth, there has been no consensus amongst
Accepted 22 October 2011
scientists and practitioners over the value of ‘‘effective depth’’ into bedrock at which the ‘‘half-space’’
Available online 21 November 2011
shear wave velocity value should be taken for modeling purposes. This paper reports an interesting and
important observation that the effective depth into bedrock is sensitive to the shear wave velocity
profile of the overlying soil sediments. A simple and heuristic method, namely Resonant Period
Equivalence (RPE) Method, is proposed herein for representing a stratified elastic bedrock of inhomo-
geneous properties by an equivalent homogeneous elastic half-space medium, which is characterized
by a single equivalent shear wave velocity (VR) value. The proposed calculation method has been
verified by extensive comparative analyses involving the use of programs SHAKE and NERA and
employing the complete shear wave velocity models of both the soil sediments and the underlying
stratified bedrock.
& 2011 Elsevier Ltd. All rights reserved.

1. Introduction those on an adjacent rock outcrop. The Peak Spectral Ratio (PSR) is
accordingly defined as follows:
In situations where a distinct soil–rock interface exists, the RSDmax
bedrock medium is commonly treated as elastic half-space and the PSR ¼ ð1Þ
RSDT g
bedrock surface as the lower boundary of the soil-column model
for site response analyses (or the lower boundary of a finite where RSDmax is the maximum response spectral displacement
element model for soil-structure interaction analyses). The shear value as indicated on the soil spectrum whereas RSDT g is the
rigidity of the assumed elastic half-space is important given that response spectral displacement value as indicated on the rock
the amount of seismic waves reflected at the interface back up into spectrum at the natural period of the site (Tg) (refer to the
the soil medium (causing wave amplification by superposition) schematic diagram in Fig. 1).
depends much on the amount of seismic impedance contrasts The effects of multiple wave reflections within the soil med-
between rock and soil. Given that site modifications of earthquake ium are well represented by the elastic displacement response
ground shaking are associated with filtering mechanisms and spectrum (Fig. 1) and the associated engineering definition of the
superposition of reflected waves within the soil sedimentary layers site amplification factor as defined by Eq. (1). This format of
overlying bedrock, the calculated response of the soil and the representing site effects is distinguished from that of conven-
superstructure in seismic conditions can be sensitive to changes tional code models.
in the shear rigidity of the modeled half-space when other The values of PSR can be calculated using computational proce-
parameters are held constant. dures for site response analyses (e.g., program SHAKE [1]). Alterna-
The effects these soil–rock modification mechanisms have tively, estimates of PSR can be obtained by a hand calculation
upon the drift demand on superstructure can be quantified by procedure, which is known as the Single-Period Approximation (SPA)
the elastic displacement response spectrum, which represents model [2,3]. Both classes of procedures have the shear rigidity, or
ground motions estimated on the soil surface in comparison with shear wave velocity (SWV), of the bedrock half-space (VR) parame-
terized. As revealed by sensitivity studies employing these proce-
dures, the value of PSR can be affected significantly by variations in
n
Corresponding author. Fax: þ852 2559 5337. the intensity of ground shaking as well as the value of VR character-
E-mail address: tsanghh@hku.hk (H.-H. Tsang). izing the bedrock materials. Fig. 2 shows the correlation of the

0267-7261/$ - see front matter & 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.soildyn.2011.10.007
90 H.-H. Tsang et al. / Soil Dynamics and Earthquake Engineering 34 (2012) 89–98

Nomenclature VR equivalent shear wave velocity of the bedrock half-


space
a impedance ratio VS weighted average of the shear wave velocity of
g shear strain amplitude in soil soil layers
z soil damping ratio (as a proportion of critical damping) Vs,30 shear wave velocity at 30 m depth
zi minimum (initial) damping ratio (characterizing small FF far-field
strain conditions) NF near-field
zmax maximum damping ratio (characterizing fully degraded PI plasticity index of soil
conditions) PSR peak spectral ratio
dj thickness of individual rock layer j QWL quarter-wavelength
DR effective depth of bedrock RP resonant period
fg fundamental resonant frequency of the site RPE resonant period equivalence
G secant shear modulus RSDmax maximum response spectral displacement value on
G0 shear modulus at small strain conditions soil spectrum
hi thickness of individual soil layer i RSDT g response spectral displacement value at natural per-
H total thickness of soil layers iod Tg on rock spectrum
n total number of soil layers RSV response spectral velocity value
N total number of rock layers considered for computing DR RSVmax maximum response spectral velocity value
T natural period of single-degree-of-freedom system RSVT g response spectral velocity value at natural period Tg
Tg fundamental resonant period of the site on rock spectrum
Ti initial fundamental resonant period of the site SPA single-period approximation
Vi shear wave velocity of individual soil layer i SWV shear wave velocity
Vj shear wave velocity of individual rock layer j

estimated values of PSR against the intensity of bedrock shaking elastic half-space, which is characterized by the value of VR based on
(which is expressed in terms of the response spectral velocity (RSV) measurements taken immediately below the overlying sediments.
of the bedrock spectrum at the site natural period). A series of curves Up until now, there has been no consensus amongst scientists and
is shown for the VR range of 500–3500 m/s. An additional hypothe- practitioners over the value of ‘‘effective depth’’ into bedrock at
tical case of bedrock rigidity VR ¼‘‘infinity’’ is also shown in the figure which the value of VR should be measured. Consequently, there are
to show the upper bound values. Clearly, PSR correlates strongly with uncertainties associated with the equivalent constant VR value for
the bedrock SWV (which in turn is a function of its shear rigidity). For characterization of the modeled half-space for site response analyses
a given intensity of input motion in bedrock (rock RSV¼100 mm/s), (or soil-structure interaction analyses).
the value of PSR varies from 1.3 to 3.8 over the considered range of VR The key objective of this paper is to introduce the Resonant Period
values. The importance of accurately estimating the value of VR for Equivalence (RPE) Method, which is a new method for determining
the calculation of the site amplification factor is now evident. the equivalent velocity value VR for any given SWV profile in rock
It is important to recognize that the shear rigidity properties and the overlying soil medium. Importantly, RPE eliminates the
(or SWV) in bedrock typically vary with depth given that crustal need for calibrating simulated against recorded strong motion data
materials are actually heterogeneous in nature, as the acoustic in the determination of suitable values of VR. Thus, the method to be
impedance increases with depth where the rock crusts become proposed herein should be of particular appeal to regions where
more compact [4–8]. This is particularly so for sedimentary rock strong motion data is unavailable, or lacking.
formation because of stratifications. Thus, there are significant This paper begins (in Section 2) with a brief review of the well-
contrasts between actual SWV profiles and assumptions asso- established Quarter-Wavelength (QWL) Method, which leads to the
ciated with the elastic half-space model. The common mistake to
make in practice is to consider the bedrock medium as a uniform
7
V s = 300 m/s ; PI = 0
Peak Spectral Ratio (PSR)

V = Infinity
5
V = 3500 m/s

V = 2000 m/s
3
V = 1000 m/s

V = 500 m/s

1
0 100 200 300 400
Rock RSV (mm/s)

Fig. 1. Engineering definition of Peak Spectral Ratio (PSR). Fig. 2. Effects of bedrock shear rigidity on Peak Spectral Ratio (PSR).
H.-H. Tsang et al. / Soil Dynamics and Earthquake Engineering 34 (2012) 89–98 91

conception of the proposed RPE Method. Application of the RPE the resonant frequency fg, or RP (Tg), of the overlying soil
method (in Section 3) for determining the equivalent SWV (VR) sediments. While H (in Eq. (2)) is measured up from the
value is illustrated by the case study of a horizontally stratified soil–bedrock interface by a quarter-wavelength, the effective
elastic bedrock medium of non-uniform shear rigidity properties. depth of bedrock (DR) is measured down from this same interface
The accuracy of the RPE Method is verified (in Section 4) by by a quarter-wavelength of the wave likewise, as defined by
comparison with results from comparative analysis of the com- Eqs. (3) and (4). As the values of SWV vary with depth due to
plete sub-surface model which incorporates SWV profiles of both stratification, the resulted values of H and DR would certainly be
the soil and stratified elastic rock models. Finally (in Section 5), different. This is the proposed RPE Method for the determination
practical issues associated with the application of the described of the DR value.
methodology (based on RPE Method) are discussed.
1 XN
dj
Tg ¼ ¼ 4 ð3Þ
fg V
j¼1 j
2. Conception of resonant period equivalence (RPE) method
X
N
QWL Method has been widely used in various disciplines, such DR ¼ dj ð4Þ
as acoustics (including musical instruments), mechanical system, j¼1
and electromagnetic system. The philosophy behind it is that a
where j is the layer number in bedrock, each having finite
standing wave can be formed when a wave is transmitted to one
thickness dj and SWV Vj. N is the total number of rock layers
end (or medium) and is reflected from the other end (or medium)
considered for computing DR. The equivalent bedrock SWV (VR)
by an impedance discontinuity (contrast). The length of the
value can be computed by dividing DR by the total traveling time
simplest standing wave that can form under this circumstance
of the seismic shear wave, using the following equation:
is one quarter-wavelength long. Standing waves are observed in
physical media such as strings and columns of air. Any waves DR
V R ¼ PN ð5Þ
traveling along the medium will reflect back when they reach the dj =V j
j¼1
end (or other medium).
The degree of reflection depends on the strength of impedance This RPE Method, which has a sound theoretical basis, is
contrast between the two media. For a sound wave transmitted considered easy to implement, and is particularly suitable for
in a solid tube filled with air under atmospheric pressure, the routine design calculations carried out by practitioners.
acoustic impedance ratio between the solid tube and air is It is noticed from Eqs. (3)–(5) that the effective depth (DR) and
(almost) infinite, hence, total reflection would be anticipated. the equivalent SWV of the bedrock (VR) are dependent on the RP of
The situation becomes more complicated when a wave front is the overlying soil profile as well as the SWV profile of the bedrock
transmitted from one medium to another, of which their acoustic itself. In other words, the value of VR can be highly variable within
impedances are comparable; the partitioning of energy occurring the same bedrock medium if the thickness and/or shear rigidity of
at the interface depends on the ratio a of the acoustic impedances the overlying sediments is variable.
of the medium on either side of the interface.
In the context of modeling seismic wave propagation, a
standing wave can be formed when a wave is transmitted from 3. Application of RPE method
the bedrock medium into the overlying soil sediments. The
fundamental resonant period (RP) of the site (Tg) can be estimated Fig. 3(a–d) illustrates the use of the RPE Method for the generic
using Eq. (2) if the thickness (hi) and SWV (Vi) of the individual rock and generic hard rock sites, of which the SWV profiles
soil layers are known, where n is the total number of soil layers. proposed by Boore and Joyner [4] are shown in Fig. 3(a). Generic
Alternatively, the value of Tg can be expressed in terms of the rock represents typical crustal conditions with sediments being
total thickness of the soil layers (H) and its weighted average SWV deposited continuously during different geological periods, and
(VS). The expression is based on taking the value of H being the its SWV profile can be modeled as a nonlinear (power) function
quarter-wavelength of the multiply reflected waves dominating [5,6]. Generic hard rock represents conditions in areas where the
the site response and has frequency fg ( ¼1/Tg). sedimentary overburden has been partially (or even entirely)
1 Xn
hi 4H stripped away due to extensive geological weathering or glacial
Tg ¼ ¼ 4¼ ð2Þ erosion [9], thus resulting in a very shallow gradient in the SWV
fg V
i¼1 i
VS
profile [5].
If the bedrock is rigid (i.e. fixed end boundary), its motion Using the proposed RPE Method (Eqs. (3)–(5)), the values of DR
would not be affected by the motions in the overlying soil. and VR are computed for the two distinct bedrock conditions
In reality, the bedrock is elastic, thus there is coupling effect considering a range of RP of soil sites, as shown in Fig. 3(b) and (c).
between the soil and bedrock media. Ground motions in rock For typical site RP values ranging from 0.4 to 2.0 s, it is found that
comprise of seismic waves with a broad spectrum of frequency for generic rock, the value of DR varies from 90 to 900 m, and the
contents radiated from the source of the earthquake while taking value of VR varies from 800 to 1800 m/s; whereas for generic hard
into account the effects of regional attenuation properties of the rock, the value of DR varies from 300 to 1600 m, and the value of
earth crust. The seismic wave (in rock) with frequency equal to VR varies only from 2900 to 3200 m/s. It is clear from this
the resonant frequency fg of the overlying soil sediments would illustration that the variation of equivalent bedrock SWV (VR)
have the strongest coupling (resonant) effect between the soil and value with the site RP is highly dependent on the gradient of the
bedrock media. Hence, the two media would experience synchro- SWV profile in rock. According to a study of the ice coverage since
nized periodic motions at the resonant frequency fg in a site the last glacial maximum (21,000 years before present) [9], the
response. lowermost reach of glaciation in America is located at latitude
Based on the phenomenon described above, there exists a 42–441N. Hence, shallow SWV gradients typifying glaciated
certain portion of bedrock that is involved in the resonance, regions are actually rare in the global context. In other words,
which is termed hereafter as the effective depth of bedrock (DR). significant variation of the value of SWV with depth is anticipated
Such effective depth is hypothesized herein to be associated with in most situations.
92 H.-H. Tsang et al. / Soil Dynamics and Earthquake Engineering 34 (2012) 89–98

SWV (s)
100 1000 10000
10 10000

Effective Depth (m)


100 1000
Depth (m)

1000 100
Generic Rock
Generic Hard Rock Generic Rock
Generic Hard Rock
10000 10
0.1 1 10
Resonant Period (s)

4000 4
Equivalent Bedrock SWV (m/s)

Peak Spectral Ratio (PSR)


3000
3

2000

2
1000
Generic Rock
Generic Rock
Generic Hard Rock
Generic Hard Rock
0 1
0.1 1 10 0 .1 1 10
Resonant Period (s) Resonant Period (s)

Fig. 3. Application of the Resonant Period Equivalence (RPE) Method. (a) SWV profiles of generic rock and generic hard rock sites (Boore and Joyner [4]) and the estimated
values of (b) DR and (c) VR for a range of RP of soil sites computed by the RPE Method (Eqs. (3)–(5)). (d) The corresponding Peak Spectral Ratio (PSR) estimated by the SPA
model [2] with the RP-dependent VR (RSVT g ¼ 100 mm=s, VS ¼ 300 m/s, PI¼ 0).

Inhomogeneous bedrock medium and its complex coupling SHAKE [1]). The accuracies of the PSR values so obtained from these
with the overlying soil sediments is represented simply by an predictions will need to be verified by a non-linear dynamic analysis
elastic half-space, which is characterized by a single equivalent VR of the complete sub-surface model comprising the SWV profiles of
value (as read off from Fig. 3(c)). The values of PSR as presented in both the stratified bedrock and the overlying soil sediments.
Fig. 1 could then be obtained by site response analyses. PSR values The non-linear dynamic analysis of the sub-surface model can
could also be obtained by inputting the value of VR into the hand- be operated either in the frequency domain (e.g., SHAKE [1]), or in
calculation procedure (SPA model) for estimation of site effects [2]. In the time domain (e.g., NERA [10]). In SHAKE, shear modulus
the illustration, the response spectral velocity on rock ðRSVT g Þ at the degradation and damping of soils are assumed to be functions of
RP of the site (Tg) is assumed to be equal to 100 mm/s, VS equal to shear strain amplitude g and are determined by iterations to be
300 m/s, plasticity index (PI) of soil equal to zero, and densities of soil consistent with the effective strain in each soil layer. On the other
and bedrock are assumed to be the same. It is shown in Fig. 3(d) that hand, NERA performs analysis based on the Iwan–Mroz non-
the value of PSR varies from 1.8 to 2.8 for generic rock sites, and from linear stress–strain model for soils. Multi-yield surface plasticity
3.5 to 3.7 for generic hard rock sites, for site RP values range from constitutive model can be used to compute the response of
0.4 to 2.0 s. Clearly, the more flexible the soil sediment overlying vertically propagating shear wave in a one-dimensional layered
bedrock (i.e. higher site RP), the higher the values of DR and VR and system. While there are usually minor differences in the results
consequently higher PSRs. The significance of an accurate estimation obtained from the time domain and frequency domain analyses,
of VR is hence evidenced, especially for typical rock SWV profiles employing both types of analyses (which are commonly used in
with a high SWV gradient. practice for site response modeling) in the verification process
serves the purpose of detecting systematic errors.
With the complete sub-surface model, the value of VR can be
4. Verification of RPE method obtained by calibration, which is much more time consuming as it
involves ground response analyses of two sub-surface models:
4.1. Objective and principle of the verification method (I) model in which the variation in the shear rigidity of the
rock crust with increasing depth is accurately represented and
The RPE Method presented in the preceding sections enables a (II) model in which bedrock is represented by a half-space
complex (inhomogeneous) bedrock medium to be simplified into an material of homogeneous shear rigidity (which is represented
(homogeneous) elastic half-space, which is characterized by a single by the value of VR). The value of VR as derived from model (II) can
equivalent value of VR. The site amplification behavior (and the value be found by trial and error until the site amplification behavior
of PSR) could then be modeled using computer programs (e.g., represented by the two models matches. The calibration procedure
H.-H. Tsang et al. / Soil Dynamics and Earthquake Engineering 34 (2012) 89–98 93

Motion B Target Motion

Multiple
soil layers Outcrop
Motion C Motion D Motion B
Motion

Multiple
Rock soil layers
Rock
(Depth = 2 km) (Depth = 2 km)
Motion C

Basement hard rock (SWV > 3000 m/s) Basement hard rock (SWV > 3000 m/s) Stage 3
(only soil layers are modelled;
Motion A Motion A
VR adjusted to match the response spectrum
Stage 1 (both soil and rock layers are modelled) Stage 2 (only rock layers are modelled) of Motion D with that of Motion B)
Response Spectral Velocity ( RSV)

RSVmax

Motion B

Motion D [VR adjusted to match closely


with the response spectrum of Motion B]

Motion C

Motion A

Period (T)

Fig. 4. Procedure for calibrating the value of VR.

for determining the value of VR (which is depicted in Fig. 4) is values (RSVmax) that are associated with Motion B and Motion D
described as follows. (refer Fig. 4(d)).
In Stage 1, a combined rock and soil column has been modeled
in a one-dimensional ground response analysis (using computer 4.2. Input motions employed in verification analyses
programs SHAKE [1] and NERA [10]). Rock layers have been
modeled down to a depth of 2 km, below which shear rigidity In order to cover a wide spectrum of ground shaking levels,
does not change significantly with depth. The input motion at the synthetic earthquake accelerograms, with RSVmax of around 20,
base of the rock layer is defined as Motion A (refer Fig. 4(a)). 100 and 300 mm/s at 2 km depth from soil–rock interface were
Computer program GENQKE [11] has been used to generate generated by stochastic simulations of the seismological model
Motion A by stochastic simulations of the seismological model. using computer program GENQKE [11]. For each level of ground
The corresponding ground response at the soil surface is defined shaking, two sets of time histories were generated: one repre-
as Motion B. This is considered as the Target Motion on the soil sents near-field (NF) (source-site distance¼50 km) ground
surface (refer Fig. 4(a)). motions, which are rich in high frequency seismic waves and
In Stage 2, only the rock layers (2 km deep) have been modeled the other represent far-field (FF) (source-site distance¼150 km)
in a one-dimensional ground response analysis. The input motion ground motions which are comparatively rich in low frequency
at the base of the rock layer is Motion A, as in Stage 1. Motion C at seismic waves. Each set contains six simulated acceleration time
the surface of the rock layers has been obtained from one- histories and a sample accelerogram from each set is shown
dimensional shear wave analysis (refer Fig. 4(b)). in Fig. 5.
In Stage 3, only the soil layers have been modeled in one-
dimensional ground response analysis. As rock outcrop motion is 4.3. Soil properties models employed in verification analyses
used as input motion at the soil–rock interface, rock outcrop
motion (Motion C) obtained from Stage 2 has been used as the Five soil columns with VS ¼100, 150, 200, 300 and 500 m/s and
input motion in Stage 3. Motion D at the surface of soil layers has four values of plasticity index (PI¼0%, 15%, 30% and 50%) have been
been obtained from the ground response analysis. The value of included in the verification analyses. There were altogether 20 soil
the input bedrock SWV value VR was then found by trial and error columns of different thicknesses H and with a wide range of initial
based on matching the response spectral properties generated by site period Ti from 0.12 to 2.4 s. This range of site period covered
Motion D and Motion B (the target motion). This calibration sandy soil sites (0.14–0.95 s) and soft soil sites (1.97–2.3 s) as
procedure is illustrated in the schematic diagram of Fig. 4(c). considered in the study by Henderson et al. [12] and Heidebrecht
Given that site effects are represented by the value of PSR at the et al. [13].
RP of the site (Tg), the value of VR can be found by calibration The nonlinear characteristics of the soil layers were captured
based on matching the maximum response spectral velocity by two strain-compatible material parameters, namely, secant
94
Acceleration H.-H. Tsang et al. / Soil Dynamics and Earthquake Engineering 34 (2012) 89–98

1
Near Field, RSV = 20 mm/s
Shear Wave Velocity (m/s)
(m/s/s)

0 100 1000 10000


1
-1
0 5 10 15 20
Acceleration

1 10
Far Field, RSV = 20 mm/s
(m/s/s)

Depth (m)
-1
0 5 10 15 20 100
Vs,30 = 500 m/s
Acceleration

2 Vs,30 = 1000 m/s


Near Field, RSV = 100 mm/s
(m/s/s)

1 1000
0 Vs,30 = 1500 m/s
-1 Vs,30 = 2000 m/s
-2
Vs,30 = 3000 m/s
0 5 10 15 20 10000
Acceleration

2
Far Field, RSV = 100 mm/s Fig. 6. Shear wave velocity (SWV) profiles of rocks.
(m/s/s)

1
0
-1 1 10
-2
0 5 10 15 20
Acceleration

4 0.75

Damping ratio  (%)


Near Field, RSV = 300 mm/s
(m/s/s)

2
0
-2 Average Rock (Schnabel et al. [1])
G/G0

-4
0 10 20 30 40 50 60 0.5 5
Worthington et al. [15] (B = 50,000: n = 0.8)

4
Acceleration

2 Far Field, RSV = 300 mm/s 0.25


(m/s/s)

0
-2
-4
0 10 20 30 40 50 60 0 0
Time (s) 0.0001 0.001 0.01 0.1 1

Fig. 5. Sample synthetic earthquake accelerograms generated by stochastic


Shear Strain  (%)
simulations of the seismological model using GENQKE [11].
Fig. 7. Degradation of shear modulus and damping curves for rocks.

shear modulus G and damping ratio z (as a proportion of critical Chandler et al. [16] reported that the degradation curve of
damping). The dynamic properties of soil adopted in this study Schnabel et al. [1] matched very closely with Eq. (6) as proposed
were obtained by Lam and Wilson [14]. by Worthington et al. [15], in which the value of B and n were
taken to be equal to 50,000 and 0.8, respectively (as shown in
4.4. Rock properties models employed in verification analyses Fig. 7). It is noted that different rock types possess different values
of B and n. However, the precision of the modeling for degrada-
A generic methodology for modeling SWV profiles in crustal tion does not seem to be critical for this study. The same values of
rock has been developed by the authors and collaborator [5]. B and n have been assumed for all rock types.
Using the developed model, SWV profiles could be defined by a series Nishi et al. [17] conducted cyclic tri-axial tests on a wide range
of logarithmic functions, each of which is assigned to a certain depth of sedimentary soft rock (having SWV ranging between 300 m/s
zone. Five hypothetical SWV profiles have been developed by adopt- and 2500 m/s), and recommended that the damping ratio can be
ing the values of the parameter Vs,30 (SWV at 30 m) at 500 m/s, estimated by the following generic expression:
1000 m/s, 1500 m/s, 2000 m/s and 3000 m/s (Fig. 6).    
G G
Models describing the shear modulus degradation and the z ¼ zi þ zmax 1 ð7Þ
G0 G0
associated damping behavior in sand and clay have been well
publicized in the literature, but similar information for rock has where z is the damping ratio (%); zi the minimum (initial) damping
been scarcely reported. Nevertheless, Schnabel et al. [1] proposed ratio (characterizing small strain conditions); zmax the maximum
some damping and degradation curves for ‘‘average rock’’. damping ratio (characterizing fully degraded conditions).
Although a clear definition for this rock type has not been given With the use of both Eqs. (6) and (7), the damping functions for
in the cited publication, a SWV value of 1500 m/s (5000 ft/s) was rocks could be obtained, along with the corresponding zi and zmax
assumed at the surface of the bedrock in the worked examples. values of 0.4% and 10%, respectively. The shear modulus degradation
More recently, a generic empirical relationship, Eq. (6), for and damping functions for rocks adopted in this study were
stiffness degradation in sedimentary rocks, characterized by two calibrated and initially proposed by Chandler et al. [16].
constants, B and n, was proposed by Worthington et al. [15],
where G0 is the shear modulus at small strain. 4.5. Results from verification analyses
n
G ½2=ðBnÞ½ð1 þ BgÞ 1 In total, 20 soil deposits having a wide range of site RP (Tg ¼
¼ ð6Þ
G0 fg þ ½1=ðBnÞ½ð1 þBgÞn 1g 0.14–2.3 s) together with five rock profiles (Vs,30 ¼500–3000 m/s)
H.-H. Tsang et al. / Soil Dynamics and Earthquake Engineering 34 (2012) 89–98 95

have been considered. These essentially cover the whole range of considered (ignoring the influence of the site RP), the estimated site
site conditions usually encountered in site response analyses. response spectra would have been in significant errors.
Fig. 8 represents the application of the proposed calibration Another observation is that a particular soil deposit (Tg) on
procedure using computer program SHAKE to obtain bedrock different rock profiles also needs different values of VR, depending
SWV (VR) for three soil deposits (Tg ¼ 0.16 s, 0.45 s and 1.0 s). Rock on the SWV of the rock layers. It is evident from Fig. 9 that soil
layers have been modeled considering Vs,30 ¼1000 m/s. It can be column of Tg ¼1.0 s requires the values of VR to be set at 1210 m/s,
observed that Motion D matches with Motion B when the values of 2135 m/s and 2745 m/s for rock profiles of Vs,30 equal to 1000 m/s,
VR are set at 750 m/s, 1050 m/s and 1210 m/s. It is noted that the 1500 m/s and 2000 m/s, respectively. This essentially proves the
value of VR has been calibrated with respect to maximum response claim that in site response analyses the value of VR should be
spectral velocity (RSVmax). Results show that had the same value of selected based on the RP of the soil site and the SWV profile of the
VR been used for the site response analyses of the three soil deposits underlying rock layers.

300
200
H = 35 m; Vs = 200 m/s; Tg = 1.0 s

Response Spectral Velocity, RSV


Response Spectral Velocity, RSV

H = 15 m; V = 500 m/s; T = 0.16 s V = 1000 m/s


V = 1000 m/s

150
200
Motion A
Motion A

(mm/s)
(mm/s)

Motion B
100 Motion B
Motion C
Motion C
100 Motion D (Vr = 1210 m/s)
Motion D (Vr = 750 m/s)
50

0 0
0 1 2 3 4 5 0 1 2 3 4 5

Period, T (s) Period, T (s)

400 300
H = 35 m; Vs = 200 m/s; Tg = 1.0 s
Response Spectral Velocity, RSV
Response Spectral Velocity, RSV

H = 25 m; V = 300 m/s; T = 0.45 s


V = 1000 m/s
Vs,30 = 1500 m/s

300
200
(mm/s)
(mm/s)

Motion A Motion A
200
Motion B Motion B
Motion C 100 Motion C

100 Motion D (Vr = 1050 m/s) Motion D (Vr = 2135 m/s)

0 0
0 1 2 3 4 5 0 1 2 3 4 5
Period, T (s) Period, T (s)

300 300
H = 35 m; Vs = 200 m/s; Tg = 1.0 s
Response Spectral Velocity, RSV
Response Spectral Velocity, RSV

H = 35 m; V = 200 m/s; T = 1.0 s Vs,30 = 2000 m/s


V = 1000 m/s

200 200
Motion A
(mm/s)
(mm/s)

Motion A
Motion B
Motion B
Motion C
Motion C
100 Motion D (Vr = 1210 m/s) 100
Motion D (Vr = 2745 m/s)

0 0
0 1 2 3 4 5 0 1 2 3 4 5
Period, T (s) Period, T (s)

Fig. 8. Calibration of the value of VR for cases with three soil deposits (0.16 s, Fig. 9. Calibration of the value of VR for a soil deposit (1.0 s) on three rock profiles
0.45 s and 1.0 s) on a rock profile (Vs,30 ¼ 1000 m/s) using SHAKE [1]. (1000 m/s, 1500 m/s and 2000 m/s) using SHAKE [1].
96 H.-H. Tsang et al. / Soil Dynamics and Earthquake Engineering 34 (2012) 89–98

The proposed RPE Method (Eqs. (3)–(5)) for the estimation of (simplified) RPE Method and from the (much more elaborate)
VR values is evaluated herein by the comparison of results calibration procedure have been demonstrated to be in agreement
obtained using the calibration procedure described above (in within 2–13%.
Section 4.1). The three soil deposits on the rock profile character- The validity of the RPE Method was further evaluated for the
ized by Vs,30 ¼1000 m/s were first considered (Fig. 8). The esti- whole range of soil deposits (Tg ¼0.14–2.3 s) and for different rock
mated VR values based on RPE Method were 765 m/s, 1075 m/s profiles (500–3000 m/s). Excellent agreement between the two
and 1390 m/s, respectively, in comparison with VR values of 750 m/s, sets of results is shown (Fig. 10) and 95% of the estimates are
1050 m/s and 1210 m/s obtained by calibration. Estimates from the within 15% agreement. In other words, the 95% confidence limits
can be taken as 715%. The average error is shown to be around
6% (refer Table 1). This level of accuracy is considered encoura-
ging, given the high level of uncertainties associated with site
response analyses. Further investigation has been made to ascer-
4000
H= 15 m; Vs,soil = 500 m/s tain whether any systematic difference exists for different soil
1.15: 1
H= 25 m; Vs,soil = 300 m/s
1:1 types (represented by the values of PI), shaking levels (repre-
sented by the values of RSVmax) and earthquake events (NF and
VR (Calibrated), m/s

3000 H= 35 m; Vs,soil = 200 m/s 0.85: 1 FF) (Fig. 11). No particular bias has been observed between
H= 50 m; Vs,soil = 150 m/s estimates made from the RPE method and that from the calibra-
H= 60 m; Vs,soil = 100 m/s tion procedure (using program SHAKE). Also, no significant
2000 difference has been observed when RPE Method is compared
with non-linear time domain analyses using NERA and frequency
domain analyses using SHAKE (Fig. 12). This essentially proves
1000 that the proposed RPE Method is equally applicable to both time-
domain and frequency-domain analyses.
In conclusions, 95% of the values estimated by the simplified
0 method based on RPE Method and that obtained by calibration
0 1000 2000 3000 4000 analysis of the full sub-surface model have been found to be within
VR (Estimated from RPE ), m/s 15% agreement. It is noted that the average error decreases with
increasing rigidity of the rock layers (refer Table 1). This is
4000 expected as the equivalent bedrock SWV in stiffer rock does not
Vs,30 = 500 m/s vary significantly with depth, or value of Tg (as illustrated in Fig. 3).
1.15: 1
Vs,30 = 1000 m/s 1:1 Nevertheless, the difference is not highly significant, considering
Vs,30 = 1500 m/s the presence of other uncertainties, and hence, does not deserve
3000
VR (Calibrated), m/s

0.85: 1 any special treatment.


Vs,30 = 2000 m/s
Vs,30 = 3000 m/s
2000
5. Discussion

In conditions pertaining to resonance behavior, the amplified


1000 response behavior of the structure (of similar natural period)
found on the surface of a flexible soil site is controlled mainly by
the site natural period [18]. Given the link between wave period
0 and wavelength, the total thickness of the soil sedimentary layers
0 1000 2000 3000 4000 overlying bedrock (i.e. quarter-wavelength of the dominant wave-
VR (Estimated from RPE ), m/s form) can be critical to the potential seismic demand behavior of
the site. Other controlling parameters are the SWV of the
Fig. 10. Bedrock shear rigidity estimated by the proposed RPE Method compared
with the calibrated value from one-dimensional shear wave analyses for (a) the
individual soil layers and their respective hysteretic (damping)
whole range of soil deposits (0.14–2.3 s) and for (b) different rock profiles properties. This study is concerned mainly with another, related,
(500–3000 m/s) considered in this study. important controlling parameter: the SWV of the bedrock, VR.

Table 1
Average error in the estimates of the bedrock SWV (VR) by RPE Method when compared with results obtained from calibration.

Average error (%)

Thickness of soil deposits H¼ 15 m H¼ 25 m H ¼30 m H ¼50 m H ¼60 m


6.5 6.3 4.0 6.5 7.6

Bedrock SWV profile Vs,30 ¼500 m/s Vs,30 ¼1000 m/s Vs,30 ¼ 1500 m/s Vs,30 ¼ 2000 m/s Vs,30 ¼ 3500 m/s
10.4 8.4 6.0 6.0 3.2

Soil plasticity PI¼ 0% PI¼ 15% PI ¼30% PI ¼50%


6.1 7.4 7.0 8.2

Shaking level RSVmax ¼ 20 mm/s RSVmax ¼ 100 mm/s RSVmax ¼300 mm/s
6.0 6.6 4.9

Earthquake event Near-field (NF) Far-field (FF)


5.6 6.1

Analysis procedure Frequency domain (SHAKE) Time domain (NERA)


5.6 6.0
H.-H. Tsang et al. / Soil Dynamics and Earthquake Engineering 34 (2012) 89–98 97

4000
4000
SHAKE 1.15: 1
1.15: 1 1:1
1:1 NERA
PI = 0%

VR (Calibrated), m/s
PI = 15% 3000
VR (Calibrated), m/s

3000 0.85: 1 0.85: 1


PI = 30%
PI = 50%

2000 2000

1000 1000

0 0
0 1000 2000 3000 4000
0 1000 2000 3000 4000
VR (Estimated from RPE ), m/s
VR (Estimated from RPE ), m/s
Fig. 12. Bedrock shear rigidity estimated by the proposed RPE Method compared
4000 with the value calibrated from one-dimensional non-linear frequency domain
1.15: 1 (SHAKE [1]) and time domain (NERA [10]) analyses.
1:1
RSVmax = 20 mm/s
RSVmax = 100 mm/s
VR (Calibrated), m/s

3000 0.85: 1
RSVmax = 300 mm/s The near-surface SWV information so obtained can be merged with
the much deeper SWV profiles obtained by traditional techniques
including seismic refraction surveys. The merging of the two sets of
2000
SWV data can be facilitated by a model developed by the authors
and collaborator [5,20]. The authors advocate a more systematic
approach employing the latest non-invasive techniques for model-
1000 ing the bedrock SWV properties in selected built-up areas and have
the information well documented for access by design engineers as
opposed to having these done on a site-by-site (per project) basis.
0 A unique (equivalent) VR value could then be obtained from the joint
0 1000 2000 3000 4000
profile of both the rock crusts and the overlying soil sediments using
VR (Estimated from RPE ), m/s
the proposed RPE Method. Standard soil dynamic analysis could
then be applied based on modeling the rock crusts as homogeneous
4000
NF elastic half-space.
1.15: 1 1:1
FF The complexity of the series of factors controlling soil reso-
nance behavior, and their measurement, as described above is
VR (Calibrated), m/s

3000 0.85: 1 clearly in contrast to the much simplified site amplification


provisions in contemporary codes of practices. For example, the
site classification schemes of the International Building Code [21],
2000 Canadian Code [22] and Eurocode 8 [23] are simply based on the
SWV in the upper 30 m of the soil layers. The playing down of the
significance of site period (or soil depth) in those provisions is
1000 defendable as amplification associated with the occurrence of site
resonance could be muted considerably by damping in both the
soil and the structure in severe conditions of ground shaking.
0 Furthermore, the soil–rock interface in some deep Californian sites
0 1000 2000 3000 4000
(which are underlain by young sedimentary rock) is hard to define
VR (Estimated from RPE ), m/s and measure.
Fig. 11. Bedrock shear rigidity estimated by the proposed RPE Method compared
Parameterizing soil depth is appropriate in geological condi-
with the value calibrated from one-dimensional shear wave analyses for different tions where soft soil sediments are underlain by a much harder
(a) soil types (PI), (b) shaking levels (RSVmax) and (c) earthquake events (NF volcanic (crystalline) rock or (older) sedimentary mudstones
and FF). [24–26]. The generally less ductile structures in regions of low
and moderate seismicity would also warrant more attention
As demonstrated in the paper, the value of VR is not unique to being paid to the risk of occurrence of resonance behavior
the bedrock material as it is also dependent on the flexibility of given the limited energy dissipation (damping) characteristics
the overlying soil sediments as well as the SWV gradient in rock. of the structure. Standards provisions in the (volcanic) country
Thus, the effective depth of the notional boundary can be tens or of New Zealand [27] and the (low-to-moderate seismicity) coun-
hundreds of meters below the physical soil–rock interface. Con- try of Australia [28] have both incorporated site period as a site
sequently, the logging of drillholes forming part of a normal site classification criterion. However, the SWV profile of the bedrock
investigation might not provide adequate information for the has never been recognized as a controlling parameter in any of
accurate determination of the boundary conditions for model- the considered codes of practices.
ing purposes. Inexpensive non-invasive monitoring techniques The above commentary is not intended to point at the inade-
employing geophones (e.g., SPAC method [19]) have been devel- quacies of contemporary code provisions given that factors, which
oped to measure SWV in both the soil sediments and the bedrock have not been parameterized, could be implicit in the empirically
medium down to a depth of some 500 m or so from the surface. derived site coefficients. What needs to be emphasized herein is the
98 H.-H. Tsang et al. / Soil Dynamics and Earthquake Engineering 34 (2012) 89–98

potential danger in the adaptation of a set of provisions to condi- [2] Tsang HH, Chandler AM, Lam NTK. Estimating nonlinear site response by
tions that are not well represented by the empirical data (when single period approximation. Earthquake Engineering and Structural Dynamics
2006;35(9):1053–76.
the underlying controlling phenomena and their associated para- [3] Tsang HH, Chandler AM, Lam NTK. Simple models for estimating period-shift
meters are not well understood). There is a widely held notion that and damping in soil. Earthquake Engineering and Structural Dynamics
site amplification factors and the associated soil classification are 2006;35(15):1925–47.
[4] Boore DM, Joyner WB. Site amplification for generic rock sites. Bulletin of the
generic. In reality, provisions for site amplification may require
Seismological Society of America 1997;87(2):327–41.
modifications to account for regional differences that are associated [5] Chandler AM, Lam NTK, Tsang HH. Shear wave velocity modelling in crustal
with the bedrock geology, the level of seismic activity and the type rock for seismic hazard analysis. Soil Dynamics and Earthquake Engineering
of structures found on the soil surface. This paper is intended to be 2005;25(2):167–85.
[6] Faust LY. Seismic velocity as a function of depth and geologic time.
used as reference by code drafters so that code provisions for the Geophysics 1951;16:192–206.
future are crafted to fully take into account local conditions. This [7] Chandler AM, Lam NTK, Tsang HH. Near-surface attenuation modelling based
paper should also be of value to engineers who are engaged in site- on rock shear-wave velocity profile. Soil Dynamics and Earthquake Engineer-
ing 2006;26(11):1004–14.
specific seismic evaluation studies, which might involve the use of
[8] Tsang HH, Sheikh MN, Lam NTK, Chandler AM, Lo SH. Regional differences in
programs such as SHAKE (in which the SWV value of the bedrock as attenuation modelling for Eastern China. Journal of Asian Earth Sciences
a half-space needs to be specified for input into the program). 2010;39(5):441–59.
[9] Peltier WR. Ice age paleotopography. Science 1994;265:195–201.
[10] Bardet JP, Tobita T. NERA: a computer program for nonlinear earthquake site
response analyses of layered soil deposits. Los Angeles, CA: Department of
6. Summary and conclusions
Civil Engineering, University of Southern California; 2001.
[11] Lam N, Wilson J, Hutchinson G. Generation of synthetic accelerograms using
1. The Peak Spectral Ratio (PSR) as a parameter for defining soil seismological modelling: a review. Journal of Earthquake Engineering 2000;
amplification of the seismically induced displacement demand 4(3):321–54.
[12] Henderson P, Heidebrecht AC, Naumoski N, Pappin JW. Site response effects
was first introduced in the paper. The significance of the shear for structures located on sand sites. Canadian Geotechnical Journal 1990;
wave velocity (SWV) of the bedrock (VR) in controlling the 27(3):342–54.
value of PSR in conditions of resonance has been demon- [13] Heidebrecht AC, Henderson P, Naumoski N, Pappin JW. Seismic response and
design for structures located on soft clay sites. Canadian Geotechnical Journal
strated. The concept of the well-known Quarter-Wavelength
1990;27(3):330–41.
(QWL) Method was then briefly reviewed. [14] Lam NTK, Wilson JL. Estimation of the site natural period from borehole
2. Recognizing the inhomogeneous nature of most stratified records. Australian Journal of Structural Engineering 1999;SE1(3):179–99.
bedrock, the Resonant Period Equivalence (RPE) Method was [15] Worthington MH, King MS, Marsden JR. Determining the damping factor of
sedimentary rocks required for seismically designed structures. International
introduced for the determination of the effective depth (DR) Journal of Rock Mechanics and Mining Science 2001;38(6):801–6.
into rock and the associated equivalent SWV value (VR). The VR [16] Chandler AM, Lam NTK, Tsang HH, Sheikh MN. Estimation of near-surface
parameter enables a complex, and inhomogeneous, bedrock attenuation in bedrock for analysis of intraplate seismic hazard. Journal of
Seismology and Earthquake Engineering 2005;7(3):159–73.
medium to be modeled as homogeneous elastic half-space in [17] Nishi K, Ishiguro T, Kudo K. Dynamic properties of weathered sedimentary
site response analysis. soft rock, soils and foundations. Japanese Geotechnical Society 1989;20:
3. Estimates of the VR values obtained from the RPE Method have 67–82.
been evaluated thoroughly by comparison with results obtained [18] Chandler AM, Lam NTK, Sheikh MN. Response spectrum predictions for
potential near-field and far-field earthquakes affecting Hong Kong: soil sites.
from a parametric study based on the dynamic analyses of Soil Dynamics and Earthquake Engineering 2002;22:419–40.
complete soil–rock models comprising of twenty soil SWV [19] Roberts J, Asten MW. Estimating the shear velocity profile of quaternary silts
profiles and five bedrock SWV profiles. Six sets of synthetic using microtremor array (SPAC) measurements. Exploration Geophysics
2005;36:34–40.
accelerograms were employed as input bedrock motions to [20] Chandler AM, Lam NTK, Tsang HH. Regional and local factors in attenuation
represent different shaking levels as well as earthquake scenarios modelling: Hong Kong case study. Journal of Asian Earth Sciences 2006;27(6):
of varying magnitudes and epicentral distances. The VR values so 892–906.
obtained from the complete soil–rock models (by calibration [21] IBC-2006. International Building Code (IBC). USA: International Code Council;
2006.
analyses) have been shown to be in excellent agreement with [22] NBCC: 2005. National building code of Canada 2005. Ottawa, Canada:
results obtained independently from the proposed RPE Method. National Research Council of Canada (NRC); 2005.
4. Practical issues associated with the measurement and modeling [23] EN1998-1:2003. Eurocode 8: design of structures for earthquake resistance
part 2: general rules, seismic actions and rules for buildings, prEN 1998-1:2003.
of SWV profiles in stratified bedrock have been discussed. European Committee for Standardization; 2003.
Implications of the research findings to codifications and engi- [24] Dobry R, Oweis I, Urzura A. simplified procedures for estimating fundamental
neering applications for the future have also been addressed in period of a soil profile. Bulletin of the Seismological Society of America
the discussion section. 1976;66(4):1293–321.
[25] Rodrı́guez-Marek A, Bray JD, Abrahamson NA. An empirical geotechnical
seismic site response procedure. Earthquake Spectra 2001;17(1):65–87.
[26] Kim DS, Yoon JK. Development of new site classification system for the
regions of shallow bedrock in Korea. Journal of Earthquake Engineering
References
2006;10(3):331–58.
[27] NZS 1170.5:2004. Structural design actions part 5: earthquake actions in
[1] Schnabel PB, Lysmer J, Seed HB. A computer program for earthquake response New Zealand. Standards New Zealand; 2004.
analysis of horizontally layered sites. Earthquake Engineering Research [28] AS 1170.4:2007. Structural design actions part 4: earthquake actions in
Center Report: EERC 72-12. USA: University of California at Berkeley; 1972. Australia. Standards Australia; 2007.

You might also like