• Embed Doc
  • Copy Link
  • Readcast
  • Collections
  • CommentGo Back
Download
 
Estimation of seismic spectral accelerationin Peninsular India
S T G Raghu Kanth
1
and
R N Iyengar
21
Department of Civil Engineering, Indian Institute of Technology, Guwahati 781 039, India.e-mail: rk@iitg.ernet.in 
2
Department of Civil Engineering, Indian Institute of Science, Bangalore 560 012, India.e-mail: rni@civil.iisc.ernet.in 
Peninsular India (PI), which lies south of 24
◦
N latitude, has experienced several devastatingearthquakes in the past. However, very few strong motion records are available for developingattenuation relations for ground acceleration, required by engineers to arrive at rational designresponse spectra for construction sites and cities in PI. Based on a well-known seismological model,the present paper statistically simulates ground motion in PI to arrive at an empirical relationfor estimating 5% damped response spectra, as a function of magnitude and source to site dis-tance, covering bedrock and soil conditions. The standard error in the proposed relationship isreported as a function of the frequency, for further use of the results in probabilistic seismic hazardanalysis.
1. Introduction
The importance of estimating seismic hazards inPeninsular India (PI), which is an intra-plateregion, needs no special emphasis. The frequentoccurrence of devastating earthquakes in this partof India has been a reminder that engineers haveto use seismological approaches to estimate regionspecific design ground motion, instead of relyingon rules of thumb and
ad hoc
seismic zones. How-ever, analytical source mechanism models are notsimple enough to be directly applicable in engineer-ing problems. There have been attempts to developsemi-empirical approaches, based on the availabledatabase, that can be projected to the future in astatistical sense. Popularly, ground motion and theconsequent hazards are described in terms of peakground acceleration (PGA). However, it is wellrecognized that PGA does not uniquely influencedamage in man-made structures. Hence, engineersprefer the response spectrum as a better descrip-tor of seismic hazard. This is a frequency domainrepresentation of the ground motion, having theadditional advantage of providing the design engi-neer with an insight into how structures madeof different materials behave under a postulatedearthquake event. The response spectrum is alsodirectly applicable in structural response analy-sis. In engineering analysis and design, the need isto know the ground motion due to all causativesources in a region of about 300km radius arounda given site. However in India, engineers have beenusing a standard spectral shape as recommendedby the code IS-1893 (2002) all over the country,modified only by a zone factor as a proxy to peakground acceleration (PGA). Such an approachneither recognizes the seismo-tectonic details of the region nor accounts for the risk associatedwith the standard response spectrum. Hence, theplanned design life of structures cannot be ratio-nally regulated by the existing earthquake haz-ard. Clearly, underestimation of hazard leads toquestionable safety margins, whereas overestima-tion makes the projects uneconomical. Thus, socialgoals suffer in either case. It is in this context thatprobabilistic seismic hazard analysis (PSHA) has
Keywords.
Attenuation; response spectra; Peninsular India; site coefficients.
J. Earth Syst. Sci.
116
, No. 3, June 2007, pp. 199–214
©
Printed in India.
199
 
200
S T G Raghu Kanth and R N Iyengar 
Figure 1. Available instrumental database in Peninsular India.
become indispensable when addressing engineer-ing safety issues in terms of quantified risk levels.The expected site PGA and the response spec-trum with a specified return period or risk canbe derived from PSHA. Such a response spectrum,which has the same return period at all frequen-cies, is known as a uniform hazard response spec-trum (UHRS). In order to obtain a UHRS, one hasto develop regional ground motion equations relat-ing spectral amplitudes to magnitude and distance.Due to lack of strong motion data, no equation forestimating ground motion was available for use inPeninsular India (PI). With this in view, Iyengarand Raghukanth (2004) investigated attenuationof PGA in PI through the stochastic seismologi-cal model of Boore (1983). Previously, for centraland eastern United States (CEUS), where strongmotion data are scarce, Boore and Atkinson (1987),Hwang and Huo (1997) have used seismologicalmodels to predict characteristics of ground motion.In the present study, this approach is applied toderive empirical equations for 5% damped responsespectra, corresponding to bedrock conditions inPI. The results of the derived equation are com-pared with instrumental data from the Koynaearthquake (
M 
w
= 6
.
5) of 11 December 1967 andthe Bhuj earthquake (
M 
w
= 7
.
7) of 26 January2001. Correction factors are also found for variousother sites defined in terms of the average shearwave velocity in the top 30 meters (
V 
30
) of thesoil. This new empirical relation will be useful inprescribing design response spectra for structuresin PI.
2. Seismological model
A critical review of the available strong motiondata in PI has been presented previously byIyengar and Raghukanth (2004). Figure 1 presentsthe available data in PI as a function of magni-tude and epicentral distance. This brings out theexisting deficiency in the database of PI from theengineering point of view. Ideally, multiple strongmotion accelerogram (SMA) data from the sameevent should be available for distances varying fromzero to 300km. In addition, magnitude values rang-ing from 4 to 8 should be covered at reasonableincrements. PI is similar to many other stable con-tinental regions (SCR) of the world where data arescarce and not representative of the existing haz-ards. Attenuation equations in such regions have tobe based on simulated ground motions instead of past recordings. The theory and application of sto-chastic seismological models for estimating groundmotion has been discussed by Boore (1983, 2003).Briefly, the Fourier amplitude spectrum of groundacceleration
A
(
f 
) is expressed as
A
(
f 
) =
CS 
(
f 
)
D
(
f 
)
P 
(
f 
)
F 
(
f 
)
.
(1)Here,
C 
is a scaling factor,
S 
(
f 
) is the sourcespectral function,
D
(
f 
) is the diminution function
 
Seismic spectral acceleration in Peninsular India 
201
Figure 2. Three sub-regions of Peninsular India with knownQ-factor.
characterizing the quality of the region,
P 
(
f 
) is afilter to shape acceleration amplitudes beyond ahigh cut-off frequency
f 
m,
and
F 
(
f 
) is the siteamplification function. In the present study, for thesource, the single corner frequency model
S 
(
f 
) = (2
πf 
)
2
M 
0
1 + (
f/f 
c
)
2
(2)of Brune (1970) is used, where the corner frequency
f 
c
, the seismic moment
M 
0
and the stress drop Δ
σ
are related through
f 
c
= 4
.
9
×
10
6
V 
s
Δ
σM 
0
13
.
(3)Here the shear wave velocity
V 
s
in the sourceregion, corresponding to bedrock conditions, istaken as 3.6km/s. The diminution function
D
(
f 
)is defined as
D
(
f 
) =
G
exp
−
πfrV 
s
Q
(
f 
)
(4)in which
G
refers to the geometric attenuation andthe remaining term denotes anelastic attenuation.
r
is the hypocentral distance and
Q
is the qualityfactor of the region. The high-cut filter in the seis-mological model is given by
P 
(
f,f 
m
) =
1 +
f f 
m
8
−
12
.
(5)Here,
f 
m
controls the high frequency fall-off of thespectrum. The scaling factor
C 
is
C 
=
R
θφ
√
24
πρV 
3
s
,
(6)where
R
θφ
is the radiation coefficient averagedover an appropriate range of azimuths and take-off angles and
ρ
is the density of the crust atthe focal depth. The coefficient
√
2 in the aboveequation arises as the product of the free surfaceamplification and partitioning of energy in orthog-onal directions. Following the work of Singh
et al 
(1999), the geometrical attenuation term
G
, for theIndian shield region, is taken to be equal to 1
/r
for
r <
100km and equal to 1
/
(10
√
r
) for
r >
100km.PI can be broadly divided into three regions, asfar as the quality factor
Q
is concerned (Iyengarand Raghukanth 2004). Mandal and Rastogi (1998)have found
Q
for the Koyna–Warna (K–W) regionto be 169
f 
0
.
77
. For the southern India (SI) region
Q
is 460
f 
0
.
83
and for the Western–Central (W–C) region
Q
is reported to be 508
f 
0
.
48
(Singh
et al 
1999). These three regions, which make upPI are shown in figure 2. The seismological modelis implemented in the time domain in each regionthrough computer simulation, consisting of threesteps (Boore 1983, 2003). First, a Gaussian station-ary white noise sample of length equal to the strongmotion duration (Boore and Atkinson 1987),
T 
=1
f 
c
+ 0
.
05
r
(7)is simulated. Second, this sample is multipliedby the modulating function of Saragoni and Hart(1974) to introduce non-stationarity and thenFourier transformed into the frequency domain.This Fourier spectrum is normalized by its root-mean-square value and multiplied by the termsof equation (1), derived from the seismologicalmodel. Third, the resulting function is transformedback into the time domain, to get a sample of acceleration time history. For calculating spectral
Table 1.
Ranges of epicentral distance
.Epicentral Number of Moment distance distancemagnitude (
r
epi
km) samples4 1–300 204.5 1–300 205.0 5–300 145.5 15–300 106.0 25–300 96.5 35–300 87.0 40–300 77.5 45–300 78.0 60–300 6
of 00

Leave a Comment

You must be to leave a comment.
Submit
Characters: ...
You must be to leave a comment.
Submit
Characters: ...