Bulletin of the Seismological Society of America, Vol. 97, No. 1B, pp. 318–330, February 2007, doi: 10.

1785/0120050127

Probabilistic Seismic-Hazard Estimation for Peninsular India

by Kishor Jaiswal and Ravi Sinha

Abstract Peninsular India (10.0⬚ N–28.0⬚ N; 68.0⬚ E–90.0⬚ E) is one of the oldest
and seismically most stable landmasses of the Indian plate. Recent seismic history,
however, shows that more than five damaging earthquakes with magnitudes greater
than Mw 6.0 have occurred in this region, highlighting the importance of seismic-
hazard assessment for the region. This article estimates the probabilistic seismic
hazard associated with peninsular India with a zoneless approach incorporating the
observed seismic activity and known geological characteristics of the region. The
seismicity parameters for hazard assessment have been estimated incorporating com-
pleteness criteria for various spans of the catalog data. The spatial and temporal
variations of seismic activity have been modeled using different source models. Seis-
mic source zones for the region have been defined on the basis of large-scale geo-
logical features, which are used for assigning the maximum possible earthquake
potential. Due to the poorly known attenuation characteristics of the study region,
three appropriate attenuation models have been used for the estimation of ground-
motion parameters. Hazard maps for peninsular India have been developed using a
convolution scheme based on weighting and incorporating various uncertainties in-
volved while modeling different parameters. The comparison of the probabilistic
seismic-hazard map developed herein with the hazard map specified in the Indian
Standards shows that the design parameters in the Indian Standards may significantly
underestimate the seismic hazard in some regions of peninsular India.

Introduction
An important prerequisite for mitigation of the devas-
tating effects of earthquakes in a region is the accurate as-
sessment of seismic hazard associated with the region so that
the potential for future damaging earthquakes can be esti-
mated. In India, the seismic regionalization studies by Tan-
don (1956) and Krishna (1959) may be considered as the
earliest efforts in the demarcation of areas of potential se-
vere, moderate, and light damage in the Indian subcontinent.
Subsequent studies include intensity-based mapping by
Guha (1962) and Gubin (1968). The problems associated
with intensity-based zoning were overcome by using prob-
abilistic hazard studies based on available data by several
researchers including Basu and Nigam (1977), Kaila and
Rao (1979), and Khatri et al. (1984). Because of the scarcity
of earthquake data in the shield region, hazard estimates in
the peninsular area (10.0⬚ N–28.0⬚ N; 68.0⬚ E–90.0⬚ E) were
based on either subjective estimates of intensity from past
earthquake data or from the known geological and seismo-
tectonic information without including the associated uncer-
tainty in the estimates (Krishna, 1992). Recent efforts in-
clude a probabilistic map developed by Bhatia et al. (1999)
under the GSHAP program and a deterministic seismic-
hazard map of India in terms of maximum displacement,
maximum velocity, and design ground acceleration pro-
posed by Parvez et al. (2003) on the basis of a compilation
of synthetic seismograms. Das et al. (2006) carried out
probabilistic seismic-hazard analysis of Northeast India
based on uniform hazard spectra for pseudospectral velocity
and a newly developed spectral attenuation model for the
region.
The seismic zoning maps prepared by the Bureau of
Indian Standards (BIS) for earthquake-resistant design spec-
ification during all its earlier phases (Krishna, 1992) and the
latest revision in 2002 (IS 1893, 2002) have been primarily
based on historical information represented by an epicentral
map superimposed by the estimated maximum Modified
Mercalli Intensities (MMIs) for large historical earthquakes,
with some upgradation for shield zones to higher seismic
zone based on recent seismogenic activity. The knowledge
of seismogenic features of the peninsular shield is limited
and has been identified only after the occurrence of large and
damaging earthquakes. The earthquakes of Koyna (1967;
Mw 6.3), Latur (1993; Mw 6.1), Jabalpur (1997; Mw 5.8), and
Bhuj (2001; Mw 7.7) have demonstrated that an urgent need
exists to make comprehensive efforts to delineate active
source zones in peninsular India and accurately define the
boundaries of high-seismogenic sources based on seismolog-
ical characteristics (Sinha et al., 2001).

318
Probabilistic Seismic-Hazard Estimation for Peninsular India
The present investigation uses the zoneless approach
proposed by Frankel (1995) for seismic-hazard estimation.
However, the spatial variation of Mmax and b are assigned
based on different identified broad regions in peninsular In-
dia. Due to unavailability of well-established attenuation re-
lations for the region, three different attenuation relation-
ships have been considered. A recently proposed attenuation
relationship based on a few recent earthquake data in India
by Iyengar and Raghukanth (2004) has been included.
Past studies have indicated that both central and eastern
United States (CEUS) and peninsular India regions share
similar features in terms of seismotectonic characteristics, as
discussed by Schweig et al. (2003), Cramer and Kumar
(2003), and Mooney et al. (2005). Studies also show that
they share similarities in the crustal Q-parameters (Bodin et
al., 2004; Singh et al., 2004). Bodin et al. (2004) studied
the aftershock data of Bhuj earthquake and found that the
seismic-wave attenuation in Kuchchh crust is very low and
is comparable to the attenuation characteristics in CEUS. The
attenuation relationships proposed by Atkinson and Boore
(1995) and Toro et al. (1997) for the CEUS have also been
included as alternate models for the hazard estimation of
peninsular India.
Seismicity Characteristics
Geology and Seismic Zonation
The Indian peninsula is known to be one of the oldest
landmasses on earth and exhibits a geology that has not ex-
perienced any major deformation since the beginning of
the Phanerozoic eon. More than one-third of the region is
covered by the basaltic flows of the Deccan Trap, which
were extruded in the Cretaceous-Eocene interval (about 60
to 135 million years ago). The rest of the peninsula mainly
consists of Precambrian crystalline rocks and sedimentary
formations of later eras (Valdiya, 1973; Naqvi et al., 1974).
Due to its complex structure with numerous associated faults
and fractures, peninsular India has been extensively studied
for earthquake phenomena associated with intraplate activity
(Rao and Murty, 1970; Chandra, 1977; Mandal et al., 2000).
The geological zonation of earthquake sources used in the
present study is primarily based on various evolutionary
units and the associated tectonic features in the peninsular
shield of India (Fig. 1).
The source zonation boundaries for the present inves-
tigation have been based on Seeber et al. (1999). Based on
tectonic features and the observed seismic activity in the
peninsular shield, Seeber et al. (1999) has proposed nine
broad seismic zones (Fig. 2), considering the characteriza-
tion of stable continental regions as either cratons or paleo-
rifts. The seismogenic characteristics of paleorifts suggest
that these regions contain large faults and have experienced
extensional deformation in their most active phase. In pen-
insular India, it includes all passive continental margins and
inactive grabens such as Cambay, Godavari, and Mahanadi
319
grabens and the active Narmada lineament (Fig. 1). The
stable southern and northern cratons have been assigned
with separate zones and a similar procedure has been
adopted for characterizing the western and eastern passive
margins.
Seismic Catalog Database and Model
Parameter Estimation
A complete earthquake catalog with a uniform magni-
tude scale for expressing the size of past earthquakes is a
prerequisite for a reliable parameterization of the magnitude
distribution used in a hazard analysis. A working catalog has
been prepared for this study based on the available infor-
mation from various sources (Jaiswal and Sinha, 2004). It
includes the published catalogs (Mw ⱖ3.0), after equivalent
moment magnitude conversion, up to 1997. Most recent
events up to 2002 are included in the catalog from the Pre-
liminary Determination of Epicenter (PDE) data (National
Earthquake Information Center [NEIC], 2003). The PDE
events are in terms of teleseismic body-wave magnitude (mb)
and hence need to be converted into equivalent Mw using
suitable empirical relationships. The relationships proposed
by Atkinson and Boore (1995) and Johnston (1996) can be
used to estimate Mw by first calculating the seismic moment
associated with the respective earthquakes. The former re-
lationship was used for preparing the earthquake catalog for
eastern North America, whereas the relationship proposed
by Johnston (1996) is based on worldwide data of stable
continental earthquakes, including data from Indian earth-
quakes. We have used the relationship proposed by Johnston
(1996) as given in equation (1) for estimating Mw. For some
of the larger events, reported Mw estimates have been used
in the preparation of the catalog.
log(M0) ⳱ 18.28 Ⳮ 0.679mb Ⳮ 0.0077mb2 . (1)
A declustering algorithm has been used for removing
dependent events of the entire catalog. The commonly used
declustering algorithms, such as Gardner and Knopoff
(1974) or Reasenburg (1985), were developed for moderate
to large events, which may be applicable for active tectonic
conditions but have not been found suitable for the Indian
peninsula because of the long window for foreshocks and
aftershocks considered in these models. We have therefore
considered a criterion based on uniform time (⬎90 days)
and space (radius ⬎50 km) windows between successive
events due to the absence of a well-defined approach appli-
cable for the Indian peninsula. The same criterion was ap-
plied for removal of dependent events by Seeber et al.
(1999). However for some of the large events (e.g., 2001
Bhuj earthquake), the aftershock activity continued for a
much longer period, hence such events have been removed
manually from the catalog to conform to uniform assump-
tions in the rate modeling. The final catalog, which includes
320
Figure 1. Major evolutionary units and associated tectonic features in the peninsular
shield of India (Gupta et al., 1995).
of 10 events of Mw ⱖ 6.0, 21 events of Mw ⱖ 5.5, and 184
events of Mw ⱖ 4.0, has been used for further analysis.
The regional earthquake recurrence activity is com-
monly expressed in terms of the Gutenberg–Richter mag-
nitude frequency relationship represented by
log10 N (ⱖM per year) ⳱ a ⳮ b M, (2)
where N is the number of events per year with magnitude
greater than or equal to M. The 10a-value is rate per unit
area per year of M earthquakes and the b-value is the decay
rate (slope) of log-linear fit that represents the relative like-
lihood of larger and smaller earthquakes. The current inves-
tigation uses Weichert’s method (1980), which gives greater
weight to the earthquake magnitude ranges with large-
number data values, is relatively insensitive to errors in mag-
nitude ranges with very few events, and is therefore particu-
larly relevant to peninsular India due to the limited data of
large earthquakes in the catalog.
For estimating seismicity parameters, the catalog data
from 1842 to 2002 have been used. The assumption of a
single threshold level of Mw 4.5 based on Stepp’s procedure
(1973) of rate uniformity gives a b-value of 0.84. The
b-value obtained using this criterion is quite similar to the
value of b ⳱ 0.85 obtained by Rao and Rao (1984) using
K. Jaiswal and R. Sinha
historical earthquake data for 170 years. To achieve the uni-
formity of activity rate of each magnitude class, the catalog
data have been divided based on the cumulative number of
events of different magnitude groups. This has led to the
selection of different completeness criteria for different
length of catalog based on observed rate uniformity in dif-
ferent time intervals; the results are shown in Table 1. Con-
sidering this completeness level, the b-value has been found
to be equal to 0.92 for the entire catalog data as shown in
Figure 3. This value is considered more representative of the
region’s seismicity and hence has been used as the basis for
rate adjustment in hazard calculations. The uncertainties as-
sociated with the fit to the Gutenberg–Richter (GR) param-
eters (a and b) are very small and hence have not been ex-
plicitly included in the present study.
Influence of Reservoir-Induced Seismicity
The Koyna earthquake of 10 December 1967 was one
of the most distinct and illustrative examples of induced seis-
micity associated with impounding of reservoirs in penin-
sular India (Gupta, 1992). Rastogi (1992) observed that more
than 50% of events that occurred in peninsular India in the
1980s are in the vicinity of major reservoirs. Seeber et al.
(1999) noted a threefold increase in seismicity for peninsular
Probabilistic Seismic-Hazard Estimation for Peninsular India 321

Figure 2. Map showing epicentral location of past earthquake data up to December

2002, along with geological zonation used in the present study for seismic-hazard
analysis of peninsular India. 1, Runn of Kuchchh (ROK); 2, Northern Craton (NC);
3, Narmada Lineament (NL); 4, Mahanadi Graben (MG); 5, Eastern Craton (EC);
6, Godavari Graben (GG); 7, Southern Craton (SC); 8, Western Passive Margin (WPM);
and 9, Eastern Passive Margin (EPM).

Table 1
Distribution of Earthquake Data for Each Decade in Different
Magnitude Ranges
Magnitude Interval Time Interval
(Mw) Completeness Period (in years)

[4.0–4.5] 1961–2002 42
[4.5–5.0] 1951–2002 52
[5.0–5.5] 1901–2002 102
[5.5–6.0] 1842–2002 160
[6.0–6.5] 1842–2002 160

India after 1960. These and other investigations raised the

issue of the overall increase in seismicity of peninsular India
due to construction of large dams and its effect on seismic
hazard, even though some recent research indicates that the
increase of seismicity in the vicinity of Koyna dam is not Figure 3. Estimation of seismicity parameters
based on completeness criteria for 1842–2002.
solely triggered by the reservoir (Gahalaut et al., 2004).
According to the analysis of catalog data the increase
in seismicity results in a b-value of 0.90 after 1960 compared model. This increase in seismicity in peninsular India has
with 0.77 for the pre-1960 data (Fig. 4). The b-value change been critically evaluated for the current hazard estimation
from 0.77 to 0.90 raises the possibility that occurrence of studies.
numerous small reservoir-induced earthquakes may increase For modeling the spatial variation of the b-value change,
the probability of larger earthquakes according to the GR the decay rates associated with the 20 known reservoirs in
322
Figure 4. Estimation of seismicity parameters for
different spans of catalog data.
the region (Gupta, 1992; and Seeber et al., 1999) have been
evaluated considering all the earthquakes that have occurred
in the 100 km ⳯ 100 km area surrounding these reservoirs.
The area of 100 km ⳯ 100 km approximately corresponds
to their catchment area and represents a conservative esti-
mate of reservoir-induced seismicity (Seeber et al., 1999).
It has been assumed that the earthquakes that are located
within this near-reservoir area are associated with the likely
effects of reservoir impoundment as a first-order assumption.
The events associated with near-reservoir areas result in a
b-value of 0.95 which is slightly higher than the b-value of
0.91 observed for the rest of the peninsular India data, ex-
cluding near-reservoir earthquakes.
The overall increase in b-value to 0.95 compared with
the pre-1960 decay rate seems to be quite similar to the rate
associated with the overall seismic activity observed for the
entire India Peninsula since 1960. This is because most of
the near-reservoir earthquakes are from a few hot spots such
as the Koyna and Bhatsa regions and not associated with the
other reservoirs in the region. Also note that the overall in-
crease in the seismicity after the 1960s is not just due to
reservoirs but also to the numerous moderate earthquakes
that occurred in the Runn of Kuchchh region, southern cra-
ton, or eastern and western passive margins (Fig. 2). It is
therefore inappropriate to assume that this threefold increase
in seismicity in peninsular India is entirely due to the in-
crease in seismic activity in the vicinity of large reservoirs.
Also note that most of the earthquakes (between Mw 3
and 4) near the Koyna region have been identified only after
installation of a better instrumentation network in the area
which took place in late 1960s, whereas most such earlier
earthquakes may have remained undetected in the time span
of catalog data (Mohan et al., 1981). To evaluate the tem-
poral variation of b-value in the rest of the Peninsula after
1960, the earthquakes associated with the 20 known reser-
voirs have been excluded from the main catalog, which gives
K. Jaiswal and R. Sinha
a higher value of b ⳱ 1.0. It indicates that small to moderate
events contribute a major share of the increase in b-value
after the 1960s. This also confirms that such an increase is
clearly experienced in large parts of peninsular India and is
not associated only with the reservoirs. The spatial variation
of b-values indicates that the decay pattern of the seismic
activity cannot be considered to be uniform all over penin-
sular India and also that the overall increase cannot be at-
tributed to reservoir-induced seismicity alone.
Spatial Characteristics of Seismicity Based
on Catalog Data
Even though the current catalog data do not cover a very
long period of earthquake history, the observed seismic ac-
tivity of small to moderate earthquakes (4 ⬍ Mw ⬍ 5.5)
clearly indicates such spatial variation in b-values. The cat-
alog data of peninsular India show that the data associated
with cratons have a higher b-value of 1.05 as compared with
rift zones such as eastern and western passive margins and
Runn of Kuchchh for which the estimated b-value is 0.85.
The smaller b-value in rift zones indicates that numerous
earthquakes are relatively larger, compared with the small
and moderate events, when compared with regions with
higher b-values. This is expected from the observation that
the peninsular shield has experienced large earthquakes at
relatively short intervals during the past many centuries. A
similar conclusion was reached by Rajendran (2000) based
on classification of recent damaging earthquakes in terms of
tectonic environment, recurrence period, maximum magni-
tude, and style of deformation. Unlike other stable continen-
tal regions (SCRs), such spatial variation in terms of decay
rates may be a peculiarity of the Indian shield and is also
ascertained based on geological and seismotectonic charac-
teristics of the region. The present investigation addresses this
issue by modeling the spatial variation in b-value estimates
in the hazard analysis using a logic tree approach. Cramer
(2001) and Cramer et al. (2002) give systematic application
of such information for incorporating various uncertainties in
terms of a logic-based approach of seismic-hazard analysis
and have been used in the current investigation.
Probabilistic Seismic-Hazard
Assessment Methodology
The zoneless seismic-hazard approach suggested by
Frankel (1995) has been used in the present investigation.
This approach is useful especially for regions where demar-
cation of distinct seismic-zone boundaries is difficult due to
a poor understanding of tectonic settings (e.g., SCRs such as
CEUS, central Asia, western Australia, and peninsular India).
This approach has also been used for development of seismic-
hazard maps of United States (Frankel et al., 1996, 2002).
When evaluating seismic hazard for peninsular India,
earthquakes that occurred in the region between 1842 and
2002 have been included. The entire peninsular region has
Probabilistic Seismic-Hazard Estimation for Peninsular India
been divided into grid cells of size 0.1⬚ ⳯ 0.1⬚, that is, ap-
proximately 11 km ⳯ 11 km area. The total number of earth-
quakes from catalog data greater than a certain cutoff value
are summed in each grid cell and then corrected in terms
of number of earthquakes per year, which represents the
a-value for that grid cell. The rate associated with certain
magnitude interval for that grid cell has been estimated using
the correction factor proposed by Herrmann (1977) and the
final corrected interval 10a-value for particular grid cells
have been spatially smoothed using a Gaussian smoothing
function. In the present investigation, the spatial variability
associated with b-values and the maximum magnitudes for
different tectonic regions of peninsular India have been in-
troduced in the analysis.
Source Model
The earthquake activity rate for each grid cell depends
on the completeness period obtained for each magnitude in-
terval of the catalog data. In peninsular India, the data com-
pleteness is very important because large areas, especially
large cratons, have rarely experienced moderate-level earth-
quakes, whereas the activity near rifting systems has been
very high. To understand the implication of such variation
of observed seismicity rates, three independent source mod-
els have been considered that are based on the completeness
of the catalog data for different magnitude ranges.
Data Model 1
In most of the stable continental regions of the world
such as the CEUS, Central Asia, and peninsular India, the
relative likelihood of moderate-level earthquakes can be in-
terpreted from the location characteristics of relatively
smaller-magnitude earthquakes. In peninsular India, it has
been observed that most of the earthquakes occur near the
continental margins and some of them are possibly induced.
It has been observed that the earthquakes in the mag-
nitude range of Mw 4.0 to 5.5 are closely associated with a
few reservoirs in the region, near the marginal areas of the
shield, and also some of the known active tectonic features
such as Runn of Kuchchh or Narmada-Son lineament. Be-
cause the region was poorly instrumented until the 1960s, it
is believed that several smaller earthquakes (Mw ⱕ4.0) may
not have been included in the catalog prior to 1960s. Data
Model 1 includes all the events greater than magnitude 4.0
since 1961 for estimating the seismicity rate. The correlation
distance of 50 km has been taken for smoothing of the
a-values to accommodate uncertainty associated with loca-
tion characteristics of individual earthquakes.
Data Model 2
In seismic-hazard analysis, it is important to evaluate
the seismicity rates associated with particular seismotectonic
features based on completeness of catalog data for different
323
magnitude intervals. For peninsular India, the records of
earthquakes of Mw ⱖ5.0 can be assumed to be complete even
for the preinstrumental period of catalog because earth-
quakes of this magnitude cause damage due to the prevalent
construction practice. The data for events greater than
Mw 5.0 between 1901 and 2002 has been found to be com-
plete and hence taken as the basis for hazard analysis in Data
Model 2. The higher correlation distance of 75 km is taken
for hazard analysis because many of the earthquakes are
from the preinstrumental era and are expected to have higher
location uncertainties.
Data Model 3
It is obviously essential to include the recurrence of
large earthquakes and these events have been better recorded
in the catalog for a longer time span. Such large earthquakes
have been found to recur frequently in most of peninsular
India. The rift zones of the peninsular shield have already
experienced one of the largest and most devastating earth-
quakes in the past few centuries. Note that the historical
earthquakes associated with the Runn of Kuchchh region
suggest that the 1819 earthquake of magnitude Mw 7.7 has
a relatively short recurrence interval of 182 years. However
in the current investigation, the catalog data are found to be
limited for estimation of recurrence of such large events.
Data Model 3 uses spatially smoothed a-values based
on Mw ⱖ6.0 since 1600. The historic rates of Mw 6.0 since
1600 are expected to underestimate the observed rate since
the available information is not complete for that period. A
scaling factor of 1.665 to the smoothed a-values has been
estimated which represents the corrected rates of Mw 6.0
based on completeness criteria. The use of this scaling factor
also allows for the correction of errors due to misinterpre-
tation of previous large events in terms of assigning intensity
or due to incomplete knowledge of recurrence characteristics
of such large events in the region. A larger correlation dis-
tance of 100 km has been used in Data Model 3. The recur-
rence of the large earthquakes is expected to cause intense
damage at large distances from their epicentral locations,
and hence it has been found necessary to model such events
separately.
While combining different data models, equal weight
has been assigned to Date Models 2 and 3, whereas Date
Model 1 is assigned with double weight. This is because
Data Model 1 is based on more precise instrumental seis-
micity data and hence assumed to be more representative of
current seismogenic activity in the region.
Convolution Technique for Seismogenesis Modeling
For seismic-hazard considerations, a convolution tech-
nique has been used in the present study to include the spatial
and temporal variation in the seismogenic activity associated
with peninsular India. These variations have been included
using a suitable weighting scheme based on relative impor-
324
tance of each of the models from known characteristics in
recent history. The uncertainty has been modeled by consid-
ering four independent cases as discussed next.
Case I: Uniform Seismicity
In Case I, a single b-value of 0.92 was obtained based
on completeness criteria (Table 1 and Fig. 3) has been as-
signed to the entire Indian peninsula. The uniform maximum
earthquake potential of Mmax 8.0 has been assigned for all
nine zones of peninsular India.
Case II: Geo-Based Seismicity
The catalog data, as discussed earlier, clearly indicates
a few broad but important temporal and spatial variation
characteristics, which are considered in this case. A uniform
b-value of 0.85 is assigned to all the rifting zones that in-
cludes Runn of Kuchchh, eastern and western passive mar-
gins and grabens, whereas a higher b-value of 1.0 is assigned
to all the cratonic regions of peninsular India. In this case,
the maximum magnitude potential in each geological zone
has been based on geological considerations proposed by
Seeber et al. (1999) following the criteria derived from the
worldwide study of stable continental region seismicity
(Frankel et al., 1996; Johnston, 1996). The maximum mag-
nitude has been assumed to be Mw 6.5 for cratons and Mw 7.5
for rifting zones; however, relatively higher value of Mw 8.0
has been taken for Cambay graben (Runn of Kuchchh zone)
based on its seismogenic characteristics and expectancy of
larger earthquake with a shorter recurrence cycle. Even
through the observed seismic activity has been very low dur-
ing the catalog period at most of the cratonic zones, the
maximum magnitude potential of Mw 6.5 has been consid-
ered as only a first order of approximation based on the data
in other SCRs and tectonic similarities. However, it can be
modified in future studies, especially in the light of the latest
version of U.S. Geological Survey (USGS) seismic-hazard
maps for CEUS (Frankel et al., 2002), the Mmax associated
with cratons has been increased to Mw 7.0.
Case III: Reservoir-Induced Seismicity
In Case III, the effect of reservoir-induced seismicity
(RIS) has been included based on observed seismicity rates
for some of the reservoirs. A uniform b-value of 0.91 has
been used for peninsular India excluding an area of 100 ⳯
100 km of 20 known reservoirs. A relatively higher b-value
of 0.95 has been estimated for the earthquakes that occurred
in the vicinity of these reservoirs. This b-value has been used
to incorporate the rate variation near the reservoirs. The ef-
fects of RIS are observed to be prominent at certain geo-
graphic locations of peninsular India, for example, Koyna
and Bhatsa dams located at Western India. Note that the
seismic data associated with these geographic locations have
not been lost or missed in other data models. Hence the
temporal variation in terms of activity rate (a-values) at these
K. Jaiswal and R. Sinha
locations has always been part of seismic-data analysis. For
example, all the small-magnitude events of Koyna region up
to Mw ⱖ4.0 since 1961 are included in Data Model 1. Sim-
ilarly, in Data Model 2, the events with Mw ⱖ5.0 since 1901
for the entire Indian peninsula have been included.
Case IV: Background Seismicity
Note that for large portions of peninsular India (the quiet
zones), historical catalog includes no information about pre-
vious earthquake activity in the region due to its relatively
short time span. It is also possible that many smaller-
magnitude earthquakes in these quiet zones may not have
been recorded because of the sparse instrumental network
even during the past several decades. However, recent earth-
quakes have clearly shown that parts of peninsular India that
were earlier assumed to be quiet have some seismic activity
and have potential of experiencing damaging earthquakes in
the future. In fact, it has been found that even cratons show
some local intraplate activity (e.g., Mw 6.3 Latur earthquake
of 1993). Hence it is more realistic to model such seismicity
in the form of background seismicity by assigning uniform
background seismicity rates for these regions. This proce-
dure has been used as one of the alternatives in the current
investigation to include such unknown possible future earth-
quakes by assigning background seismicity rate, which, in
the present case, is taken as seismicity rates prior to 1960
acting uniformly over the entire Indian peninsula. This was
achieved by using a single b-value obtained for pre-1960
data, whereas the a-values of pre-1960s data were uniformly
distributed for all of peninsular India and included in the
hazard analysis.
The four cases discussed previously are expected to
broadly cover overall seismicity-rate variation in the differ-
ent regions of peninsular India. Note that the weighting
scheme can be further fine-tuned in future studies based
on additional earthquake data or regional seismotectonic in-
formation.
Hazard Assessment
The probabilistic estimation of seismic hazard at any site
consists of the estimation of the annual exceedance rate of
any ground-motion parameter above a certain specified level.
The mean peak horizontal ground acceleration (PGA) has been
used for specifying exceedance of ground motion and the re-
sults are portrayed in terms of hazard maps at the desired
exceedance level for peninsular India. The exceedance level
for mean hazard estimates has been taken as 10% probability
of occurrence in 50 years, giving an average return period of
approximately 475 years or a recurrence rate of 0.0021 per
year. The earthquake PGA for this hazard level (often known
as design-basis earthquake) is generally considered for esti-
mation of seismic forces in earthquake-resistant design prac-
tices. However, recent revisions in some buildings codes, such
as the U.S. code, have moved away from 10% probability of
Probabilistic Seismic-Hazard Estimation for Peninsular India 325

exceedance in 50 years to 2% probability of exceedance in

50 years (or a return period of 2475 years). This decrease in
the probability of exceedance, and the corresponding increase
in PGA, is from the consideration of a 475-year return period
being too low for stable continental regions. However, in
Indian codes such reduction in the probability of exceedance
for design-basis earthquake is not being considered.
For hazard assessment, all three attenuation relations as
discussed earlier have been used with different weights. The
Iyengar and Raghukanth (2004) relationship for PGA is
based on recent large earthquakes in peninsular India. It is
considered to be more representative of the attenuation char-
acteristics of the region and hence given twice the weight as
compared with the other two relationships, that is, Atkinson
and Boore (1995) and Toro et al. (1997). The use of these Figure 5. Distribution of peak ground acceleration
(PGA) for hard-rock conditions for different attenua-
three attenuation relationships with different weights in the tion relationships.
logic tree incorporates the uncertainties involved in the
ground-motion modeling. Both CEUS attenuation relation-
ships have been available for hard-rock site conditions (i.e., PGA for uniform hard-rock site conditions, the comparison
shear-wave velocity in the range of 2.8 km/sec), whereas the of these three relationships at different epicentral distances
relationship proposed by Iyengar and Raghukanth (2004) is is illustrated in Figure 5. The hypocentral distance has been
applicable for the hard-rock category in the range of 1.5 to used in the calculation of distance R used for Iyengar and
3.6 km/sec. Raghukanth (2004) and Atkinson and Boore (1995) relation-
All three attenuation relationships can be assumed to be ships for hypocentral depth of 10 km, whereas the Joyner–
appropriate (as a preliminary estimate) for estimating the Boore distance (which is the closest horizontal distance to the

Figure 6. Block representation of combination rules for convolution scheme.

326
rupture surface if this rupture surface is projected to the
ground surface) has been used for the Toro et al. (1997)
attenuation relationship.
The estimated a-values in each of the data models rep-
resent the total number of earthquakes in each grid greater
than a certain lower-bound value based on completeness cri-
teria. For Data Model 1, the estimated total number of earth-
quakes above Mw 4.0 per annum is 3.81. The lower limit of
magnitude greater than Mw 4.0 for seismicity-rate estimation
of each individual grid indicates contribution from moderate-
level earthquakes in terms of likely hazards in the regions such
as Southern India where there have been no large earthquakes
during the catalog period.
For Data Model 2, the estimated total number of earth-
quake above Mw 5.0 per annum is 0.353. This activity rate
indicates regions of potential large earthquakes in peninsular
India and hence is taken as the basis for rate adjustment
while combining the different models in hazard analysis.
In Data Model 3, the observed number of earthquakes
above Mw 6.0 is less than estimated rates based on the criteria
discussed earlier; hence, a correction factor of 1.665 is ap-
plied to have a corrected rate of 0.0621. The estimated haz-
ard from Data Model 3 indicates contribution of only large
earthquakes in all of peninsular India.
Thus Data Models 1, 2, and 3 are taken in such a way
that the earthquake activity rate (a-values) are representative
Figure 7. Probabilistic seismic-hazard map of peninsular India, in terms of mean
peak ground acceleration (%g) for a return period of 475 years using Data Model 2.
K. Jaiswal and R. Sinha
of the entire span of catalog data considering its complete-
ness and accuracy. The combination model consists of dif-
ferent models framed to incorporate earthquake activity rate
associated with different magnitude ranges based on com-
pleteness criteria for the catalog data. The combination rule
considers equal weight given to individual cases of uniform,
induced, geo-based, and background seismicity as shown in
Figure 6.
The typical results of estimated PGA for peninsular In-
dia using Data Model 2 and the three attenuation relation-
ships is shown in Figure 7. Since Data Model 2 includes
only large earthquakes, which are few in number, the PGA
estimates show the typical bulls-eye seismicity pattern, in-
dicating localized recurrence of moderate to large earth-
quakes. It is seen that the estimated PGA for some regions
the Narmada Lineament (NL) (Fig. 2) is of the order of
0.12g, whereas for some pockets of the Eastern Passive Mar-
gin (EPM), the estimated PGA is 0.18g, illustrating the in-
fluence of more severe seismic activity in EPM than in NL.
The analysis results for all data models are combined as
described previously to obtain the combined PGA map.
The new seismic-zoning map prepared based on the
combination rule of Figure 6 is shown in Figure 8. The zon-
ing map represents the average of the estimated ground mo-
tion in the contour interval so that the possible variation in
terms of zoning boundaries are included while estimating
Probabilistic Seismic-Hazard Estimation for Peninsular India
Figure 8. Probabilistic seismic-zoning map of peninsular India, in terms of mean
PGA (%g) for a return period of 475 years.
the zoning factors for each combination. The seismic-hazard
map given in the IS code (Fig. 9) has also used a similar
averaging scheme of estimated damage intensities. The
Medvedev–Sponheuer–Karnik (MSK) damage intensity of
VI or less corresponds to seismic zone II; damage intensity
VII corresponds to seismic zone III; damage intensity VIII
corresponds to seismic zone IV, and damage intensity IX
and greater corresponds to seismic zone V.
The suggested zoning map (Fig. 8) based on PGA esti-
mates with 10% probability of exceedance in 50 years needs
to be interpreted in terms of zoning parameters mentioned
in the current seismic-hazard map of India. The mean hori-
zontal PGA for the Koyna region is found to be between
0.20g and 0.36g. This level of ground motion is associated
with seismic zone V of the seismic zoning map of India (IS
1893, 2002). However this region is currently placed in seis-
mic zone IV (Fig. 9). For Runn of Kuchchh, the estimated
PGA of 0.20g to 0.36g is similar to that associated with seis-
mic zone V. Similarly, large portions of the Narmada Lin-
eament region extending up to the northern portion of Go-
davari graben show higher estimated hazard than the current
IS code specification. The current analysis suggests that it
should be included in seismic zone IV, which is also sup-
ported by the fact that the damaging Jabalpur earthquake of
1999 (maximum MSK intensity of VIII) occurred in this
region. Similarly most of the southern portion of India, es-
327
pecially the Eastern Passive Margin, shows higher PGA es-
timation than its assigned seismic zone III. Most of the east-
ern part of PI consisting of the western portion of Bengal,
Orissa, and southern portion of Madhya Pradesh show very
low ground motions and are considered to constitute the
main portion of the stable shield of peninsular India.
Conclusions
The investigation presented in this article evaluates the
seismic hazard associated with peninsular India. The most
recent knowledge on seismic activity in the region has been
used to evaluate the hazard incorporating uncertainty asso-
ciated with different modeling parameters as well as spatial
and temporal uncertainties.
In this article, the data-related uncertainties that can be
associated with existing seismicity catalogs as well as the
uncertainties associated with estimated seismicity rates have
been discussed. The uncertainties have been incorporated
through a combination of data models, spatial and temporal
variation in seismicity rates, and variation of ground-motion
characteristics through convolution of such information by
assigning appropriate weights in the logic-tree analysis.
Ground motions depicted in terms of PGA have been
prepared using contours of specific range to facilitate the
easy comparison with the seismic-hazard map of India. The
328
Figure 9. Seismic-zoning map of India (IS 1893, 2002).
method for obtaining zonal boundaries in the IS code
seismic-zoning map are based on the damage intensity data
of past earthquakes and hence shows lower hazard for re-
gions without records of damaging earthquakes in the recent
past.
Based on the current investigation, following conclu-
sions have been obtained:
1. The lack of knowledge about the occurrence of earth-
quakes in peninsular India, a SCR, has been addressed
using probabilistic analysis that accounts for uncertainties
in assessments of different parameters as well as com-
binations of different models using the logic-tree scheme.
The probabilistic seismic-hazard assessment of peninsu-
lar India has been carried out which results in a more
realistic seismic-hazard map for the region.
2. Recent seismogenic characteristics in peninsular India
clearly indicate the possible temporal and spatial varia-
tion in terms of seismicity rates that needs to be incor-
porated in the hazard analysis. The use of a single b-value
for the entire Indian peninsula is not an appropriate
approach.
K. Jaiswal and R. Sinha
3. The effect of reservoir-induced seismicity has been eval-
uated and seismicity parameters have been obtained for
data in the vicinity of 20 known reservoirs and their rate
patterns have been included. It is shown that the increase
in seismic activity after 1960 is not solely attributed to
the effect of reservoirs or other triggering mechanisms
near reservoirs, but is a manifestation of real increase in
rates in peninsular India.
4. The zoneless seismic-hazard estimation approach has been
suitable for peninsular India to incorporate spatial and tem-
poral variations in terms of b-values. The seismic hazard
for different regions of peninsular India clearly indicates
higher levels of hazard for most of the rifting zones.
5. The seismic-hazard analysis is based on certain assump-
tions in the form of seismicity modeling, both back-
ground or reservoir-induced seismicity, and prediction of
ground motion for uniform hard-rock conditions. The
weighting scheme proposed in the present investigation
is based on judgment and can be refined. We feel that the
current analysis broadly reflects the first-order seismic
hazard of peninsular India by using probabilistic model-
Probabilistic Seismic-Hazard Estimation for Peninsular India
ing, and the refinements by improving the assumptions
will result in higher-order changes near the boundaries of
the contours. We recommend that this map should be
used within the framework of the assumptions.
Acknowledgments
We thank Arthur Frankel, Charles Mueller, and Mark Petersen for
their inspiration and cooperation in different phases of this work. We spe-
cially thank Chris Cramer and Ivan Wong for their helpful suggestions and
for critically reviewing the manuscript. We also thank Freidemann Wenzel
and Vladimir Sokolov of Geophysical Institute, University of Karlsruhe,
Germany, for their cooperation and help during the initial phase of this
investigation.
References
Atkinson, G. M., and D. M. Boore (1995). Ground motion relations for
eastern North America, Bull. Seism. Soc. Am. 85, 17–30.
Basu, S., and N. C. Nigam (1977). Seismic risk analysis of Indian Penin-
sula, in Proc. 6th World Conf. on Earth. Eng., New Delhi, India,
Vol. I, 782–788.
Bhatia, S. C., M. Ravi Kumar, and H. K. Gupta (1999). A probabilistic
seismic hazard map of India and adjoining regions, Ann. Geofis. 42,
no. 6, 1153–1166.
Bodin, P., L. Malagnini, and A. Akinci (2004). Ground motion scaling in
the Kachchh basin, India, deduced from the aftershocks of the 2001
Mw 7.6 Bhuj earthquake, Bull. Seism. Soc. Am. 94, 1658–1669.
Chandra, U. (1977). Earthquakes of peninsular India, a seismotectonic
study, Bull. Seism. Soc. Am. 65, no. 5, 1387–1413.
Cramer, C. H. (2001). The New Madrid seismic zone: capturing variability
in seismic hazard analyses, Seism. Res. Lett. 72, no. 6, 664–672.
Cramer, C. H., and A. Kumar (2003). 2001 Bhuj, India, earthquake engi-
neering seismoscope recordings and Eastern North America ground
motion attenuation relations, Bull. Seism. Soc. Am. 93, 1390–1394.
Cramer, C. H., R. L. Wheeler, and C. S. Mueller (2002). Uncertainty anal-
ysis for seismic hazard in the Southern Illinois basin, Seism. Res. Lett.
73, no. 5, 792–805.
Das, S., I. D. Gupta, and V. K. Gupta (2006). A probabilistic seismic hazard
analysis of Northeast India, Earthquake Spectra 22, no. 1, 1–27.
Frankel, A. (1995). Mapping seismic hazard in the Central Eastern United
States, Seism. Res. Lett. 66, no. 4, 8–21.
Frankel, A., C. Mueller, T. Barnhard, D. Perkins, E. V. Leyendecker, N.
Dickman, S. Hansom, and M. Hopper (1996). National seismic hazard
maps: documentation, U.S. Geol. Surv. Open-File Rept. 96-532, 69.
Frankel, A., M. Petersen, C. Mueller, K. Haller, R. Wheeler, E. Leyen-
decker, R. Wesson, S. Harmsen, C. Cramer, D. Perkins, and K. Ruks-
tales (2002). Documentation for the 2002 update of the national
seismic hazard maps, U.S. Geol. Surv. Open-File Rept. 02-420.
Gahalaut, V. K., V. K. Kalpana, and S. K. Singh (2004). Fault interaction
and earthquake triggering in the Koyna-Warna region, India, Geo-
phys. Res. Lett. 31, no. 11, 1–4.
Gardner, J. K., and L. Knopoff (1974). Is the sequence of earthquakes in
southern California, with aftershocks removed, Poissonian?, Seism.
Soc. Am. 64, 1633–1638.
Gubin, I. E. (1968). Seismic zoning of the Indian Peninsula, Bull. Int. Inst.
Seism. Earthquake Eng. 5, 109–139.
Guha, S. K. (1962). Seismic regionalization of India, in Sec. Symp. Earth-
quake Eng., University of Roorkee, 191–207.
Gupta, H. K. (1992). Reservoir-Induced Earthquakes, Elsevier Scientific
Publishing, Amsterdam, 355.
Gupta, I. D., M. I. Todorovska, V. K. Gupta, V. W. Lee, and M. D. Trifunac
(1995). Selected topics in probabilistic seismic hazard analysis, Re-
329
port No. CE95-08, University of Southern California, Los Angeles,
304.
Herrmann, R. (1977). Recurrence relations, Earthquake Notes 48, nos. 1–2,
47–49.
IS 1893 (Part 1) (2002). Indian standard criteria for earthquake resistant
design of structures, Fifth Revision, Bureau of Indian Standards, New
Delhi.
Iyengar, R. N., and S. T. G. Raghukanth (2004). Attenuation of strong
ground motion in peninsular India, Seism. Res. Lett. 75, 530–540.
Jaiswal, K., and R. Sinha (2004). Webportal on earthquake disaster aware-
ness in India, www.earthquakeinfo.org (last accessed 10 December
2004).
Johnston, A. C. (1996). Seismic moment assessment of earthquakes in sta-
ble continental regions, Part-II: Instrumental seismicity, Geophys. J.
Int. 124, 381–414.
Kaila, K. L., and N. M. Rao (1979). Seismic zoning maps of the Indian
subcontinent, Geophys. Res. Bull. 17, 293–301.
Khatri, K. N., A. M. Rogers, and S. T. Algermissen (1984). A seismic
hazard map of India and adjacent areas, Tectonophysics 108, 93–134.
Krishna, J. (1959). Seismic zoning of India, in Earthquake Eng. Seminar,
Roorkee University, India, 32–38.
Krishna, J. (1992). Seismic zoning maps of India, Curr. Sci. 79, nos. 1–2,
17–23.
Mandal, P., B. K. Rastogi, and H. K. Gupta (2000). Recent Indian earth-
quakes, Curr. Sci. 79, nos. 9–10, 1334–1346.
Mohan, I., M. V. D. Sitaram, and H. K. Gupta (1981). Some recent earth-
quakes in peninsular India, J. Geol. Soc. India 22, 292–298.
Mooney, W. D., V. V. Rao, P. R. Reddy, G. S. Chulick, and S. T. Detweiler
(2005). Comparison of the deep crustal structure and seismicity of
North America and the Indian subcontinent, Curr. Sci. 88, no. 10,
1639–1651.
Naqvi, S. M., V. Divakara Rao, and H. Narain (1974). The protocontinental
growth of the Indian shield and the antiquity of its rift valleys, Pre-
crambian Res. 1, 345–398.
National Earthquake Information Center (2003). PDE reportings, http://
neic.usgs.gov/neis/epic (last accessed 21 March 2003).
Parvez, I. A., F. Vaccari, and G. F. Panza (2003). A deterministic seismic
hazard map of India and adjacent areas, Geophys. J. Int. 155, 489–508.
Rajendran, C. P. (2000). Using geological data for earthquake studies: a
perspective from peninsular India, Curr. Sci. 79, nos. 9–10, 1251–
1258.
Rao, B. V., and B. V. Satanaranaya Murty (1970). Earthquakes and tecton-
ics in peninsular India, J. Indian Geophys. Union 7, 1–9.
Rao, B. R., and P. S. Rao (1984). Historical seismicity of peninsular India,
Bull. Seism. Soc. Am. 74, 2519–2533.
Rastogi, B. K. (1992). Seismotectonics inferred from earthquakes and earth-
quake sequences in India during the 1980s, Curr. Sci. 62, nos. 1–2,
101–108.
Reasenberg, P. (1985). Second-order moment of central California seis-
micity, 1969–1982, J. Geophys. Res. 90, 5479–5495.
Schweig, E., J. Gomberg, M. Petersen, M. Ellis, P. Bodin, L. Mayrose, and
B. K. Rastogi (2003). The Mw 7.7 Bhuj earthquake: Global lessons
for earthquake hazard in intra-plate regions, J. Geol. Soc. India 61,
277–282.
Seeber, L., J. G. Armbruster, and K. H. Jacob (1999). Probabilistic assess-
ment of earthquake hazard for the State of Maharashtra, Unpublished
Report, Government of Maharashtra, Earthquake Rehabilitation Cell,
Mumbai.
Singh, S. K., J. F. Pacheco, B. K. Bansal, P. Campos, R. S. Dattatrayam,
and G. Suresh (2004). A source study of the Bhuj, India, earthquake
of 26 January 2001 (Mw 7.6), Bull. Seism. Soc. Am. 94, 1195–1206.
Sinha, R., A. Goyal, R. Shaw, H. Arai, J. Saita, K. Pribadi, M. Chaudhari,
and K. Jaiswal (2001). The Great Bhuj earthquake of January 26th
2001, consequences and future challenges, a joint report based on
reconnaissance studies prepared by IIT Bombay and EdM Japan,
http://www.civil.iitb.ac.in/BhujEarthquake/OpenPage.htm (last ac-
cessed 10 June 2005).
330 K. Jaiswal and R. Sinha

Stepp, J. (1973). Analysis of completeness of the earthquake sample in Weichert, D. H. (1980). Estimation of the earthquake recurrence parameters
Puget Sound area. Seismic zoning edited by S. T. Harding NOAA for unequal observation periods for different magnitudes, Bull. Seism.
Tech. Report ERL 267-ESL30, Boulder, Colorado. Soc. Am. 70, 1337–1346.
Tandon, A. N. (1956). Zoning of India liable to earthquake damage, Ind.
J. Meteorol. Geophys. 10, 137–146.
Department of Civil Engineering
Toro, G. N., N. Abrahamson, and J. Schneider (1997). Model of strong Indian Institute of Technology Bombay
ground motions from earthquakes in central and eastern North Amer- Powai, Mumbai 400 076 India
ica: Best estimates and uncertainties, Seism. Res. Lett. 68, 41–57. rsinha@civil.iitb.ac.in
Valdiya, K. S. (1973). Tectonic framework of India: A review and inter-
pretation of recent structural and tectonic studies, Geophys. Res. Bull.
11, 79–114. Manuscript received 16 June 2005.