Sie sind auf Seite 1von 27

Probabilistic earthquake hazard assessment

for Peninsular India

Ashish ᅟ, C. Lindholm, I. A. Parvez &


D. Kühn

Journal of Seismology

ISSN 1383-4649
Volume 20
Number 2

J Seismol (2016) 20:629-653


DOI 10.1007/s10950-015-9548-2

1 23
Your article is protected by copyright and all
rights are held exclusively by Springer Science
+Business Media Dordrecht. This e-offprint
is for personal use only and shall not be self-
archived in electronic repositories. If you wish
to self-archive your article, please use the
accepted manuscript version for posting on
your own website. You may further deposit
the accepted manuscript version in any
repository, provided it is only made publicly
available 12 months after official publication
or later and provided acknowledgement is
given to the original source of publication
and a link is inserted to the published article
on Springer's website. The link must be
accompanied by the following text: "The final
publication is available at link.springer.com”.

1 23
Author's personal copy
J Seismol (2016) 20:629–653
DOI 10.1007/s10950-015-9548-2

ORIGINAL ARTICLE

Probabilistic earthquake hazard assessment


for Peninsular India
Ashish · C. Lindholm · I. A. Parvez · D. Kühn

Received: 7 September 2015 / Accepted: 15 December 2015 / Published online: 13 January 2016
© Springer Science+Business Media Dordrecht 2016

Abstract In this paper, a new probabilistic seismic in 50 years observed in Koyna is 0.4 g and in the
hazard assessment (PSHA) is presented for Peninsular Kutch region of Gujarat up to 0.3 g. With respect to
India. The PSHA has been performed using three dif- spectral acceleration at 1 Hz, estimated ground motion
ferent recurrence models: a classical seismic zonation amplitudes are higher in Gujarat than in the Koyna
model, a fault model, and a grid model. The develop- region due to the higher frequency of occurrence of
ment of a grid model based on a non-parameterized larger earthquakes. We discuss the higher PGA levels
recurrence model using an adaptation of the Kernel- for Koyna compared Gujarat and do not accept them
based method that has not been applied to this region uncritically.
before. The results obtained from the three mod-
els have been combined in a logic tree structure in Keywords Hazard · Peninsular India · KERGRID ·
order to investigate the impact of different weights PSHA
of the models. Three suitable attenuation relations
have been considered in terms of spectral accelera-
tion for the stable continental crust as well as for the 1 Introduction
active crust within the Gujarat region. While Peninsu-
lar India has experienced large earthquakes, e.g., Latur Peninsular India, usually considered as a stable conti-
and Jabalpur, it represents in general a stable conti- nental part of the Indian subcontinent, has experienced
nental region with little earthquake activity, as also damaging earthquakes of magnitude ∼6 and more
confirmed in our hazard results. On the other hand, in the last couple of decades. Earthquakes occurred
our study demonstrates that both the Gujarat and the in various places in Peninsular India: Latur (1993,
Koyna regions are exposed to a high seismic hazard. Mw 6.1), Jabalpur (1997, Mw 5.8), and most recently
The peak ground acceleration for 10 % exceedance Bhuj (2001, Mw 7.6), claiming thousands of lives and
causing huge economic losses due to the damage to
infrastructure. Although the frequency of occurrence
Ashish () · I. A. Parvez
of large earthquakes is low, their impact on society
CSIR Centre for Mathematical Modelling and Computer is high. Thus, it becomes important to quantify the
Simulation (CMMACS), NAL Belur Campus, seismic hazard for Peninsular India for future events
Bangalore, India in terms of potential ground shaking, while at the
e-mail: ashish@cmmacs.ernet.in
same time acknowledging difficulties and the inher-
C. Lindholm · D. Kühn ent uncertainties. The seismicity and major geological
NORSAR, Kjeller, Norway features are illustrated in Fig. 1.
Author's personal copy
630 J Seismol (2016) 20:629–653

Fig. 1 Geology of
Peninsular India derived
from the Geological Map of
India (GSI 2000), and
locations of historical
earthquakes (circles). Color
and diameter of circles
indicate depth and
magnitude, respectively.
Major events that occurred
during the past 50 years are
indicated by red stars

Early work on seismic hazard in India by Tondon few earthquake hazard studies were carried out for
(1956) and Krishna (1959) is limited to the identi- Peninsular India and most studies concentrate on
fication of potential earthquake areas. These studies small sub-regions, e.g., Seeber et al. (1999) provide
demarcate the Himalayas, the Chaman fault region, a detailed hazard map for the state of Maharash-
and the Indo-Burmese arc as exhibiting a higher haz- tra; Menon et al. (2010) provides for Tamilnadu and
ard, the Indian shield to feature a lower hazard and the Vipin et al. (2009) provides for South India. Jaiswal
enclosed regions to pose a moderate damage poten- and Sinha (2007) estimate probabilistic seismic hazard
tial. Other hazard studies carried out by Guha (1962) for Peninsular India applying the zonefree approach
and Gubin (1968) are based on microseismic inten- proposed by Frankel (1995) including multiple mod-
sity. Subsequent studies include probabilistic hazard els with appropriate weight. In another study, Nath
by Basu and Nigam (1977), Kaila and Rao (1979) and and Thingbaijam (2012) estimated probabilistic seis-
Khatri et al. (1984). Khatri et al. (1984) provide a mic hazard for Indian subcontinent, applicable to B-C
detailed peak ground acceleration (PGA) hazard map soil types (VS30 ∼ 760 m/s), using similar zone-free
describing a 10 % annual probability of exceedance approach used by Jaiswal and Sinha (2007) along with
within 50 years. Recent efforts include a probabilis- areal zones in a logic tree. A detail review of seis-
tic map developed by Bhatia et al. (1999) under the mic hazard has been published by Yanger Walling
Global Seismic Hazard Programme (GSHAP). Since and Mohanty (2009) and Verma and Bansal (2013).
the aim of GSHAP was to produce hazard maps on a Verma and Bansal (2013) review is largely focussed
global scale, the GSHAP project includes only a few on microzonation studies conducted in various Indian
sources for Peninsular India focusing on the inter-plate cities.
region along the Himalayan belt. Parvez et al. (2003) Recently, the National Disaster Management
carried out the first neo-deterministic seismic hazard Authority of India (NDMA) published a report on
assessment for the Indian subcontinent on the basis of the comprehensive probabilistic seismic hazard for the
simulation and computation of synthetic seismograms. Indian Subcontinent (NDMA 2011). In this report,
Due to the rare occurrence of larger earthquakes India has been subdivided into 32 seismic zones with
and the resulting scarcity of catalogue data, only Peninsular India comprising 9 zones applying a classical
Author's personal copy
J Seismol (2016) 20:629–653 631

zonation approach (Cornell 1968) and not allowing The NSL has undergone several periods of reactiva-
for gaps between zones. The study also provides and tion since its evolution in the Archean (Choubey 1971;
applies ground motion prediction equations (GMPEs) Verma and Banerjee 1992) and remains seismically
newly developed for seven of those regions based on active. Most of the earthquakes throughout Peninsular
strong motion data recorded on the Indian subconti- India occur at shallow depths (0–20 km); earthquakes
nent. A similar study also carried out by Sitharam and with depths between 10 and 20 km are mostly located
Kolathayar (2013) using a similar approach except the in a belt crossing northern Peninsular India from west
use of 104 seismic source zones which is much higher to east.
than (NDMA 2011).
The present study expands the classical methods
used largely by NDMA (2011) and Sitharam and 3 Preparation of earthquake catalogues
Kolathayar (2013) to generate probabilistic seismic
hazard maps for Peninsular India using a logic tree A comprehensive earthquake database including the
approach that includes both the classical approach locations, origin times, and magnitudes of past earth-
(area and line sources) and the zone free Kernel quakes is required in order to reliably model the
method that applies the gridded seismicity model by seismicity to be used in a probabilistic seismic haz-
Woo (1996). ard analysis (PSHA). Thus, a common catalogue is
constructed for the Peninsular region (5◦ N to 26◦
N and 67◦ E to 90◦ E) with priorities according to
2 Geology and tectonics of Peninsular India the list of catalogues based on the expected catalogue
reliability, completeness, catalogue homogeneity, and
The Indian Peninsula is one of the oldest landmasses the size of seismic network constituting the basis for
on earth and consists of a Precambrian craton as well the catalogue. No double reporting was also ensured.
as mobile belts disrupted by recent episodes of vol- The same priorities were applied to all information
canism and rifting (Fig. 1). Most of Peninsular India extracted from the catalogues (magnitude, location,
is built up of Precambrian crystalline rocks and sed- and depth). The list of priorities is as follows:
imentary formations and a third region is covered by
1. The ISC reviewed bulletin since 1960 (www.isc.
the Deccan trap basaltic flows, which were extruded in
ac.uk),
the Cretaceous-Eocene interval (Valdiya 1973; Naqvi
2. The PDE catalogue since 1973 (USGS National
et al. 1974). Archean rocks, high grade metamorphic
Earthquake Information Center http://earthquake.
rocks, cover the southern part of the Indian penin-
usgs.gov/∼regional/neic/)
sula, the Southern Granulite Province (SGP). To the
3. The IMD (India Meteorological Department) cat-
north of the SGP, the Dharwar craton exposes 2.5
alogue since 1898 (http://www.imd.gov.in),
Ga old Clospet granite, a N-S lineament, dividing the
4. The catalogue prepared by the NDMA from var-
Dharwar craton in two parts: the eastern and west-
ious sources for their report (between −2474 to
ern Dharwar craton, the eastern Dharwar craton being
2008) (http://www.ndma.gov.in),
younger than western Dharwar. North of the west-
5. The National Oceanic and Atmospheric Adminis-
ern Dharwar craton, the Indian Peninsula is covered
tration (NOAA) catalogue (www.noaa.gov)
by the Deccan basalts. The eastern side of the Indian
6. Events listed by Oldham (1883), Tondon and
Peninsula is segregated by the Godavari and Mahanadi
Chaudhury (1968), and Guha et al. (1970). For
rift basins resulting from the breakup of Gondwana.
historic events, Ramalingeswara Rao and Sitap-
The Godavari rift separates the Dharwar craton from
athi Rao (1984) provide a magnitude scale based
the archaean rocks of the Baster and Singbhum cra-
on intensities.
tons, which are divided by the Mahanadi rift. Further
to the east, archaean rocks are covered by quaternary Figure 2 shows a comparison between various
alluvium deposits of Bengal basin. To the northwest, earthquake catalogues. In general, the geographical
the Aravali craton and the minor Bundelkhand craton pattern exhibits a low level of seismic activity over
are located, which are separated from the Deccan vol- the last 50 years, but with activity concentrated in the
canic region by the Narmada Son Lineament (NSL). Bhuj, Latur, Koyna, and Bay of Bengal regions as
Author's personal copy
632 J Seismol (2016) 20:629–653

Fig. 2 Comparison of earthquake catalogues: a Indian cata- catalogue since 1973; d ISC catalogue since 1960; background:
logue compiled in this study from various sources (see sym- Shuttle Radar Topography Mission (SRTM) elevation map
bol colors); b similar to (a), but excluding events introduced (USGS 2014); size of the triangles corresponds to magnitude
by Ramalingeswara Rao and Sitapathi Rao (1984); c PDE

well as along the Narmada Son structural lineament. more uniform earthquake catalogue. We will, how-
Whereas the PDE (Fig. 2c) and ISC (Fig. 2d) cata- ever, calibrate our recurrence model parameters by the
logues agree in geographical distribution of the earth- recurrence of larger historic events and qualitatively
quakes, the distribution of events in Fig. 2a including by the earthquake distribution in earlier catalogues.
the events listed by Ramalingeswara Rao and Sitap- Another special occurrence was the December 10,
athi Rao (1984) is more regular and earthquakes are 1967, MW 6.3 Koyna earthquake that was proba-
more numerous. Thus, we decided to quantitatively bly triggered by water reservoir impoundment (Gupta
derive seismic parameters from the common catalogue 1992) of the Shivajisagar Lake reservoir confined
starting with the year 1960, when the establishment by the Koyna dam (just north of Koyna earthquake
of a larger worldwide seismic network guarantees a epicentre) constructed in 1964. Until today, the occur-
Author's personal copy
J Seismol (2016) 20:629–653 633

rence of small and moderate magnitude earthquakes addition, as the overestimation of the number of fore-
in this region is high. Seeber et al. (1999) notice a and aftershocks may lead to the underestimation of
threefold increase in seismic activity for Peninsular seismic hazard, we applied a semi-automatic approach
India since 1960 and Rastogi (1992) states that more considering the magnitude of the main event, the inter-
than 50 % of the seismic activity within Peninsu- event distance, the difference in magnitude between
lar India occurs within the region around the Koyna main shock and fore-/aftershocks as well as the differ-
reservoir. Recent studies, though, indicate that the ence in origin time. The spatial window for occurrence
increase of seismic activity in the vicinity of the Koyna of aftershocks is determined according to Christo-
reservoir is not solely triggered by water reservoir phersen and Smith (2000) and, assuming a circular
impoundment (Gahalaut et al. 2004), but possibly also area, converted to a radius as a measure for the maxi-
includes a tectonic component. However, including mum inter-event distance. The time window for occur-
reservoir-induced seismicity into the seismic hazard rence of aftershocks is set to 0.5 years and the min-
analysis may lead to an overestimation of seismic haz- imum magnitude difference between main and after-
ard within this region (see more detailed discussion in shock is chosen as 1.2 following Båth’s law (Richter
Section 8). 1958; Bȧth 1965) leading to a relatively soft filter.
In order to homogenize magnitudes to a com- All events that were removed from the catalogue were
mon magnitude scale (i.e., moment magnitude controlled manually. Due to the laborious process,
MW ), we established regression relations using an the catalogue was declustered removing only after-
orthogonal regression method (Das et al. 2013) shocks for events with magnitude 6 or above (no fore-
employing data from the Harvard CMT catalogue shocks could be identified). Totally 170 events were
(www.globalcmt.org/CMTsearch.html) recorded noticed as dependent and removed, which accounts for
within the region of interest. The conversion relations 15 % of the total events. The maximum magnitude
are summarized in Table 1. For regression between removed was 6.1 as an aftershock related to 2001 Bhuj
ML and MW , only 17 data points are used which is earthquake. The final catalogue used as basis for the
fairly less data point to estimate a robust regression. analysis of seismicity is displayed in Fig. 1.
Therefore, conversion from ML and MW has been
carried out in two steps using ML to mb and mb to
MW relations 4 Seismicity models
Probabilistic seismic hazard assessment relies on
the assumption of earthquakes following a stationary The seismic catalogue needs to be transformed into
Poisson process, i.e., being independent from each source models that can be used as input for the seismic
other. To this end, the catalogue has to be declus- hazard assessment, providing essentially a recurrence
tered, i.e., separating main shocks from fore- and model for the number of expected events at differ-
aftershocks. However, since all declustering methods ent magnitude levels and for different source regions.
(deterministic and stochastic) rely on a conceptual Three models have been generated for the computa-
model of how to define a mainshock, there is no pre- tion of the hazard map: classical seismic zones, a grid
ferred method as such (Van Stiphout et al. 2012). In model, and a fault model. The grid model is based on

Table 1 Magnitude conversion relations estimated from the Harvard CMT catalogue for the Indian peninsula

Conversion Relation Data points

mb ↔ ML ML = 0.9930(±0.023)mb + 0.0550(±0.097) 441


mb ↔ MW MW = 1.3488(±0.051)mb − 1.6520(±0.274) 194
ML ↔ MW MW = 0.6044(±0.337)ML + 2.5938(±1.615) 17
M0 ↔ MW MW = 0.7570(±0.011)M0 + 1.4299(±0.066) 98
MS ↔ MW MW = 0.5667(±0.006)MS + 2.6046(±0.034) 125

mb body wave magnitude, ML local magnitude, MS surface wave magnitude, MW moment magnitude, M0 other magnitude
Author's personal copy
634 J Seismol (2016) 20:629–653

a non-parameterized recurrence model with the use of of the physical potential. This clearly is a simplifi-
kernel methods (Woo 1996) using a power-law decay cation with respect to a more continuous seismicity
of the kernel proposed by Vere-Jones (1992). This distribution, but on the other hand, the method avoids
methodology has not been applied previously for com- over-interpretation of an earthquake catalogue only
putation of hazard maps for the Indian subcontinent. covering a short time window compared to return peri-
ods of larger events. To this end, the seismicity should
4.1 Seismic zonation be reasonably uniform within each area zone and the
zonation should be broadly consistent with regional
The zonation has the purpose of dividing the seismic- geology and tectonics. Based on the above general
ity into distinct source zones reflecting the tectonic principles, we subdivided Peninsular India into 11
conditions in order to subsequently evaluate the seis- seismogenic zones (Fig. 3):
mic potential of each zone. The main principle applied
in the zonation process is to represent the seismicity – Zone 1 corresponds to the Gujarat region, which
and tectonics as balanced as possible. Seismic activity is tectonically different from the remaining part
within each zone is modeled as homogeneous, and this of the peninsula. The Allahbund earthquake of
should be justifiable as a reasonable approximation 1819 and the January 26, 2001 Bhuj intra-plate

Fig. 3 Map of Peninsular India displaying the eleven seismic source zones. The background illustrates the faults digitized from the
seismotectonic atlas (GSI 2000)
Author's personal copy
J Seismol (2016) 20:629–653 635

earthquake (MW 7.6) occurred within this zone, – Zone 11 comprises the Koyna-Warna zone (see
the focal mechanism given by Harvard CMT indi- Fig. 3).
cates reverse faulting with a strike-slip compo-
In order to assess the magnitude of completeness,
nent.
time-magnitude plots of the final catalogue have been
– Zone 2 represents largely the Narmada Son Linea-
generated for three different time periods (i.e., 1340–
ment (NSL) including the Son-Narmada and Tapti
2010, 1800–2010, and 1960–2010) as shown in Fig. 4.
North faults but portion of NSL continues into
Very few events have been reported before the year
zone 4 also.
1800 and a significant increase in the number of events
– Zones 3 and 6 contain the Bundelkhand and Bas-
is recognized after 1960, the advent of instrumentally
tar cratons, respectively, both exhibiting very low
recorded data. Thus, a completeness value is derived
seismicity; zone 6 might even be described as
as M > 4.5 for the time period 1960 to the present.
aseismic.
The seismic activity of a region is characterized
– Zone 4 comprises the Singhbhum craton as well
by the Gutenberg-Richter (G-R) recurrence relation
as the Bengal basin.
(Gutenberg and Richter 1944):
– Zones 5, 8, and 10 are coastal zones and corre-
spond to the eastern, north-western, and south- log10 N(m) = a − bm, (1)
western passive margins, respectively.
– Zone 7 represents the Godavari graben. where N(m) is the number of earthquakes greater than
– Zone 9 covers the Dharwar craton. or equal to magnitude m, a denotes the seismicity

Fig. 4 Time-magnitude
distribution of earthquakes
within the final catalogue
for three different time
periods; top: 1340–2011,
middle: 1800–2011, bottom:
1960–2011
Author's personal copy
636 J Seismol (2016) 20:629–653

Table 2 Comparison between a and b values gained from the Gutenberg-Richter regression for the time period 1960–2011 for the
entire catalogue, zone 1 (Gujarat) and a megazone constructed from zones 2–10. N(4.8) represents the annual number of earthquakes
of magnitude 4.8 or greater

Region 960–2011

b a N(4.8)

Entire catalogue 1.02 5.3 2.5351


Zone 1 (Gujarat) 0.80 3.64 0.6309
Zone 2-10 0.85 4.11 1.0954
Zone 11 (Koyna) 1.11 5.22 0.7798

rate computed by the logarithm of the average num- for zone 1 (Gujarat), 7.0 ± 0.4 for zone 2 (NSL),
ber of earthquakes of magnitude Mmin or larger, and b 7.5 ± 0.4 for zone 4, and 6.8 ± 0.4 for the remaining
characterizes the proportion of large earthquakes rel- zones based on historical observations (see Table 4).
ative to small earthquakes. For tectonic earthquakes,
the b value is expected to be confined within the range 4.2 Grid model
0.8 < b < 1.2.
Zone 1, representing the Gujarat region, is consid- The aim of using a kernel-based grid model is to
ered to be tectonically different from the remaining obtain seismic activity parameters smoothly varying
zones. The seismic events occurring within zone 11, over the grid based on the historic catalogue data
representing the Koyna region, differ in so far as to while avoiding extensive use of expert judgement. In
include not only tectonic, but also triggered events the zonation model, size and boundaries of zones are
leading to a higher seismicity rate as well as a higher adjusted manually, in addition, the zonation model
b-value (Table 2, Fig. 5). The number of earthquakes assumes a uniform distribution of seismicity in each
with M > 4.5 in the remaining zones is rather low, zone. In reality, the geographical distribution of seis-
impeding the regression analysis by rendering the b micity within Peninsular India is rather inhomoge-
values unstable. Thus, the catalogue has been sepa- neous. The grid model, on the contrary, consists of
rated into zone 1 (Gujarat), zone 11 (Koyna), and a a non-parametric activity distribution on an equally
‘megazone’ consisting of the remaining source zones spaced grid of 0.25◦ × 0.25◦ .
(2–10) exhibiting consistent tectonic features, result- The zone free method proposed by Woo (1996)
ing in a b value of 0.85 for the composite zone, has until now not been widely used with exception
1.11 for the Koyna, and 0.80 for the Gujarat region of Bommer et al. (1998) and Molina et al. (2001),
(see Table 2 for regression results, and Fig. 5a, b but has gained interest recently (Beauval et al. 2006;
for the G-R plot of the full catalogue and the Koyna Ornthammarath et al. 2008; Crespo Álvarez et al.
region, respectively). The regression has been per- 2014). Beauval et al. (2006) compared probabilistic
formed using the maximum likelihood method as seismic hazard estimation from zoning and smoothing
implemented in the ZMAP software (Wiemer 2001). approach and demonstrate a similar hazard estimates
In order to obtain stable seismicity rates for zones from both approach in low seismicity region. The
2–10 despite the low number of earthquakes having fundamental principle consists of a direct use of the
occurred in most of them, the percentage of earth- earthquake catalogue rather than applying simplified
quakes with M ≥ 4.8 within each zone compared recurrence parameters. Each earthquake is treated as
to the total number of earthquakes has been used to the center of a seismic source and a Kernel probabil-
compute the a value of these zones (Table 3). ity function of occurrence rate is constructed around it
In this study, Mmax has been estimated using con- (see kernel example in Fig. 6). Furthermore, the nor-
ventional increment of 0.3 and 0.5 in maximum mal completeness estimate of a catalogue (which most
observed magnitude (Mmax obs ). The base value of Mmax often excludes older data from being quantitatively
has been fixed to 6.8 for stable continental region. We used) is substituted with a magnitude probability for
estimate the maximum magnitude Mmax as 8.0 ± 0.3 the reporting function such that a matrix is constructed
Author's personal copy
J Seismol (2016) 20:629–653 637

Fig. 5 G-R regression a for


the entire catalogue of
Peninsular India and b for
the Koyna region

for different time windows and the time intervals of gridded activity rates is computed with the kernel-
indicate the probability of detectability. With this based method for a discrete set of magnitudes suitable
approach, an effective cumulative observability can be for the general formulation of gridded seismicity in
established, and the full catalogue is used to the max- CRISIS. The result is a highly flexible combination
imum of its information value. The original approach of the two approaches, gridded activity rates provided
of Woo (1996) was later adapted in the KERGRID in a mesh and based on the zone free kernel method
software and recently, an interface to the CRISIS soft- that can be entered into a logic tree together with other
ware (Ordaz et al. 2014) was established: A distinct set (more conventional) approaches.
Author's personal copy
638 J Seismol (2016) 20:629–653

Table 3 A values obtained for zones 2–10: n ≥ 4.8 is a count of the number of earthquakes with M ≥ 4.8 within each zone and
N(4.8) is the annual occurrence of earthquakes with M ≥ 4.8

Zones b (megazone) n ≥ 4.8 Per. (%) a N(4.8)

2 0.85 8 11.76 3.18 0.1289


3 0.85 3 4.41 2.75 0.0483
4 0.85 25 36.76 3.68 0.4027
5 0.85 8 11.76 3.18 0.1289
6 0.85 1 1.47 2.28 0.0161
7 0.85 3 4.41 2.75 0.0483
8 0.85 1 1.47 2.28 0.0161
9 0.85 18 26.47 3.53 0.2900
10 0.85 1 1.47 2.28 0.0161

Sum 68 100

For each grid point, a kernel-based probability dis- recurrence period, an expert opinion is considered to
tribution of occurrence is computed following the give a reasonable estimate. With this approach the
approach by Woo (1996) employing discrete mag- effective observability period is derived for different
nitude steps from 4.5 to 7.5 with a step width of magnitudes (Table 5).
0.25 and a spatial kernel width of 25 to 200 km. A zonefree method proposed by Frankel (1995)
The frequency of occurrence of events in the histor- also used grid method where the ‘a’ value is calculated
ical catalogue contributes to the activity rate, which based on the number of earthquakes above a certain
depends on the event’s effective period of observabil- threshold magnitude and smoothed over a region and
ity. A 50-year time window is used, and the percentage then assigned to each grid center along with a region-
of events computed for different time interval and ally determined b value. The kernel method (Woo
the obtained numbers were used as detectability esti- 1996) is different on two accounts: (1) the construc-
mates by assuming that in last 50 year (after the tion of the spatial probability density function (the
advent of instrumental seismology), the probability of kernel) around each grid point and (2) the use of effec-
detectability of events greater than M 5.0 is 100 %. tive observability period not only for one threshold
For the large earthquake (M > 6.5), which has a long magnitude but for the full magnitude range. In the

Table 4 Seismic activity parameters used for the classical zonation model for each zone (β = log10 (b) and λ = N). λnorm is
normalized to a 100 x 100 km2 area

Zones λ β Mmax Mmax (obs.) λnorm

1 0.631 1.842 8.0±0.3 7.7 0.0228


2 0.129 1.953 7.0±0.4 6.5 0.0069
3 0.048 1.953 6.8±0.4 6.0 0.0019
4 0.403 1.953 7.5±0.3 7.0 0.0077
5 0.129 1.953 6.8±0.4 6.0 0.0031
6 0.016 1.953 6.8±0.4 3.9 0.0008
7 0.048 1.953 6.8±0.4 5.9 0.0041
8 0.016 1.953 6.8±0.4 5.9 0.0033
9 0.290 1.953 6.8±0.4 6.2 0.0054
10 0.016 1.953 6.8±0.4 5.0 0.0009
11 0.735 2.556 6.8±0.4 6.3 0.1576
Author's personal copy
J Seismol (2016) 20:629–653 639

Fig. 6 Example of a typical


kernel adapted after
(Molina et al. 2001)

kernel-based approach, a generalized model is used and formation of 4–6-m high fault scarps (Bilham
rather than a simplified Gutenberg-Richter distribu- et al. 1999). The KMF with a length of 125 km is a
tion as used by Frankel (1995) in his computations. reverse fault and exhibits a total structural displace-
ment of 2–3 km. The 2001 Bhuj earthquake occurred
4.3 Fault model along a hidden fault situated about 25 km north of
the KMF (Rastogi 2001). The ABF (F01) and KMF
Unfortunately, there is not much information avail- (F02) are E-W trending major active faults, whereas
able on faults within Peninsular India (apart from the the Marginal fault (F03) is N-NW trending and less
Gujarat region) and even less on the potential seismic active.
activity of the mapped structures. Based on the fault Another major feature is the Narmada Son Linea-
model displayed (GSI 2000) and the geographic distri- ment (NSL), one of the most prominent geological
bution of earthquakes within the catalogue, 11 major features of India apart from the Himalaya. It passes
fault zones have been defined (Table 6). In order to through broach on the west coast of Peninsular India in
account for irregular fault geometry, fault’s dip with NNE-SSW direction cutting across the whole of Cen-
angles lower than 90◦ and unseen (fractal) branches tral India and has been periodically reactivated since
of the fault, a buffer zone is created around the fault Precambrian times (Choubey 1971). The NSL (F04)
lines with a width of 0.2◦ (Fig. 3). Three major faults stays mostly in zone 2, although it cuts across and
are located within zone 1, namely the Allahbund fault extend a bit in zone 4 as well.
(ABF), the Kutch Mainland fault (KMF), and the The Gundlakamma fault (F07) is located at a crustal
Marginal faults (MF) of the Cambay basin. The 1819 junction between a ridge in the north and a depres-
earthquake occurred on the ABF, resulting in uplift sion in the south. Seismicity is reported mainly along

Table 5 Effective observability period (in percent) for each period of the earthquake catalogue compiled in this study to be applied in
the kernel based grid computation

Century of observation 5.0 5.5 6.0 6.5 7.0 7.5 8.0

1900–2011 0.55 0.76 0.85 0.90 0.90 1.00 1.00


1800–1900 0.10 0.21 0.60 0.60 0.70 0.80 0.90
1700–1800 0.03 0.07 0.35 0.40 0.50 0.60 0.90
1600–1700 0.01 0.20 0.25 0.30 0.40 0.80
1500–1600 0.05 0.05 0.10 0.20 0.50
1400–1500 0.01 0.01 0.03 0.10 0.25
1300–1400 0.01 0.01 0.05 0.10
1200–1300 0.01 0.01 0.05
1100–1200 0.01 0.02
1000–1100 0.01
900–1000 0.01
800–900 0.01

Effective cumulative observation period 68 105 206 222 259 317 455
Author's personal copy
640 J Seismol (2016) 20:629–653

Table 6 Faults included in the model and their seismic activity parameters employed in the hazard computations

Zone No Fault λ β Mmax

F01 Allahbund fault 0.076 1.842 8.0 ± 0.3


1 F02 Kutch Mainland fault 0.101 1.842 8.0 ± 0.3
F03 Marginal fault 0.076 1.842 8.0 ± 0.3

2 F04 Narmada Son Lineament 0.103 1.953 7.0 ± 0.4

4 F05 Eocene Hinge Zone 0.161 1.953 6.8 ± 0.4

5 F07 Gundlakamma fault 0.052 1.953 6.8 ± 0.4

7 F06 Kaddam fault 0.019 1.953 6.8 ± 0.4

8 F08 West Coast fault 0.006 1.953 6.8 ± 0.4

9 F11 Cauvery fault 0.116 1.953 6.8 ± 0.4

10
F09 Chiplun fault 0.156 2.556 6.8 ± 0.4
11
F10 Warna fault 0.156 2.556 6.8 ± 0.4

the NW-SE trending Gundlakamma river fault, which the characteristic rupture model, Wesnousky (1994)
runs for nearly 100 km from inland to the east examined the shape of the magnitude-frequency dis-
coast. Focal mechanisms indicate strike-slip faulting tribution along the major strike-slip faults of southern
(Murthy et al. 2010). California, and found that the frequency distribution
The Koyna-Warna fault zone is associated with two along the San Jacinto fault satisfies the Gutenberg-
NNW-SSE left-lateral strike-slip faults (F09) and a Richter relationship which may reflect the superpo-
normal fault (F08). The West Coast fault (F08) is also sition of seismicity along a number of distinct fault
NNW striking and controlling the seismic hazard at segments, each of which displays a characteristic
the west coast around Mumbai city. earthquake distribution. In order to derive reasonable
Since for most of the faults, no information is a- and thus λ values, 40 % of the total seismic activ-
available on style of faulting, fault depth, and annual ity of the corresponding zone has been assigned to
slip, the seismic activity parameters (a, b) have been the fault located within the zone. In case more than
assigned to each fault buffer zone based on the seis- one fault is situated within a zone, the seismic activ-
micity of the corresponding zone of the zonation ity has been attributed to the faults based on the
model (Table 3). Many researchers have the view length of each fault. The NSL is an exception, since
that individual major active geological faults do not its buffer zone covers more than 50 % of the total
exhibit seismicity patterns that is conformed to a G- area of zone 2 and in addition, 80 % of the seis-
R type magnitude-frequency distribution (e.g., Youngs micity of this zone occurs within the NSL buffer
and Coppersmith 1985); however, it is assumed in zone. Thus, 80 % of the seismicity rate of zone 2
general that for the fault sources, the maximum mag- is assigned to the NSL fault. However, it has to be
nitude is well estimated based on the fault dimensions noted that the hazard away from the faults will be
(length and area), and Wells and Coppersmith (1994) relatively damped by this approach. Table 6 gives an
developed a series of empirical relationships between overview of activity parameters employed for the fault
moment magnitude and rupture length. Contrasting to model.
Author's personal copy
J Seismol (2016) 20:629–653 641

5 Ground motion prediction equations (GMPEs) attenuation relations have been estimated, e.g., by
Parvez et al. (2001) for the Himalaya as well as
For the estimation of physical ground motion param- Iyengar and Raghukant (2004) and Raghukant and
eters at a specific site, local and regional properties of Iyengar (2007) for Peninsular India. Parvez et al.
the subsurface conditions play a major role. Whereas (2001) use strong motion acceleration (SMA) data
local site amplification will not be taken into account, for the eastern section of India known to be
selection of appropriate GMPEs is applied within a part of the Indian shield and demonstrate three
logic tree computational framework in order to reflect times higher amplitudes compared to the west-
regional conditions. ern Himalaya. Unfortunately, these relations are
Since only little strong motion data are available available only for peak ground acceleration (PGA)
for the Indian shield, only a few regional GMPEs and no information is provided in terms of spec-
are available. Exceptions are the Koyna-Warna region tral accelerations except in Raghukant and Iyen-
(Gupta et al. 1992; Baumbach et al. 1994), the 1997 gar (2007) which extended the work of Iyengar
Jabalpur earthquake (Singh et al. 1999), and the 2001 and Raghukant (2004) and provide spectral accel-
Kutch earthquake (Singh et al. 2003). Semi-empirical eration using the stochastic seismological model of

Fig. 7 Comparison
between GMPEs for stable
continental regions. All
GMPE are displayed for
magnitude M 7: a peak
ground acceleration vs.
focal distance, b response
spectrum. Spectral
acceleration is not given in
Parvez et al. (2001) and
Iyengar and Raghukant
(2004)
Author's personal copy
642 J Seismol (2016) 20:629–653

Boore (1983). However, it is noted that attenua- full range of possibilities (Atkinson et al. 2014). Since
tion in Peninsular India is comparatively low. In the our choices are limited and we consider the local data
recent hazard study carried out by NDMA (NDMA to be insufficient in quality and quantity to constrain
2011), attenuation relations are constructed in terms attenuation, we tried to overcome this obstacle by
of PGA and spectral acceleration for different tectonic comparing the GMPE relations developed explicitly
regimes, in-contrast to the previous work (Iyengar and for Indian regions with GMPE relations recommended
Raghukant 2004; Raghukant and Iyengar 2007) which for different geotectonic regions on a global scale.
provide single attenuation relation for the Peninsular The most recent GMPEs have been examined and
India, at least partly developed from local data. There- compared within the Global Earthquake Model pro-
fore, we employ their relations for Peninsular India gram (GEM 2014). For stable continental regions
(zones 5, 8, 9, 10, 11), Central India (zones 2, 3, 4, 6, (zones 2–11), their recommendations comprise the
7), and the Gujarat region (zone 1). Atkinson and Boore (2006), Pezeshk et al. (2011)
GMPEs should be chosen to be tectonically appro- and Toro (1997) relations (Fig. 7). For active
priate in order to cover both the center body and the crustal regions—appropriate for zone 1 representing

Fig. 8 Comparison
between GMPEs for regions
comprising active crust. All
GMPEs are displayed for a
magnitude M 7: a peak
ground acceleration vs.
focal distance, b response
spectrum. Spectral
acceleration is not given in
Parvez et al. (2001) and
Iyengar and Raghukant
(2004)
Author's personal copy
J Seismol (2016) 20:629–653 643

Fig. 9 Graphic representation of the logic tree structure


Author's personal copy
644 J Seismol (2016) 20:629–653

Gujarat—GEM (2014) recommends the Akkar and SIS 2014 in the computation uses circular extended
Bommer (2010), Chiou and Youngs (2008) and Zhao sources with a radius ‘r’ in a plane defined by the
et al. (2006) relations. source geometry: Radius (r) depending on magnitude
A comparison of GMPE’s for stable continental of earthquake in the form [A = K1 .exp(K2 .M)].
and active crustal regions is shown in Figs. 7 and K1 and K2 are constants given by Brune (1970)
8, respectively. For comparison of different GMPE’s, (K1 = 0.003809 and K2 = 1.1513). For hazard com-
Rrup and J&B distances are converted to epicentral putations of stable continental regions (zone 2–11), we
distances (point source for smaller earthquake); CRI- include NDMA (2011), Atkinson and Boore (2006),

Fig. 10 Hazard maps of peak ground acceleration com- a seismic zonation model, b fault model, and c grid model for
puted for 10 % probability of exceedance in 50 years peak ground acceleration (PGA); see color bar for values
(return period 475 years) for different seismicity models;
Author's personal copy
J Seismol (2016) 20:629–653 645

and Pezeshk et al. (2011) in our study. We discard (2007) attenuation relation is already incorporated in
Toro (1997), since this relation results in signifi- NDMA (2011), it is not included again to avoid dupli-
cantly higher amplitudes (Fig. 7). For the active crustal cation of GMPEs. Please note that for the grid model,
region (zone 1), NDMA (2011), Akkar and Bommer due to the derivation of a single model for Peninsu-
(2010), and Chiou and Youngs (2008) GMPE’s are lar India, no differentiation into stable continental and
employed and Zhao et al. (2006) is discarded due to active crustal regions could be applied. This limitation
the significantly lower response spectrum amplitudes could effect hazard levels in the Gujarat region (zone
than other GMPE’s. Since the Raghukant and Iyengar 1) which is considered as an active crustal region.

Fig. 11 Hazard maps at spectral acceleration of 1 Hz computed for 10 % probability of exceedance in 50 years (return period
475 years) for different seismicity models; a seismic zonation model, b fault model, and c grid model; see color bar for values
Author's personal copy
646 J Seismol (2016) 20:629–653

Fig. 12 Hazard maps computed for 10 % probability of models; displayed at a PGA and b a spectral acceleration of
exceedance in 50 years (return period 475 years), using a 1 Hz; see color bar for values
logic tree approach with equal weights for all three seismicity

6 Logic tree computations various input parameters, e.g., the description of seis-
mic sources and choice of ground motion prediction
Logic trees were first introduced by Kulkarni et al. model. Figure 9 shows the structure of the logic tree
(1984) and since then have been applied frequently used in this analysis reflecting three different recur-
in probabilistic seismic hazard analysis to capture rence models used to describe the seismic sources as
and quantify epistemic uncertainties associated with well as the different GMPE models. The recurrence

Fig. 13 Hazard maps computed for 2 % probability of displayed at a PGA and b a spectral acceleration of 1 Hz; see
exceedance in 50 years (return period 2475 years), using a logic color bar for values
tree approach with equal weights for all three seismicity models;
Author's personal copy
J Seismol (2016) 20:629–653 647

Fig. 14 Hazard maps computed for 10 % probability of zonation, fault, and grid model, respectively; displayed at a
exceedance in 50 years (return period 475 years), using a PGA and b a spectral acceleration of 1 Hz; see color bar for
logic tree approach with weights of 0.4, 0.2, and 0.4 for the values

models are the seismic zonation described in Section being used for the branches. In both weight schemes,
4.1, the grid model specified in Section 4.2, and the equal weights have been used for all GMPE’s incor-
fault model described in Section 4.3. The GMPE porated in the logic tree. For the first scheme, equal
models arise from the permutations of the models weighs have been applied to the all seismic models.
selected for the stable continental and the active crust In the second scheme, different weights have been
(Section 5). Two different combinations of weights are applied to the seismic source models: 0.4, 0.2, and

Fig. 15 Hazard maps computed for 2 % probability of fault, and grid model, respectively; displayed at a PGA and b a
exceedance in 50 years (return period 2475 years), using a logic spectral acceleration of 1 Hz; see color bar for values
tree approach with weights of 0.4, 0.2, and 0.4 for the zonation,
Author's personal copy
648 J Seismol (2016) 20:629–653

0.4 for zonation model, fault model, and grid model, and active crust) giving rise to nine different combi-
respectively, due to the insufficient information avail- nations. Results are valid for hard-rock sites (VS30 ≥
able on faults. 1100 m/s). Figures are presented as contour plots for
10 and 2 % probability of exceedance in 50 years cor-
responding to return periods of 475 and 2475 years,
7 Results respectively. Individual results from each seismicity
model (zonation, fault, and grid model) are presented
Across Peninsular India, ground motion values are in Figs. 10 and 11 for PGA and 1 Hz, respectively.
estimated for PGA and spectral acceleration by using Weighted combined results are shown in Figs. 12 and
the most recent knowledge on seismic activity in the 13 for equal weights between seismicity branches and
region and the uncertainties associated with differ- Figs. 14 and 15 for the unequal weight scheme both
ent modelling parameters addressed by the formu- for PGA and 1 Hz estimated for return period of 475
lation of a logic tree (Fig. 9) consisting of three and 2475, respectively.
branches of seismicity models and three branches of In Figs. 10 and 11, the hazard maps correspond-
GMPE’s for two tectonic settings (stable continental ing to the three seismicity models show significantly

Table 7 Estimated peak ground acceleration (PGA) for 10 % probability of exceedance in 50 years at the sites of selected major cities
and comparison of values to listings of the Bureau of India Standard (BIS 2002), GSHAP (Bhatia et al. 1999), (NDMA 2011) and
other studies

City Peak ground acceleration (g)

(Latitude, Longitude) BIS GSHAP NDMA (2011) Our Study# Others

Bangalore (12.98◦ N, 77.58◦ E) 0.05 (II) 0.05 0.02 0.057 0.11 (Nath and Thingbaijam 2012),
0.10 (Jaiswal and Sinha 2007),
0.15 (Anbazhagan et al. 2009),
0.06 (Sitharam and Kolathayar 2013)
Bhuj (23.25◦ N, 69.66◦ E) 0.18 (V) 0.20 0.12 0.212 0.42 (Nath and Thingbaijam 2012),
0.25 (Jaiswal and Sinha 2007),
0.2-0.7 (Petersen et al. 2004)
Chennai (13.00◦ N, 80.18◦ E) 0.08 (III) 0.05 0.03 0.032 0.12 (Nath and Thingbaijam 2012),
0.10 (Jaiswal and Sinha 2007),
0.08 (Sitharam and Kolathayar 2013)
Hyderabad (17.45◦ N, 78.46◦ E) 0.05 (II) 0.05 0.02 0.045 0.09 (Nath and Thingbaijam 2012),
0.15 (Jaiswal and Sinha 2007),
0.03 (Sitharam and Kolathayar 2013)
Jabalpur (23.20◦ N, 79.95◦ E) 0.08 (III) 0.10 0.08 0.080 0.19 (Nath and Thingbaijam 2012),
0.15 (Jaiswal and Sinha 2007)
Kolkata (22.60◦ N, 88.45◦ E) 0.08 (III) 0.10 0.09 0.129 0.15 (Nath and Thingbaijam 2012),
0.10 (Jaiswal and Sinha 2007),
0.08 (Sitharam and Kolathayar 2013)
Koyna (17.40◦ N, 73.75◦ E) 0.12 (IV) 0.25 0.12 0.402* 0.47 (Nath and Thingbaijam 2012),
0.25 (Jaiswal and Sinha 2007)
Mumbai (19.11◦ N, 72.85◦ E) 0.08 (III) 0.10 0.09 0.067 0.16 (Nath and Thingbaijam 2012),
0.15 (Jaiswal and Sinha 2007),
0.08 (Sitharam and Kolathayar 2013)

∗ See Section 8 for detailed discussion


# Values extracted from Fig. 14a
Author's personal copy
J Seismol (2016) 20:629–653 649

different patterns reflecting model characteristics. A the numerous occurrence of earthquakes in the Koyna
common feature is the presence of the highest haz- region, their magnitudes are moderate with less poten-
ard levels in the Koyna and Gujarat regions, though. tial for large magnitude earthquakes compared to
The hazard map for the seismic zonation model Gujarat. In addition, the higher ground motion values
(Fig. 10a) shows uniform hazard inside the zones and visible in southern India for the grid model are lower
sharp changes in hazard values at their edges. The map at 1 Hz than at PGA as well.
indicates high PGA values in Koyna (0.36 g) followed In Figs. 12, 13, 14, and 15, we display our final
by Gujarat (0.16 g). It does not reflect main geological hazard maps employing two different weight schemes
structures and shows a smooth and simplified hazard (discussed in Section 6). Although the fault model was
over Peninsular India. The hazard map corresponding integrated in the logic tree with the lowermost weight,
to the fault model (Fig. 10b) clearly reflects the under- the traces of the Narmada Son lineament and the
lying fault model, therefore ground motion values Kaddam fault are discernible in the combined haz-
are minimized away from structures. Parameteriza- ard map, probably illustrating the fact that seismicity
tion was difficult to conduct for the fault systems; within the historical catalogue—constituting the foun-
the degree of fault activity was assigned based on dation of the grid model—is also concentrating along
expert judgement and is thus mirrored in the hazard these fault lines. Furthermore, we observe a slight
map. This model shows higher hazard in the Gujarat dominance of the hazard pattern resulting from the
region (0.25 g) and ground motion values compara- grid model because of the higher resulting values of
ble to the ones from the seismic zonation model in ground motion.
the Koyna region (0.36 g). The hazard map resulting In both maps (Figs. 12 and 14), the highest ground
from the grid model (Fig. 10c) shows smoothly vary- motion value (at PGA and for 10 % of exceedance
ing hazard levels reflecting the historical distribution in 50 years) is observed in the Koyna region (0.28 to
of earthquakes occurrence. However, the PGA levels 0.39 g) followed by the Gujarat region (0.15 to 0.29 g).
computed for Koyna (0.43 g) and Gujarat (0.37 g) are The peak ground acceleration (PGA) values for the
significantly higher than for both other models. most populated cities in seismically active regions are
In general, the hazard maps generated at 1 Hz reported in Table 7 along with PGA values reported
(Fig. 11) show the highest hazard within the Gujarat by previous studies in the Peninsular Indian region.
region and a lower hazard level at Koyna, since despite Uncertainties of PGA values are also estimated using

Fig. 16 Plot showing uncertainties associated with estimation of peak ground acceleration along with mean (blue circle) and median
(orange circle) values. Uncertainties are only plotted for the locations described in Table 7
Author's personal copy
650 J Seismol (2016) 20:629–653

the values obtained from different branches of the Our study shows higher hazard levels than sev-
logic tree and presented in Fig. 16. The recurrence eral earlier works (Bhatia et al. 1999; NDMA 2011)
period of large earthquakes in the stable continental particularly in regions with high earthquake potential
region is very high. Therefore, in Figs. 13 and 15, (Table 7). The National Disaster Management Author-
results are presented for 2 % of exceedance in 50 years ity (NDMA 2011) conducted a PSHA analysis for
for a return period of 2475 years as well. In both the whole Indian subcontinent. They applied a classi-
maps, for 2 % of exceedance in 50 years, PGA val- cal seismic zonation approach and divided the Indian
ues in Gujarat (0.7 to 1.07 g) and Koyna (0.81 to subcontinent into 32 seismic zones. Furthermore, they
1.09 g) region show high and comparable hazard lev- estimated GMPE relations for seven tectonic regimes,
els. The results are broadly consistent with previous three of which were adopted by the present study.
studies in terms of spatial variation. In the subsequent A potential reason for the discrepancy in results
section, the results obtained for 10 % of exceedance is that their attenuation relations scale quite differ-
in 50 years will be largely used for discussion and ently with magnitude. Our findings seem to be more
comparison with other hazard studies in the region. in agreement with the results of the GSHAP study
(Bhatia et al. 1999) except for the Gujarat and Koyna
regions. Here, the GSHAP study employs the Joyner
8 Discussion and conclusions and Boore (1981) attenuation relation without giving
any explanation.
In this study, four hazard maps corresponding to two One notable recent PSHA study has been per-
different weight schemes are presented (discussed formed by Jaiswal and Sinha (2007) using a zonefree
in Section 6), for 10 and 2 % of exceedance, as approach and incorporating ground motion prediction
examples for discussion. It is understood that there equations by Toro (1997), Atkinson and Boore (1995)
exist several more weight schemes that are equally and Iyengar and Raghukant (2004) in a logic tree
defensible and may result in different ground motion structure. A recent work by Nath and Thingbaijam
values. (2012) provides a probabilistic seismic hazard map
In all maps, the Koyna region is prominently dis- for the Indian subcontinent adapting the seismic zona-
playing high levels of ground motion due to the pres- tion approach as well, but differentiating for various
ence of reservoir induced seismicity except the map source depths along with a zonefree approach simi-
for 2 % g in 50 years. Therefore, the G-R relationship lar to Jaiswal and Sinha (2007) and applied in a logic
derived from an earthquake catalogue of lower mag- tree structure. They implemented the zonefree method
nitude may not hold true for larger magnitude events. using the grid model in CRISIS and used 4 GMPEs for
After Koyna, the Gujarat region shows the highest stable continent region (Toro 2002; Campbell 2003;
hazard level focussed around the area of Kutch. This Atkinson and Boore 2006; Raghukant and Iyengar
region has lower seismic activity compared to Koyna, 2007). Both studies deliver consistently higher hazard
but a higher potential for larger earthquakes. The max- levels over Peninsular India, (Nath and Thingbaijam
imum PGA value for this region is 0.29 g computed 2012) showing even higher ground motion values than
with the unequal weight scheme, which is considered (Jaiswal and Sinha 2007), potentially due to use of
to be reasonable with regard to the historical earth- B-C soil types.
quake catalogue. Latur indicates around 10 % g (0.09 Seeber et al. (1999) prepared a PSHA map for the
g) and this does not contradict the tragic experience of State of Maharashtra including Mumbai City applica-
the Latur earthquake (1993; MW 6.1) ble to B-C type sites recognised by NEHRP (BSSC
A number of studies have been carried out pre- 2001). Raghukant and Iyengar (2007) carried out a
viously for Peninsular India by various authors and PSHA study for Mumbai city, resulting in similar haz-
agencies (see Table 7 for a direct comparison of val- ard levels as computed by Seeber et al. (1999). Except
ues). In general, it is noted how strong variations for the GSHAP report (Bhatia et al. 1999), most stud-
between different studies may be; for example, PGA ies deliver higher PGA values for Mumbai city than
for Bhuj is given as 0.12 g by NDMA (2011), 0.2 g by the present study. Most likely, those values represent
Bhatia et al. (1999), 0.42 g by Nath and Thingbaijam an overestimation due to the inclusion of the Koyna
(2012) and as high as 0.7 by Petersen et al. (2004). region in the same zone as Mumbai.
Author's personal copy
J Seismol (2016) 20:629–653 651

The Koyna region has attracted nationwide and more work, but certainly improved the manuscript. Finally, we
international attention since the 1967 earthquake. thank the Royal Norwegian Embassy in India, New Delhi for
funding the Indo-Norwegian (EQRisk) research project, R-1-
Thereafter, the region has been carefully monitored
165. Ashish and I. A. Parvez also acknowledge ARiEES, project
leading to an extensive earthquake catalogue. While of CSIR.
there is little doubt that tectonic earthquakes are gen-
erated close to the Koyna reservoir, it is generally
agreed that reservoir-induced seismicity is the main References
cause of activities (Rastogi 1992; Gahalaut et al.
2004). So far, there is no consensus in the scien- Akkar S, Bommer JJ (2010) Empirical equations for the predic-
tific community on how to deal with triggered and tion of PGA, PGV, and spectral accelerations in Europe, the
induced seismicity within a seismic hazard study. Mediterranean Region, and the Middle East. Seismol Res
Lett 81(2):10.1785/gssrl.81.2.195
In addition, the proof that seismicity is either trig-
Anbazhagan P, Vinod JS, Sitharam TG (2009) Probabilis-
gered or induced is still under discussion (see Dahm tic seismic hazard analysis for Bangalore. Nat Hazards
et al. 2013 for a more detailed discussion on the dis- 48:145–166
crimination of seismicity). Thus, we have not tried Atkinson G, Boore D (1995) New ground motion relations for
to exclude events that were triggered or induced in eastern North America. Bull Seism Soc Am 85:17–30
Atkinson G, Boore DM. (2006) Earthquake ground-motion pre-
this study, but instead chose an additional seismic diction equations for Eastern North America. Bull Seism
zone around the Koyna region in order to limit the Soc Am 96:2181–2205. doi:10.1785/0120050245
influence of these earthquakes on the ground motion Atkinson GM, Bommer JJ, Abrahamson NA (2014) Alternative
values in surrounding areas. Nevertheless, this leads approaches to modeling epistemic uncertainty in ground
motions in probabilistic seismic-hazard analysis. Seismol
to values of peak ground acceleration that are high
Res Lett 85(6):1141–1144. doi:10.1785/0220140120
compared to the Gujarat region and the reasons for Basu S, Nigam NC (1977) Seismic risk analysis of Indian
this behavior should be kept in mind as discussed Peninsula. In: Proceedings 6th World Conf. on Earth Eng.,
above. pages 782–788, New Delhi India
In conclusion, the work described in this study adds Bȧth M (1965) Lateral inhomogeneities in the upper mantle.
Tectonophysics 2:483–514
value to the hazard maps from previous studies. A Baumbach M, Grosser H, Schmidt HG, Paulat A, Rietbrock A,
set of GMPEs have been used to describe the atten- Ramakrishna Rao CV, Solomon Raju P, Sarkar D, Mohan I
uation of different regions within Peninsular India (1994) Study of foreshocks and aftershocks of the intraplate
unlike most earlier studies employing only a single Latur earthquake of September 30, 1993, India. In: Gupta
HK (ed) Latur Earthquake, pp 33–63. New Delhi India,
attenuation relation (Bhatia et al. 1999; Menon et al. Geol Soc India Mem
2010). Multiple seismicity models used in the logic Beauval C, Scotti O, Bonilla F (2006) The role of seismicity
tree approach including area, grid, and fault models models in probabilistic seismic hazard estimation: compar-
were not considered in previous studies except for ison of a zoning and a smoothing approach. Geophys J Int
162(2):584–595
(Nath and Thingbaijam 2012).
Bhatia SC, Kumar MR, Gupta HK (1999) A probabilistic seis-
For future studies, it should be considered to mic hazard map of India and adjoining regions. Ann Geofis
include also offshore faults in a seismic hazard anal- 42:1153–1166
ysis. Unfortunately, there is little knowledge available Bilham R, Stewart IS, Vitafinzi C (1999) Slip parameters for
on potentially active fault systems and their recurrence the Rann of Kachchh, India, 16 June 1819 earthquake
quantified from contemporary accounts Coastal Tectonics,
parameters. Once the geometry of the faults, stress pp 295–319. Geol. Soc. London Spec. Publ.
field, and kinematics is known, the stress variations Bommer J, McQueen C, Salazar W, Scott S, Woo G (1998) A
caused by expected earthquakes on a particular fault case study of the spatial distribution of seismic hazard (El
can be better included into the PSHA scenarios. Salvador). Nat Hazards 18:147–168
Boore DM (1983) Stochastic simulation of high-frequency
ground motion based on seismological models of the radi-
Acknowledgments A number of colleagues have helped to ated spectra. Bull Seism Soc Am 73:1865–1894
improve our work. Particularly, we would like to thank Mario Brune JN (1970) Tectonic stress and the spectra of seismic shear
Ordaz for helping us in exploring the new capabilities of waves from earthquakes. J Geophys Res 75(26):4997–5009
CRISIS2014 and Gordon Woo for providing the KERGRID BSSC (2001) NEHRP Recommendations, Part 1: Provisions;
code and helping with implementing the grid model into CRI- prepared by the Building Seismic Safty Council for the Fed-
SIS2014. We gratefully appreciated three anonymous reviews eral Emergency Management Agency, Report No FEMA
and the detailed review of Julian Bommer, which caused us (368), Washington, D. C., USA.
Author's personal copy
652 J Seismol (2016) 20:629–653

Campbell KW (2003) Prediction of strong ground motion using Jaiswal K, Sinha R (2007) Probabilistic seismic-hazard estimation
the hybrid empirical method and its use in the development for peninsular India. Bull Seism Soc Am 91(1B):318–330
of ground-motion (attenuation) relation in eastern North Joyner WB, Boore DM (1981) Peak horizontal acceleration
America. Bull Seism Soc Am 93:1012–1033 and velocity from strong-motion records including records
Chiou BS-J, Youngs RR (2008) An NGA model for the average from the 1979 Imperial Valley, California, earthquake. Bull
horizontal component of peak ground motion and response Seism Soc Am 71:2011–2038
spectra. Earthq Spectr 24:173–215. doi:10.1193/1.2894832 Kaila KL, Rao NM (1979) Seismic zoning map of the Indian
Choubey VD (1971) Narmada-son lineament. Nature Phys Sci subcontinent. Geophys Res Bull 17:293–301
232:38–40 Khatri KN, Rogers AM, Algermissen ST (1984) A seismic
Christophersen A, Smith EGC (2000) A global model for after- hazard map of India and adjacent areas. Tectonophysics
shock behaviour. In: Proceedings of the 12th World confer- 108:93–134
ence on Earthquake Engineering, Auckland, New Zealand, Krishna J (1959) Seismic zoning of India Earthquake Eng
2000. 30th January–4th Seminar. Roorkee University, India, pp 32–38
Cornell CA (1968) Engineering seismic risk analysis. Bull Kulkarni RB, Youngs RR, Coppersmith KJ (1984) Assessment
Seism Soc Am 58(5):1583–1606 of confidence intervals for results of seismic hazard anal-
Crespo Álvarez MJ, Martinez Cutillas FJ, Marti Rodriguez J ysis. In: Proceedings of the Eighth World Conference on
(2014) Seismic hazard of the Iberian Peninsula: evaluation Earthquake Engineering, volume 1, pages 263–270, San
with kernel functions. Nat Hazards Earth Syst Sci 14:1309– Francisco
1323 Menon AT, Corigliano Ornthammarath M, Lai CG (2010) Prob-
Dahm T, Becker D, Bischoff M, Cesca S, Dost B, Fritschen abilistic seismic hazard macrozonation of Tamil Nadu in
R, Hainzl S, Klose CD, Kühn D, Lasocki S, Meier Southern India. Bull Seism Soc Am 100(3):1320–1341.
Th, Ohrnberger M, Rivalta E, Wegler U, Husen S doi:10.1785/0120090071
(2013) Recommendation for the discrimination of human- Molina S, Lindholm CD, Bungum H (2001) Probabilistic seis-
related and natural seismicity. J Seismol 17(1):197–202. mic hazard analysis: zoning free versus zoning methodol-
doi:10.1007/s10950–012–9295–6 ogy. Bollettino di Geofisica Teorica ed Applicata 42:19–
Das R, Wason HR, Sharma ML (2013) General orthog- 39
onal regression relations between body-wave and Murthy KSR, Subramanyam V, Subramanyam AS, Murty GPS,
moment magnitudes. Seismol Res Lett 84(2):219–224. Sarma KVLNS (2010) Landocean tectonics (LOTs) and
doi:10.1785/0220120125 the associated seismic hazard over the Eastern Continental
Frankel A (1995) Mapping seismic hazard in the Central and Margin of India (ECMI). Earthquake Hazards 55(2):167–
Eastern United States. Seismol Res Lett 66(4):4–8 175
Gahalaut VK, Kalpna VK, Singh SK (2004) Fault interaction
Naqvi SM, Divakara Rao V, Narain H (1974) The protocon-
and earthquake triggering in the Koyna-Warna region. India
tinental growth of Indian shield and antiquity of its rift
Geophys Res Lett 31(1–4)
valleys. Precambrian Res 1:345–398
GEM (2014) Global Earthquake Model, 2014. http://www.
Nath SK, Thingbaijam KKS (2012) Probabilistic seismic
globalearthquakemodel.org. Last accessed 22
hazard assessment of India. Seismol Res Lett 83(1).
GSI (2000) Seismotectonics Atlas of India and its Environs,
doi:10.1785/gssrl.83.1.135
Geological Survey of India
NDMA (2011) Development of probabilistic hazard map
Gubin IE (1968) Seismic zoning of Indian Peninsula. Bull
of India. http://ndma.gov.in/ndma/disaster/earthquake/
Seism Soc Am 5:109–139
PSHATechReportMarch%202011.pdf. Report
Guha SK, Gosavi PD, Varma MM, Agarwal SP, Padale JG,
Marwadi SC (1970) Recent seismic distribution in the Oldham T (1883) A catalogue of Indian earthquakes from the
Shivajisagar Lake area of the Koyna hydroelectric project, earliest times to the end of 1869 A.D. MGSi, XIX: Part. 3
Maharastra, India, comprehensive report up to June 1969. Ordaz M, Aguilar A, Arbdeda J (2014) Program for com-
Technical report, Khadakwasla, Poona-24, India puting seismic hazard. Instituto de Ingenieria, Instituto de
Guha SK (1962) Seismic regionalization of India, Sec. Symp. Ingenieria, UNAM, Mexico, Ver. 1.1.
Earthquake Eng. pp 191–207, University of Roorkee Ornthammarath T, Lai CG, Menon A, Corigliano M,
Gupta HK (1992) Reservoir-Induced Earthquakes. Elsevier Sci- Dodagoudar GR, Gonavaram K (2008) Seismic hazard at
entific Publishing, Amsterdam, p 335 the historical site of Kancheepuram in Southern India. In:
Gupta ID, Joshi RG, Rambabu V (1992) Analysis of Some Sig- The 14th World Conference on Earthquake Engineering, p
nificant Accelerograms of Koyna Dam Earthquakes Using 1–8, Beijing, October 12-17
Improved Data Processing Techniques. Technical memo- Parvez IA, Gusev AA, Panza GF, Anatoly GP (2001) Prelim-
randum,Ministry of Water Resources, Central Water and inary determination of the interdependence among strong-
Power Research Station motion amplitude, earthquake magnitude and hypocentral
Gutenberg B, Richter CF (1944) Frequency of earthquakes in distance for the Himalayan region. Geophys J Int 144:577–
California. Bull Seism Soc Am 34:185–188 596
Iyengar RN, Raghukant STG (2004) Attenuation of strong Parvez IA, Vaccari F, Giuliano FP (2003) A deterministic seis-
ground motion in peninsular India. Seismol Res Lett mic hazard map of India and adjacent areas. Geophys J Int
75(4):530–540 155:489–508
Author's personal copy
J Seismol (2016) 20:629–653 653

Petersen MD, Rastogi BK, Echweig ES, Harmsen SC, Homberg USGS (2014) Shuttle Radar Topography Mission (SRTM).
JS (2004) Sensitivity analysis of seismic hazard for the Available at http://earthexplorer.usgs.gov/. Last accessed 11
northwestern portion of the state of Gujarat. India Tectono- March 2014
physics 390(1-4):105–115 Valdiya KS (1973) Tectonic framework of India: a review
Pezeshk S, Zandieh A, Tavakoli B (2011) Hybrid empiri- and interpretation of recent structural and tectonic studies.
cal ground-motion prediction equations for Eastern North Geophys Res Bull 11:79–114
America using NGA models and updated seismologi- Van Stiphout T, Zhuang J, Marsan D (2012) Seismicity
cal parameters. Bull Seism Soc Am 101(4):1859–1870. declustering, Community Online Resource for Statistical
doi:10.1785/0120100144 Seismicity Analysis. Available at http://www.corssa.org.
Raghukant STG, Iyengar RN (2007) Estimation of seismic doi:10.5078/corssa-52382934
spectral acceleration in Peninsular India. J Earth Syst Sci Vere-Jones D (1992) Staticle methods for the description and
116(3):199–214 display of earthquake catalogues. In: Walden AT, Guttorp
Ramalingeswara Rao B, Sitapathi Rao P (1984) Historical seis- P (eds) Statistics in the Environmental and Earth Sciences,
micity of Peninsular India. Bull Seism Soc Am 74(6):2519– pp 220–246. Edward Arnold, London
2533 Verma M, Bansal BK (2013) Seismic hazard assessment and
Rastogi BK (2001) Ground deformation study of mw 7.7 bhuj mitigation in India: an overview. Int J Earth Sci 102:1203–
earthquake of 2001. Episodes 24(3):160–165 1218
Rastogi BK (1992) Seismotectonics inferred from earthquakes Verma RK, Banerjee P (1992) Nature of continental crust along
and earthquake sequences in India during the 1980s. Curr the Narmada-Son Lineament inferred from gravity and deep
Sci India 62(1–2):101–108 seismic sounding data. Tectonophysics 202(2):375–397.
Richter CF (1958) Elementary Seismology. WH Freeman & Co, doi:10.1016/0040–1951(92)90121–L
San Francisco, USA Vipin KS, Anbazhagan P, Sitharam TG (2009) Estimation of
Seeber L, Armbruster JG, Jacob KH (1999) Probabilistic assess- peak ground acceleration and spectral acceleration for south
ment of earthquake hazard for the state of Maharashtra. India with local site effects: probabilistic approach. Nat
Unpublished Report, Government of Maharashtra, Earth- Hazards 9:865–878
quake Rehabilitation Cell Mumbai Wells DL, Coppersmith KJ (1994) New empirical relation-
Singh SK, Bansal BK, Bhattacharya SN, Pacheco JF, ships among magnitude, rupture length, rupture width,
Dattatrayam RS, Ordaz M, Suresh G, Kamal, Hough SE rupture area and surface displacement. Bull Seism Soc Am
(2003) Estimation of ground motion for Bhuj (26th January 84(4):974–1002
2001. Mw = 7.6) and for future earthquakes in India. Bull Wesnousky SG (1994) The Gutenberg-Richter or characteris-
Seism Soc Am 93(1):353–370 tic earthquake distribution, which is it Bull Seism Soc Am
Singh SK, Ordaz M, Dattatrayam RS, Gupta HK (1999) A 86(6):1940–1959
spectral analysis of the May 21, 1997, Jabalpur, India earth- Wiemer S (2001) software package to analyze seismic-
quake (Mw 5.8) and estimation of ground motion from ity:ZMAP. Seismol Res Lett 72(2):374–383
future earthquakes in the Indian shield region. Bull Seism Woo G (1996) Kernel estimation methods for seismic hazard
Soc Am 86(6):1620–1630 area source modelling. Bull Seism Soc Am 86(2):353–362
Sitharam TG, Kolathayar S (2013) Seismic hazard analysis of Yanger Walling M, Mohanty WK (2009) An overview on the
India using areal sources. J Asia Earth Sci 62:647–653 seismic zonation and microzonation studies in India. Earth-
Tondon AN (1956) Zoning of India liable to earthquake dam- sci Rev 96:67–91
age. Ind J Meteorol Geophys 10:137–146 Youngs RR, Coppersmith KJ (1985) Implication of fault slip
Tondon AN, Chaudhury HM (1968) Koyna earthquake of De- rates and earthquake recurrence models to probabilistic
cember Scientific report no. 59, India Meteorol. Dept., 1968 seismic hazard estimates. Bull Seism Soc Am 75(4):939–
Toro GN (2002) Modification of the Toro et al. (1997) atten- 964
uation equations for large magnitudes and short distances. Zhao XJ, Zhang J, Asano A, Ohno Y, Oouchi T, Takahashi T,
Risk Engineering, http://www.riskeng.com/PDF/atten toro Ogawa H, Irikura K, Thio HK, Somerville PG, Fukushima
extended.pdf Y, Fukushima Y (2006) Attenuation relations of strong
Toro GN (1997) Model of strong ground motion in eastern and ground motion in Japan using site classification based
central North America: best estimates and uncertainties. on predominant period. Bull Seism Soc Am 96:898–913.
Seismol Res Lett 68:41–57 doi:10.1785/0120050122

Das könnte Ihnen auch gefallen