Bulletin of the Seismological Society of America, Vol. 92, No. 2, pp.

555–569, March 2002

Ground Motion in Delhi from Future Large/Great Earthquakes in

the Central Seismic Gap of the Himalayan Arc
by S. K. Singh, W. K. Mohanty, B. K. Bansal, and G. S. Roonwal

Abstract We estimate ground motions in Delhi from possible future large/great

earthquakes in the central seismic gap in the Himalayan arc. The closest distance
from the rupture areas of such postulated earthquakes to Delhi may be about 200
km. We have used two methods to synthesize the expected ground motions. In the
first, recordings in Delhi (three on soft sites and one on a hard site) of the 1999
Chamoli earthquake (Mw 6.5; epicentral distance, ⬃300 km), which was located in
the gap, are used as empirical Green’s functions (EGFs). The ground motion during
the target event is synthesized by random summation of the EGFs. In the second, the
stochastic method, the motions have been estimated from the expected Fourier spec-
trum of the ground motion in Delhi through the application of Parseval’s theorem
and results from random vibration theory. We apply two versions of the stochastic
method: the first assumes a point source while the second considers the source to be
finite. The predictions from the two methods are in reasonable agreement for Mw ⱕ
7.5. For Mw ⬎ 7.5 events, the finiteness of the source becomes important. Several
rupture scenarios are considered in the application of the finite-source stochastic
method. The largest ground motions are predicted in Delhi for rupture occurring
between the main boundary thrust and main central thrust and the hypocenter located
at the northeast edge of the fault. For this rupture scenario and a postulated Mw 8.0
earthquake, the maximum expected horizontal acceleration (Amax), and velocity
(Vmax) at soft sites in Delhi range between 96 and 140 gal and 8 to 19 cm/sec,
respectively. For Mw 8.5 event, the corresponding values range between 174 and 218
gal and 17 to 36 cm/sec. Amax at the hard sites are 3 to 4 times less than at the soft
sites. The differences are somewhat smaller for Vmax, which are roughly 2 to 3 times
at soft sites as compared to the hard site. The horizontal Amax and Vmax estimated by
Khattri (1999) for Mw 8.5, using a composite source model, are remarkably similar
to those estimated here. The seismic hazard in Delhi may be especially high to the
east of Yamuna river because the area is underlain by recent fluvial deposits. More
extensive earthquake recordings, microzonation studies, research on liquefaction po-
tential of the fluvial deposits, and further work on the estimation of expected ground
motions in Delhi area are urgently needed.

Introduction
The relative velocity of Indian plate with respect to the cussion). The region has experienced several great earth-
Eurasian plate near Delhi is about 5 cm/yr in the direction quakes in the past hundred years or so (1897 Assam; 1905
of N13⬚E (NUVEL-1A model of DeMets et al. [1994]). The Kangra; 1934 Bihar-Nepal; 1950 Assam). The magnitudes
collision of these continental plates results in crustal short- of these earthquakes are listed in Table 1. The Himalayan
ening along the northern edge of the Indian plate. This pro- geodynamics and the occurrence of great earthquakes are
cess has given rise to three major thrust planes (e.g., Gansser, well summarized by Seeber and Armbruster (1981), Khatttri
1964; Molnar and Chen, 1982): the Main Central Thrust (1999), and Bilham and Gaur (2000). During the last episode
(MCT), the Main Boundary Thrust (MBT), and the Main of strain release, a 750-km-long segment, which lies between
Frontal Thrust (MFT) (Fig. 1). The relative role of these the eastern edge of the 1905 rupture zone and the western
thrust planes in the Himalayan seismicity remains a matter edge of the 1934 earthquake, remained unbroken (Fig. 1).
of debate (see, e.g., Rajendran et al. [2000] for a brief dis- This segment, called the central seismic gap, continues to be

555
556 S. K. Singh, W. K. Mohanty, B. K. Bansal, and G. S. Roonwal

Figure 1. (Top) Tectonic map of the region

(modified from Seeber and Armbruster, 1981).
Hatched areas denote intensity greater than or
equal to VIII. The segment between the rupture
areas of the 1905 and 1934 earthquakes is
known as the central seismic gap. MCT, main
central thrust; MBT, main boundary thrust. Lo-
cations and focal mechanisms of 1991 Uttar-
kashi and 1999 Chamoli earthquakes are
shown. (Bottom) A closer view of the area in
the box shown in the top figure (modified from
Yu et al., 1995). Triangles, sites where ground
motions during the 1999 Chamoli were re-
corded; stars, epicenters of 1991 and 1999
earthquakes. Target rupture areas are either
centered at the Chamoli hypocenter or confined
within MBT and MCT with center at 30⬚ N,
79.2⬚ E. Rectangle, rupture area of a Mw 8.0
earthquake in the region of the Chamoli.
Dashed rectangle, area used by Khattri (1999)
to compute ground motions from a postulated
Mw 8.5 earthquake. A, B, and C indicate three
possible hypocenters.

under high strain. In 1803 and 1833 large earthquakes oc-

curred in this seismic gap but the magnitudes of these earth-
quakes were less than 8, and, hence, they were not gap-filling
events (Khattri, 1999; Bilham, 1995). Based on these con-
Table 1 siderations and a shortening rate of 20 mm/yr across the
List of Great Earthquakes in the Himalayan Arc Himalayas (Lyon-Caen and Molnar, 1985; Avouac and Tap-
ponnier, 1993; Gahalaut and Chander, 1997; Bilham et al.,
Date Magnitude 1998), Khattri (1999) has estimated the probability of oc-
12 June 1897 m 8*; Ms⬎8.2† currence of a great Mw 8.5 earthquake in the gap in the next
4 April 1905 MGR 8‡; MD 8.6; Ms 8.1|; Ms 7.5# 100 yr to be 0.59.
15 January 1934 MGR 8.3‡; MD 8.4§; Ms 8.3**; Mw 8.1†† A large/great earthquake in the central seismic gap is
15 August 1950 MGR 8.6‡; MD 8.7§; Ms 8.6** likely to cause great loss of life and severe damage to con-
*Gutenberg (1956). struction in and near the epicentral zone. This emphasizes
†
Kanamori and Abe (1979) the need for realistic estimation of ground motion from fu-
‡
Gutenberg and Richter (1954) ture earthquakes in the gap. Delhi, a city of more than 10
§
Duda (1965) million inhabitants, lies approximately 200 km from MBT
|;
Abe and Noguchi (1983a)
#
Abe and Noguchi (1983b)
and 300 km from MCT (Fig. 1). The city now extends over
**Geller and Kanamori (1977) swamps and recent fluvial deposits on the banks of Yamuna
††
Chen and Molnar (1977) river (Fig. 2). For these reasons, there is an increasing con-
Ground Motion in Delhi from Future Large Earthquakes in the Central Seismic Gap of the Himalayan Arc 557

We apply two methods to synthesize the expected

ground motions. One method uses the recordings of Chamoli
mainshock and an aftershock as empirical Green’s functions
(EGFs). The second, called the stochastic method (Hanks and
McGuire, 1981; Boore, 1983), is based on the spectrum of
the ground motion that, in turn, is described by a physically
reasonable seismological source spectrum, modified by path
and site effects. One significant advantage of the EGF
method comes from the fact that the wave propagation and
the site effects are included in the recordings. We note, how-
ever, that the EGF method as applied here assumes that the
rupture can be approximated by a point source, an assump-
tion that may be valid for Mw ⱕ 7.5 earthquakes but may
not be tenable for larger events. We apply two versions of
the stochastic method. The first assumes a point source
(henceforth called the point-source stochastic method),
whereas the second considers the source to be finite (hence-
forth called the finite-source stochastic method).
Recently, Khattri (1999) has synthesized expected
strong ground motions at nine critical sites, including Delhi,
from a postulated Mw 8.5 earthquake in the central seismic
gap. The method assumes a composite source model (Zeng
et al., 1993) and uses theoretical Green’s functions. It is not
known whether the crustal structure used in computing the
Green’s functions corresponds to a generic site in Delhi. For
this reason, the estimated ground motions by Khattri need
validation. Furthermore, Khattri limits his study to a target
event of Mw 8.5. It is clearly important to estimate ground
motions during postulated earthquakes of magnitude less
than 8.5.

Data
The digital recordings of the mainshock, which were
Figure 2. Map of Delhi showing surface geology available to us for the analysis, are summarized in Table 2.
(modified from Bhatnagar and Firozuddin, 1979) and The recordings consist of accelerograms from networks op-
stations which recorded the mainshock (Ridge Ob-
servatory [RO]; CSIR; IHC; CPCB) and the after- erated by the Department of Earthquake Engineering (DEQ),
shocks (University of Delhi, South Campus, UDSC). University of Roorkee; the Central Building Research Insti-
tute (CBRI), Roorkee; and seismograms from Ridge Obser-
vatory (RO), Delhi, a station operated by the India Meteor-
ological Department (IMD). We have also used seismograms
cern about the seismic hazard in the capital, especially after of the early aftershocks recorded at a station located in the
the Bhuj earthquake of 26 January 2001 (Mw 7.6) that caused University of Delhi, South Campus (UDSC). These after-
damage to the city of Ahmedabad located ⬃300 km from shocks are listed Table 3. Tables 2 and 3 give characteristics
the epicentral zone. The sparse strong-motion data set pres- of the recording system, the sampling rate, and the peak
ently available in Delhi region, however, makes the esti- values of the recorded accelerations (Amax) and the velocities
mation of ground motion during future earthquakes quite a (Vmax).
challenge. Two moderate earthquakes have recently oc- Since the recordings from Delhi are critical to our study,
curred in the western part of the central seismic gap: the we briefly mention some relevant features of this data set.
Uttarkashi earthquake of 19 October 1991 (Mw 6.8) and the The accelerograms of the mainshock were recorded by dig-
Chamoli earthquake of 28 March 1999 (Mw 6.5). Fortu- ital accelerographs (Kinemetrics, model K2, sampling rate
nately, the Chamoli earthquake and some its aftershocks 200 Hz) at three sites in Delhi: CSIR, Rafi Marg; IHC, Lodhi
were recorded in Delhi. In this study, we take advantage of Road; and CPCB, Arjun Nagar (Fig. 2). A description of the
these recordings in the estimation of ground motions in instrumentation and the site characteristics is given in a Cen-
Delhi from future large/great earthquakes in the western part tral Building Research of India report, henceforth referred
of the central seismic gap. to as CBRI (1998). The stations are situated on the top of
558 S. K. Singh, W. K. Mohanty, B. K. Bansal, and G. S. Roonwal

Table 2
Peak Accelerations and Velocities during the Chamoli Earthquake
Amax (gal) Vmax (cm/sec)
Distance
(R, km) H1§ H2§ Z H1 H2 Z Station

22.6 199.0 359.0 156.0 22.55 45.30 7.50 Gopeshwar*

29.8 71.0 63.0 41.00 3.180 8.940 2.72 Joshimath*
37.7 91.0 96.0 47.00 6.850 5.450 4.05 Ukhimath*
76.9 73.0 83.0 39.0 3.310 4.080 2.00 Ghansiali*
91.0 54.0 62.0 34.0 4.600 5.380 2.11 Tehri*
96.9 27.0 28.0 27.0 2.350 1.893 1.270 Almora*
96.9 5.0 6.0 11.0 0.175 0.214 0.248 Lansdowne*
101.8 54.0 64.0 23.0 3.550 4.670 1.580 Uttarkashi*
107.9 52.0 45.0 49.0 3.120 3.300 3.140 Chinyalisaur*
125.6 17.0 23.0 19.0 0.830 1.240 0.766 Barkot*
160.0 56.0 47.0 17.0 — — — Roorkee*
265.0 45.61 28.0 9.43 — — — Panipat*
268.0 13.5 — 8.23 — — — Baghpat*
291.8 10.92 8.86 5.78 1.299 0.690 0.546 CSIR, Delhi†
286.8 11.55 14.32 5.59 2.010 1.645 0.634 CPCB, Delhi†
293.3 9.67 11.41 8.23 1.745 1.240 0.673 IHC, Delhi†
287.5 2.69 3.30 2.29 0.617 0.978 0.394 Ridge Obs., Delhi‡

*Accelerograph operated by Department of Earthquake Engineering (DEQ), University of Roorkee. Data

available at 50 samples per sec.
†
Accelerograph (19 bit K2) operated by Central Building Research Institute, Roorkee. Data available at 200
samples per sec.
‡
Seismograph (RefTek 24 bit digitizer connected to 1-sec natural period L-4C-3D seismometer) operated by
India Meteorological Department (IMD). Data available at 50 samples per sec.
§
Except for DEQ stations, H1 and H2 refer to north–south and east–west component, respectively.

Table 3 in Delhi (CSIR, IHC, and CPCB) are about 4 to 5 times

Peak East–West Ground Motions during Chamoli Aftershocks greater than the value at the hard site of RO.
at University of Delhi, South Campus (UDSC)* The aftershocks were recorded at a hard site located at
M:D Hr:Min Mw Amax (gal) Vmax (cm/sec)
UDSC (Fig. 2), where a Quanterra 24-bit digitizer connected
to WRI seismometer is operated. The response of the system
03:29 08:49 3.78 0.106 0.0043 is flat for velocity in the frequency range of 0.02–20 Hz. The
03:30 15:52 3.71 0.067 0.0027
03:29 21:02† 4.63 0.331 0.0166
mainshock recording at UDSC was clipped on the S-wave.
04:06 19:37 4.63 0.265 0.0111 Only east–west and Z components of the aftershocks were
04:06 20:47 4.27 0.107 0.0055 recorded, as the north–south component malfunctioned. The
04:07 15:47 4.38 0.155 0.0073 station is located on tightly folded quarzites of Alwar series
*Recorded by a broadband seismograph (Quanterra 24-bit digitizer
(Fig. 2). We estimated the seismic moments of the after-
connected to a WRI seismometer) at 20 samples per sec. shocks from the long-period spectral level of the S-wave
†
Event used as empirical Green’s function. displacement spectra (see subsequent section).

sandy silt (Fig. 2). The bedrock lies at a depth of about 100 Synthesis of Ground Motion
m. The CPCB site, which is located to the east of Yamuna
river (Fig. 2), is also underlain by recent fluvial deposits. A A brief description of the two methods used in ground-
summary and a brief and preliminary analysis of the strong- motion synthesis is given in a following section; here we
motion data recorded during the mainshock is given in an note that the EGF method used by us requires the specifi-
IMD monograph (2000). Ridge Observatory (RO) is located cation of the seismic moment, M0, and the stress drop, Dr,
on top of quartzites of Alwar series (Fig. 2). The seismo- of both the EGF and the target event. The point-source sto-
graph consisted of 24-bit RefTek digitizer connected to a chastic method requires the specification of the stress drop
three-component L-4C-3D seismometer with a natural pe- of the target event and knowledge of the path and the site
riod of 1 Hz. effects. In the following we take M0 of the Chamoli earth-
Figures 3a and 3b show plots of Amax and Vmax versus quake as 7.7 ⳯ 1025 dyne cm, which is the scalar seismic
hypocentral distance R, respectively. The large dispersion moment reported in the Harvard CMT catalog. In the next
seen in the observed peak values is most probably due to section we use the recordings of the Chamoli earthquake to
variable local site effects. We note that the Amax at soft sites infer the parameters required in the synthesis of the ground
Ground Motion in Delhi from Future Large Earthquakes in the Central Seismic Gap of the Himalayan Arc 559

Figure 3. Observed peak ground motion (triangles) versus hypocentral distance R

during the Chamoli earthquake. Continuous curves show estimated values at hard site
for three values of stress drops (50, 100, and 200 bars), based on the point-source
stochastic model (see text). (a) Amax; (b) Vmax.

Table 4
Focal Parameters of the 28 March 1999 Chamoli Earthquake
Latitude Longitude Depth
Source (⬚N) (⬚E) (km) Mo (dyne cm) Strike Dip Rake

IMD 30.41 79.42 21 2.8 ⳯ 10 25

— — —
Harvard CMT 30.38 79.21 15 7.7 ⳯ 1025 280⬚ 7⬚ 75⬚

motion. These same parameters, along with a few others, are C = Rθϕ FP(2π )2 /(4πρβ 3 ), (2)
needed in the application of the finite-source stochastic
method. and S(f), the source acceleration spectrum, may be written
as
Chamoli Earthquake and Delhi Data S(f) = f 2 M0 ( f ), (3)
The focal parameters of the earthquake are listed in Ta-
where M0(f ) is the moment-rate spectrum. For an x2-source
ble 4. Our analysis is based on the location given by IMD
model,
and the seismic moment reported in Harvard CMT catalog.
S(f) = f 2 fc2 M0 /( f 2 + fc2 ). (4)
Source Spectrum, Q, and Dr
To estimate the stress drop, Dr, of the Chamoli earth- For Brune’s source model (Brune, 1970), f c, the corner fre-
quake and the quality factor, Q, between the source region quency, is given by
and Delhi, we analyze the spectra of the recordings at Delhi.
This analysis is based on following considerations. The far- fc = 4.9 × 106 × β (∆σ / Mo)1 / 3 , (5)
field Fourier acceleration spectral amplitude of the intense
part of the ground motion at a distance R from the source, where b is in km/sec, Mo is in dyne cm, and Dr, the stress
A(f, R), can be written as drop, is in bars. In the previous equations, b is shear-wave
velocity (3.6 km/sec), q is density (2.85 gm/cm3), Q(f) is the
A(f,R) = C S(f) e −π fR / β Q / G(R), (1) quality factor, Rh␾ is the average radiation pattern (0.55), F
is the free surface amplification (2.0), P takes into account
where the partitioning of energy in the two horizontal components
560 S. K. Singh, W. K. Mohanty, B. K. Bansal, and G. S. Roonwal

(1 / 2 ), and M0 is the seismic moment of the earthquake. this figure. We note that the observed source spectrum now
G(R) in equation (1) is the geometrical spreading term, significantly deviates from the theoretical source spectrum.
which may be taken as G(R) ⳱ R for R ⱕ Rx and G(R) ⳱ The misfit strongly suggests significant site effects at IHC,
(RRx)1/2 for R ⬎ Rx. The form of G(R) implies dominance CSIR, and CPCB between 0.4 and 10 Hz. As mentioned
of body waves for R ⱕ Rx and of surface waves for R ⬎ Rx. previously, a strong site effect at these stations is expected
Herrmann and Kijko (1983) showed that Rx is roughly twice from the subsoil profile and from the fact that horizontal Amax
the crustal thickness. In this study Rx has been taken as values at these stations are about 4 to 5 times greater than
100 km. at the hard site of RO.
Figure 4 (left) shows source displacement spectrum Our further analysis is based on high-pass-filtered rec-
(continuous curve), M0(f ), and source acceleration spectrum ords. All CBRI accelerograms were decimated to 50 Hz and
(dashed curve), f 2M0(f ), determined from the RO data using high-pass filtered at 0.2 Hz. The velocity traces were ob-
equations (1)–(3). In correcting the spectrum we have taken tained by integration of the accelerograms. The recordings
Q(f) ⳱ 508f 0.48, a relationship found by Singh et al. (1999) at the RO were first corrected for the instrumental response
for the Indian shield region. For an x2-source model, the to obtain acceleration traces. These traces were then pro-
source acceleration spectrum is flat at frequencies greater cessed in a similar fashion as the CBRI accelerograms.
than the corner frequency, f c. As seen from Figure 4 (left),
this is, indeed, the case for f ⬎ 1 Hz. This suggests that Q(f) Site Characteristics of Stations in Delhi
for the Indian shield region may also provide a reasonable Our predictions of future ground motions in Delhi rely
approximation to the Q value for the region of interest, at heavily on the recordings of the Chamoli earthquake. Thus,
least within the framework of the x2-source model. Both the it is useful to quantify the site effects at the stations where
low- and the high-frequency levels of the spectrum are well these recordings were obtained. Figure 4 (left) suggests that
fit by x2-source model with M0 ⳱ 7.7 ⳯ 1025 dyne cm and the RO station may be considered free from local site effects
f c ⳱ 0.132 Hz (corresponding to a stress drop, Dr, of 60 and may conveniently be taken as a reference site. The site
bars, equation 5) (smooth curves in Fig. 4, left). The source effect is clearly visible in Figure 5 which shows that the
displacement and acceleration spectra determined from the high-frequency ground motions (e.g., accelerations) at “soft”
recordings at IHC, CSIR, and CPCB sites are shown in Fig- sites (IHC, CSIR, CBCB) are about 3 to 4 times greater than
ure 4 (right). The x2-source spectrum corresponding to M0 that at the “hard” RO, while the difference is less for low-
⳱ 7.7 ⳯ 1025 dyne cm and Dr ⳱ 60 bars is also shown in frequency signals (e.g., velocity traces). The figure also

Figure 4. Source displacement (continuous curve) and acceleration spectra (dashed

curve), M0(f ), and f 2M0(f ), of the Chamoli earthquake from data recorded in Delhi.
Median and Ⳳ one standard deviation curves for each spectra are shown. (Left) Data
from the hard Ridge Observatory site. Note that the spectra are well fit with M0 ⳱ 7.7
⳯ 1025 dyne cm; an x2-source model, Q ⳱ 508f 0.48; and a stress drop, Dr, of 60 bars
(continuous smooth curve). This is not the case for the spectra from CSIR, IHC, and
CPCB (right), suggesting a strong local site effect at these sites.
Ground Motion in Delhi from Future Large Earthquakes in the Central Seismic Gap of the Himalayan Arc 561

shows the synthesized ground motions at the UDSC site, us- Estimation of Ground Motion in Delhi from Large/
ing an aftershock an EGF. We discuss this synthesis in a later Great Earthquakes in the Central Seismic Gap
section.
Figure 6 shows spectral ratios of the horizontal com- Point-Source Approximation
ponents at the “soft” sites with respect to RO. The spectral Random Summation of Empirical Green’s Functions. We
ratios suggest that the amplification of seismic waves at use a technique for random summation of EGF to synthesize
“soft” sites of Delhi may reach a factor of up to about 20 at ground motions from future large earthquakes proposed by
the dominant frequencies of the sites, which are roughly 1.4, Ordaz et al. (1995). This scheme obeys the x2-source scaling
1.0, and 2.2 Hz at sites CPCB, IHC, and CSIR, respectively. law at all frequencies. The method requires specification of
Except for the IHC site, the estimations of the dominant only the seismic moments and the stress drops of the EGF
frequencies obtained here agree with those given in CBRI and the target event. The details of the method are given in
(1998). Ordaz et al. (1995).

Figure 5. East–west component of acceleration and velocity traces (high-pass fil-

tered at 0.2 Hz) at sites in Delhi during the Chamoli earthquake. The simulated ground
motion at UDSC, using an aftershock of the Chamoli earthquake as an EGF, is also
shown (top traces).

Figure 6. Spectral ratios of soft sites to Ridge Observatory site during the Chamoli
earthquake. Continuous curve, north–south component; dashed curve, east–west com-
ponent. The traces were high-pass filtered 0.2 Hz before computing the spectral ratios.
562 S. K. Singh, W. K. Mohanty, B. K. Bansal, and G. S. Roonwal

We note that if only peak ground motion parameters are we choose a butterworth filter given by [1 Ⳮ (f /f m)8]ⳮ1/2.
desired, then the computation of the time histories is not In our calculations we have set f m to 15 Hz. Figure 8 shows
needed; the Fourier spectrum along with an estimation of expected Amax and Vmax for Q ⳱ 508f 0.48 and Dr ⳱ 60, 100,
duration (TR) of the intense part of the ground motion and and 200 bars. The aftershock data at UDSC and the main-
application of results from random vibration theory (RVT) shock data at RO are well explained by Dr between 50 and
suffices (see Appendix B of Ordaz et al. [1995] for relevant 200 bars.
formulas). In Figure 8 we notice that the estimated Amax value for
In the synthesis, we have taken Dr ⳱ 60 bars for both Mw ⱖ7.0 using the EGF technique (crosses) lies above the
the target and the EGF event. The duration TR in sec is given: predicted curve from the point-source stochastic method
TR ⳱ f cⳮ1 Ⳮ 0.05R, where f c is the corner frequency (equa- with Dr ⳱ 100 bars. However, the corresponding estimated
tion 5) and R is the hypocentral distance in km (Herrmann, Vmax values lie below the predicted curve for Dr ⳱ 100 bars.
1985). Our tests show that there are two reasons for this. First, the
The expected values of Amax and Vmax as function of Mw Chamoli EGF recordings have been high-pass filtered at 0.2
at all Delhi sites and for all components are shown in Figure Hz. The contribution to Vmax from low-frequency waves be-
7. For Mw 8.0, the expected Amax and Vmax on horizontal come important for large earthquakes. Signals below 0.2 Hz
components at soft sites range between 40 to 60 gal and 5 are missing in simulations using the EGF technique but not
to 13 cm/sec, respectively. The corresponding values for the in the synthesis using the point-source stochastic method.
Z component range between 18 to 40 gal, and 4 to 5 cm/sec, Second, Chamoli Amax and Vmax data cannot be simulta-
respectively. As expected, Amax at the hard site of RO is 3 neously explained by a single Dr even when the EGF is not
to 4 times less than at the soft sites. The difference is smaller high-pass filtered.
for Vmax, which is roughly half at the hard site as compared From Figure 8 we conclude that the predicted ground
to the soft sites. motions at hard sites in Delhi from the two techniques are
Figure 8 summarizes observed and estimated horizontal within a factor of about 2 of each other. It is useful to com-
Amax and Vmax as function of Mw at Delhi sites. In this figure, pare the observed peak values during the Chamoli earth-
we have included the peak values from aftershocks recorded quake over the entire distance range with the predicted ones
at UDSC (Table 3), which are shown by large plusses. The from the point-source stochastic method for a Mw 6.5 earth-
smaller plusses indicate estimated values using the after- quake. The comparison shows that, at some sites, the ob-
shock of 29 March 1999 (Mw 4.63) as the EGF (Table 3) and served horizontal Amax and Vmax values are much higher than
the random summation technique described previously. In those predicted even with Dr ⳱ 200 bars (Fig. 3a). For
these simulations we have again used Dr ⳱ 60 bars. Ex- example, this is the case at Gopeshwar, Ghansiali, Tehri,
amples of synthesized traces for a Mw 6.5 earthquake are Uttarkashi, Roorkee, Panipat, and at the three soft sites in
shown in Figure 5. The estimated Amax and Vmax values at Delhi. We attribute this difference to the effect of local sur-
UDSC for Mw 6.5 earthquake are within a factor of 2 of the face geology. As we verified earlier, this is the case of the
recorded value at RO during the Chamoli earthquake (Figs. soft sites in Delhi.
5 and 8). In Figure 8 the peak values corresponding to
Mw⬎6.5 are taken from Figure 7, which were synthesized Simulation of Ground Motion Using Finite-Source
using the Chamoli recordings as the EGFs. Stochastic Method
The estimation of ground motion above is based on the
Simulation of Ground Motion Using Point-Source Stochas- assumptions that the source follows x2 scaling in the far field
tic Method. This method was first proposed by Hanks and irrespective of the size of the earthquake and that the point-
McGuire (1981) and later extended by Boore (1983). Hanks source approximation is valid (i.e., the source dimension and
and McGuire (1981) related root mean square (rms) accel- the wavelength of interest are smaller than the distance to
eration to x2-source spectrum modified by attenuation, the observation). Rupture area, A, of an earthquake may be
through Parseval’s theorem. The expected peak amplitude is estimated from the relation: log A ⳱ Mw ⳮ 4.0, where A is
obtained from the rms amplitude using equations of the RVT in km2 (Wyss, 1979; Singh et al., 1980). Assuming that the
(Cartwright and Longuet-Higgins, 1956). Boore (1983) ex- width, W, of the fault along the Himalayan arc does not
tended these results to predict Vmax and response spectra by exceed 80 km, the length, L, for Mw 8.0 and 8.5 earthquakes
generating time series of filtered and windowed Gaussian are 125 and 400 km, respectively. As the closest distance
noise whose amplitude spectrum approximated the acceler- from the postulated earthquake to Delhi may be between 200
ation spectrum. Here, we briefly outline some relevant as- and 300 km, the point-source approximation for such earth-
pects of the method. quakes is grossly violated. To account for the finiteness of
For an x2-source model, the source spectrum of an the source, we use a modification introduced by Beresnev
earthquake is completely specified by its seismic moment and Atkinson (1997, 1998, 1999, 2001) to the point-source
and the stress drop (equations 1–5). To simulate the observed stochastic method. A description of the computer program
spectra at a site, the right-hand side of equation (1) needs to is given in Beresnev and Atkinson (1998). The fault plane
be multiplied by a high-cut filter. Following Boore (1986), is divided in subfaults whose size, Dl, in km, is given by log
Ground Motion in Delhi from Future Large Earthquakes in the Central Seismic Gap of the Himalayan Arc 563

Figure 7. Predicted peak ground motions at sites in Delhi from postulated future
large earthquakes. High-pass filtered (.02 Hz) Chamoli earthquake recordings have been
used as empirical Green’s functions (EGFs). Stress drop is 60 bars for the both the EGF
and the target event. Crosses, Ridge Observatory; triangles, IHC; open circle, CPCB;
square, CSIR.

Dl ⳱ 0.4 Mw ⳮ 2.0. The subfaults are stochastic x2 sources. In finite-source calculations we consider two locations
The subevent time history at a site is generated following of the rupture area (Fig.1b). In one, the rupture area is cen-
the procedure of Boore (1983). The rupture propagates ra- tered at the hypocenter of the Chamoli earthquake. In the
dially from a specified hypocenter. A standard technique other, this area lies between MCT and MBT, centered at
sums the contribution from each subfault. Randomness is 30.0⬚ N, 79.2⬚ E. In both cases, the center of the fault is at
introduced in the subevent rupture times. A stress parameter, a depth of 16 km. The second location corresponds to the
which relates subfault moment and its size, is fixed at 50 scenario considered by Khattri (1999).
bars. A free parameter, called the strength factor, which con- The strike and the dip of the fault have been taken as
trols the level of high-frequency radiation, needs to be spec- 300⬚ and 7⬚, respectively. These values are similar to those
ified (see Beresnev and Atkinson, 1997, 1998). In the present chosen by Khattri (1999). Simulations were performed for
case, we assume a standard earthquake (strength factor ⳱ three hypocenter locations: the northeast edge of the fault,
1.0). All other required parameters are the same as in the the center of the fault, and the center of the downdip edge
case of point-source stochastic method. The program permits of the fault (points A, B, and C in Fig. 1b). Assuming W ⱕ
site effect to be included in the computation. We assume no 80 km, the expected L ⳯ W for Mw 7.0, 7.5, 8.0, and 8.5
site effect at RO. The logarithmic average of north–south are 32 ⳯ 32 km, 56 ⳯ 56 km, 125 ⳯ 80 km, and 400 ⳯
and east–west spectral ratios (Fig. 6) between 0.4 Hz and 15 80 km, respectively. Khattri (1999) takes a rupture area of
Hz is taken as the site effect at CBCP, IHC, and CSIR. We 240 ⳯ 80 km in his simulations of the target earthquake of
assume that below 0.4 Hz and above 15 Hz the spectral ratio Mw 8.5. We performed computations for both rupture areas
has the same value as at 0.4 and 15 Hz, respectively. for this earthquake. The results do not differ significantly.
564 S. K. Singh, W. K. Mohanty, B. K. Bansal, and G. S. Roonwal

Figure 8. Observed and predicted horizontal Amax and Vmax as function of Mw at Delhi
sites. Results from both the EGF technique and the point-source stochastic method are
shown. EGF predictions are based on Dr ⳱ 60 bars; for Mw ⬎6.5 they are taken from
Figure 7. The figure also includes the peak values from aftershocks recorded at UDSC
shown by large plusses. The smaller plusses indicate predicted values using an aftershock
as the EGF (see text). Continuous curves show predictions at hard sites in Delhi from
the stochastic method with Q ⳱ 508f 0.48 and Dr values of 60,100, and 200 bars.

Table 5 summarizes Amax and Vmax values for different rup- between MBT and MCT and hypocenter at A, the predicted
ture scenarios. Each value corresponds to an average of 15 Amax and Vmax for Mw 8.0 at the hard site of RO are about
simulations. The results given in Table 5 for Mw 8.5 are 28 gal and 6 cm/sec, respectively. The corresponding values
based on a rupture area of 240 ⳯ 80 km. at soft sites range between 96 and 140 gal and 8 to 19 cm/
Figure 9 shows estimated peak ground motions from sec. For Mw 8.5, the values of Amax and Vmax at RO are about
finite-source simulations (open and solid circles). The hy- 52 gal and 13 cm/sec, respectively. The corresponding val-
pocenter is located at B, the center of the fault. The predicted ues at soft sites range between 174 and 218 gal and 17 to
values from EGF technique are also shown in the figure 36 cm/sec. It is interesting to note that horizontal Amax and
(crosses). These values should be compared with finite- Vmax estimated by Khattri (1999) for Mw 8.5 range between
source calculations for rupture area centered at the hypocen- 160 to 210 gal and 28 to 32 cm/sec, respectively. These
ter of the Chamoli earthquake (open circles). The predictions values are remarkably similar to those given in Table 5 at
from the finite-source stochastic model for Mw 6.5 earth- soft sites.
quake are in rough agreement with the observed data at all
four sites, giving us confidence that the parameters chosen Simulated Ground-Motion Time Series,
in the simulations are reasonable. Figure 9 also suggests that and Response Spectra
the predictions from the point-source EGF summation tech- Samples of simulated horizontal accelerations, as well
nique may be acceptable for target events in the Chamoli as velocities, for Mw 7.5, 8.0, and 8.5 at sites CPCB and RO
region of magnitude less that about 7.5. For larger earth- are shown in Figure 10. These computations correspond to
quakes, the finiteness of the source can not be ignored. the rupture area lying between MBT and MCT and the hy-
Table 5 shows that the expected peak ground motions pocenter located at A (Fig. 1b). The figure also illustrates
are larger for a postulated rupture area confined between the east–west component of the Chamoli earthquake record-
MBT and MCT as compared to the rupture centered at the ing. The expected horizontal pseudoacceleration response
Chamoli hypocenter. This is expected since the former rup- spectra (5% damping), Sa, from these postulated earth-
ture area is closer to Delhi than the latter one (Fig. 1b). quakes at CPCB and RO sites (average of 15 simulations)
Generally, the largest ground motions in Delhi result from are shown in Figure 11. For comparison, the figure also in-
hypocenter A, located at the northeast edge of the fault. This cludes Sa computed by Khattri (1999) for a Mw 8.5 earth-
is because of the directivity effect. For a rupture confined quake. At CPCB, Sa presented by Khattri is nearly equal to
Ground Motion in Delhi from Future Large Earthquakes in the Central Seismic Gap of the Himalayan Arc 565

Table 5
Predicted Horizontal Amax and Vmax in Delhi from Earthquakes in the Central Seismic Gap of the Himalayan Arc Using
the Finite-Source Stochastic Method

Ridge Obs. CPCB IHC CSIR

Rupture Amax Vmax Amax Vmax Amax Vmax Amax Vmax
Location* Hypocenter* Mw† (gal) (cm/sec) (gal) (cm/sec) (gal) (cm/sec) (gal) (cm/sec)

Centered at A 7.5 11.9 2.4 59.6 7.2 55.8 8.9 43.8 3.8
Chamoli 8.0 23.1 5.6 110.2 15.2 103.1 16.8 76.4 6.7
hypocenter 8.5 35.6 11.9 165.4 26.4 162.1 27.0 114.8 12.7
B 7.5 11.8 2.3 48.7 6.0 50.4 8.0 39.5 3.4
8.0 19.6 5.2 88.1 13.4 91.5 13.5 66.8 6.5
8.5 32.9 11.9 151.3 27.8 144.2 24.4 112.4 13.7
C 7.5 11.9 2.3 52.4 5.9 56.2 8.1 54.9 4.1
8.0 20.7 5.1 99.8 13.5 90.2 14.2 95.8 8.0
8.5 36.0 13.0 171.6 26.2 137.7 24.4 173.9 17.1
Between A 7.5 17.3 2.9 81.1 9.4 70.2 10.9 54.9 4.1
MBT-MCT 8.0 28.3 6.2 139.6 16.7 121.4 19.4 95.8 8.0
with 8.5 51.9 12.9 218.0 32.9 208.6 36.2 173.9 17.1
center at B 7.5 15.4 2.4 72.5 8.4 69.8 10.6 51.8 3.7
30.0⬚, 79.2⬚ 8.0 27.7 6.2 120.7 14.5 116.7 17.6 89.0 8.1
8.5 47.6 14.3 202.5 30.8 198.6 37.0 159.9 18.9
C 7.5 18.5 2.8 72.5 8.2 70.0 10.8 55.8 4.1
8.0 26.9 6.2 125.6 16.1 120.2 19.3 94.4 8.1
8.5 50.1 16.8 189.2 33.0 219.3 40.7 158.1 17.2

*See Figure 1b.

†
Rupture area computed from the relation log A ⳱ Mw ⳮ 4.0, where A is in km2. Width is not allowed to exceed 80 km. For Mw 8.5, length ⳱ 240
km, width ⳱ 80 km.

the Sa estimated here for Mw 8.5 at frequencies between 4 puted from finite-source stochastic method for several of
and 10 Hz. At lower frequencies the two spectra deviate: the rupture scenarios. A comparison of Figures 7–9 suggest that
Sa given by Khattri is smaller by a factor of 2 to 4 between point-source approximation may be valid for Mw less than
0.8 and 2 Hz as compared to our estimation. The Sa given about 7.5. Thus, Figures 7 and 8 may be useful in estimating
by Khattri clearly overestimates the expected Sa predicted Amax and Vmax for Mw ⬍ 7.5 from target events located in
here for hard sites in Delhi, such as the RO site. the Chamoli region. The largest ground motion in Delhi re-
sult from rupture areas confined between MBT and MCT,
with rupture initiating at the northeast edge of the fault. In
Discussion and Concluding Remarks this case, the calculations based on the finite-source sto-
Figures 7–9 summarize our predictions of peak ground chastic method predict horizontal Amax and Vmax at soft sites
motions in Delhi from future earthquakes in and near the in Delhi between 174 and 218 gal and 17 to 36 cm/sec,
Chamoli region of the central seismic gap. These predictions respectively for Mw 8.5 (Table 5). These values are remark-
are for two groups of sites: hard sites that may be similar to ably similar to those reported by Khattri (1999). However,
the Ridge Observatory (RO) and University of Delhi, South there are significant differences in the pseudoacceleration
campus (UDSC) sites, and soft ones whose characteristics response spectra (5% damping) (Fig. 11), reflecting the ef-
may be similar to CSIR, IHC, and CPCB sites. There must fect of local surface geology that is included in our study
be zones in Delhi that do not fall in either of the two groups. but is missing from Khattri’s estimations.
A microzonation of the region will be needed to identify Our point-source predictions are based on a stress drop
them, to determine their transfer functions with respect to a of 60 bars, and the finite-source estimations correspond to a
reference hard site, such as the RO or UDSC, and to estimate standard earthquake (see Beresnev and Atkinson, 1998). The
ground motions in these areas. estimated peak ground motions are uncertain by a factor of
The predictions in Figure 7 and 8 are based on point- at least 2 due to uncertainty in these assumptions and the
source approximation using random summation of EGF and deviation of the source spectrum from x2-source model.
the stochastic method. The results from the EGF technique These are only the model-inherent uncertainties, the total
is, of course, valid for target events located in the Chamoli uncertainty, of course, is likely to be larger.
area only. The rupture areas of future earthquakes may lie Perhaps the area of highest seismic hazard in Delhi is
closer to Delhi, delimited by the main boundary thrust the one located to the east of Yamuna river, which is un-
(MBT) and main central thrust (MCT) (Fig. 1b). Figure 9 derlain by recent fluvial deposit (Fig. 2). The CPCB site is
and Table 5 summarize expected peak ground motions com- located in this area. For postulated Mw ⱖ 8.0 earthquakes,
566 S. K. Singh, W. K. Mohanty, B. K. Bansal, and G. S. Roonwal

Figure 9. Predicted peak ground motions at

sites in Delhi from postulated earthquakes us-
ing finite-source stochastic method. The hy-
pocenter is taken at the center of fault (Fig. 1b).
Open circles, rupture areas centered at the
Chamoli hypocenter; solid circles, the rupture
confined between MBT and MCT (Fig. 1b).
Crosses at Mw 6.5 represent recorded peak mo-
tions during the Chamoli earthquakes; those
corresponding to larger magnitudes represent
predicted values using high-pass filtered (0.2
Hz) Chamoli earthquake recordings as empir-
ical Green’s functions. (a) Ridge Observatory,
(b) CPCB site, (c) IHC site, and (d) CSIR site.
Ground Motion in Delhi from Future Large Earthquakes in the Central Seismic Gap of the Himalayan Arc 567

Figure 10. Samples of simulated horizontal ground motions at CPCB and RO, using
finite-source stochastic method. Chamoli recording (east–west component) is also shown.

the predicted Amax values exceed 100 gal at CPCB, as they needed in the synthesis of ground motion during large/great
do also at IHC and CSIR. At this level of ground motion earthquakes, using different techniques.
significant liquefaction may be expected in areas underlain
by the fluvial deposits (see, e.g., Kramer, 1996, chapter 9). Acknowledgments
Clearly, there is an urgent need to record earthquakes si-
multaneously at many representative sites in the Delhi area We are grateful to Central Building Research Institute (CBRI), Roor-
and to carry out a microzonation of the city. This study and kee; Department of Earthquake Engineering (DEQ), University of Roorkee;
for providing us with the Chamoli earthquake recordings. The India Me-
the work of Khattri (1999) provide a preliminary and rough teorological Department (IMD) kindly made the seismograms recorded at
estimation of ground motions in Delhi from future large Ridge Observatory, Delhi, available to us. We thank I. Beresnev for pro-
earthquakes in and near the Chamoli region. More work is viding us the finite-source program; K. N. Khattri, R. K. Midha, and H. K.
568 S. K. Singh, W. K. Mohanty, B. K. Bansal, and G. S. Roonwal

Figure 11. Pseudoacceleration response spectra (5% damping), Sa, at CPCB and
RO sites obtained from finite-source stochastic method for Mw 7.5, 8.0, and 8.5. For
comparison, the Sa curves (N–S and E–W components) reported by Khattri (1999) for
an Mw 8.5 earthquake are also shown.

Gupta for their interest and encouragement during the course of this re- Bilham, R., F. Blume, R. Bendick, and V. K. Gaur (1998). The geodetic
search; Mario Ordaz for fruitful discussions; Joan Gomberg for a careful constraints on the translation and deformation of India: implications
revision of the manuscript; and Javier Pacheco and Lars Ottemoeller for for future great Himalayan earthquakes, Current Sci. 74, 213–229.
helping us in casting some of the data in a form that we could easily use. Boore, D. M. (1983). Stochastic simulation of high-frequency ground mo-
Comments by G. Atkinson and an anonymous reviewer forced us to con- tions based on seismological models of radiated spectra, Bull. Seism.
sider some important issues that we had previously ignored. V. K. Gahalaut Soc. Am. 73, 1865–1884.
participated in the early stage of the research. Lourdes Godinez helped us Boore, D. M. (1986). The effect of finite bandwidth on seismic scaling
in preparing some figures. The research was partially supported by DGAPA, relationships, in Earthquake Source Mechanics, S. Das, J. Boatwright,
UNAM Project IN109598. and C. Scholz (Editors), American Geophysical Monograph 37, 275–
283.
Brune, J. N. (1970). Tectonic stress and the spectra of seismic shear waves
References from earthquakes, J. Geophys. Res. 75, 4997–5009.
Cartwright, D. E., and M. E. Longuet-Higgins (1956). The statistical dis-
Abe, K., and S. Noguchi (1983a). Determination of magnitudes of large
tribution of maxima of a random function, Proc. R. Soc. London Ser.
shallow earthquakes, 1897–1912, Phys. Earth Planet. Interiors 32,
A. 237, 212–232.
45–59.
Central Building Research Institute (CBRI) (1998). Strong Motion Seismic
Abe, K., and S. Noguchi (1983b). Revision of magnitudes of large shallow
earthquakes, 1897–1912, Phys. Earth Planet. Interiors 33, 1–11. Instrumentation in and around Delhi Region, Project Report, Central
Avouac, J., and P. Tapponnier (1993). Kinematic model of active defor- Building Research Institute, Roorkee, Report No. G(S) 012.
mation in central Asia, Geophys. Res. Lett. 20, 895–898. Chen, W. P., and P. Molnar (1977). Seismic moments of major earthquakes
Beresnev, I. A., and G. Atkinson (1997). Modelling finite fault radiation and the average rate of slip in central Asia, J. Geophys. Res. 82, 2945–
from the x(??)” spectrum, Bull. Seism. Soc. Am. 87, 67–84. 2969.
Beresnev, I. A., and G. Atkinson (1998). FINSIM: a Fortran program for DeMets, C., R. G. Gordon, D. F. Argus, and S. Stein (1994). Effect of
simulating stochastic acceleration time histories from finite faults, recent version to the geomagnetism reversal time scale on estimates
Seism. Res. Lett. 69, 27–32. of current plate motions, Geophys. Res. Lett. 21, 2191–2194.
Beresnev, I. A., and G. Atkinson (1999). Generic finite-fault model for Duda, S. (1965). Secular seismic energy release in the circum-Pacific belt,
ground motion prediction in eastern North America, Bull. Seism. Soc. Tectonophysics 2, 409–452.
Am. 89, 608–625. Gahalaut, V. K., and R. Chander (1997). On interseismic elevation changes
Beresnev, I. A., and G. Atkinson (2001). Subevent structure of large earth- and strain accumulation for great thrust earthquakes in the Nepal Him-
quakes: a ground-motion perspective, Geophys. Res. Lett. 28, 53–56. alaya, Geophys. Res. Lett. 24, 1011–1014.
Bhatnagar, N. C., and T. G. Firozuddin (1979). Hydrologic conditions and Gansser, A. (1964). Geology of the Himalayas, Interscience, New York,
possibilities of further ground water development in the Union Ter- 289.
ritory of Delhi, unpublished CGWB report, 39. Geller, R. J., and H. Kanamori (1977). Magnitude of great shallow earth-
Bilham, R. (1995). Location and magnitude of the Nepal earthquake and quakes from 1904–1952, Bull. Seism. Soc. Am. 67, 587–598.
its relation to the rupture zones of the contiguous great Himalayan Gutenberg, B. (1956). Great earthquakes 1896–1903, EOS 37, 608–614.
earthquakes, Current Sci. 69, 101–128. Gutenberg, B., and C. F. Richter (1954). Seismicity of the Earth and As-
Bilham, R., and V. K. Gaur (2000). The geodetic contribution to Indian sociated Phenomena, Second Ed., Princeton University Press, Prince-
seismotectonics, Current Sci. 79, 2000. ton, New Jersey.
Ground Motion in Delhi from Future Large Earthquakes in the Central Seismic Gap of the Himalayan Arc 569

Hanks, T. C., and R. K. McGuire (1981). The character of high-frequency spectral analysis of the May 21, 1997, Jabalpur, India earthquake (Mw
strong ground motion, Bull. Seism. Soc. Am. 71, 2071–2095. ⳱ 5.8) and estimation of ground motion from future earthquakes in
Herrmann, R. B. (1985). An extension of random vibration theory estimates the Indian shield region, Bull. Seism. Soc. Am. 89, 1620–1630.
of strong ground motion at large distances, Bull. Seism. Soc. Am. 75, Wyss, M. (1979). Estimating maximum expectable magnitude of earth-
1447–1453. quakes from fault dimensions, Geology 7, 336–340.
Herrmann, R. B., and A. Kijko (1983). Modelling some empirical vertical Yu, G., K. N. Khattri, J. G. Anderson, J. N. Brune, and Y. Zeng (1995).
component Lg relations, Bull. Seism. Soc. Am. 73, 157–171. Strong ground motion from the Uttarkashi, Himalaya, India, earth-
India Meteorological Department (IMD) (2000). Chamoli earthquake of quake: comparison of observations with synthetics using the com-
March 29, 1999 and its aftershocks, a consolidated document, in Seis- posite source model, Bull. Seism. Soc. Am. 85, 31–50.
mology, Meteorological Monograph, India Meteorological Depart- Zeng, Y., J. G. Anderson, and G. Yu (1993). A composite source model
ment, Government of India, New Delhi, no. 2. for computing realistic synthetic strong motions, Geophys. Res. Lett.
Kanamori, H., and K. Abe (1979). Reevaluation of the turn-of-the-century 21, 725–728.
seismicity peak, J. Geophys. Res. 84, 6131–6139.
Khattri, K. N. (1999). An evaluation of earthquakes hazard and risk in Instituto de Geofı́sica, U.N.A.M.
northern India, Himalayan Geology 20, 1–46. Ciudad Universitaria
Kramer, S. L. (1996). Geotechnical Earthquake Engineering, Prentice Hall, 04510 México, DF, Mexico
New Jersey, 653. krishna@ollin.igeofcu.unam.mx
Lyon-Caen, H., and P. Molnar (1985). Gravity anomalies, flexure of the (S.K.S.)
Indian plate, and structure, support, and evolution of the Himalaya
and Ganga basin, Tectonics 4, 513–538. Department of Civil Engineering
Molnar, P., and W. P. Chen (1982). Seismicity and mountain building, in Indian Institute of Technology
Mountain Building Processes, K. Hsu (Editor), Academic, New York, Hauz Khas, New Delhi 110016, India
41–57. (W.K.M.)
Ordaz, M., J. Arboleda, and S. K. Singh (1995). A scheme of random
summation of an empirical Green’s function to estimate ground mo- Department of Science and Technology
tions from future large earthquakes, Bull. Seism. Soc. Am. 85, 1635– Government of India
1647. New Mehrauli Road
Rajendran, K., C. P. Rajendran, S. K. Jain, C. V. R. Murty, and J. N. Arlekar New Delhi, 110016, India
(2000). The Chamoli earthquake, Garhwal Himalaya: field observa- (B.K.B.)
tions and implications for seismic hazard, Current Sci. 78, 45–51.
Seeber, L., and J. G. Armbruster (1981). Great detachment earthquakes Centre for Geosources
along the Himalayan arc and long-term forecasting, in Earthquake University of Delhi, South Campus
Prediction: An International Review, Maurice Ewing Series 4, Amer- Benito Juarez Road
ican Geophysical Union, Washington, D.C., 259–277. New Delhi 110021, India
Singh, S. K., E. Bazan, and L. Esteva (1980). Expected earthquake mag- (G.S.R.)
nitude from a fault, Bull. Seism. Soc. Am. 70, 903–914.
Singh, S. K., M. Ordaz, R. S. Dattatrayam, and H. K. Gupta (1999). A Manuscript received 20 March 2001.