Pure Appl. Geophys.
Ó 2011 Springer Basel AG
DOI 10.1007/s00024-011-0339-6
Pure and Applied Geophysics
Homogenization of Earthquake Catalog for Northeast India and Adjoining Region
RANJIT DAS,1 H. R. WASON,1 and M. L. SHARMA1
Abstract—A catalog for northeast India and the adjoining
region for the period 1897–2009 with 4,497 earthquakes events is
compiled for homogenization to moment magnitude Mw,GCMT in
the magnitude range 3–8.7. Relations for conversion of mb and Ms
magnitudes to Mw,GCMT are derived using three different methods,
namely, linear standard regression, inverted standard regression
(ISR) and orthogonal standard regression (OSR), for different
magnitude ranges based on events data for the catalog period
1976–2006. The OSR relations for Ms to Mw,GCMT conversion
derived in this paper have significantly lower errors in regression
parameters compared to the relations reported in other studies.
Since the number of events with magnitude C7 for this region is
scanty, we, therefore, considered whole India region to obtain the
regression relationships between Mw,GCMT and Ms,ISC. A relationship between Mw,GCMT and Mw,NEIC is also obtained based on 17
events for the range 5.2 B magnitude B 6.6. A unified homogeneous catalog prepared using the conversion relations derived in
this paper can serve as a reference catalog for seismic hazard
assessment studies in northeast India and the adjoining region.
Key words: Moment magnitude, catalog, homogenization,
orthogonal regression, Northeast India.
1. Introduction
Earthquake catalogs describing the historical
seismicity of a seismic region in general, are heterogeneous in magnitude types, whereas a homogeneous
earthquake catalog is a basic requirement for studying
the earthquake occurrence patterns in space and time
and seismic hazard estimates for any seismic region. In
order to compile a homogeneous earthquake catalog
for a seismic region, the regression relations used for
conversion of different magnitude types to a preferred
magnitude scale are of critical importance since any
1
Department of Earthquake Engineering, Indian Institute of
Technology Roorkee, Roorkee, India. E-mail: ranjit244614@
gmail.com
bias introduced during the conversion process propagates errors in the parameters of the frequency
magnitude distribution and consequently in the seismic
hazard estimates. A majority of such regression relations are, however, derived based on the assumption
that one of the magnitudes (independent variable) is
error free. When both the magnitude types contain
measurement errors, the use of the standard leastsquares regression procedure is found to be inadequate.
In such a case, the use of orthogonal regression analysis
is more appropriate to estimate regression relationships
between different magnitude types (CASTELLARO et al.,
2006).
Some reported regression relations also make use
of other approximations, such as taking averages of
conversions from mb and Ms to moment magnitude,
which limits the accuracy of the converted magnitudes (DAS and WASON, 2010). Because of the
inherent limitations of different magnitude scales in
accurately representing the size of an earthquake and
the fact that they also tend to saturate at higher
magnitude levels, it is better to use the moment
magnitude Mw. As is well known, Mw has two main
advantages over other magnitudes; firstly, Mw is
physically meaningful because it is derived from
seismic moment which is directly related to earthquake source physics (slip, fault area, rigidity) and
secondly, the Mw scale does not saturate for large
earthquakes which is the limitation of all other
magnitude scales. In this study, we also adopt Mw as
the homogeneous size estimate requiring assignment
of Mw to all events of the catalog through appropriate
magnitude conversions.
Northeast India and the adjoining region is seismically one of the most active regions in the world.
The seismicity in this region is considered to be
related to the collision of the Indian plate with Tibet
in the north and the Burmese landmass towards the
R. Das et al.
Pure Appl. Geophys.
east. This region has witnessed two great earthquakes
in the recent past. The Shillong earthquake with Mw
8.1 (BILHAM and ENGLAND, 2001) on 12 June 1897 and
the Assam earthquake with Mw 8.7 (THINGBAIJAM
et al., 2008) on 15 August 1950, caused extensive
damage and destruction over a very large area.
Available earthquake catalogs for northeast India and
the adjoining region covering the seismicity from
historical times (1897 onwards) to the period up to
1962 are heterogeneous in magnitude types (e.g.,
GUPTA et al., 1986; CHANDRA, 1992).
Recently, THINGBAIJAM et al. (2008) obtained
generalized orthogonal regression (GOR) relationships for conversion of Ms,ISC and mb,ISC to Mw,GCMT.
As significant dispersion was observed in the scaling
relation between mb,ISC and Mw,GCMT, mb,ISC magnitudes were first scaled to Ms,ISC and subsequently
converted to Mw,GCMT. In another study for this
region, standard regression (SR) relations have been
reported by YADAV et al. (2009) for conversion of Ms
and mb magnitudes to Mw,GCMT, but their Ms,ISC and
Ms,NEIC conversion relations differed significantly.
In this study, regression relations for conversion
of mb and Ms magnitudes to Mw,GCMT using OSR, SR
and ISR approaches have been derived. Earthquake
occurrence data for 4,497 events in the Mw magnitude
range 3.0–8.7 pertaining to the study region (lat. 20°–
32°N and long. 87°–100°E) for the period
1897–2009, combining the historical and instrumental periods, has been compiled from ISC, NEIC and
GCMT databases. A unified homogeneous catalog
prepared using the conversion relations derived in
this study can serve as a reference catalog for seismic
hazard estimates and other seismicity studies in
northeast India and the adjoining region.
folded belt in the east and also the uplift of Shillong
plateau. The eastern Himalayas and the Arakan-Yoma
mountain arc meet and define the Assam syntaxis
encasing between them the upper Assam petroliferous
Tertiary basin. The Shillong massif stands out as a
plateau with an average elevation of 1,500 m at the
SW mouth of this basin. The Shillong massif was the
seat of the great earthquake of 12 June 1897
(Mw = 8.7). On 15 August 1950, another great earthquake (Mw = 8.1) occurred further northeast in the
vicinity of the India–China border. These great earthquakes are considered to be the responses of the India–
Asia convergence zone to their continued relative
motion (e.g., SEEBER and ARMBRUSTER, 1981; KHATTRI
and TYAGI, 1983; MOLNAR, 1987; MOLNAR and PANDEY,
1989). Further, this region has experienced several
damaging large earthquakes with magnitudes [7.0.
In this study, earthquake occurrence data for the
period 1897–2009 has been compiled from different
sources. For the historical seismicity period
1897–1962, events are taken from the catalog by
GUPTA et al. (1986). Those earthquake events which
do not have any specific magnitude unit assigned in
this catalog are taken as Ms,ISC following THINGBAIJAM
et al. (2008) and YADAV et al. (2009). For the period 1964 to May 2007, events data has been
compiled from International Seismological Center
(ISC), UK (http://www.isc.ac.uk/search/Bulletin),
National Earthquake Information Center (NEIC),
USGS, USA (http://neic.usgs.gov/neis/epic/epicglobal.htm) and HRVD (HRVD is presently addressed as GCMT http://www.globalcmt.org/CMTsearch.
html) earthquake data bulletins. Data for the year
1963 has been adopted from International Seismological Summary (ISS).The complete catalog period
(1897–2009) contains a total of 4,497 events out of
which 14 events have M [ 7 and 223 events are with
M C 6. Conversion relations have been developed
using the data for the period 1964–2006 only. The
seismicity of the region for Mw magnitudes in the
range 3–8.7 is shown in Fig. 1.
2. Study Region and Data Sources
Northeast India and the adjoining region encompasses a very active seismic region bounded by
latitudes 20°–32°N and longitudes 87°–100°E. The
seismicity in this region is related to the collision of
the Indian Plate with Tibet towards the north and the
Burmese landmass towards the east. This collision
resulted in the formation of the Himalaya thrust front
in the north, Arakan-Yoma, Naga Hills and Tripura
3. Regression Procedures
For seismological applications including homogenization of earthquake catalogs, it is important to
Homogenization of Earthquake Catalog
Figure 1
A seismotectonic map of Northeast India and Adjoining Region on GIS platform depicting seismicity for Mw C 3.0 from the earthquake
catalog prepared in this study
know how different magnitude determinations compare with each other and the associated measurement
errors. It is widely in use to assume one of the
magnitudes to be error free, and thus obtain regression conversion relationships using standard linear
least-squares approach. Inverted standard leastsquares regression is similar to the standard leastsquares regression but instead minimizes the
horizontal offsets to the best fit line. In this regression
the role of the dependent and independent variables
gets reversed.
When both the magnitude types have measurement errors, it is more appropriate to use OSR
approach (CASTELLARO et al., 2006, 2007; THINGBAIJAM et al., 2008; RISTAU, 2009). However, this
regression procedure requires the knowledge of the
R. Das et al.
Pure Appl. Geophys.
error variance ratio between the two magnitude types.
An advantage of the OSR approach is that the computations yield predicted values for both the
variables. The procedure for orthogonal standard
regression is described in detail in the literature
(MADANSKY, 1959; FULLER, 1987; KENDALL and
STUART, 1979; CARROLL and RUPPERT, 1996) and is not
included here.
4. Magnitude Conversion Relations
For conversion of surface and body wave magnitudes to moment magnitudes, we have considered
events in the period 1964–2006 in specified magnitude ranges. Ms,ISC and Ms, NEIC values with focal
depths B50 km only are considered for deriving the
conversion relations. OSR relation for conversion of
surface and body wave magnitudes to moment magnitudes requires the value of the error variance ratio
g. For the events data considered in this study, we use
the standard deviations 0.15, 0.20 and 0.25 for Mw,
Ms and mb, respectively as considered by THINGBAIJAM
et al. (2008). Software developed by the authors in
another study (DAS et al., 2011) has been used to
derive the regression relations.
4.1. Relationship between MS,ISC/MS,NEIC
and Mw, GCMT
For the regression relations between Ms,ISC and
Mw,GCMT we used 124 events (compared to 114
considered by YADAV et al., 2009) within the range
4.1 B Ms, ISC B 6.9, and the derived OSR, SR and
ISR conversion relations are given in Table 1 and
Fig. 2a. Similarly, for conversion of Ms,NEIC to
Mw,GCMT, we considered 97 events (compared to 16
considered by YADAV et al., 2009) for the range
4.4 B Ms, NEIC B 6.3, and the derived OSR, SR and
ISR conversion relations are given in Table 1 and
Fig. 2b.
The OSR and SR conversion relations obtained
above have lesser uncertainties in the regression
parameters compared to the corresponding OSR
relation obtained by THINGBAIJAM et al. (2008) and
the SR relations of YADAV et al. (2009).
Since the higher magnitude events in this region
for the year 1964–2006 are scarce, we instead
considered the whole India region for the range
7.0 B magnitude B 8.7. Based on 13 events, the
regression relationships between Mw,GCMT and Ms,ISC
are obtained as given in Table 1 and Fig. 2c. The Mw
estimates given by the OSR and SR relationships are
Table 1
Regression parameters for different magnitude conversion relations
Regression relation
Magnitude range
Slope
Error (slope)
Intercept
Error (intercept)
R2
r (SD)
Ms, ISC to Mw, GCMT (OSR), g = 1
Ms,ISC to Mw,GCMT (SR)
Ms,ISC to Mw,GCMT (ISR)
Ms,NEIC to Mw,GCMT(OSR), g = 1
Ms,NEIC to Mw,GCMT(SR)
Ms,NEIC to Mw,GCMT(ISR)
Ms,ISC to Mw,GCMT (OSR), g = 1
Ms,ISC to Mw,GCMT(SR)
Ms,ISC to Mw,GCMT(ISR)
mb,ISC to Mw,GCMT (OSR), g = 0.36
mb,ISC to Mw,GCMT (SR)
mb,ISC to Mw,GCMT (ISR)
mb,NEIC to Mw,GCMT(OSR), g = 0.36
mb,NEIC to Mw, GCMT(SR)
mb,NEIC to Mw,GCMT(ISR)
Mw,NEIC to Mw,GCMT(OSR,) g = 1
Mw,NEIC to Mw,GCMT(SR)
Mw,NEIC to Mw,GCMT(ISR)
4.1 B Ms,ISC B 6.9
0.638
0.608
1.397
0.66
0.623
1.375
1.42
1.324
0.708
1.40
1.060
0.688
1.37
0.983
0.707
0.998
0.962
0.966
0.0006
0.025
0.053
0.0007
0.03
0.057
0.05
0.104
0.055
0.0043
0.05
0.032
0.006
0.054
0.039
0.014
0.069
0.069
2.20
2.355
-2.518
2.13
2.281
-2.412
-3.35
-2.628
2.343
-1.98
-0.151
1.537
-1.77
0.216
1.473
0.03
0.232
0.177
0.016
0.119
0.291
0.0201
0.135
0.314
2.73
0.793
0.415
0.122
0.263
0.177
0.1567
0.286
0.211
0.456
0.387
0.389
0.85
0.109
0.146
0.151
0.09
0.125
0.13
0.108
0.175
0.18
0.17
0.238
0.21
0.133
0.201
0.2
0.055
0.106
4.4 B Ms,NEIC B 6.3
7.0 B Ms,ISC B 8.7
4.7 B mb,ISC B 6.6
4.6 B mb,NEIC B 6.8
5.2 B Mw,NEIC B 6.2
0.86
0.93
0.73
0.69
0.92
Homogenization of Earthquake Catalog
Figure 2
Regression relations (OSR, SR, ISR) for different magnitude conversions; a Ms,ISC | Mw,GCMT relation, b Ms,NEIC | Mw,GCMT relation, c Ms,ISC |
Mw,GCMT relation (for larger events), d mb,ISC | Mw,GCMT relation, e mb,NEIC | Mw,GCMT relation, f Mw,NEIC | Mw,GCMT relation
found to be much closer to the directly measured Mw
values compared to those given by the global relation
of SCORDILIS (2006), and hence should be preferred
for conversion of regional events. These relations
also match with the global trend between Ms and Mw
as given by HEATON et al. (1986).
R. Das et al.
Pure Appl. Geophys.
The square root of the error variance ratio
between Ms and Mw,GCMT measurements is found to
lie between 0.7 and 1.8 for the events data considered
in this study. However, we prefer OSR relationships
with g = 1, which yields better results as suggested
by CASTELLARO and BORMANN (2007).
However, standard linear regression is not a good
estimator when both the dependent and independent
magnitude variables contain errors of non-negligible
size. OSR relations for conversion of mb and Ms
magnitudes to Mw,GCMT for this region are reported
only in one study (THINGBAIJAM et al., 2008). As
significant dispersion was observed in the scaling
relation between mb,ISC and Mw,GCMT, mb,ISC magnitudes were first scaled to Ms,ISC and subsequently
converted to Mw,GCMT. The main purpose of the
present study is to fill up the gap for reliable OSR
relations towards catalog homogenization for northeast India and the adjoining region.
In this study, earthquake occurrence data for
4,497 events pertaining to northeast India and the
adjoining region (lat. 20°–32°N and long. 87°–
100°E) for the time period 1897–2009, combining
the historical and instrumental periods, has been
compiled from ISC, NEIC and GCMT databases.
The seismicity of the study area (Mw C 3) is shown
in Fig. 1.
Ms,ISC to Mw,GCMT and Ms,NEIC to Mw,GCMT conversion relations have been derived for magnitude
ranges 4.1 B Ms, ISC B 6.9 and 4.4 B Ms,NEIC B 6.3,
respectively. The quality of OSR relations obtained is
found to be better than the corresponding SR relations
obtained by YADAV et al. (2009) as the standard
deviation and the absolute average of the differences
between the directly measured and the Mw,GCMT
values estimated by the OSR are relatively smaller
compared to the SR relation. Also, the regression
relations for Ms conversion are found to be in good
agreement as expected (UTSU, 2002; SCORDILIS, 2006;
DAS and WASON, 2010).
The data for strong earthquakes with Ms,ISC C 7
for the region being scarce, OSR and SR conversion
relations are instead derived for the whole Indian
region using 13 data points. The derived relations are
the first such relations for conversion of large Ms
events for the Indian region and provide better
Mw,GCMT estimates compared to the global relation of
SCORDILIS (2006). These relationships also match well
with the global trend of Ms and Mw (HEATON et al.,
1986). Further, regression relationships between
Mw,GCMT and Mw,NEIC are also obtained which show
that the Mw estimates by GCMT and NEIC can be
treated as equivalent.
4.2. Relationship between mb,ISC/mb,NEIC
and Mw, GCMT
With respect to mb determinations by ISC and
NEIC, it is observed in several studies that the two
determinations are not equivalent (NATH and THINGBAIJAM, 2010; DAS et al., 2011). For the conversion of
mb,ISC to Mw,GCMT in the magnitude range
4.7 B magnitude B 6.6, and mb,NEIC to Mw,GCMT in
the magnitude range 4.6 B magnitude B 6.8, we
follow the same approach using data sets of 171
and 149 events, respectively, for the period
1964–2006.
We derived OSR by taking g = 0.36 and also SR,
ISR relations, but the ISR relation is preferred for
conversion of mb to Mw,GCMT since ISR is found to
perform better compared to OSR and SR relations if
pffiffiffi
g\ 0:7 (CASTELLARO and BORMANN, 2007). The
regression parameters obtained for these conversion
relations are given in Table 1 and the plots are shown
in Fig. 2d, e, respectively.
4.3. Relationship between Mw,GCMT and Mw,NEIC
In order to obtain regression relationships
between Mw,NEIC and Mw,GCMT, we considered 17
events for the range 5.2 B magnitude B 6.6, during
the period 1964–2006. The OSR, SR and ISR
relationships between Mw,GCMT and Mw,NEIC are
given in Table 1 and the plot is shown in Fig. 2f.
The OSR relation obtained is matches well with the
global relation given by SCORDILIS (2006).
5. Discussion and Conclusions
The regression relations available for the study
region for conversion of different earthquake magnitude types to the unified Mw,GCMT are mostly based
on the standard linear least-squares (SR) approach.
Homogenization of Earthquake Catalog
Further, ISR, OSR and SR relations have been
derived for mb,ISC to Mw,GCMT conversion using 171
events data for the magnitude range 4.7 B mb,ISC B
6.6. Similarly, mb,NEIC to Mw,GCMT conversion relations for the range 4.6 B mb,NEIC B 6.8, are based on
149 events data. The derived relations have significantly lower errors in their regression parameters than
the corresponding relations derived by YADAV et al.
(2009) and THINGBAIJAM et al. (2008). We prefer ISR
relation for conversion of mb to Mw,GCMT., since
pffiffiffi
g\ 0:7: No regression relation is attempted for
smaller body wave magnitudes, i.e. mb \ 4.6, as the
data is very scarce. No conspicuous nonlinearity is
noticed in the derived OSR relation as observed
in some other studies (CASTELLARO et al., 2006;
THINGBAIJAM et al., 2008).
This study presents improved OSR and ISR conversion relations for the study region as evidenced by
lower uncertainties in the regression parameters. A
homogenized catalog prepared using these relations
shall imply lower uncertainties in unified moment
magnitude estimates, which is an important input for
seismic hazard estimation and other seismological
studies.
Acknowledgments
The critical review and constructive suggestions
given by the reviewer helped improve the manuscript
significantly. The first author is thankful to MHRD,
Govt. of India for research fellowship and AICTE,
Govt. of India for award of National Doctoral
Fellowship.
REFERENCES
BILHAM, R. and ENGLAND, P. (2001), Plateau ‘pop up’ in the great
1897 Assam earthquake, Nature. 410, 806-809.
CARROLL, R. J. and RUPPERT, D. (1996), The use and misuse of
orthogonal regression in linear errors-in-variables models, Am.
Stat. 50, no. 1, 1–6.
CASTELLARO, S., MULARGIA, F., and KAGAN,Y.Y. (2006), Regression
problems for magnitudes, Geophys. J. Int. 165, 913-930.
CASTELLARO, S. and BORMANN, P. (2007), Performance of different
regression procedures on the magnitude conversion problem,
Bull. Seism. Soc. Am. 97, 1167-1175.
CHANDRA, U. (1992), Seismotectonics of Himalaya, Current Science, 62 (1&2), 40-71.
DAS, R. and WASON, H. R. (2010), Comment on ‘‘A homogeneous
and complete earthquake catalog for Northeast India and the
adjoining region’’, Seism.Res.Lett. 81, 232-234.
DAS, R., WASON, H. R. and SHARMA, M.L. (2011), ‘‘Global
regression relations for conversion of surface wave and body
wave magnitudes to moment magnitude’’, Natural Hazards
(accepted for publication). doi:10.1007/s11069-011-9796-6.
FULLER, W. A. (1987), Measurement Error Models, Wiley, New
York.
GUPTA, H.K., RAJENDRAN, K. and SINGH, H.N. (1986), Seismicity of
the North-East India region, Part I: the data base, J. Geol. Soc.
India, 28, 345-365.
HEATON, T., TAJIMA, F. and MORI, A.W. (1986), Estimating ground
motions using recorded accelerograms, Surv. Geophys. 8, 25-83.
KENDALL, M. G. and STUART, A. (1979), The Advanced Theory of
Statistics, (Vol. 2, 4th ed.), Griffin, London.
KHATTRI, K. N. and TYAGI, A.K. (1983), Seismicity patterns in the
Himalayan plate boundary and identification of the areas of high
seismic potential, Tectonophysics, 96, 281-297.
MADANSKY, A. (1959), The fitting of straight lines when both
variables are subject to error, Am. Statist. Assoc. J. 54, 173-205.
MOLNAR, P. (1987), The distribution of intensity with the great
Assam earthquake and bounds on the extent of the rupture zone,
J. Geol. Soc. India, 30, 13–27.
MOLNAR, P. and PANDEY, M.R. (1989), Rupture zones of great
earthquakes in the Himalayan Region. Proc. Ind. Acad. Sci.
(Earth Planet. Sci.), 98, 61–70.
NATH,S.K and THINGBAIJAM, K.K.S. (2010), Comment on ‘‘Estimation of seismicity parameters for India’’, Seism. Res. Lett. 81,
1001-1003.
RISTAU, J. (2009), Comparison of Magnitude Estimates for New
Zealand Earthquakes: Moment Magnitude, Local Magnitude,
and Teleseismic Body-Wave Magnitude, Bull. Seism. Soc. Am.
99, 1841-1852.
SCORDILIS, E.M. (2006), Empirical global relations converting MS
and mb to moment magnitude, J. Seism. 10, 225-236.
SEEBER, L. and ARMBRUSTER, J.G. (1981), Great detachment and
earthquakes along the Himalayan arc and long-term forecasting,
in Earthquake Prediction, Simpson D. W. Richards P. G., Editors
A.G.U., Washington, D.C., 259-277.
THINGBAIJAM, K.K.S., NATH, S.K., YADAV, A., RAJ, A, WALLING,
M.Y. and MOHANTY, W.K (2008), Recent seismicity in Northeast
India and its adjoining region, J. Seism. 12, 107-123.
UTSU, T. (2002), Relationships between magnitude scales, in
International Handbook of Earthquake and Engineering Seismology, Part A, W. H. K. Lee, H. Kanamori, P. C. Jennings and
C. Kisslinger (Editors), Academic Press, Amsterdam, 733-746.
YADAV, R.B.S., BORMANN, P., RASTOGI, B.K., DAS, M.C., and CHOPRA, S. (2009), Homogeneous and complete earthquake catalog
for Northeast India and the adjoining region, Seism.Res.Lett. 80,
609-627.
(Received July 21, 2010, revised March 23, 2011, accepted May 9, 2011)