Tectonophysics 532-535 (2012) 215–222

Contents lists available at SciVerse ScienceDirect

Tectonophysics
journal homepage: www.elsevier.com/locate/tecto

Spatial and temporal evolution of b-values before large earthquakes in Taiwan

Chung-Han Chan a, Yih-Min Wu a,⁎, Tai-Lin Tseng a, Ting-Li Lin b, Chien-Chih Chen c
a
Department of Geosciences, National Taiwan University, Taipei 10617, Taiwan
b
Department of Earth Sciences, National Cheng Kung University, Tainan 701, Taiwan
c
Department of Earth Sciences and Graduate Institute of Geophysics, National Central University, Jhongli 32001, Taiwan

a r t i c l e i n f o a b s t r a c t

Article history: We investigate the spatial and temporal variations of b-values before twenty-three earthquakes with ML ≥ 6.0
Received 25 November 2010 in the Taiwan region from 1999 to 2009. We estimate the spatial distribution of b-values within a one-year
Received in revised form 13 January 2012 period before the occurrence of investigated earthquakes. It shows that the epicenters of those earthquakes
Accepted 5 February 2012
are located predominately in the regions with low b-values relative to the entire study area. In terms of tem-
Available online 18 February 2012
poral evolutions, the b-values are slightly lower during the year prior to the investigated earthquakes in com-
Keywords:
parison to the periods two to five years earlier. We propose that these anomalies may offer spatiotemporal
Gutenberg–Richter relation constraints for the earthquake forecasting in Taiwan in the future.
b-value © 2012 Elsevier B.V. All rights reserved.
Molchan diagram
Taiwan

1. Introduction (Smith, 1981), even though the validity of earthquake prediction exper-
iments remains controversial.
The frequency–magnitude distribution (Ishimoto and Iida, 1939), Due to collisions between the Philippine Sea Plate and the Eurasian
also known as the Gutenberg–Richter relation (Gutenberg and Plate, numerous earthquakes have been taking place inland and off
Richter, 1954), in a seismogenic volume can be expressed as the east coast of the Taiwan Island. Most significant events that oc-
curred in the past decades in Taiwan were well recorded and carefully
log10 N ¼ a−bM; studied by governmental bureau and research scientists. The 1999 ML
7.3 Chi-Chi earthquake (earthquake No.1 in Fig. 1), for example, is the
where M is magnitude, N is the cumulative number of earthquakes with largest inland earthquake ever recorded in Taiwan. After the devastat-
magnitude larger than M, and a and b are constants. The b-value repre- ing Chi-Chi event, the 2003 ML 6.4 Chengkung earthquake (earthquake
sents the ratio of small to large earthquakes. In general, the b-value in No.13 in Fig. 1) is the first event that occurred on the Longitudinal Valley
broad seismic regions is close to 1.0 from global statistics (Lay and fault of east coast with magnitude larger than 6. Interestingly, both
Wallace, 1995). For regions in smaller scale over a few to tens of kilome- events were reported to be preceded by seismicity precursors such as
ters, the temporal and spatial variations in b-values may be significantly noticeable decreases in regional seismicity rate and b-value (Chen,
large, ranging from 0.5 to 1.5 in Japan (Ogata and Katsura, 1993) and 2003; Chen et al., 2005; Wu and Chiao, 2006; Wu et al., 2008b). Those
from 0.5 to 1.3 near Parkfield of the San Andreas fault (Schorlemmer temporal decreases in regional seismicity rate and b-value may be
et al., 2004). explained by the quiescence in the seismicity with small magnitudes
Variation of b-value in a region is commonly correlated with charac- (Huang, 2006; Huang and Nagao, 2002; Huang et al., 2001). A detailed
teristics of regional seismic activities. For example, high seismicity in study by Wu et al. (2008b) also showed that the rupture region of the
magma chambers may be characterized by high b-values (Sanchez Chengkung earthquake had low b-value, similar to the observations in
et al., 2004; Wiemer and McNutt, 1997). The initial rupture of large California (Wyss et al., 2000) and Iceland (Wyss and Stefansson,
earthquakes, by contrast, was found to occur in regions where b- 2006). Those results motivated us to comprehensively investigate the
values are low (Wyss and Stefansson, 2006; Wyss et al., 2000). Previous spatial and temporal evolutions of b-values associated with the occur-
studies (Schorlemmer et al., 2005, and references therein) pointed out rence of large earthquakes in the whole Taiwan region.
that the region with low b-value implies large differential stress and
suggests its being toward the end of the seismic cycle. Such relationship 2. Seismic network and catalogue
can be applied for evaluating seismic hazard and earthquake forecasting
The Central Weather Bureau Seismic Network (CWBSN) is respon-
⁎ Corresponding author. Tel.: + 886 2 2362 0054; fax: + 886 2 2363 6095. sible for monitoring regional earthquakes in Taiwan. Since 1991, real-
E-mail address: drymwu@ntu.edu.tw (Y.-M. Wu). time digital recording has been performed; about 18,000 events each

0040-1951/$ – see front matter © 2012 Elsevier B.V. All rights reserved.
doi:10.1016/j.tecto.2012.02.004
216 C.-H. Chan et al. / Tectonophysics 532-535 (2012) 215–222

25˚

24˚

23˚

22˚

120˚ 121˚ 122˚ 123˚

Fig. 1. Topographic map showing distribution of the 23 investigated earthquakes around Taiwan. We select earthquakes with ML greater or equal to 6.0 and focal depths less than
40 km during 1999 and 2009. Each of earthquakes is assigned a number according to the chronologic order of occurrence times. Gray lines represent the Longitudinal Valley fault,
which is the boundary between Eurasia Plate and Philippine Sea Plate.

year in a roughly 400 km × 550 km region are recorded (Wu et al., 3. Distribution of b-values
2008a). The network currently consists of a central recording system
with 71 telemetered stations (Fig. 2). The CWBSN instruments were In this study, we investigate the b-value distributions before large
operated in a triggered-recording mode until the end of 1993 when events with ML greater than or equal to 6.0 and focal depths less than
continuous recording began. P and S waves arrival times are picked 40 km in the Taiwan region during 1999–2009. To avoid using earth-
manually for determination of earthquake parameters, i.e. hypocenter quakes from the same sequence, we keep solely the first shock in a se-
and Richter local magnitude (ML) (Shin, 1993). quence as the investigated earthquake and exclude all other events
The CWBSN has greatly enhanced the earthquake monitoring ca- that occurred within one month after and within inter-epicentral dis-
pability since the end of 1993 (Wu and Chiao, 2006). To evaluate tances of 40 km. There is a total of 23 earthquakes that fulfill (Fig. 1).
the reliability of the catalogue, we calculate spatial distribution of We first calculate the regional b-values for events with ML ≥ Mc
magnitude of completeness (Mc) by using the maximum curvature and focal depths less than 40 km in the region bounded by latitudes
approach (Wiemer and Wyss, 2000). We consider the catalogue in from 21.5°N to 25.5°N and longitudes from 119.5°E to 123.0°E. We di-
the period from the beginning of 1994 to the end of 2009 for shallow vide our study region into 0.2° × 0.2° grids and search for the events
earthquakes (with focal depth ≤ 40 km). We divide our study region within a circle of 30 km in radius, which is comparable to the rupture
into 0.2° × 0.2° grids and search for the events within a circle of size for an event with magnitude 6.0 and above and is larger than the
30 km in radius (Fig. 2). The pattern of Mc simply reflects the density uncertainty in location of those earthquakes in the CWBSN catalogue.
of the seismic stations (Mignan et al., 2011). The station densities are There are several methods proposed for computing b-values (e.g. Aki,
the highest in the northern and southwestern Taiwan, where the Mc 1965; Shi and Bolt, 1982). In this study, we use a maximum likelihood
can be as low as about 1.4. The Mc for the inland region is generally approach to estimate the b-values and the standard deviation (Aki,
less than 2.0, whereas the Mc in the offshore region is between 2.5 1965) in the grids. The standard deviation (σ) of b-value is as follows:
and 3.2 due to poor network coverage. The spatial distribution and
−1=2
magnitudes estimated in this study is consistent with those obtained σ ¼ bN ;
from the Bayesian magnitude of completeness method (Mignan et al.,
2011). To achieve reliable b-values, it is important that we re- where N is number of the events used for b-value estimation.
evaluate Mc according to the catalogue fulfill the same criterion of We calculate the spatial distributions of b-values one year prior to
each b-value calculation. the 23 investigated events (Fig. 3). To avoid temporal incompleteness
 C.-H. Chan et al. / Tectonophysics 532-535 (2012) 215–222 217

Magnitude completeness
during 1994-2009
25˚

24˚

23˚

22˚

CWBSN station

120˚ 121˚ 122˚ 123˚

M c according to seismic catalog from CWBSN

1.4 1.6 1.8 2.0 2.2 2.4 2.6 2.8 3.0 3.2 3.4

Fig. 2. Distribution of the magnitude completeness (Mc) according to the catalogue acquired from the Central Weather Bureau Seismic Network (CWBSN). The white triangles show
the locations of the CWBSN stations. Note that only cells with more than 100 events within the radius of 30 km are reported for reliability.

of catalogue after large earthquakes, we keep all events but exclude considered, 12 out of 19 (63%) earthquakes are in the regions with
only the earthquakes within 24 h following the 23 studied earth- lower b-values than the average in the study region.
quakes (Fig. 1) for b-value calculation. In order to acquire reliable re- We further employ the Molchan diagram to evaluate statistical
sults, only cells with more than 100 events within the radius of 30 km significance of the argument that large earthquakes tend to occur in
are reported. Due to insufficient number of seismicity, no b-values are low b-value regions (Fig. 4). This diagram is originally designed for
reported in the grids where earthquakes Nos. 7, 12, 15, and 21 initiat- evaluation of earthquake forecasting ability and is presented as a frac-
ed. After excluding these disqualified cases, the epicentral b-values of tion of space occupied by expectation against fraction of failure to
17 out of the remaining 19 (89%) events are lower than 1.0. We fur- predict (Molchan, 1990, 1991). We present “fraction of space occu-
ther compare the epicentral b-value of each investigated earthquake pied by expectation” as sorted events according to the proportion of
with those of the entire study area. For each event, we estimate the study area having a b-value equal to or lower than those at the epi-
spatial fraction (relative to total area in percentage) that contains b- centers (presented in parentheses of each panel in Fig. 3). Thus, we
values equal to or lower than the epicentral b-value. In totality, we define “expectation” as a presumed threshold of the proportion of
find that 16 out of 19 (84%) earthquakes are located in the low b re- study area having lower b-values within one year prior to occurrences
gions that account for less than 50% of the study area (Fig. 3). For ex- of investigated events. The “fraction of failure to predict” indicates the
ample, the 2003 Chengkung earthquake (earthquake No. 13) took proportion of the earthquakes with higher fraction of space occupied
place in the grid with b-value of 0.99 and there is 28% of study area by expectation. That is to say, when the data points closely distributed
having lower b-values than 0.99 (Fig. 3). Another representative is along the diagonal line, it suggests distribution of the investigated
earthquake No. 22 that occurred on July 13th, 2009 offshore Taiwan. earthquakes is independent of b-values; when a convex is presented,
Its epicentral b-value is 0.73 and only 3% of study area has b-values it suggests most of the earthquakes occurred in the regions with
lower than 0.73 (Fig. 3). Furthermore, even though when both stan- higher b-values in comparison to the entire area; whereas a concave
dard deviations of the b-values for epicentral and study region are suggests that most of the earthquakes occurred in the area with
218 C.-H. Chan et al. / Tectonophysics 532-535 (2012) 215–222

Fig. 3. Distribution of b-values estimated in one-year prior to the occurrence of 23 investigated earthquakes. Stars show the epicenter of earthquakes. The b-value with corresponding
standard deviations in epicenter (lower line) and the study region (upper line) are presented in lower-left corner of each panel. The cells with less than 100 events in radius of 30 km
are colored in grey reporting null b. Note that the areas with null b are excluded for calculation. The proportion of study area having a b-value equal to or lower than those at the epicenters
of the earthquakes are presented in parentheses of each panel.
 C.-H. Chan et al. / Tectonophysics 532-535 (2012) 215–222 219

Fig. 3 (continued).

lower b-values. We confirm that such relationship of the low epicen- available. In other words, we do not consider earthquakes Nos. 7, 12,
tral b-values exists when the standard deviations are taken into 15, and 21, which took place in grids without reported b-values due to
account. insufficient number of seismicity. In the Molchan diagram (Fig. 4), the
In order to present a more reliable result in the Molchan diagram, epicentral b-values of investigated earthquakes are significantly lower
we consider only the earthquakes in the grids where b-values are in comparison to the rest of study area. All of earthquakes are located
220 C.-H. Chan et al. / Tectonophysics 532-535 (2012) 215–222

Fig. 4. Molchan diagram for investigating corresponding b-value at each epicenter compared with the entire study area. Numbers in the rectangles represent the corresponding
number of the earthquake as shown in Fig. 1. The error bar for each earthquake represents corresponding fraction of space for the standard deviation of the b-value calculation
(shown in Fig. 3).

in the region with lower b-values than at least 25% of the entire study larger when higher magnitude cutoffs are considered. It can be attrib-
area (i.e. 75% of study area having a b-value equal to or lower than uted to fewer data for b-value calculations. When lower magnitude
those at all of the epicenters of the earthquakes). For the 1999 ML7.3 cutoffs are considered, systematical lower b-values are acquired,
Chi-Chi earthquake (earthquake No. 1), if we further consider earth- which can be associated with incomplete catalogue. It should be men-
quakes within the true ruptured region of ca. 40 km× 80 km (Johnson tioned that since the Mc in this study is acquired by the maximum
and Segall, 2004) (Fig. 3), the b-value of the entire rupture zone be- curvature approach (Wiemer and Wyss, 2000), Mc is seldom overes-
comes 0.92 (c.f. the epicentral b-value of 1.00 for the Chi-Chi earth- timated, i.e. minimum magnitude should not be smaller than Mc.
quake), which suggests only 40% of the study area having lower b- Our observations of the relatively low spatial b-values for the epi-
value than in the rupture zone. central areas are in good agreement with the results presented by
To examine temporal variation of the b-values around epicentral Wyss et al. (2000) and Wyss and Stefansson (2006), who argued
regions, we consider five different time periods prior to the occur- that the initiations of large earthquakes may be mapped by low b-
rences of the earthquakes (Fig. 5). In order to acquire reliable results, values. Many studies have considered the b-value as a monitoring
we only consider the investigated earthquakes with more than 100 index related to forthcoming large earthquakes (e.g. Ogata and
events within the radius of 30 km in the time windows. We also re- Katsura, 1993; Papadopoulos et al., 2010). Main et al. (1989) showed
evaluated Mc according to the catalogue in the corresponding periods. that b-value varies between 0.5 and 1.5 during the earthquake cycle
It clearly demonstrates that the b-values are very stable (0.90-0.93) and proposed that critical failure in a large earthquake should occur
during the time windows from five to one year(s) prior to the occur- in regions with b = 0.5. They also discussed b-value evolution for lab-
rence of earthquakes. It then drops by ca. 5% within a year before oratory creep experiments and seismicity during the 1984 Western
earthquakes. By considering standard deviations, however, the tem- Nagano, Japan (Figs. 1d and 3 in Main et al. (1989)). Their studies ex-
poral evolutions of the b-values become trivial. amined decreases in the b-value as stress increases before failure of
faults. Further investigation for the southern Californian region also
4. Discussions and conclusions supports that seismicity preceding large main shocks can be associat-
ed with low b-values (Henderson et al., 1994).
For the Gutenberg–Richter relation, magnitude cutoffs may play From previous analysis on the seismicity in the Taiwan region
an important role for determination of b-value. In order to test the re- (Wu and Chiao, 2006; Wu et al., 2008b), the relatively low seismicity
liability of our results, we present biases of the b-values by consider- rate and the temporal decrease in the b-values before large earth-
ing different magnitude cutoffs for the calculation (Fig. 6). When the quakes in Taiwan may be considered as a forecasting indicator associ-
differences between magnitude cutoff and Mc are in between −0.1 ated with the quiescence in the seismicity with small magnitudes.
and +0.5, the results show that change of the b-values are smaller Our results further show that the 2003 Chengkung earthquake
than 0.03. In contrast with the averaged standard deviation of 0.04 (earthquake No. 13) took place in a grid with a b-value of 0.99,
for b-values, the biases imparted by considering different magnitude which is lower than 72% of the region (Fig. 3), and is consistent
cutoffs are insignificant when the cut-off magnitude is close or with the observations reported in Wu et al. (2008b). We also found
above Mc. Note that the deviations of the b-value changes become that several earthquakes, both inland and offshore of Taiwan, have
 C.-H. Chan et al. / Tectonophysics 532-535 (2012) 215–222 221

Fig. 5. Average of epicentral b-values (solid lines) and standard deviations (dash lines) for the 23 investigated earthquakes in different time windows. In order to acquire reliable
results, we only consider the investigated earthquakes with more than 100 events within the radius of 30 km in the time windows.

fairly low epicentral b-values (Fig. 3). One of the lowest values is Segall (2004), we find a somewhat low b-value of 0.92, which corre-
found to be 0.73 for the earthquake No. 22 as shown in Fig. 3. Overall, spond to 40% of study area with lower b-value. This implicitly sug-
the epicentral b-values for 17 out of 19 (89%) events are lower than gests that decreases in b-value, which imply an increase in stress
1.0. Through the Molchan diagram (Fig. 4), we show that the low b- (Main et al., 1989), may take place in the entire rupture fault zones
values in epicenter regions are statistically significant. of large earthquakes, but not at the rupture initiations.
It is worth noting that the epicenter of the 1999 ML7.3 Chi-Chi We also evaluate temporal variations of the b-values before the
earthquake is in a region with an unexpectedly high b-value of 1.00 occurrences of investigated earthquakes (Fig. 5). It shows that epicen-
in the year prior to the shock (earthquake No. 1 in Fig. 3). If we eval- tral b-values (0.90–0.93) are already lower than in broad seismic re-
uate the b-value in its rupture area, which is acquired by Johnson and gions during the time windows of five to one year prior to the

Fig. 6. Change of epicentral b-values caused by different magnitude cutoffs for the investigated earthquakes. The solid line and dash lines represent mean and standard deviation
(0.03) of b-value calculation when magnitude cutoff is equal to Mc, respectively. Open circles represent b-value differences using different magnitude cutoffs for the investigated
earthquakes. Solid circles and error bars represent average difference and standard deviation of the change, respectively, using different magnitude cutoffs.
222 C.-H. Chan et al. / Tectonophysics 532-535 (2012) 215–222

target earthquake (where b-value from global statistics is close to 1.0 Lay, T., Wallace, T.C., 1995. Modern Global Seismology. Academic Press, p. 393.
Main, I.G., Meredith, P.G., Jones, C., 1989. A reinterpretation of the precursory seismic
(Lay and Wallace, 1995)). Further drop of b-values by 4% is observed b-value anomaly from fracture mechanics. Geophysical Journal 96, 131–138.
for one year before the investigated earthquakes, although this evolu- Mignan, A., Werner, M.J., Wiemer, S., Chen, C.-C., Wu, Y.-M., 2011. Bayesian estimation
tion is relative small in comparison with the standard deviation. This of the spatially varying completeness magnitude of earthquake catalogs. Bulletin of
the Seismological Society of America 101, 1371–1385. doi:10.1785/0120100223.
suggests that monitoring the evolution of b-values may provide spa- Molchan, G.M., 1990. Strategies in strong earthquake prediction. Physics of the Earth
tial and temporal constraints for the forthcoming strong earthquakes and Planetary Interiors 61, 84–98.
consistently in both of long-term and short-term periods. Molchan, G.M., 1991. Structure of optimal strategies in earthquake prediction. Tectono-
physics 193, 267–276.
Ogata, Y., Katsura, K., 1993. Analysis of temporal and spatial heterogeneity of magni-
Acknowledgments tude frequency distribution inferred from earthquake catalogues. Geophysical
Journal International 113, 727–738.
Papadopoulos, G.A., Charalampakis, M., Fokaefs, A., Minadakis, G., 2010. Strong fore-
This research was supported by the Central Weather Bureau and
shock signal preceding the L'Aquila (Italy) earthquake (Mw 6.3) of 6 April 2009.
the National Science Council (NSC99-2811-M-002-123, NSC99- Natural Hazards Earth System Science 10, 19–24. doi:10.5194/nhess-10-19-2010.
2119-M-002-022, NSC99-2627-M-002-015, NSC100-3114-M-002- Sanchez, J.J., McNutt, S.R., Power, J.A., Wyss, M., 2004. Spatial variations in the frequen-
001, and NSC99-2116-M-002-006) of the Republic of China. We also cy–magnitude distribution of earthquakes at Mount Pinatubo volcano. Bulletin of
the Seismological Society of America 94, 430–438.
appreciate the constructive comments provided by editor and two Schorlemmer, D., Wiemer, S., Wyss, M., 2004. Earthquake statistics at Parkfield: 1. Stationarity
anonymous reviewers. of b-values. Journal of Geophysical Research 109, B12307. doi:10.1029/2004JB003234.
Schorlemmer, D., Wiemer, S., Wyss, M., 2005. Variations in earthquake-size distribu-
tion across different stress regimes. Nature 437 (22), 539–542.
References Shi, Y., Bolt, B.A., 1982. The standard error of the magnitude-frequency b value. Bulletin
of the Seismological Society of America 72, 1677–1687.
Aki, K., 1965. Maximum likelihood estimate of b in the formula logN = a - bM and its Shin, T.C., 1993. The calculation of local magnitude from the simulated Wood–Ander-
confidence. Bulletin Earthquake Research Institute University Tokyo 43, 237–239. son seismograms of the short-period seismograms. Terrestrial, Atmospheric and
Chen, C.C., 2003. Accelerating seismicity of moderate-sized earthquakes before the Oceanic Sciences 4, 155–170.
1999 Chi-Chi, Taiwan, earthquake: testing time-prediction of self-organizing spi- Smith, W.D., 1981. The b-values as an earthquake precursor. Nature 289, 136–139.
nodal model of earthquakes. Geophysical Journal International 155, F1–F5. Wiemer, S., McNutt, S., 1997. Variations in frequency–magnitude distribution with
Chen, C.C., Rundle, J.B., Holliday, J.R., Nanjo, K.Z., Turcotte, D.L., Li, S.C., Tiampo, K.F., depth in two volcanic areas: Mount St. Helens, Washington, and Mt. Spurr, Alaska.
2005. The 1999 Chi-Chi, Taiwan, earthquake as a typical example of seismic activa- Geophysical Research Letters 24, 189–192.
tion and quiescence. Geophysical Research Letters 32, L22315. Wiemer, S., Wyss, M., 2000. Minimum magnitude of completeness in earthquake cata-
Gutenberg, B., Richter, C., 1954. Seismicity of the Earth and Associated Phenomena, 2nd logs: examples from Alaska, the Western United States, and Japan. Bulletin of the
ed. Princeton Univ. Press, Princeton, N.J. (310 pp.). Seismological Society of America 90 (4), 859–869.
Henderson, J., Main, I.G., Pearce, R.G., Takeya, M., 1994. Seismicity in north-eastern Brazil — Wu, Y.M., Chiao, L.Y., 2006. Seismic quiescence before the 1999 Chi-Chi, Taiwan Mw7.6
fractal clustering and the evolution of the b-value. Geophysical Journal International earthquake. Bulletin of the Seismological Society of America 96, 321–327.
116, 217–226. Wu, Y.M., Chang, C.H., Zhao, L., Teng, T.L., Nakamura, M., 2008a. A comprehensive relo-
Huang, Q., 2006. Search for reliable precursors: a case study of the seismic quiescence cation of earthquakes in Taiwan from 1991 to 2005. Bulletin of the Seismological
of the 2000 western Tottori prefecture earthquake. Journal of Geophysical Re- Society of America 98, 1471–1481. doi:10.1785/0120070166.
search 111, B04301. doi:10.1029/2005JB003982. Wu, Y.M., Chen, C.C., Zhao, L., Chang, C.H., 2008b. Seismicity characteristics before the
Huang, Q., Nagao, T., 2002. Seismic quiescence before the 2000 M = 7.3 Tottori earth- 2003 Chengkung, Taiwan Mw6.8 earthquake. Tectonophysics 457, 177–182.
quake. Geophysical Research Letters 29, 1578. doi:10.1016/j.tecto.2008.06.007.
Huang, Q., Sobolev, G.A., Nagao, T., 2001. Characteristics of the seismic quiescence and Wyss, M., Stefansson, R., 2006. Nucleation points of recent mainshocks in Southern Ice-
activation patterns before the M = 7.2 Kobe earthquake, January 17, 1995. Tecto- land, mapped by b-values. Bulletin of the Seismological Society of America 96,
nophysics 337, 99–116. 599–608.
Ishimoto, M., Iida, K., 1939. Observations of earthquakes registered with the microseis- Wyss, M., Schorlemmer, D., Wiemer, S., 2000. Mapping asperities by minima of local re-
mograph constructed recently. Bulletin Earthquake Research Institute Tokyo Uni- currence time: the San Jacinto-Elsinore fault zones. Journal of Geophysical Re-
versity 17, 443–478. search 105, 7829–7844.
Johnson, K.M., Segall, P., 2004. Imaging the ramp—décollement geometry of the Che-
lungpu fault using coseismic GPS displacements from the 1999 Chi-Chi, Taiwan
earthquake. Tectonophysics 378, 123–139.