Bulletin of the Seismological Society of America, 91, 5, pp.

1218–1228, October 2001

Near Real-Time Mapping of Peak Ground Acceleration and

Peak Ground Velocity Following a Strong Earthquake
by Yih-Min Wu, Tzay-Chyn Shin, and Chien-Hsin Chang

Abstract During a disastrous earthquake, the early assessment and timely re-
porting of the peak ground acceleration (PGA) and peak ground velocity (PGV) maps
will be crucial in an effective emergency response operation. In this study, we first
derive an empirical relationship between ML and MW. The PGA and PGV attenuation
relationships are deduced with data from the Taiwan Strong Motion Instrumentation
Program (TSMIP) and the Taiwan Rapid Earthquake Information Release System
(TREIRS). Site corrections of the attenuation relationships for shallow and large earth-
quakes in Taiwan region are also obtained. Peak values of earthquake strong ground
motion can be well determined in Taiwan as soon as the earthquake location is
determined, and magnitudes are calculated by the TREIRS. This peak ground motion
value information can be immediately turned into the calculated PGA and PGV maps
that can be issued within two minutes of the earthquake origin time. During any
disastrous earthquake, these maps are found to be very useful for immediate seismic
damage assessment and dispatching of emergency response missions.

Introduction
Real-time seismic monitoring, especially for strong ground acceleration (PGA) and peak ground velocity (PGV)
earthquakes, is an important tool for seismic hazard miti- in the potentially damaged area. In 1995, the Central
gation (Kanamori et al., 1997; Teng et al., 1997). It provides Weather Bureau (CWB) developed the T system by a real-
valuable near–real-time information for rapid earthquake time strong-motion network for intensity observations, mag-
emergency response, thereby mitigating the loss. With the nitude, and hypocenter determination routinely after felt
importance of rapid earthquake information for seismic haz- earthquakes in the Taiwan region (Teng et al., 1997; Wu et
ard mitigation being recognized, efforts to design and im- al., 1997).
plement systems to provide a broad range of rapid earth- The T system consists of 75 telemetered strong-motion
quake information have recently been expanded (Heaton, stations in Taiwan (Fig. 2). Three-component force-balanced
1985; Nakamura, 1988, 1989; Espinosa-Aranda, et al., 1995; accelerometer (FBA) digital signals are continuously tele-
Gee et al., 1996; Wu et al., 1997, 1998, 1999; Wald et al., metered to the headquarters of the CWB in Taipei via leased
1999a,b). During the 1999 Chi-Chi, Taiwan, earthquake a telephone lines. The FBA signal is digitized at 50 samples
severe test was put on the Taiwan Rapid Earthquake Infor- per sec at a 16-bit resolution. The full recording range is
mation Release System (TREIRS) system (Shin et al., 2000; Ⳳ2g. The interstation spacing of the T system is about 30
Wu et al., 2000). For the purpose of briefness, we use the km. This spacing is still too large for the damage assessment
term T system to mean the TREIRS in this article. We also due to complex geology in Taiwan (Lee et al., 2001). On
report our experience and thoughts in the development and the other hand, the Taiwan Strong Motion Instrumentation
application of the T system. Program (TSMIP) was successfully implemented six years
The most common information available immediately ago by the CWB, with about 650 modern digital accelero-
following a damaging earthquake is its magnitude and epi- graphs at free-field sites (Fig. 2). The TSMIP signals are dig-
center. However, the damage pattern is not a simple function itized at 200 samples per sec or higher and at 16-bit or higher
of these two parameters alone. More detailed information is resolution. Most accelerometer sensor recording ranges are
needed for emergency response agencies to assess the situ- Ⳳ2g. The TSMIP interstation spacing is about 5 km in met-
ation for better details and accuracy. For example, for the ropolitan areas. It offers much more detailed description of
20 September 1999 Chi-Chi, Taiwan, earthquake, the city of ground shaking for damage assessment. But they are not
Tung-Shih Town is in the region with the worst damage, continuously telemetered. In this study we combine the data
even though it is about 50 km from the epicenter (Fig. 1). from the T system and the TSMIP network to determine the
Thus, it is highly desirable to map out distributions of peak site corrections. Then, the database of TSMIP site corrections

1218
Near Real-Time Mapping of PGA and PGV Following a Strong Earthquake 1219

Figure 1. The Chi-Chi mainshock epicenter, the Figure 2. Distribution of the 75 TREIRS stations
Chelongpu fault rupture, and topographical settings and the 650 TSMIP stations.
of Taiwan.

will be integrated into the T system to estimate the PGA and Relationship of the ML and MW for Shallow
PGV values at the TSMIP sites for the new earthquake. We Earthquakes in Taiwan
are aiming at the issuance of these estimated peak values of
ground motion within 2 min of the occurrence of a strong We first use the T system accelerograms to simulate the
earthquake in Taiwan region. Wood-Anderson seismograms from which we determine the
local magnitude ML (Shin, 1993). In our data set, 32 events
have MW reported by Harvard University Seismology Cen-
Data ter. Therefore we use those 32 events to find the relationship
Big and shallow earthquakes often cause serious dam- between ML and MW and use this relationship to convert ML
age in heavily populated areas. Thus, we selected 60 large to MW for the following analysis. Figure 4 shows that ML
and shallow events in Taiwan for this study (Fig. 3 and Table and MW are correlated reasonably well, even though the
1). The selection criteria are ML greater than 5.0 and focal number of data points for large events is limited and consid-
depth less than 35 km. All events were well recorded by ering ML is saturated in the large magnitude portion. Thus,
both the T system and TSMIP network. These events oc- ML and ln(MW) are used in the regression process. The em-
curred in 1995 to 1999 and were widely felt in Taiwan. pirical equation of the corresponding least-squares line is
These earthquakes are relocated in this study by using both
the T system and the TSMIP records. A total of 1369 T sys- ML ⳱ 4.53 ⳯ ln(MW) ⳮ 2.09 Ⳳ 0.14,
tem records and 8262 TSMIP records are used for this study. for 5.0 ⱕ ML ⱕ 7.1 or 4.8 ⱕ MW ⱕ 7.6. (1)
1220 Y.-M. Wu, T.-C. Shin, and C.-H. Chang

Table 1
Earthquake Recordings Used in This Study.
Origin Time (UTC)
Date Time Lat. Long. Depth ML MW
(yy/mm/dd) (hr:min:sec) (N) (E) (km) (This Study) (Harvard)

95/01/10 07:55:19.71 23.677 121.412 11.25 5.18 5.1

95/02/10 02:21:59.78 23.779 121.894 20.24 5.03 –
95/02/23 05:19:01.79 24.169 121.807 22.03 6.27 6.2
95/04/03 11:54:40.09 23.977 122.334 5.00 5.79 5.6
95/05/02 22:48:20.71 23.840 121.950 22.96 531 –
95/05/09 01:03:51.37 23.012 121.373 22.22 5.03 –
95/05/27 18:11:11.92 23.040 121.367 21.98 5.60 5.7
95/06/28 14:14:53.64 22.220 121.354 6.87 5.20 5.3
95/07/07 03:04:48.45 23.888 121.092 13.57 5.45 –
95/07/14 16:52:46.93 24.361 121.754 9.29 5.57 –
95/10/14 03:01:36.86 24.622 122.409 5.00 5.04 –
95/10/31 22:27:06.61 23.281 120.368 18.07 5.10 –
95/12/18 16:17:54.49 24.047 121.643 29.48 5.39 5.2
95/12/25 19:05:28.23 22.838 121.343 5.00 5.19 –
96/03/05 14:52:27.98 23.964 122.298 8.87 6.24 6.3
96/03/05 17:32:08.10 23.934 122.252 5.00 5.93 5.8
96/03/29 03:28:52.91 23.985 122.309 2.38 5.72 5.7
96/05/28 21:53:23.10 24.075 121.502 19.89 5.23 –
96/07/27 00:26:39.51 24.329 122.051 10.13 5.02 –
96/08/10 06:23:05.70 23.885 122.649 5.57 5.81 5.6
96/09/05 23:42:07.73 22.003 121.366 9.92 6.49 6.8
96/09/06 02:04:56.39 21.920 121.366 7.98 5.53 –
96/09/06 11:34:32.06 21.754 121.285 5.00 5.37 5.4
96/11/14 01:39:11.24 23.418 122.086 12.80 5.16 –
96/12/18 11:20:23.77 22.840 121.314 16.75 5.02 –
97/01/05 10:34:18.63 24.670 122.320 5.00 5.52 5.2
97/01/16 01:07:20.32 22.007 121.306 6.55 5.26 –
97/03/24 23:32:19.97 24.157 121.725 31.13 5.09 –
97/05/03 02:46:14.56 22.520 121.364 8.50 5.39 5.0
Figure 3. Epicentral distribution of the 60 events 97/06/22 09:36:04.34 22.157 121.390 4.57 5.37 5.2
used in this study. 97/07/04 18:37:29.68 23.059 120.792 1.83 5.11 5.1
97/10/22 11:16:26.28 22.409 121.490 13.82 5.36 –
97/11/14 04:29:51.14 24.176 121.732 3.37 5.29 –
98/01/18 19:56:51.53 22.761 121.009 6.97 5.17 5.2
It is apparent that for 4.8 ⱕ MW ⱕ 5.5, the ML is overesti- 98/03/11 17:21:54.84 22.443 122.261 4.07 5.25 –
mated by about 0.2 unit; for 5.5 ⱕ MW ⱕ 6.5, the ML is 98/07/17 04:51:15.02 23.503 120.664 5.71 5.90 5.7
close to MW. It is clear that ML begins to saturate at about 98/07/24 18:44:02.57 21.602 121.839 5.39 6.03 6.1
98/10/09 12:52:51.65 22.196 121.584 1.40 5.53 –
magnitude 6.5. 98/11/03 07:06:37.43 22.132 121.127 25.04 5.28 –
98/11/17 22:27:31.84 22.828 120.793 27.01 5.65 5.3
Attenuation Relationships for PGA and PGV 98/12/14 00:59:03.50 24.290 122.157 34.22 5.10 –
99/06/10 07:17:59.88 23.797 121.558 32.06 5.06 –
The Chi-Chi earthquake sequence is very energetic and 99/07/07 13:53:12.86 23.305 120.757 18.88 5.10 –
99/09/20 17:47:15.85 23.853 120.815 8.00 7.06 7.6
provides a wealth of the digital data for large- and near- 99/09/20 17:57:15.46 23.923 121.038 5.93 6.05 –
source observations. We used those data to determine the 99/09/20 18:03:42.05 23.806 120.850 9.97 6.26 –
attenuation relationships for PGA and PGV. The basic linear 99/09/20 18:11:52.00 23.899 121.031 5.29 6.35 –
regression model used is as following (Liu, 1999): 99/09/20 18:16:17.31 23.866 121.059 8.36 6.33 –
99/09/20 21:46:36.86 23.564 120.904 4.39 6.22 6.4
99/09/22 00:14:40.77 23.826 121.047 15.59 6.47 6.4
log10(Y) ⳱ C1 Ⳮ C2MW ⳮ log10(rrup Ⳮ h) Ⳮ C3rrup , (2) 99/09/22 00:49:42.39 23.746 121.006 6.03 5.97 5.8
99/09/22 11:17:20.94 23.718 121.003 18.06 5.54 5.2
where Y is either PGA or PGV, and rrup is the source-to-site 99/09/23 12:44:33.83 23.910 121.082 19.71 5.36 5.2
99/09/25 08:43:28.92 23.642 120.978 7.38 5.43 5.1
distance in terms of the closest distance to the rupture sur-
99/09/25 23:52:47.28 23.826 120.994 20.90 6.44 6.5
face. If the surface was not defined for an event, then epi- 99/10/01 12:54:09.35 23.684 120.943 8.00 5.26 5.2
central distance was used as the source-to-site distance. The 99/10/04 12:26:12.60 23.788 120.955 11.24 5.01 –
variable h is the saturation term of the PGA and PGV for 99/10/18 16:00:39.56 23.714 121.061 25.42 5.32 5.3
near-source observation. Here we use square root of rupture 99/10/22 02:18:56.55 23.496 120.422 23.14 6.07 5.8
99/10/22 03:10:16.35 23.505 120.436 19.78 5.86 5.5
area for h. The coefficients C1, C2, and C3 are to be deter-
Near Real-Time Mapping of PGA and PGV Following a Strong Earthquake 1221

log10(PGA) ⳱ 0.00215 Ⳮ 0.581MW ⳮ log10(rrup

Ⳮ 0.00871 ⳯ 100.5MW) ⳮ 0.00414rrup (4)

log10(PGV) ⳱ ⳮ2.49 Ⳮ 0.810MW ⳮ log10(rrup

Ⳮ 0.00871 ⳯ 100.5MW) ⳮ 0.00268rrup (5)

where the PGA unit is cm/sec2, the PGV unit is cm/sec, and
rrup is rupture distance in km. The comparisons between the
observed and predicted of PGA and PGV values are shown
in Figures 5 and 6, respectively. The standard deviations of
the residuals between the observed and predicted values (i.e.,
ln(observed) ⳮ ln(predicted)) are 0.79 and 0.75 for PGA and
PGV (Figs. 7, 8), respectively. According to the regression
results, the PGV fit is better than that of the PGA. Generally,
smaller earthquakes can produce high PGA in a near-source
area, but they cannot produce high PGV in a near-source
area. Obviously, a smaller earthquake in a near-source area
with a high PGA is caused by a more higher-frequency signal
than is a bigger earthquake. Thus, it does not produce high
PGV in the near-source area. The attenuation relationships
Figure 4. Relationship between MW and ML for
shallow earthquakes in the Taiwan region. are determined for a magnitude range from MW 4.8 to 7.6.
Therefore, it is appropriate for damage assessment applica-
tion in a rapid reporting system.

mined by regression analysis. A relation between earthquake Estimating the TSMIP Site Ground Motions
fault rupture area and magnitude can expressed as (Wyss, The TSMIP stations cover almost all of the populated
1979) areas of the Taiwan. Therefore, estimating the TSMIP site
ground motions will be very important for damage assess-
M ⳱ log10(Arup) Ⳮ C, (3) ment. Generally, site effect is one of the important factors
for predicting ground motion. The TSMIP sites still have not
where M is magnitude, Arup is fault rupture area in km2, and been well classified (Lee et al., 2001). However, those sta-
C is a constant. For the case of the Chi-Chi earthquake, MW tions have recorded many earthquakes. Thus, the TSMIP site
is 7.6, the fault area is 80 ⳯ 40 km2 (Kikuchi et al., 2000), correction, S, can be determined empirically by averaging
and then C can be determined to be 4.12, and h can be ex- the residuals between the observed and predicted values as
pressed as 0.00871 ⳯ 100.5MW. This value is used in our following:
regression calculations. A total of 1941 records are used in
regression process. Data selection criteria is as follows: n
(1) all of the T system records of the 60 events in Table 1,
and (2) source-to-site distance less than 30 km for the TSMIP
S ⳱ exp 冢1
n
兺 ln(Di /D¯ i)冣,
i⳱1
(6)

records of the 60 events in Table 1.

In the regression process we only used the source-to- where Di is either the observed PGA or PGV value, and D̄i
site distance less than 30 km of the TSMIP records. The is either the predicted PGA or PGV value obtained by the
reasons are that recording is likely more complete in the near attenuation relationships. Thus, the TSMIP site peak ground
source areas because the TSMIP instruments are maintained motion can be expressed as S ⳯ D̄i. Figures 9 and 10 show
in trigger mode. Thus, more distant instruments may not be the PGA and PGV site correction contour maps of the TSMIP
triggered when seismic ground motion decay to below a trig- stations. The results show that the western coastal plain, Tai-
ger threshold. Only instruments in the large site amplifica- pei basin, and Pingtung and Lanyang plains are places of
tion areas will be triggered. On the contrary instruments in high amplification (Fig. 1). On other hand, the Central
small site amplification areas will not be triggered. This phe- Mountain Range and the eastern Taiwan areas are places of
nomenon will cause over estimated of the attenuation curve low amplification. The results agree reasonably well with
in more distant zones. the surface geology from published maps. After applying the
The resulting attenuation relationships for PGA and PGA site amplification correction in the regression process, the
are given by standard deviations of the residuals between the observed
1222 Y.-M. Wu, T.-C. Shin, and C.-H. Chang

Figure 5. The 1941 observed PGAs and the predicted curves.

and predicted values significantly decreased to 0.66 and 0.61 at the nearest T system station, and Tcal is the predicted value
for PGA and PGV (Figs. 7, 8), respectively. by the attenuation relationships with site correction at that
Immediately after a strong earthquake, PGA and PGV T system station. Generally, the attenuation relationships and
values of all of the T system stations are available at the site corrections represent a statistically averaged effect.
CWB headquarters. The peak ground motion P at a TSMIP Every event possesses its own characteristics, such as the
site can be estimated by source radiation patterns and directivity effect. And the T
system observed values imply such characteristics. Since our
¯ ⳯ S ⳯ Tobs ,
P⳱D (7)
primary objective is to estimate the TSMIP site ground mo-
Tcal tions within 2 min after occurrence of a felt earthquake in
the Taiwan region (Wu et al., 2000), the procedure presented
where D̄ is either the predicted PGA or PGV value obtained in this article represents a first-order measure of ground-
by the attenuation relationships, Tobs is the observed value motion intensity.
Near Real-Time Mapping of PGA and PGV Following a Strong Earthquake 1223

Figure 6. The 1941 observed PGVs and the predicted curves.

Validation Tests on Estimation of Peak Values estimation scheme outlined above, it gives a PGA map much
more close to the observed PGA map given by Figure 11D.
We use the extensive recordings of the Chi-Chi earth- We conclude that our schemes are satisfactory in estimating
quake to test the estimation scheme outlined above. Figure the PGA values. However, Figure 9C still underestimates the
11D gives the reality—the PGA map based on both the T PGA values to the north of the epicenter where the Che-
system and the TSMIP observations. If only the T system lungpu rupture is terminating. This discrepancy mainly
data are used, we have a PGA map that gives the basic and comes from insufficient source correction. However, the re-
rough structure of Figure 11D. If only the calculated TSMIP sults are acceptable as first cut for practical applications.
PGA values are used, Figure 11B gives an inferior PGA map Figure 12A and B shows the Chi-Chi mainshock distri-
because the attenuation relationship is mainly based on a bution of the calculated and observed PGV values, respec-
point source that deviates from the reality of the long Chi- tively, at the TSMIP sites. Generally, the fitting is reasonably
Chi earthquake rupture. Figure 11C shows the result of the close except for a special high PGV area toward the north of
1224 Y.-M. Wu, T.-C. Shin, and C.-H. Chang

Figure 7. PGA residuals between the observed and Figure 8. PGV residuals between the observed and
predicted values in the regression process. predicted values in the regression process.

Figure 9. The PGA site correction factor contour Figure 10. The PGV site correction factor contour
map of the TSMIP stations. map of the TSMIP stations.
Near Real-Time Mapping of PGA and PGV Following a Strong Earthquake 1225

Figure 11. Maps A, B, C, and D show the Chi-Chi mainshock PGA distribution
from the T system records, the predicted values from point source without the T system
records, the predicted values, and the observed values, respectively.
1226 Y.-M. Wu, T.-C. Shin, and C.-H. Chang

Figure 12. Maps A and B show the Chi-Chi mainshock PGA distribution of the
predicted values and observed values, respectively.

the epicenter. This discrepancy again can be attributed to lution (that can provide the earthquake location and mag-
extensive length of the Chelungpu rupture and insufficient nitude in 1 min). This will be incorporated into the current
site corrections. However, the principal features in Figure CWB seismic network operation. We expect that the calcu-
12A are not too far off for emergency response purpose. We lated PGA and PGV maps will be useful for earthquake emer-
have performed a similar validation test using data from a gency response operations.
large (MW 6.5) Chi-Chi aftershock that occurred on 25 Sep-
tember 1999. The calculated and observed PGA values are
given in Figure 13A and B, respectively. The calculated and Acknowledgments
observed PGV values are given in Figure 13C and D. For
events without extended rupture, the calculated peak values We wish to thank Prof. Ta-liang Teng and Prof. Yi-Ben Tsai for
greatly improving this article and providing many thought-provoking com-
more closely approximate the observations.
ments. We also wish to thank Dr. Lupei Zhu for reviewing and greatly
improving this article. This research was supported by the National Science
Conclusions Council of the Republic of China under Grant No. NSC89-2625-Z-052-
023.
From a large set of strong-motion data of 60 shallow
large earthquakes recorded over a relatively short period of
time (5 yr), we have derived the PGA and PGV attenuation References
relationships for shallow earthquakes in the Taiwan region.
Using these attenuation relationships and the large amount Espinosa-Aranda, J., A. Jiménez, G. Ibarrola, F. Alcantar, A. Aguilar, M.
Inostroza, and S. Maldonado (1995). Mexico City seismic alert sys-
of strong-motion data, especially from those densely distrib- tem, Seism. Res. Lett. 66, 42–53.
uted 650 or so TSMIP stations, we have derived the site effect Gee, L. S., D. S. Neuhauser, D. S. Dreger, M. E. Pasyanos, R. A. Uhrham-
corrections for all strong-motion station sites in Taiwan. mer, and B. Romanowicz (1996). Real-time seismology at UC Berke-
With the attenuation relationships and the site corrections ley: The Rapid Earthquake Data Integration Project, Bull. Seism. Soc.
database, we have developed and tested a practical procedure Am. 86, 936–945.
Heaton, T. H. (1985). A model for a seismic computerized alert network,
for the calculating PGA and PGV values as soon as (in 2 min) Science 228, 987–990.
a strong earthquake happens in Taiwan. The only input this Kanamori, H., E. Hauksson, and T. Heaton (1997). Real-time seismology
peak values calculation procedure needs is the T-system so- and earthquake hazard mitigation, Nature 390, 461–464.
Near Real-Time Mapping of PGA and PGV Following a Strong Earthquake 1227

Figure 13. Maps A, B, C, and D show a Chi-Chi aftershock (25 September 1999)
(MW 6.5) predicted and observed PGA distribution, and predicted and observed PGV
distribution, respectively.
1228 Y.-M. Wu, T.-C. Shin, and C.-H. Chang

Kikuchi, M., Y. Yagi, and Y. Yamanaka (2000). Source process of the Chi- Wald, D. J., V. Quitoriano, T. H. Heaton, H. Kanamori, C. W. Scrivner,
Chi, Taiwan earthquake of September 21, 1999 inferred from tele- and C. B. Worden (1999b). TriNet “ShakeMaps”: rapid generation of
seismic body waves, Bull. Earthquake Res. Inst. Univ. Tokyo 75, instrumental ground motion and intensity maps for earthquakes in
1–13. Southern California, Earthquake Spectra 15, 537–556.
Lee, C. T., C. T. Cheng, C. W. Liao, and Y. B. Tsai (2001). Site classifi- Wu, Y. M., T. C. Shin, and Y. B. Tsai (1998). Quick and reliable deter-
cation of Taiwan free-field strong motion stations, Bull. Seism. Soc. mination of magnitude for seismic early warning, Bull. Seism. Soc.
Am. (submitted). Am. 88, 1254–1259.
Liu, K. S. (1999). Attenuation relationships for strong ground motion in Wu, Y. M., C. C. Chen, T. C. Shin, Y. B. Tsai, W. H. K. Lee, and T. L.
Taiwan, Ph. D. Dissertation, Institute of Geophysics, National Central Teng (1997). Taiwan Rapid Earthquake Information Release System,
University, Chung-Li, Taiwan, 240 pp. Seism. Res. Lett. 68, 931–943.
Nakamura, Y. (1988). On the urgent earthquake detection and alarm system Wu, Y. M., J. K. Chung, T. C. Shin, N. C. Hsiao, Y. B. Tsai, W. H. K.
(UrEDAS), in Proc. of the 9th World Conference on Earthquake En- Lee, and T. L. Teng (1999). Development of an integrated seismic
gineering, Tokyo-Kyoto, 2–9 August, Japan. early warning system in Taiwan, TAO 10, 719–736.
Nakamura, Y. (1989). Earthquake alarm system for Japan railways, Japa- Wu, Y. M., W. H. K. Lee, C. C. Chen, T. C. Shin, T. L. Teng, and Y. B.
nese Railway Engineering 109, 1–7. Tsai (2000). Performance of the Taiwan Rapid Earthquake Informa-
Shin, T. C. (1993). The calculation of local magnitude from the simulated tion Release System (RTD) during the 1999 Chi-Chi (Taiwan) earth-
Wood-Anderson seismograms of the short-period seismograms, TAO quake, Seism. Res. Lett. 71, 328–333.
4, 155–170. Wyss, M. (1979). Estimating maximum expectable magnitude of earth-
Shin, T. C., K. W. Kuo, W. H. K. Lee, T. L. Teng, and Y. B. Tsai (2000). quakes from fault dimensions, Geology 7, 336–340.
A preliminary report on the 1999 Chi-Chi (Taiwan) earthquake,
Seism. Res. Lett. 71, 24–30.
Teng, T. L., L. Wu, T. C. Shin, Y. B. Tsai, and W. H. K. Lee (1997). One
Central Weather Bureau
minute after: strong-motion map, effective epicenter, and effective
Taipei, Taiwan, R.O.C.
magnitude, Bull. Seism. Soc. Am. 87, 1209–1219.
Wald, D. J., V. Quitoriano, L. Dengler, and J. W. Dewey (1999). Utilization
of the internet for rapid community seismic intensity maps, Seism.
Res. Lett. 70, 680–693. Manuscript received 1 December 2000.