Some Seismological Aspects of the 1999 Chi-Chi Earthquake in Taiwan

The Chi-Chi earthquake is the largest inland earthquake in Taiwan during the 20th century. It caused severe damage in central Taiwan. Three possible induced earthquakes triggered by a mainshock are being exam­ ined from the strong ground motion data. The occurrence of these three­ triggered events can be used to explain the difference between preliminary isoseismal map, based on the real-time data, and final one. They can also explain why a wide area of aftershock distribution was observed. On the other hand, the overshooting of ground motion at those stations located near the northern end of Chelungpu fault is probably the reason for the complex surface rupture observed between Feng Yuan and TungShih. (


INTRODUCTION
On September 21, 1999, at 1:47 AM local time (17:47 Sep.20 UT), an earthquake of magnitude ML =7.3 took placed in central Taiwan.The epicenter is located near the town of Chi-Chi, from which the event was named.This event is the largest earthquake on the island in the past hundred years.The Central Weather Bureau (CWB) located the epicenter at 120.82°E, 23.85°N with a focal depth of 1 km which was initially determined by an automatic process.
The focal depth was later refined to 8.0 km after more data became available.As of October 13, 1999, the officially released casualty figures were 2333 dead, 10,002 injured, with more than 100,000 houses destroyed.About 80 km of rupture was observed along the Chelungpu fault with the largest measured vertical offset reaching more than 5 meters.
The purpose of this article is to discuss some seismological aspects of the Chi-Chi earth quake sequence and to present the strong motion data recorded by accelerographs which were installed under the Taiwan Strong-Motion Instrument Program (TSMIP) between 1992 and 1997, a network that covers the entire island.

HISTORICAL SEISMICITY AND LOCAL GEOLOGICAL SETTINGS
The Philippine sea plate and Eurasian plate collide with each other with the suture zone located along the Longitudinal Valley and give rise to a very complex structure in the Taiwan area.The seismicity and tectonics of eastern Taiwan are interesting simply because of the nature of the tectonic processes associated with the plate subduction and plate collision.With regard to damage, the historical records show that moderate to large earthquakes in western Taiwan have caused severe damages because they are shallow and occurred along active faults that are very close to the heavily populated areas.The Chi-Chi earthquake is a remarkable example.Figure 1 shows the locations of earthquakes with magnitude greater than 6.0 occur ring in the Taiwan area, as obtained from the CWB's Seismology Center catalog from 1900 to 1999.The solid asterisks are the earthquakes after 21 Sep., 1999 and the solid circles are those before the Chi-Chi earthquakes.It is recognized historically in the Chi-Chi area where only two earthquakes with magnitude 7 .1 were located.The Chi-Chi earthquake occurred in a region only with relatively low recent seismicity as comparing to the southwest of Taiwan prior to the earthquake� In the west-centr a l part of Taiwan, where the Chi-Chi earthquake occurred, there are at least three major Quaternary faults --the ShuangTung fault, the Chelungpu fault and the ChangHua fault --from east to west, respectively, and striking nearly north-south with an eastward dip.The ShuangTung fault is thrusting over the Quaternary Toukoshan Formation and the Chelungpu fault is thrusting over the younger alluvium deposits of the Tai Chung basin.The hanging wall of the Chelungpu fault is mainly composed of the Chinshiu shale or the Kueichulin Formation.The Chelungpu fault is thuscategorized as the major geological structure in this area.The ChangHua fault is the frontal thrust fault in the west and is partially covered by recent alluvium deposits.

MODERN SEISMIC MONITORING
At the beginning of 1990, the CWB began to install a new seismic network (CWBSN) which eventually included 73 stations in the Taiwan area (Fig. 2).Digital telemetry and digital recording of conventional three-component short-period velocity sensors were used for seis mic monitoring operation.In 1992, the CWBSN added three-component high-quality force balanced accelerometers to all CWBSN stations, and these are shown as asterisks following the station codes in Fig. 2. Their signals are also simultaneously digitally telemetered in real time to the CWB headquarters in Taipei.The new system is called as RTD, and with maxi mum ± 2g capacity, has been used for the real-time recording and processing operation.Also, an effort to develop a T a iwan Rapid Earthquake Information Release System (TREIRS) has been successfully accomplished (Lee and Shin 1997;Teng et al. 1997;Wu et al. 1997).Cur rently, the system can routinely release the location and magnitude of a strong earthquake, as well as the intensity distribution, about 60 seconds after the occurrence of an inland earth quake.This capability makes Taiwan one of the leaders in rapid earthquake response around the world.
Longitude(E)  initially as potentially heavy damage areas.It turns out that the estimation was fairly accurate.
Figure 3 shows the comparison of the isoseismal map issued preliminarily and the final one which was revised about one month later following collection of all the observed strong mo tion data.It is worth noticing that the area of intensity VI elongated toward north and south.
On examining the strong motion data at these two tip areas, we have found that at least three induced earthquakes occurred within one minute after the mainshock.The signals from the induced event are mixed with the mainshock signal.They can hardly be distinguished from one and another to be used to determine their location and magnitude (Fig. 4).The bottom-left trace is the vertical strong motion data at station CHY080 at about 32 km south of mainshock.
The curred in the northern part was placed at 121.01°E; 24.69°N and the magnitude was given as around 6.0.These magnitudes were estimated by comparing the peak ground motion with similar shallow event that occurred at a similar distance in the past.The occurrence of these three induced events can explain the difference between isoseismal maps, and the distribution of aftershocks as well.In addition, the peak ground acceleration recorded by the stations lo cated in the area of the induced earthquakes is shown on the NS component while most of other stations have recorded the peak ground on the EW component.It turns out that the mechanisms of the induced earthquakes are different from that of the Chi-Chi earthquake.
Combining the first motion polarity of short-period seismograms and strong motion data, the focal mechanism of the mainshock shows two almost north-south striking planes.The east-dipping plane, with dip 30°, strike N20°E, and rake 85°, is consistent with the geologi cally observed rupture outcrop and aftershock distribution of the area.

AFTERSHOCKS
The felt event, including aftershocks of the Chi-Chi earthquake, and the locations as de termined by the RTD system, are shown in The earthquakes at the northern and southern ends of the aftershock distribution area are due to the induced earthquakes.These events should be treated as individual earthquake sequences.
In addition, a clustering distribution in Chia Yi county is the location of the 22 October, 1999 Chia Yi earthquake.

STRONG GROUND MOTION DATA
The strong motion data recorded at many stand-alone stations in the source region were collected during the two weeks following the mainshock.Figure 6 presents two sets of accel eration data recorded at station TCU129 and TCU089.They show the highest acceleration above lG for the EW components.At station TCU089, the relatively low frequency signal shown for horizontal components may be associated with the site characteristics at TCU089, which is located on the slope of a hill.Figure 3 shows that the contours of the strong ground shaking appear to elongate along the fault line and the ground motion on the hanging wall side is stronger than that on the side at the foot of the wall.This agrees with the distribution of severely damaged areas.The result is also consistent with the focal mechanisms determined from the first motion solution as well as from the field investigation of the Chelungpu fault.
The displacement of ground motion can be obtained directly from a double integration of    ment is gradually turning from east-west to north-west northward along the fault.In addition, the uplift in the hanging wall decreases eastward away from the fault line and becomes a subsidence in the Puli area.The relatively small southeastward movements and subsidence are also observed on the foot wall side.The hanging wall exhibits larger ground movement than those on the foot wall side.The displacement signal at TCU068 integrated directly from accel eration shows the offset increasing to about 9.2 meters northwestward with a 1 meter over-shoot before returning to its final state.The similar phenomena, indicated by rebound arrows in Fig. 8, can be found in stations at the northern end of the fault.

CONCLUSIONS
1.The TREIRS of the CWB is capable of performing rapid earthquake reports for inland earthquakes in about one to three minutes.The information includes origin time, location and depth, as well as isoseismal mapping.Taiwan currently leads the world in the accom plishment rapid earthquake reporting which will ultimately be evolved toward an early warning system for significant earthquakes.The CWB target is trying to provide a 5-10 second warning time before the strong-shaking arrival of large events from the east coast.
2. The ruptural processes of mainshock are more complicated than expected.According to the analysis of strong motion data under TSMIP, at least three induced earthquakes were trig gered within one minute of the mainshock.The induced events can explain the wide distri bution of aftershocks.
3. The final displacements derived from strong motion data show good agreement with the results of fault rupture from field investigation.The overshoot shown at the northern end of the Chelungpu fault might explain the complex surface rupture observed between Feng Yuan and TungShih where some back-thrust ground failures were seen.

Fig. 2 .
Fig. 2. Station distribution of CWBSN.The 73 stations are all equipped with three component velocity sensors.The 63 stations with asterisks follow ing the station codes also having real-time acceleration digital output for seismic monitoring.

Fig. 3 .Fig. 4 .
Fig. 3. Comparison of observed and predicted isoseismal maps.Predicted isos eismal map issued three minutes after the occurrence of the mainshock was constructed using real-time acceleration data and attenuation curve (Shin 1998).The observed isoseismal map was plotted after collection of all strong motion data.The differences are discussed in the text.

Fig. 5 .
The aftershocks of the Chi-Chi earthquake were distributed over an area of about 40 by 100 square kilometers.Most of the aftershocks occurred on the eastern side of the Chelungpu fault because the fault is an east-dipping thrust fault.It is interesting to observe that the aftershocks disappeared on the eastern side against the Central Mountain Range (CMR).The three aftershocks with magnitude greater than 6.5 are all located in this area, while some smaller aftershocks occurred on the other side of the CMR It appears that the ductile material beneath the CMR plays a very important role in preventing the occurrence of earthquake, but it can transfer released energy from one side to the other.

Fig. 5 .Fig. 6 .
Fig. 5.The distribution map of felt events after the Chi-Chi earthquake.The earthquakes including the aftershocks of the Chi-Chi mainshock are those only triggered by RTD system.A clustering distribution in Chia Yi County is the place where the 22 October, 1999 Chia Yi earthquake occurred.

-Fig. 7 .
Fig. 7. Example of data processing.The acceleration records for station TCU068 are integrated to velocity and displacement shown from up to down.

�Fig. 8 .
Fig. 8. Vectorized final displacement distribution.In general, the final displace ment distribution is consistent with the satellite images and field investi gations.The values appeared in the parentheses are horizontal displace ment and its direction is counter clockwise from north.The rebound dis placement and direction for two stations at the northern end of Chelungpu fault are also shown.