Source Parameters of Regional Earthquakes in Taiwan : 1999-2000 Including the Chi-Chi Earthquake Sequence

279 We report source parameters of 277 earthquakes that occurred during 1999-2000 in the Taiwan region, a time period dominated by the Chi-Chi earthquake sequence. We used an improved CMT inversion algorithm to handle high background noise and structural heterogeneity from complex tectonic settings near Taiwan. We also searched the best fitting solutions at an increment of 1 km of focal depths for all events that occurred after the Chi-Chi main shock on September 20, 1999. In general, central and northern Taiwan was seismically quiet before the Chi-Chi main shock. Aftershocks of this sequence dominated the regional seismicity for approximately 3 months afterward. The Chi-Chi sequence consists of a wide range of focal mechanisms, but the overall pattern is a west-verging thrust block bound by strike-slip tear faults near the northern and southern termini of the surface rupture the Chelungpu fault. Seismicity outside of the source zone of the Chi-Chi sequence generally agrees with known active tectonics from previous studies. Nonetheless, under high elevations of the eastern Central Ranges, there are several intriguing cases of changing focal mechanisms from normal faulting near the surface to thrust at depth. To facilitate further research of the Chi-Chi sequence, results of our inversion and original waveforms from the Broadband Array in Taiwan for Seismology (BATS) of 170 events within this sequence are ready for distribution on CD-ROM. The dataset is also online at the BATS web site (http://bats.earth.sinica. edu.tw/Chi-Chi_CMT).


INTRODUCTION
Development of the centroid-moment-tensor (CMT) inversion in the early 1980's allowed the systematic determination of source parameters for many large-and moderate-sized earth quakes on a global scale (e.g., Dziewonski et al. 1981;Kawakatsu et al. 1995;Sipkin, 1982).Later on, knowledge of seismic velocity structures on regional scale and establishment of regional broadband networks facilitated the routine application of CMT inversion to regional earthquakes (e.g., Dreger and Helmberger, 1993;Lay et al. 1994;Zhu and Helmberger, 1996;Pasysnos et al. 1996).
Taiwan has one of the highest seismicity in the world.Since late 1994, the Broadband Array in Taiwan for Seismology (BATS, Fig. 1) has enabled the systematic determination of reliable source parameters for regional earthquakes using the CMT technique (Kao et al. 1998, 200 1;Kao andJian, 1999, 2001).
This paper is a continuing effort to systematically report source parameters of regional events using waveforms recorded by the BATS.This report includes earthquakes that oc curred in 1999 and 2000 with most of them associated with the devastating (Mw -7.5) Chi-Chi earthquake sequence of September 20, 1999.
In response to high interest in seismic data from the Chi-Chi sequence, the Data Manage ment Center of the Institute of Earth Sciences (DMC-IES), Academia Sinica, has put in con siderable efforts to prepare event-oriented files according to the Standard for Exchange of Earthquake Data (SEED) of the Federation of Digital Seismographic Network (FDSN).To this end, this paper also serves the purpose of announcing the availability of SEED volumes recorded by BA TS through either CD-ROM or the Internet.We made this dataset complimen tary to the public in the hope to facilitate further research of the Chi-Chi sequence.

DATA AND ANALYSIS
BATS has been in constant expansion, with 12 permanent stations established prior to the occurrence of the Chi-Chi earthquake.Currently, it has 13 permanent stations over an aperture of -45 0 km by -350 km (Fig. 1).Each station consists of a state-of-the-art 24-bit data logger (Quanterra Q-series) with two sensors, one very-broadband Streckeisen STS-1 or STS-2 seis mometer and one Terra Technology SSA-320 accelerograph.Each station sends continuous data stream to the nearest Internet node via either dedicated phone line or microwave.The data packets are then transmitted back to DMC-IES in Taipei for further processing.The BATS website (http://bats.sinica.edu.tw)describes detailed configuration of the array and its current operations.
The DMC-IES is responsible for processing, quality control, storage/archiving, and data distribution from BATS.We also deposit a replica to the Data Management Center of the Incorporated Research Institutions for Seismology (DMC of IRIS) for their distribution.Currently, the DMC-IES is collaborating with the Data Center of the Ocean Hemisphere Project (OHP), Tokyo University, to establish a joint interface that offers direct data request and re trieval from both BATS and OHP data servers.Technical aspects of this interface, named NINJA, are fully described in a recent report (Takeuchi et al. 2002).Seismology (BATS).Solid and gray squares show stations currently in operation and under construction, respectively.In addition to the perma nent stations, BA TS includes 20 mobile units that can be deployed for specific tasks.The topographic contours are in 2000 m interval.CenR: the Central Range.
Our CMT algorithm begins with a background-noise evaluation to determine the fre quency band of the inversion.By comparing the power spectra of data in two windows span ning 300 s before and after the P first arrival, the low-frequency cut-off is usually at -0.02 Hz (Kao et al. 1998).We set the high-frequency cut-off at 0.05-0.08Hz to reduce the scattering effect and to avoid the calculation of high frequency portion of the synthetics.We also apply different velocity models for different station-source pairs to allow lateral heterogeneity (Kao et al. 1998;Kao and Jian, 1999).
Previously, the CMT inversion seeks the best-fitting focal depth at an interval of 3 km.We further refined this interval to 1 km for all events that occurred after the main shock of the Chi-Chi sequence, so uncertainties in focal depths can be fully explored.For all source parameters, we explore the model space through forward modeling.The uncertainties reported in Table 1 correspond to values that cause a 5-10% increase of the overall root-mean-square (RMS) misfit between observed and synthetic waveforms.To characterize the quality of inversion, we organize each re sult by assigning a letter (A F) and a number (1-4) based on the value of two parameters: the waveform misfit (E) and the amount of CLVD component (£), respectively.Thus, Al and F7 respectively represent the best and worst possible quality.Table 2 shows the range of E and E that defines each category (Kao et al. 1998;Kao and Jian, 1999).Kao and Jian (1999), we only report source parameters if (1) 3-component waveforms from at least three stations are available for inversion, and (2) the quality of result are higher than C4.Table 1 shows the results of 277 earthquakes that satisfy these conditions and Figs 2-8 show the corresponding best double-couple solutions with focal depths.Only 36 events occurred before the Chi-Chi mainshock, and most of the remaining 241 events are aftershocks of the Chi-Chi sequence.

Following previous work by
For brevity, full results of inversion including observed and synthetic waveforms appear as an appendix (in electronic form) of this article and are available at the BA TS web site (http:/ /bats.earth.sinica.edu.tw/CMT_Solutions/cmtF1999.html and http://bats.earth.sinica.edu.tw/CMT_Solutions/cmtF2000.html).Corresponding tables (including Table 1 of this study and those for earthquakes between 1995 and 1998) are also available at the same site (http://bats.earth.sinica.edu.tw/CMT_Solutions).
In this section, we briefly discuss the re sults in seven periods (Figs 2-8) with each corre sponding to a natural break in the pattern of seismicity.For the Chi-Chi sequence and the concentrated seismicity beneath the southwestern fold-and-thrust belt, Kao and Chen (2000) and Chen and Kao (2000a and b) discussed the overall spatial patterns of earthquakes in detail, which include newly discovered seismogenic structures, kinematics of faulting, and tectonic significances.
Table 2. Classification Describing the Quality of Results.3.1 Time Period Before the Main Shock of the Chi-Chi Sequence (1/1/1999 to 9/20/1999) Our results reveal two regions of frequent seismicity: one is off the eastern seaboard, particularly in a region between the southernmost Ryukyu arc and NE Taiwan.The other is beneath southwestern Taiwan adjacent to the southern end of the surface rupture during the Chi-Chi sequence.In contrast, the entire central and northern Taiwan was seismically quies cent with only one significant event beneath the Central Range (No. 14).
Focal depths of all onshore events are shallower than 25 km (shown by crosses in Fig. 2), whereas a few offshore earthquakes occurred at much greater depths, and are likely to be within the subducted Philippine Sea slab (e.g., No. 7 at 102 km, Fig. 2).The majority of events are thrust faulting.Earthquakes associated with normal faulting are either related to the open ing of the Okinawa trough (e.g., No. 13), or within the Manila trench-Luzon arc subduction system (e.g., No. 12 and 22).

3.2
The First Three Days of the Chi-Chi Sequence (9/20/1999 to 9/22/1999) The source parameters of Chi-Chi mainshock have been studied in detail using teleseismic body waveform inversion (Kao and Chen, 2000), and are not included in the present report.At the onset of this sequence, nearly all aftershocks occurred to the east of the surface rupture (the Chelungpu fault).However, the region immediately to the east of the fault, where several meters of slip occurred during the main shock, lacked any significant aftershocks (Fig. 3).The events occurred north of -23.7°N seem to form two N-S-trending bands that delineate a curi ous aseismic zone beneath the Central Range.Background seismicity prior to 1999 had al ready defined this aseismic zone (e.g., Wang et al, 1994), a persistent feature even the Chi-Chi sequence did not breach (Chen and Kao, 2000a).
A cluster of aftershocks south of -23.7°N extends NNW-SSE from the southern terminus of the Chelungpu fault into the Central Range (Fig. 3).Most of these events show strike-slip faulting.Kao and Chen (2000) and Chen and Kao (2000a) interpreted this pattern reflecting slip on tear fault(s) that marks the southern end of the west-verging Chelungpu thrust system.
To the east of the aseismic zone and beneath the eastern flank of the Central Range, several events of normal faulting mechanism occurred at depths shallower than 10-15 km (e.g., No. 41,53,and 73).Events at greater depths (-20 km and more), on the other hand, consistently showed thrust faulting (e.g., No. 69 and 76; Fig. 3).

3.3
The Next Week (9/23/1999-9/30/1999) In this time period (Fig. 4), the basic pattern of seismicity is similar to that shown in Fig. 3 except that (1) the footwall of the Chelungpu fault began to show some seismicity (No. 93, 10 1, and 110), and (2) both the northern and southern ends of the Chelungpu fault zone showed concentrated aftershocks.The region immediately to the east of the surface rupture remains absent of aftershocks except at the southernmost tip of the rupture where the strike bends from N-S to NE-SW.
Notice that aftershocks reflecting strike-slip faulting are now abundant along both north ern and southern boundaries of the Chelungpu thrust block (Fig. 4 ), further substantiate the � i) i) � Ci  �� Fig. 3. Epicenters and source parameters of regional earthquakes that occurred within the first two days after the Chi-Chi mainshock (up to 9/22/1999).
Layout is the same as that in Fig. 2.
role of tear faults as termination of this thrust system.
Under the eastern flank of the Central Range, two interesting pairs of events (No. 87 and 88, No. 106 and 107) confirmed the change from normal faulting on top (focal depths -12-15 km) to thrust faulting at depth (focal depths -27-28 km).Although the physics for such a systematic variation is not immediately clear, there is little doubt that the state of strain in upper crust is distinct from that at deeper depth beneath the eastern Central Range., <iS l f;) 110( 13) " 1()) 120{17) 1 � ) � Fig. 4. Epicenters and source parameters of regional earthquakes that occurred during 9/23/1999-9/30/1999.Layout is the same as that in Fig. 2.

291
The entire northern Taiwan including the northern half of the Chelungpu thrust system was seismically quiet during this time period.Most aftershocks of the Chi-Chi sequence scat tered south of -24°N (Fig. 5).
Fig. 5. Epicenters and source parameters of regional earthquakes that occurred during 10/1/1999-11117 /1999.Layout is the same as that in Fig. 2. Fig. 5).Earthquakes in this cluster showed either thrust or strike-slip mechanisms with most P-axes trending -E-W or NW-SE.Chen and Kao (2000b) showed that those events are mainly associated with high-angle reverse faulting beneath the southwestern foothills, a tectonic do main distinct from the source region of the Chi-Chi sequence.
of 27 Km on June 10, 2000 (No. 218).In fact, both layers of the dual, out-of-sequence thrust remained active at this late stage of the Chi-Chi sequence leaving little doubt about the impor tance of the deeper layer in seismogenesis and tectonics (Kao and Chen, 2000).
Meanwhile, high activity continued beneath the southwestern fold-and-thrust belt (No. 187 and 225;No. 192,197,200,and 202;No. 178,182,184,and 208 ; Fig. 6) with the same pattern of faulting as in previous time periods.
Background seismicity is also high off the eastern shore (Fig. 6).All earthquakes seem to reflect known patterns of active deformation such as: normal faulting (No. 220) in the opening of the Okinawa trough , downdip extensional of the subducting Philippine Sea plate at inter mediate-depths (No. 205 and 224), and interplate slip and shortening of the fore-arc region along the Ryukyu and Luzon arcs (e.g., No. 190,194,20 1,206;No. 180,191,193,214,216,respectively).

3.6
The Waning Chi-Chi Sequence (7/1/2000-8/31/2000) The pattern of regional seismicity in this time period is no longer dominated by after shocks of the Chi-Chi sequence (Fig. 7).Instead, most noticeable activity occurred offshore, immediately to the east of Hualien (-23.6-24°N;Fig. 7).Nearly all these events show thrust faulting, which are consistent with vigorous E-W collision in this region.
As in previous periods, the shallow portion of the eastern Central Range is characterized by normal faulting with T-axes sub-parallel to the overall structural trend of [246][247][248][249] .Another normal faulting event marks the opening near the western terminus of the Okinawa trough (No. 228).
One lone event occurred off the southwestern coastline (No. 261).Its mechanism of ob lique strike-slip faulting indicated N-S compression or equivalently E-W extension.Since E W /NW-SE compression dominates the western Taiwan orogen, and an m,, -6.5 normal fault ing earthquake consistent with N-S extension occurred inside the Taiwan Strait in 1994 (e.g., Kao and Wu 1996), event No. 261 may have occurred in a transition zone between contraction to extension regimes of strain (i.e., strain permutation, Hu et al. 1997)." � () � a> Fig. 7. Epicenters and source parameters of regional earthquakes that occurred during 7/1/2000-8/3 1/2000.Layout is the same as that in Fig. 2.

295
sequence are now freely available.For the user' s convenience, we have prepared a CD-ROM of event-oriented files with our results of inversion (both graphics and tables) and original waveforms recorded by BATS for 170 aftershocks of the sequence shown in Table 1.These 170 events occurred within one year after the main shock and fell between events No. 36 and

257.
For each event in the CD-ROM, we performed quality control on available BATS stations and then extract a 15-minutes segment of data from each broadband channel, commence at 300 s before and terminate at 600 s after the first P arrival.Together with the corresponding instrument and network information, we then create a SEED volume by combining all ex tracted waveforms for each event.Depending on the number of available waveforms, the file size for each event varies from -0.9 to -2.5 MB .
This dataset on CD-ROM and an electronic version of Tables 1-2 and Figs 2-8 are also readily available through the Internet at the BATS website (http ://bats.earth.sinica.edu.tw/Chi-Chi_ CMT) .Epicenters and source parameters of regional earthquakes that occurred during 9/1/2000-12/3 112000.Layout is the same as that in Fig. 2.

SUMMARY
We report a systematic study of centroid-moment-tensor (CMT) solutions including re fined focal depths from our previous reports for regional earthquakes that occurred near Tai wan during 1999-2000.We primarily used waveform data recorded by the Broadband Array in Taiwan for Seismology (BATS) in the inversion.For the largest events, our results have been corroborated by the inversion of teleseismic broadband data (Kao and Chen 2000;Chen andKao 2000a, 2000b ).Among a total of 277 earthquakes in this report, most of them belong to the Chi-Chi earthquake sequence except the 36 events occurred prior to the main shock of this sequence.
To improve estimations on uncertainty of focal depths, we performed inversion at a re fined interval of 1 km for focal depth for all events that occurred after the Chi-Chi mainshock.As in previous reports (e.g., Kao et al. 1998), we make realistic estimations of uncertainties by searching the model space through forward modeling.

Fig. 1 .
Fig. 1.Map showing the configuration of the Broadband Array in Taiwan for Seismology (BATS).Solid and gray squares show stations currently in operation and under construction, respectively.In addition to the perma nent stations, BA TS includes 20 mobile units that can be deployed for specific tasks.The topographic contours are in 2000 m interval.CenR: the Central Range.

Fig. 2 .
Fig. 2. Epicenters and source parameters of regional earthquakes that occurred in 1999 before the Chi-Chi sequence.Event numbers correspond to those in Table 1.Symbols show different ranges of focal depths: crosses for 0-25 km, solid dots f o r 25-50 km, and solid triangles for >50 km.For reference, the open star and solid curve mark the epicenter of the main shock and the surface break (i.e., the Chelungpu fault), respectively.The corresponding focal mechanisms (lower-hemisphere projection with dark ened areas showing quadrants with compressional P wave first motions) appear below the map.Above each fault plane solution, we show the event number and the focal depth (in parenthesis, unit in km).