Tectonic Implications of 1998 , Ruey-Li , Taiwan , Earthquake Sequence

We carried out a seismotectonic study on the 1998 Ruey-Li, Taiwan earthquake. This ML = 6.2 event occurred in a close neighborhood to and about one year before the Mw = 7 .6 disastrous Chi-Chi, Taiwan earthquake of 1999. Both the Ruey-Li and the Chi-Chi events share a common re­ gional stress system. From the relocated hypocenters, focal mechanisms and stress inversion of the Ruey-Li sequence, we find that the spatial distri­ bution of aftershocks forms two distinct groups, one is consistent with a planar thrust, the other gives a peculiar 3 km X 5 km x 15 km of nearly vertical columnar distribution made principally of left-lateral strike-slip faulting. A regional block rotation model is proposed to explain the rup­ ture process; this model is supported by the well-documented GPS data. (

The damaging Ruey-Li, Taiwan earthquake, magnitude M� = 6.2, occurred on July 17, 1998 in southwestern Taiwan, roughly 25 km to the northeast of the large city of Chia-Yi.This event was followed by an energetic but unusually short (about one week) aftershock sequence with a peculiar spatial distribution.About one year later, the disastrous Mw = 7.6 Chi-Chi, Taiwan earthquake initiated a short distance north of the Ruey-Li epicenter with a similar but thrust mechanism with a 80-km long surface break coming close to the Ruey-Li epicenter.According to the Central Weather Bureau Seismic Network (CWBSN) report, the number of aftershocks was hundreds per day for the first 3 days, but decreased rapidly to several per day and ended abruptly in about a week.The epicenter of the main-shock was located in the CWBSN report at 23° 30.16' N and 120° 39.75' E, with a focal depth of 2.8 km.
The fault plane solution for the mainshock is an oblique thrust; for the larger aftershocks (Chen et al. 1999), two types of focal mechanism are shown: strike-slip in one larger cluster and thrust in the smaller cluster.A shallow event, the Ruey-Li earthquake caused landslides, rock falls and damage to buildings, and injured more than 25 people.
The source area of the 1998 Ruey-Li earthquake is located on the western margin of the Western Foothills province.The geological survey shows a thick sequence of shallow marine to shelf elastic sediments ranging in time from late Oligocene and Miocene to early Pleis tocene in this area (Huang 1980).Figure 1 shows the major fault system of the study area as mapped by surface geology.The dominant structure is a combination of asymmetric folds and low-angle thrust faults striking northeast and dipping southeast, a result of deformed rocks in the Western Foothills (Ho 1976;Suppe 1980).The roughly parallel thrust faults from west to east are the Tachienshan fault, the Chukou fault, the Shihkuping fault, the Luku fault and the Shechiunhu fault.There are two northwest-trending left-lateral strike-slip faults: the Shuisheliao and Neipang.The southern part of the Tachienshan fault and the northern part of the Neipang fault show significant variation in strike and slip direction.The Tachienshan fault, northern segment of the Chukou fault, separates the area into two distinct geological structures (Liu and Lee 1998).The structural manifestation to the west of the Tachienshan fault is simple with gentle folds and fewer faults, while those to the east are complex with tighter folds and many faults (Keng 1986).The faults trending northeast were probably formed later in the event and were followed by northwest-southeast striking and left-lateral strike-slip faults (Tsan and Keng 1962;Liu and Lee 1998).The folds striking in the northeast may have been formed during the earlier Penglai Orogeny of Plio-Pleistoce time (Liu and Lee 1998).
However, no surface break was observed.We have relocated 98 events (Table 1) that occurred within the period of 24 hours after the mainshock.Improved locations give two tighter clusters than those given by the CWBSN.This paper focuses on the analysis of focal mechanisms and stress inversion for a discussion on the seismotectonic significance of the Ruey-Li earthquake vis-a-vis the regional structural and deformation background.We have also calculated the stress induced by the mainshock of the earthquake sequence to discuss whether the slip of the mainshock affected the distribution of aftershocks, and whether it trig gered some nearby faults, as indicated by aftershocks with different focal mechanisms and locations.

Earthquakes Relocation
The Ruey-Li earthquake sequence was well recorded by the CWBSN short-period net work and strong-motion telemetered stations.The dense nearby free-field stations of the Tai wan Strong-Motion Instrumentation Program (TSMIP) also contributed data of both the rela tive arrival times and P-polarities.The hypocenters of the Ruey-Li earthquake sequence had been routinely obtained by 1-D crustal model by CWBSN (e.g., CWB report 1999); improve ments can be made by using an appropriate 3-D crustal model to account for lateral velocity variations.We firstly relocated the mainshock with a 3-D crustal model for central Taiwan derived from a tomographic study by Ho and Shin (1994) in order to improve the hypocenter location in a region of strong lateral velocity heterogeneity.Besides the first arrival times of the P-and S-wave of the mainshock given by the CWBSN, we further obtained the S-P times from the numerous nearby TSMIP free-field strong-motion stations to properly constrain the hypocenter solution.An earthquake location procedure by Virieux et al. (1988) using 3-D ray tracing was performed with the CWBSN 1-D solution as the initial guess.Far smaller RMS of travel time residuals (0.13 sec.) of the 3-D solution than that of 1-D solution (0.35 sec.)shows improvement on the hypocenter location determination.The relocated hypocenter of the mainshock shows small changes but confirms that the mainshock was a shallow (2.4 km) event.
To relocate the aftershocks, we used a method to determine high-resolution hypocenter locations over large distances developed by Waldhouse and Ellsworth (2000).This method determines hypocenters by minimizing residuals between observed and theoretical travel-time differences for pairs of earthquakes (or double-differences) at each station while linking to gether all observed event-station pairs.It permits the recognition of segment boundaries and fault bends that are believed to play important roles in the initiation and arrest of a rupture.182 aftershocks occurred roughly within the first 24-h period (except for 3 events) after the mainshock was relocated (Table 1) using the double-difference method.These 182 after shocks were selected, each having at least 8 nearby P and S arrival-time readings.The relo- cated hypocenters give a picture of 3-D event distribution of the Ruey-Li earthquake sequence that is somewhat different from the solutions obtained based on 1-D velocity structure.The differences between the CWBSN 1-D and the relocated 3-D solutions resulted from the strong lateral variation of crustal velocities.That the western part of the research area shows positive values of the station correction that implies lower velocity in the Western Coastal Plain than in the Western Foothills region and is consistent with the surface geology.
Figures la and b illustrates the mainshock and those 182 aftershocks studied, before and after the relocation, respectively.An insert in the lower right of both figures gives the location map and nearby seismic stations.Also shown in the insert is the outline (dashed curve) of the Pekang High (PKH) which is the Tertiary basement that comes close to the surface of the western Taiwan sedimentary basin.The epicenter distribution of the relocated aftershocks shows one linear cluster and one round cluster: the linear cluster (Group A, in black dots in Fig. 1 b) is extended along the direction of roughly N45°E, and the 3 km x 5 km round cluster (Group B, in open circles in Fig. 1 b) is concentrated at the eastern side of the linear cluster.The vertical profile along the E-W direction (Fig. 2a) also shows two spatial alignments of hypocenters.Both Group A and Group B show narrow distributions of events with roughly vertical apparent dips.The mainshock shown by a star occurred at the upper part of these clusters.In the vertical profile along the N-S direction (Fig. 2b), Group B events still show a narrow distribution with nearby vertical apparent dip, and Group A events scatter about in the profile.These profiles demonstrate that the Group A events and the mainshock occurred over a rupture plane of nearly vertical dip.The Group B events are not distributed over a planar surface, instead, they congregate in a columnar volume of 3 km x 5 km x 15 km in dimension.The mainshock is located near the top of this "event column".This columnar distribution of the aftershocks is very peculiar.We shall discuss its nature later.

Focal Mechanisms
Focal mechanisms of the mainshock and 21 aftershocks (Table 2) of the Ruey-Li earth quake sequence were studied with the fault-plane solutions obtained using the first P-motion data from both short-period and strong-motion instruments.This procedure increased both the quantity and the reliability of the first motion data.The azimuths and the take-off angles of the ray paths of the direct P waves were calculated with the same dynamic ray tracing technique and the 3-D velocity model used for the event relocation.The FPFIT software (Reasenberg and Oppenheimer 1985) was used to obtain the fault plane solution.Figure 3 shows the fault plane solutions of the mainshock and 21 aftershocks, the lower right insert gives the location map and nearby seismic stations.Also shown in the insert is the outline (dashed curve) of the Pekang High (PKH) which is the Tertiary basement that comes close to the surface of the western Taiwan sedmentary basin.The presence of PKH introduces a further complication to the local stress field.A large number of Group A events, including the mainshock, show oblique thrust or thrust-type faulting, and most of the fault-plane solutions in Group B show left-lateral strike-slip fault type faulting.We surmise that the mainshock and Group A events are associated with a nearly vertical east-dipping rupture plane.The Group B events are asso ciated with a short segment of the Shuisheliao left-lateral strike-slip fault.The columnar

Stress Inversion
Although there are several methods for inverting focal mechanisms for the stress tensor The first is that it deals with first motions directly rather than via the intermediate step of focal mechanisms.This allows the use of the numerous smaller aftershocks rather than the more limited set of larger aftershocks for which we have focal mechanisms.The second is that it invokes the Coulomb failure criterion and provides us with a more realistic view of rupture generation.There are 625 clearly determined polarities collected from the 183 events in the earthquake sequence, including events that give no fault-plane solutions.The results are given in Fig. 4, which show the horizontal a 1 -and a 3 -axes.The best stress tensor explains 95 % of the polarity observations.This orientation of the regional stress tensor is quite close to that inferred from shear-wave splitting (Chen and Yen 1998) and other geological and geophysical observations (Lee and Chu 1991).The 95 % confidence level fora 1anda 3is obtained from 1000 resamples using a method by Michael (1987).The two stress axes are indicated by the two black dots.The tighter 95 % confidence limit area of a 1 implies the compression axis is

TECTONIC IMPLICATIONS AND CONCLUSIONS
The distribution of Group A and Group B events implies that these events did not occur by the GPS velocity (open) arrows, the local deformation is represented by the deformation ellipse in response to a maximum compressional stress measured by fault-plane solutions and stress inversion.As a result, oblique slips are found on both frontal thrust and tear faults (block models at center).Note that the principle compressional stress direction observed from geodetic surveys in this area does not have to be parallel to cr 1 , which is inferred from fault plane solutions and stress inversion.For the GPS results reflect the finite strain accumulation, while the cr 1 -axis is determined by the dynamic motion induced by sudden rupture.In addition, the rotational motion (or the rotation tensor), which cannot be detected by the seismic sensors, should exist and would affect the finite strain accumulation measured in a large area by GSP of other geodetic surveys.
The existence of rotational strain in the study area is a key point to the above interpretation.With the geodetic deformation velocity data (Yu et al. 1997), we calculated the horizontal component of the block rotation in this area.By applying the cubic-spline technique to the 2-D GPS data, we obtained the spatial derivatives of deformation velocity for grid points.Then, the angular velocity of rotational tensor on the horizontal plane for all points were calculated according to the rotation tensor:

37
The results (Fig. 6) show that the portion with fast clockwise rotation movement extends as a band-like area with a NE trend in which the Ruey-Li earthquake sequence occurred.This explains how the block rotation process plays a role in the regional tectonics of Taiwan defor mation front area.The process of fault block rotation can leave a fault in an unfavorable orientation relative to the stress field and has been observed in the Mojave area (Ron 2001).
This process will lead to increasing normal stress and decreasing shear stress acting across those faults (Nur et al. 1986) until the faults lock.As the crustal deformation is continuing under the block rotation process, neighboring faults that have not moved, such as Chelungpu fault, will bear strong stress and, admittedly a hind sight now, would more likely be triggered into motion.

CONCLUSION
The Ruey-Li earthquake sequence provided us an opportunity to better understand the seismotectonics of the complex fold and thrust belt.With the analysis of events location, focal mechanisms and the direction of tectonic stress, we infer that a block rotation process is pro ceeding in the study area, which is supported by the well-attested GPS data.This process explains how the block rotation process plays a role in the regional tectonics of Taiwan defor mation front area and leads to increasing normal stress and decreasing shear stress acting across those faults.In this case, large thrust movements on neighboring faults, such as the Chi-Chi earthquake, could possibly be triggered.

Fig. 1 .
Fig. 1. (Upper) The epicenter distribution of Ruey-Li earthquake sequence included mainshock (star) and aftershocks that occurred during 24 hrs after mainshock de rived from 1-D velocity structure (CWB).The major geological fault system is also shown.(Lower) The relocated epicenter distribution.The solid triangles in the figure index indicate the stations used for relocation.

Fig. 2 .
Fig. 2. The vertical profiles ofrelocated events along the directions of East-West and North South respectively.

(
Gephart and Forsyth 1984; Rivera and Cistemas 1990; Horiuchi et al. 1995), we applied the method described by Robinson & Mcginty(2000)  to invert the observed polarities for the orientation of principal stress axes.This method was applied to this study for two reasons.
better determined than a 3 -axis.The two groups of distribution of a 3 indicate two fault types with the same a 1-axis, accounting for the change of a strike-slip to a thrust faulting: the strike slip type is dominant and the other is oblique faulting is less so.However, considering that the mainshock is an oblique rupture, the principal deformation is still dominated by a thrust-type motion.The "beach balls" show the two obtained fault planes (indicated by short bars) from inversion with traditional focal mechanism projection.The loweer one is the optimal fault plane for having fewer polarity mismatches.The preferred fault plane shows a vertical left lateral strike-slip fault type and is consistent with both the distribution of the Group B events and the strike of the nearby Neipang fault.

Fig. 3 .Fig. 4 .
Fig. 3. Focal mechanisms of selected (22) aftershocks.The solid and open circles represent events of Group A and Group B respectively.

Fig. 5 .
Fig. 5.A schematic diagram illustrates the bloc k model and the relation ship between the seismic and GPS observations.Please see text for further detail.

Fig. 6 .
Fig. 6.The contour map of element values ofrotation tensor (vertical axis) around the source area of the Ruey-Li earthquake.

Table 1 .
The source parameters of 183 earthquakes of the Ruey-Li earthquake sequence that occurred roughly Number Year Month Day Hour Minute Sceond Longitude Latitude Depth ML Number Year Month Day Hour Minute Sceond Longitude Latitude Depth (TableI.continued)

Table 2 .
The fault plane solutions of 22 selected aftershocks.