Deformation of the Philippine Sea Plate Under the Coastal Range, Taiwan: Results From an Offshore-Onshore Seismic Experiment

The ongoing orogeny in Taiwan is a result of the collision of the Philip­ pine Sea and the Eurasian plates. While the structure of the continental crust on the Eurasian plate (EP) is now mostly known, that of the oceanic crust on the Philippine Sea plate (PSP) is not well mapped. Using offshore­ onshore refraction data, collected during the RN Ewing cruise of 1995, we investigate the nature of the transition between the EP and the PSP in the vicinity of the southern Coastal Range of Taiwan. The data were produced by the air-gun array of the RN Ewing along a WNW-ESE trending line off the coast of central Taiwan (MCS/OBS line 23). The refracted P waves were collected by 19 ReITek recorders with L-28 sensors placed along the southern cross-island highway of Taiwan, which is the onshore extension of line 23. Because of high noise levels in the Coastal Plain, we were only able to retrieve usable data from 11 stations located from the east coast through the Central Range. We performed forward modeling of the first arriving waves at these stations; our strategy was to search for the simplest model that fitted all the data. Our results indicate that the crust thickens gradu­ ally toward Taiwan, from about 9-12 km thick in the Huatung basin to 1518 km thick off the eastern coast of Taiwan, with a 4-8 degree dip of the Moho. Continuing westward, the crust thickens more rapidly to 27-32 km thick under the east coast of Taiwan, with a 26-32 degree dip of the Moho. (


INTRODUCTION
It is commonly recognized that Taiwan was created as the former Luzon volcanic arc collided with the Eurasian continental shelf (Biq, 1972;Chai, 1972). But the timing of the various collision events and the mechanisms and nature of the collisions are continually being explored (e.g., Biq, 1981;Suppe, 1987;Davis et al., 1983;Teng, 1990Teng, , 1996Wu et al., 1997).
1 Department of Geological Sciences, State University of New York at Binghamton, Binghamton, NY While the evolution of the orogeny can be constructed from geological mapping, the mechani cal modeling has to include the three dimensional features of the orogen. Two main hypoth eses have been proposed regarding the mechanics of the orogeny. On the one hand, the "thin skinned" hypothesis (Suppe, 1987) views the building of the Central Range of Taiwan as a result of the westward overthrust of the Philippine Sea plate (PSP) onto the Eurasian plate (EP). In this model, the mountains are built entirely out of Tertiary sediments, with the PSP acting as a "rigid indenter" or a "bulldozer" and the EP subducting underneath the PSP (Davis et al., 1983). On the other hand, the "lithospheric collision" model  proposed that the entire lithospheres of the PSP and EP are in collision, and thus shortening and defor mation should occur in both the PSP and EP.
The deformation associated with the collisional mountain building has been studied mostly on the basis of onshore geophysical and geological surveys (e.g., Ho, 1986;Wu, 1978;Suppe, 1981Suppe, , 1987. Although marine surveys around the island have been conducted since the early 1970's (e.g., Bowin et al., 1978), the deep crustal structures have not been mapped. It is evident that the structure of the oceanic crust in the vicinity of the collision can provide a strong constraint for modeling the collision. The crustal thickness represents an integrated effect of the deformation since the collision began, and its changes allow us to determine the crustal response of the PSP in the collision. In the common depiction of the thin-skinned model, the oceanic lithosphere should overlie the continental upper-crust of the EP. Alterna tively, the PSP could be considered rigid and therefore the crustal thickness would remain unchanged. Tomography using the earthquake data recorded by the Taiwan Seismic Network has already shown the presence of a root and the relatively high velocity rocks at shallow depths under the Central Range (Rau and Wu, 1995;Rau 1996). Based on these and other observations, Wu et al. (1997) concluded that the thin-skinned hypothesis is inadequate in the modeling of Taiwan, and, in its stead, proposed the-lithospheric collision model.
The marine-based data east of Taiwan can be used to further test the newly proposed model as well as the thin-skinned model. In particular, the data gathered during the 1995 RIV Ewing (cruise EW9509) offshore-onshore experiment (Figure 1 ), provide new information on the crustal structure of the PSP and the transition zone near Taiwan. The seismic waves from the air-gun shots were recorded up to offsets of about 200 km. A preliminary model of OBS line 23 by Wang et al. (1996) showed a gradually thickening of the crust from the Huatung Basin toward Taiwan. The OBS-airgun geometry yielded essentially split profiles along the line and could be modeled with little difficulty. However, since the profile stops short of the Taiwan coast, it does not cover the transition zone where changes in crustal structure are likely to be most rapid. On the other hand, the onshore stations recorded waves crossing this transi tion. Preliminary analyses Wu, 1996, 1997;Shih et al., 1996Shih et al., , 1997Lin et al., 1997) of these data show that their quality is sufficient to resolve several important features in the crust. Our present study is a more detailed analysis of the southern onshore profile, along the extension of MCS/OBS line 23 ( Figure 1). The onshore extension lies in the middle of the main collision zone of Taiwan, roughly between 22°N and 23.5°N. It is in this section that orogenic activity is the most active . The eastern end of our onshore line is the Coastal Range, which is separated from the Central Range by the Longitudinal Valley (LV; Figure 1). The LV is evidently the contact between the EP and the PSP at the surface in this study are labeled in solid black numbers. The circles refer to end points of cross-sections presented in this study, and profile E-E' is a to mography profile determined by Rau and Wu (1995). Trench. (Chai, 1972;Biq, 1972;Bowin et al., 1978;Wu, 1978;Angelier et al., 1986;Barrier and Angelier, 1986;Ho, 1986;Huchon et al., 1986). The transition zone here should be represen tative of the orogen. Unfortunately, at stations in the Coastal Plain, the signals from the airguns did not rise above the noisy background and therefore the root of the Central Range is not well sampled. Although wide-angle reflections can be used for imaging the root, they will be analyzed by us in a subsequent study and are discussed by Yeh et al. in this issue.

2.DATA
In planning the 1995 RIV Ewing offshore/onshore study east of Taiwan three MCS/OBS lines were charted, and two of them were placed along the extensions of two land instrument deployments as shown in Figure 1. The onshore lines were constrained by the layout of the highways on the island, and by utilizing two cross-island highways these lines were nearly perpendicular to the structural trend of the island. Although a reversal of the southern line was planned, using airgun shots in the Taiwan Strait, it was canceled due to operational difficul ties. Altogether 70 stations were established onshore. The 20 (8470 cu. in.) air-gun array of the RIV Ewing was fired every 40 seconds, while steaming southwest with velocity such that shot points were spaced nominally every 100 m. The onshore stations included 37 three chan nel RefTek recorders with L-28 sensors (PASSCAL, 1994). In this study, we used the vertical component records from the 19 stations on the southern profile (onshore extension of Line 23; The signals recorded at stations in the Coastal and the Central Ranges were quite clear for offsets up to about 210 km (shot 1250; see Figure 4a). The signal/noise ratios were generally enhanced when filtered with a two-pass, Butterworth band-pass filter ( Figure 3; Oppenheim and Schafer, 1989). To enhance the resolution, all time windows with high noise levels were muted after filtering. For stations in the Western Foothills, the signals were not very obvious on individual traces, but when plotted as a section the waves could be readily traced ( Figure   4b). However, for stations in the Coastal Plain area the noise level was usually quite high so that no arrivals could be discerned. It appears that the cultural noise in the Coastal Plain was mainly responsible for this problem. For example at station 231 (we use the serial numbers of the RefTek recorders to denote the stations) in the Coastal Plain, the noise level was generally higher than the signals at a neighboring station (396) on the edge of the Central Range.
In this study, we concentrate on the earliest arriving direct and refracted seismic waves, because in a complex structure, the analysis of later arrivals is made difficult by the possible presence of out-of-plane reflections and refractions. The picking of the arrivals on the section plots is relatively easy (Figures 4, 5). We picked the first arrival times on common receiver gathers (CRG). Based on source and receiver reciprocity (Lay and Wallace, 1995), the CRG are east-looking profiles (i.e. equivalent to a profile with source fixed at the receiver site with

369
We relied on forward modeling for this study. Our principle strategy was to start with the simplest layered model that could satisfy the observed data, and fit the calculated arrival times to the observed data with reasonable perturbations of the model. We added structural com plexities (i.e., more layers or lateral blocks) to the model, only if the calculated times could not be fitted with minor perturbations of the initial model. Although we could not avoid non-uniqueness problems, we ensured that the structural complexities added were reasonable. For each velocity model, we calculated travel times using an algorithm adapted from Ocola (1972), which is based on simple geometric ray considerations. We isolated several acceptable mod els and used the ray tracing package TRAMP (Zelt and Smith, 1992) to determine one average model, which simultaneously modeled the arrivals recorded at all stations. In this paper, we will give enough detail so that the results can be reproduced by other investigators, and natu rally, the end product of this study can be further tested with new analysis or when new data becomes available.
We first investigated the general features of the oceanic crust by examining the arrivals recorded at the eastern stations (554 and 338; Figure 5). Based on the apparent velocities on the CRG and CSG, we initially took the oceanic crust structure to consist of three layers (in addition to the water), with velocities in the range of 4.7-5.3, 6.2-6.6 and 6.9-7.7 km/sec, overlying a mantle with velocity of 8. 1 km/sec. We made all velocity discontinuities west dipping and planar. The mantle velocity was held fixed throughout all calculations and was based on the OBS analysis by Wang et al. (1996). We calculated travel times for the CRG from stations 554 and 338 for over 5000 combinations of boundary depths and dips. We isolated 77 combinations of depths and dips which produced a good least-squares fit of calcu lated to observed travel times, and then isolated and adjusted the model which achieved the best fit. We also compared the calculated refracted arrivals off the lower mid-crust layer and the Moho to the arrivals recorded at stations 418 and 552. We found that in order to obtain a close fit for stations 418-338, we had to introduce a dip, just east of the coast of Taiwan, on the boundary between the second and third layer ( Figure 6). We chose the final velocities to be 5.0, 6.3 and 7.3 km/sec as representative of oceanic crust. However, we could not determine these velocities uniquely with these unidirectional data alone. The first boundary was deter mined to be 8.5 ± 0.5 km below sea level (b.s.l.) and dipping 0.57 ± 0.25 degrees to the west below station 554. The second boundary was 18.5 ± 1.0 km b.s.l. and dipping 4.6 ± 0.6 degrees to the west below station 554, and 13.3 ± 1.0 km b.s.l. and dipping 2.0 ± 0.6 degrees to the west at 140 km from station 396. We also found that by introducing a gradual increase of dip in the Moho towards Taiwan, we could better model the Pn arrivals recorded at stations 418-338 ( Figure 6). However, this gradual dip in Moho was necessitated by analysis of the stations west of the LV.
To constrain the geometry of the Moho, we modeled the Pn and refracted mantle waves  Figure 7a). When we tried to explain this delay by a lower average crustal velocity under the western stations, we found that the average crustal velocity needed to be at least 1.3 km/sec lower, in addition to lowering the Moho velocity to 7 .85 km/sec under the island. Since the needed average crustal velocity change is over a distance of less than 20 km, which is unrealistic, we concluded that the Moho must be increasing dip to the west.
To determine the overall geometry of the Moho, we modeled the Moho as a smooth, continuous function with increasing dip to the west, and approaching horizontal in the Philip pine Sea basin. We calculated travel-times for more than 1000 Moho configurations, and iterated the average continental crust velocity from 4.0 km/sec to 6.5 km/sec in steps of 0.5 km/sec. We found that there were several configurations which simultaneously fit the ob served data, recorded at the western and central stations (Figure 7b). For a crustal velocity of 6.0 km/sec, the set of acceptable Moho configurations is shown in Figure 8. of Taiwan to be 6.0 km/sec, which is consistent with the tomography results of Rau and Wu (1995). We selected a particular Moho configuration which lies within the range of possible Moho configurations (Figure 8) whose Pn and refracted mantle arrivals fit data at all of the stations ( Figure 9). Obviously, this model is still a generalization, but it shows the overall structures of the crust well.

CONCLUSION AND DISCUSSION
We analyzed over 10,000 refracted P wave arrivals (Figures 4,5) from an offshore-on shore seismic experiment. The result of modeling these first arri vals leads us to conclude that under the eastern margin of Taiwan, along the onshore profile across the Southern Cross Island highway, the crust thickens from the PSP to under the east coast of Taiwan. Although the current data set allows us to resolve only a range of crustal models, there is no question that the dip of the Moho steepens significantly toward the west near Taiwan. Furthermore, this conclusion is supported by the analysis of wide-angle reflections along this same profile by Yeh et al. (this issue). The tomographic result of Rau and Wu (1995) and the interpretation of seismicity in terms of the rheology  did provide indirect evidence that the oceanic crust must have thickened in the vicinity of Taiwan, but this is the first time such phenomenon is observed directly (Figure 10).
That the crust under the Coastal Range is thickened relative to the adjacent ocean is not surprising. The Coastal Range is composed mainly of trench sediments and island-arc volcanics, evidently a foreshortened oceanic crust (Ho, 1988), with a shortening of perhaps one hundred kilometers. Such shortening would be obviously accompanied by thickening. The mecha nism for the thickening is most probably related to thrust-faulting as revealed by recent earth quakes in this region (Salzberg, 1996). With the Coastal Range at an elevation of 1 km, the presence of thickened crust to support it is also reasonable. It is interesting to note that the Bouguer gravity anomaly is positive at nearly 100 mgal at the Coastal Range. This is evi dently achieved, despite the thickened crust, by raising higher density oceanic crust to shallow     Rau and Wu (1995). See Figure 1 for locations of the cross sections.
depths, similar to what was shown by Wu et al. (1997) for the root of the Central Range.
The thickening of PSP is an integral part of the lithospheric collision hypothesis , and its existence therefore supports the hypothesis. However, by itself, the thicken ing of the PSP does not argue against the thin-skinned hypothesis, for one can modify it by incorporating a non-rigid indenter. However, the depth at which this thickening occurs is far greater than the depth of the Tertiary sedimentary wedge in question.
The other question one may raise is whether the thickening can be interpreted as west ward subduction. Since no other evidence favors the presence of a subduction zone here, crustal thickening alone is not sufficient to argue for it. However, if the underthrusting of denser oceanic lithospheric materials creates an instability, similar to the recognized process of creating a normal subduction zone, then subduction may occur in the future.