Crustal Velocity Variation of the Wes tern Philippine Sea Plate From TAICRUST OBS/1\1CS Line 23

The aims of this research are to understand the deformation of the shallow structures ( <6 km depth), the crustal velocity variation of the west­ ern Philippine Sea Plate (PSP) and their relation to the arc-continent colli­ sion using Ocean Bottom Seismometers (OBS) and a multi-channel seismic (MCS) survey (together known as TAICRUST) offshore southeastern Tai­ wan. A seismic line, Line 23 which covers the Luzon Arc, the Huatung Ba­ sin, the Taitung Canyon and the western edge of the Gagua Ridge, is inves­ tigated. Prominent reflected and refracted arrivals from the sediment, the oceanic basement and the Moho can be seen in the OBS data. By applying the travel-time inversion of the stacked MCS data and the OBS data, the velocity-depth model is built sequentially from the shallow to the deep struc­ tures. Low RMS travel-time error and numerous travel-time picks demon­ strate the accuracy and the high resolution of the model, respectively. A deep basement beneath OBS stations 30 and 31, a narrow basement trough beneath the Taitung Canyon, and the long-wavelength bending of the oceanic crust near the western edge of the Gagua Ridge are found in the velocity model. Anomalously low velocity in the upper crust is also identi­ fied beneath the Taitung Canyon and near the Gagua Ridge. The former may result from the strike-slip fault while the latter may be generated from the uplift of the Gagua Ridge. According to the variation of the crustal thickness, the velocity model is divided into three portions with the dis­ tance larger than 74 km, between 23 and 74 km, and less than 23 km from its northwest end. The crustal thickness in the southeast portion (>74 km) is almost uniform at about 12 km. Similarly, the thickness of the upper crust in the central model (23-7 4 km) and the thickness of the lower crust in the northwest portion ( <23 km) remain uniform at about 4 km and 8 km, respectively. However, the lower crust in the central portion and the upper crust in the northwest portion gradually thicken northwestward. The maxi­ mum crustal thickness is about 24 km at the northwest end of the velocity model. The variations of the crustal thickness and the lateral velocity at a 1 Institute of Applied Geophysics, National Taiwan Ocean University, Keelung, Taiwan, 20224, ROG 379 380 TAO, Vol. 9, No. 3, September 1998 distance of 23 km from the northwest end of the model imply the eastern edge of the Luzon Arc. Furthermore, northwestward dipping of the Moho in the velocity model is consistent with other studies. The mechanism of crustal thickening in the western PSP is probably related to intra-plate de­ formation, thrust faulting and/or future subduction of the western PSP beneath the Luzon Arc. (

distance of 23 km from the northwest end of the model imply the eastern edge of the Luzon Arc. Furthermore, northwestward dipping of the Moho in the velocity model is consistent with other studies. The mechanism of crustal thickening in the western PSP is probably related to intra-plate de formation, thrust faulting and/or future subduction of the western PSP beneath the Luzon Arc.  Figure 1. At present, offshore southeastern Taiwan, the western PSP converges with the EP at a rate of 7.1 cm/yr. and along ar. azimuth of 310° (Seno et al., 1993). Collision of the western PSP and the margin of the EP has resulted in both the northward extension of the Luzon volcanic arc and the emergence of the Coastal Range (Ho, 1988). Deformation may have accumulated within the Luzon Arc owing to arc-continent collision, but how has it fur ther affected the basement of the western PSP and where is the eastern edge of the Luzon Arc?
More geophysical data are required to determine the crustal structure beneath the western PSP and for a better understanding of the collision of the Luzon Arc and the EP.
Numerous geophysical surveys have been carried out offshore southeastern Taiwan. Earth quake tomography (Rau and Wu, 1995) provided a general trend of the tectonic structures without good constraints of the marine structures. By considering the high-resolution bathy metric data within the Huatung Basin, both the depth of the Taitung Canyon (extending from Lanyu and Lutao islands to the Ryukyu Trench for a distance of 170 km) and the elevation of the Gagua Ridge (along the 123°E meridian) were identified (Schntirle et al., 1998). Accord ing to the gravity signature, the Gagua Ridge appears to be an uplifted block of the oceanic crust formed under highly oblique convergence (Karp et al., 1997;Deschamps et al., 1997;Hsu et al., 1998) that may explain the bending of the oceanic crust near the Gagua Ridge.
However, bathymetry and gravity modeling are generally insufficient to interpret the deep structures of the western PSP. Two fracture zones trending NS along 122.5°E and beneath the Taitung Canyon (as an active right-lateral strike-slip fault) have been inferred, respectively, from magnetic lineations (Hsu et al., 1998) and from focal mechanisms (Sibuet and Hsu, 1997). However, more evidence is required to better define these fracture zones in crustal structures. Contour interval of the bathymetry data (Hsu et al., 1996) is 500 meter.

381
out in September 1995 to investigate the deep structures and geodynamic processes of the arc continent collision zone offshore southeastern Taiwan (Liu et al., 1995;Wang et al., 1996b;Yang et al., 1996;Lin et al., 1997;Yeh et al., 1998). The preliminary results of this survey show that the oceanic basement of the western PSP beneath the Huatung Basin generally deepens westward east of the Luzon Arc (Chen, 1996;Yang et al., 1996;Hetland and Wu, 1998;Yeh et al., 1998). The aims of this paper are to fu rther understand deformation of the shallow structures, the crustal velocity variation of the western PSP, and their relation to the arc-continent collision from the seismic survey offshore southeastern Taiwan.

OBS REFRACTION SURVEY
In the summer of 1995, a combined Ocean Bottom Seismometer (OBS) and multi-chan nel seismic (MCS) survey was conducted offshore eastern Taiwan with OBSs deployed from RIV Ocean Researcher I and an air-gun arr ay and a multi-channel streamer from RN" Maurice Ewing. The seismic survey was designed to investigate the tectonic structures of the south western Ryukyu subduction zone (Wang et al., 1996a and1996b), the Ryukyu forearc system (Wang and Chiang, 1998;Mcintosh and Nakamura, 1998), the western PSP (Yang et al., 1996) and their interactions.
An OBS/MCS line (Line 23), lying in a NW-SE direction and having 7 OBSs (black circles in Figure 1), was acquired for exploring the crustal structures of the western PSP. The seismic line crosses the Luzon Arc, Huatung Basin and Taitung Canyon, and ends near the western edge of the Gagua Ridge with the total length of about 150 km. The north-south variation of the crustal structure beneath the Huatung Basin can be seen from the southern portion of OBS Line 1 (Wang et al, 1996a). OBS stations along Line 1 are denoted by white circles in Figure 1. The intersection of Line 1 and Line 23 at OBS stations 13 and 33 can provide more constraints for building the velocity models along both lines.
In this paper, we describe a velocity-depth model derived from OBS and MCS data and discuss its implications for the Luzon Arc and the western PSP offshore southeastern Taiwan.
It is hoped that, by combining the velocity structures from both OBS Line 23 in this study and the land stations (white squares in Figure 1) across southern Taiwan (Chen, 1996;Lin et al., 1997;Yeh et al., 1998), we can provide better constraints on the crustal structures of the Taiwan Orogeny and new insights into the mountain building processes of Taiwan.

OBS DATA PROCESSING
OBS data processing in this study includes first arrival picking, velocity model building, ray tracing and travel-time inversion. These processing procedures are described as follows.

Initial Processing
By using an interactive software, OBSTOOL (Christeson, 1995), OBS raw data are pro

Building a Starting Velocity Model
It is sufficient to build a shallow velocity model, based on the bathymetry ( Figure 1) and the stacked MCS data (Figure 3), along the seismic profile. Several horizontal layers, with increasing velocity for deeper layers, are set in the starting model of the shallow structures.
Four interfaces are readily identified from Figure 3. The uppermost interface, the seabed, shows that the depth of the sea floor increases toward the southeast. The two-way time of the water-bottom reflections increases from 2.5 sec at the northwest end to 6.5 sec at a distance of

Ray Tracing and the Travel-time Inversion
We comput travel times of tracing the reflected and refracted rays in the starting velocity model. Differences of the calculated and picked travel times are then considered in the inver sion (Zelt and Smith, 1992) to update the velocity model. If the calculated and picked travel times do not fit well after several inversions, we modify the starting model, re-pick the far offset travel times and re-start the inversion. The modifications are especially needed for deep structures and far-offset travel times because they are less certain.
In this paper, the velocity model is built from the shallow to the deep structures succes sively to ensure good travel-time fits at each step.

RESULTS AND DISCUSSIONS
According to the OBS velocity model of the shallow structures ( Figure 5), the travel times of the normal incidence from the shallow interfaces are calculated as three solid lines in Figure  3 that generally fit the stacked MCS data well. In addition, the velocity gradient shown in Figure 6 is obtained from the velocity model in Figure 7(a) by applying the Fourier transform along the seismic profile. The strong gradient contrast in Figure 6 is comparable with the interfaces of the velocity model as displayed by the solid lines. In view of the stacked MCS data in Figure 3, the contrast of the velocity gradient in Figure 6 and the strong velocity con trast in Figures 5 and 7(a), four velocity contours at 1.51, 3, 4.5 and 7.8 km/sec are labeled as the interfaces in the velocity models of Figures 5 and 7(a).
Based on a water-wave velocity of 1.5 km/sec, the velocity contour of 1.51 km/sec in the model indicating the sea floor is consistent with the water-bottom reflections in the MCS data. It is suggested that velocity contours of 3 and 4.5 km/sec in Figure 5 are, respectively, the bottoms of the sediment and the compacted sediment that match two of the lower interfaces in Figure 3. In addition, the observed PmP arrivals, illustrated by the green bars in Figure 4 (Hsu et al., 1996) in Figure 1, the sedimentary section beneath Line 23 comes mainly from the Taiwan mountain-belt through the Taitung Canyon.

Volcaniclastics and Compacted Sediment (Vp = 3-4.5 km/sec)
Since the Neogene depositions in the Coastal Range are mainly marine and partly volcaniclastic (Ho, 1988), the velocity layer between 3-4.5 km/sec is possibly composed of "volcaniclastics" and older sediments at distances of 0-70 km in Figure 5. The thickness of this section at the northwest side of the model is about 4 km and becomes thinner southeast ward. Deformation of the sea floor and the deeper section within a distance of 23 km from the northwestern end are seen in Figure 5 and may result from the collision of the Luzon Arc and the EP. In addition, a deep basement between OBS stations 30 and 31 was found 7 km below the sea level as shown in Figure 5. The bottom and the vertical offset of the deep basement are 15 km wide and 1 km deep, respectively. Since shallow and deeper sedimentary sections above the deep basement are not deformed, it is implied that the deep basement was formed before the collision, probably during the Eocene/Oligocene. Hence, the volcaniclastics in the deep basement were filled by the sediment probably after 41 to 33 Ma, when the oceanic basement was formed owing to the collision of the western PSP and the EP (Hsu et al., 1998).
No volcaniclastics exists southeast of OBS station 31 except for two compacted sedi ments with a thickness of 1.5 km below Taitung Canyon and a thickness of 1 km southeast of OBS station 33 as illustrated in Figure 5. These compacted sediments are interpreted to have been deposited in a narrow basement trough beneath the Taitung Canyon and the long-wave length bending of the oceanic crust owing to the uplift of the Gagua Ridge (Karp et al., 1997;Deschamps et al., 1997). The narrow basement trough beneath the Taitung Canyon may result from the right-lateral strike-slip fault of an old N-S trending fracture zone in the Huatung Basin (Lee and Hilde, 1984;Hsu et al., 1996;Sibuet and Hsu, 1997;Hsu et al., 1998). 4.3 Crust (Vp = 4.5-7.8 km/sec) In Figure 5, we found three depressions in the crust that may result in continuous deposi tion of the compacted sediments below OBS stations 30, 32 and 33. These phenomena have been confirmed in this paper by adjusting various velocities and depths below the depressions of the compacted sediment for the best travel-time fits. Furthermore, the three pairs of nega tive and positive gradients (denoted, respectively, by blue and green in Figure 6) demonstrate the existence of three depressions in Figure 5.
The crustal thickness of the western PSP increases northwestward from about 10 km near the Gagua Ridge to about 26 km below the Luzon Arc in Figure 7(a). The velocity contour of 7 km/sec is chosen as the boundary between the upper and lower crust because the velocity contours of less than 7 km/sec are dense as shown in Figure 5. In addition, according to Figure  6, the velocity contour of 7 km/sec fits well as the bottom of the anomalously low velocity in the upper crust. The bottom of the upper crust drops abruptly from a 10 km depth (between OBS stations 28 and 29) to about a 22 km depth (at the northwest end) northwestward in Figure 7(a). Similarly, the depth of the velocity contour of 7.5 km/sec (between the velocity contours of 7 and 7 . 8 km/sec in Figure 7(a)) increases northwestward from 12 km depth (be neath OBS station 31) to 28.5 km depth (at the northwest end).  (Chen, 1996;Hetland and Wu, 1998;Yang et al., 1996;Yeh et al., 1998).

CONCLUSIONS
High-quality OBS data have been acquired in a regional study comprising 7 OBS deploy ments along OBS Line 23 offshore southeastern Taiwan. By using travel-time inversion of OBS data and MCS data, we have built a high-resolution velocity model as shown in Figure   7(a). Figure 7(b) displays the normalized ray density of the travel-time modeling. The ray distribution is high for depths less than 20 km, but few rays can penetrate through and reflect from the Moho as illustrated by Pn and PmP arrivals in Figures 2(a)-(c), so that ray coverage is sparse in the lower crust of Figure 4(c).
Based on the velocity model established in this study, thick sediment from the Coastal Range have been deposited in the Huatung Basin. Three depressions (the deep basement at the east of the Luzon Arc, the narrow basement trough beneath the Taitung Canyon and the bend ing near the Gagua Ridge) have been found in Figure 5. Since the sediment is not deformed along the OBS line 23, the three depressions probably existed before the arc-continent colli sion.
On the basis of travel-time inversion in the paper, the oceanic crust around and below the sedimented depressions appears to have anomalously low velocity beneath the Taitung Can yon and near the Gagua Ridge. The former may be generated by the strike-slip fault (Hsu et al., 1998;Schntirle et al., 1998) while the latter may result from the uplift of the Gagua Ridge (Karp et al., 1997;Deschamps et al., 1997). The crustal thickness is almost uniform in the southeastern model (>74 km) of Figure 7(a). In the central model, the crustal thickness gradu-ally thickens northwestward mainly from the lower crust. The upper crust at the northwest side ( <23 km) is responsible for the thickening that the maximum crustal thickness reaches about 24 km. The variation of the crustal thickness in Figure 7(a) and the strong contrast of the velocity gradient in Figure 6 at a distance of 23 km imply the eastern edge of the Luzon Arc.
Furthermore, the northwestward dipping of the Moho in the velocity model is also consistent with other studies (Chen, 1996;Yang et al., 1996;Hetland and Wu, 1998;Hsu et al., 1998;Yeh et al., 1998). The mechanism of the crustal thickening in the western PSP is probably related to thrusting (Salzberg, 1996), intra-plate deformation or future subduction (Hetland and Wu, 1998) of the western PSP beneath the Luzon Arc.