Velocity Structure Near the Northern Manila Trench:an OBS Refraction Study

1992 to 1993. Although the newly established system is still in its early stages, t.he data collected make it possible to define a crustal structure in the vicinity of the northern Manila Trench. The arrivals of deep refraction with apparent velocity at about 8.1 km/sec help to define the depth of the Moho discontinuity 'vhich is about 1 2 km below sea level. A thin crust of merely 5 to 6 km in thickness underlying the sediments to the west of the trench in the study area suggests an oceanic origin of the crust there. The thickness of the sediments is maintained over 3 km throughout the area, but a thicker crust to the east of the trench than that to the west is demon­ strated.


INTRODUCTION
Located at the boundary bet�'een the Eurasian and Philippine Sea plates, the island of Taiwan� including its surrounding areas, is one of the most perplexing regions in the world in tern1s ot� tectonic complexity ( Figure 1 ). Because man)' key structural features of the offshore region are cove1�ect with thick sediments and have not been fully investigated, se\leral conflict ing hypotheses regarding the northern extension of the Manila Trench have bee. n proposed but have not been backed up with solid data. For instance, Big ( · 19 · 77) proposed that the trench might run northv�rard near 20.5°N and extend onshore th1�ough . Fongliao Canyon in southern Tai wan; Lin and Tsai ( 1981) argued that it might terminate some.where near 21 °N and shift eastward to 121°E by a transfor1n fault parallel to 2l.2°N approximately. Recently, Lundberg et al. ( 1991) suggested that on the basis of six-channel seismic data acquired in offshore southern Taiwan the Manila Trench extends northward to at least 22.5°N . It might turn northeasterly the1·e and further connect to the major thrust fault near Tainan (Huang et al., 1992) or turn to the northeast near 22°N and continue its onshore extension to the Meilin thrust fault near Kaohsiung (Lee et al., 1993). The different sc.enarios regarding the northern extension of the tre.nch demonstrate only a small portion of the geological problems and complexities Fig. 1. Regional map ()f Taiwan and its st1rrounding area. Tai'A1an is located at the plate boundary betw·een the Eurasian and Philippine Sea plates. The Phil ippine Sea plate moves northwest\�l<:lrd at ab()Ut 7 cm/ye,lr re lative to the Eu rasian plate . . Bathyme. tric data are provided by C. S. Li u (Liu et czl, 1996). pertaining to this convergent zone. In order to trace the trench and delineate the plate bound ary between the South China Sea Basin and the Philippine block in southwestern offshore Tai\:van (Seno and Kurita, 1978), the sedimentary and crustal structures would have. to be established first. A study of� the crustal structure can provide information to answer some basic questions. For example as what is the nature of the crust in southwestern offshore Taiwan? Is it oceanic as part of the South China Sea Basin, or continental, as part of the Eurasian plate? How is the tectonic evolution of Taiwan related to this part of the trench where subduction disappears and collision takes place? A spectacular relationship of plate movement enhances the enigma of geometric configuration: while a portion of the Eurasian plate-the South China Sea Basin is subducting eastward along the Manila Trench, the Philippine Sea plate maintains its subduction rate of 7 cm/year northwest"'1ard underneath the Ryukyu Trench.
Earthquake studies near the Manila Trench (Lin and Tsai,198 ·1 ) need further refining so that location ambiguity caused by the incomplete coverage by a seismic network confined on land can be resolved. Other geophysical means, such as gravity and magnetics have. provided some estimates as to the depth to the magnetic basement , yet neither have they detailed the structure nor qualified the nature of the crust (e.g., Liu et al., 1992). Multi-channel seismic profiling aimed at the dee.p structures in this region was not established until after the 1996 TAICRUST (Tai°"''an International Collaborative Research for Understanding Subduction-col lision system in Taiv v ·an) survey when RN Maurice Ewing conducted a seismic survey cover ing the areas e. astern and southern offshore Taiwan. Howeve.r, those re.suits have not been released yet. On the other hand, dee.p-penetrating, large-offset offshore seismic experiments using ocean-bottom seismographs (OBS) had alread)' been carried out a few years be. fore the TAICRUST.
OBS experiments ,�rere conducted from 1991 to 1993 in offshore Taiwan during the RN Ocean Researclier I cruise.s 277, 282, 312 and 347, hereafter referred as ORI-277, ORI-282, ORI-3 12 and ORI-347, respectively (Chen et al., 1992). The large-offset seismic lines that are presented here ran either across or parallel to the presumed northern extension of the Ma nila Trench between latitudes 21° and 22°N ( Figure 2). These. surve)'S were limited in both the number of instruments used and in the capacity of the seismic sources available. Ho\V·ever, preliminary results provide some constraints to construct velocity models and reveal some interesting properties of' the crustal structure.

TECTONIC SETTING
Ll nder the frame'A1ork of' plate tectonics, the Eurasian plate interacts with the Philippine Sea plate at a con\rergent rate of about 7 cm/year along a southeast-northwestern direction (Seno, 1977;Sena and Kurita, 1978;Yu and Chen, 1994). The oblique collision of these two plates has caused intensive fo lding and severe faulting of the Taiwan mountain belt, thus resulting in one of the most rapid uplift rates in the world (Wu, 1978;Jahn et al., 1986;Banier and Angelier, 1986 ). Further to the south, however, an accretionary prism 70-1 10 km wide to the east of the Manila Trench (e.g., Bowin et al., 1978;Suppe, 1988;Huang et al., 1992;Reed et al, 1992) is bounded by an incipient collision along the. North Luzon Trough . Base.d on the study of slip vectors of shallow earthquake events along the eastern margin of the Philippines, Seno ( 1977) proposed that the Philippine block is a microplate.  sitting t() the east (1t' the Eurasian plate. The l\1anila Trench represents the boundary between the South China Se<. 1 Basin and the Philippine block. The South China Se.a Basin is being subducted underneath the Philippine block. The Manila Trench can be recognized south of 2 I 0N, but its tl)pographic exp14ession }()Ses its cha14acteristic t � eatures further nl)rth because of the change in the natu1-e c)f, the. tectc)nic contact has from subduction to arc-continent collision (_B iq, 1973;St1ppe, 1988 ).

DAT A ACQUISITION
The instruments (OBS) used in the experiments of this study we. re duplicated from the Texas OBS (Nakamura et c1l., 1987� Chen et czl., I 994�). Unfortunately, because of a fai lure to recover one of the t\\lO OBSs deplO)'ed in cruise O . RI-3 12, and a loss of another set ot' data due to a tape recording problem i11 one of the OBSs de.plo)'ed in cruise ORI-347, no reversed prot,ile data were acquired. ORI-3 12G, however, was equivalent to a split spread line. The coordinates and water depths of each OBS deployed and data successfully analyze. ct, are shown in Table ' I. The size of the ai1�-gun source varie. d on each c1·uise; an air-gun array of 15.1 liters (920 in 3 ) in total volume was used on ORI-3 12, �7 hile a t\\lo-gun array with a total volume of 16.5 liters (1000 in 3 ) \\las used on ORI-347 . The air gun was shot at 3()-second intervals \\lhile cruising at about 5 knots, giv·ing a shot spacing of about 80 m. The surve)' lines ran through water depths ranging f� rom 1800 m to 3600 m. While ORI-3 120 and ORI-347 were approxi mately perpendicular to the postulated exte.nsion of the i\t1ani1a Trench, ORI-3 l 2A and -3 l 2C were parallel to it ( Figure 2). The Global Positioning Sy·stem (GPS) (Ashte.ch XII GPS ) �'as used primarily for nav·igation, 'W'hile dit� ferential GPS (DGPS) data we14e av·ailable for the calibration ot� the navigatil)n data in ORI-347 . An ech() sotinder pro\1ided the water depth profile along each seismic line. The seismic signals were recorded at a 4-ms sampling rate by three geophone cha11nels installed in a gimbal mount. The data have an excellent signal-to noise ratio despite some \1ariation in source ''olume. The amount of data acquired along each line is about 25 M bytes f' or a period of about 4 to 5 hours of� shooting.
The major data sets obtained during the OBS survey included ra\v OBS fi eld data re corded on cartridge tapes, shot-time logs, navigation logs and bathymetry logs. Supplemen tary data recorded included clock-calibration files before deployme: nt· and after recovery of instruments, time, coordinates and wate.r depths of' OBS deplo)7ment and recovery, source positions with respect to navigation antenna, shot delay bet\.\i·een logged shot time and actual air-gun firing, echo sounder depth and the time-depth conversion factor.

DATA PROCESSING
The OBS field data need pre-processing treatments to correct f'o r errors and eliminate irregularities before they are processed. The calibrations <.)f the OBS clock, shot time log, sh ip track, bath)'Jnetry and sampling rates are alsc1 necess<:try. Because ot' the inherent diffe re nces in the source-receiver geometry and the data content� the processing of OBS data is some�1hat different from that ()f' con\'entional seis1nic refl ection. Some impo1·tant differences are ad dressed in the following: Shot locations: The precise locations of shots and OBS are important in a large-offs et seismic survey. How·ever, because raw na,1igation data are ot'ten tluctuant and n<.)t suft'iciently stable, • some additional processing of' the navigation data, such as smoothing to eliminate glitches, is 282 TAO, VrJ!. 7, No. 3, Septeniber 1996 necessary. Each shot location was computed in this study with the smoothed OPS (ORI-3 12) or DGPS (ORI-347) data based on a calibrated shot time. log.
OBS location: The actual position of an OB S on the sea floor is slightl)' different from the location where it \�as deployed because of· ocean currents encountered during its descent to the sea floor. Here, the direct water-w·ave arri val times were measured from the near range. data and horizontal polari zation was computed from the amplitude ratio H 1 /H 2 of the two horizontal components ()f the. gec)phones The ac.tual location, its orientation and the. precise clock corrections could be determined simultaneously· on the basis of arri val time of the direct water-wave signals, OBS deploy·ed location an.d the clock calibrations one to two hours before deployment and after recovery of the OBS (Nakamura et al., 1987).
Clock correction: Since each OBS works \\lith its own internal clock which is independent of the clock on the. shc. )oting ship, the calibration of all clocks relative to a refere.nce clock is necessary. Final clock adjustment can be made t'rom the clock drift rates and the clock correc tion obtained \\1 ith the OBS location as mentioned above. In turn, the shot delay and clock drift rate during data acquisition can be estimated.
A Standard SEG-Y t' ormat tape was generated by combining demultiplexed, decoded and time-corrected OBS seismic data \V ith properly corrected shot-location and shot-time data. Horizontal-component data w·ere rotated to radial and transverse directions based on the. deter mined instrun1ent orientation. Further processing of the data could be performed using avail able software packages. The record se.ctions sho\\1n in Figures 3 to 6 were produce.ct b) ' stacking traces into unit' ormly spaced ot't'set bins, band-pass f'iltering the trace.s and adjusting the gain.

ANAL\'.l'SES AND IN1'ERPRETATIONS
The line segments of the time-distance curve interpreted in the record sections (Figures 3 to 6) were used to deduce the velocity and thickness of each layer under the assumption that the velocity was constant and the refractor was horizontal in each layer. The slope of each line segment was the reciprocal C)f the \1elocity associated \\1 ith the medit1m just below the bound ary·. When estimating the velocity and thickness of each layer by employing the. tra\'el-time equation of the head wave (e.g., Telford et al., 1976), it was assumed that the \1elocit)1 in the lower layer was higher than that in the upper la)'er. With only one OBS receiv·e14 generating useful seismic data set on each surve)1 line w·e acquired, it was impossible to derive. a unique structural solution from the unreversed data set. Nevertheless, the data did shed some light on certain aspects of' the crustal structure which are discussed below.
i Allen T. Chen and YL-tng-Se11 Jai-�· 287 layers are. sorne\vhat deeper in the eastern side than that in the western prot'i le. While the OBSs on both the pr{1f'i les ORI-3 I 2A '1 nd ORI-3 l 2C were deployed near the southern end, another OBS was located near the cente14 of the. profile ORI-3 l 2G. The refracted seismic arri' \lals from layer boundaries in either side of the trench along line ORI-3 12G is asymmetric ( Figure 5). This n1ay suggest thLlt either the dipping layers were present or the local ve.locity structures were dif'fei·ent along the line. In addition to arrivals from sedimentary layers, oceanic basement, and deep crustal bound aries, Moho ret'ractio11 was identif'ied at a distance of about 23 km from the OBS along the profile ORI-3 I 2C (Figure. 4 ) . The corresponding \1elocit)' converted from the slope of the tra\lel time cur\'e . exceeds 8. 1 km/sec (Table 2 and Figure 10). The mantle velocity refracted at a depth of abot1t 12 kn1 below sea level provides crucial constraint in estimating the. crustal thickness . To refine the preliminar)' one-dimensional v·elocity model of Table 2 and to extend it into a two-dimensional one, a 2-D ray tracing technique was employed (Cerveny et al., 1977;Luergert, 1993). Three multi-channel seismic profiles MCS320-37, MCS320-39 and MCS320-41, \\1ere acquired duri11g cruise ORI-320 in the study· area. The. locations of these profiles almost coincided i11 loc<: 1tions with the OBS lines ORI-3 12G, C and A, respecti\rely (Figure 2).
The prot'i ]es l \11CS32 ()-39 e: 1 nd 320-4 1 were located at opposite sides of the Manila Trench. Their interpreted ti1ne sections show· sedimentary layers dipping to the south (Figure 7 and 8). Localized t, an-shape deposits \\/e14e a comml1n feature near the surface, deeper reflections were rather flat and C()ntinuous in MCS320-39. On the other hand, hummocky clinoforms charac terized the seismic C()nfi guratio11 in MCS320-4 1 and the seismic signals are \\'eaker and less continuous throughout the sedimentary section. The profile MCS320-37, which crossed the trench, sho\\iS reflected horizons dipping steeper toward the arc (Figure 9). The acoustic base ment \\t'as deeper than 6.5 se. c and \V as pr(1gressiv·ely deeper tow·ard the east.    Table 2.

290
TA O, V(J/. 7, J"t\/() . 3, Septe111ber 1996 Sedimentary st1·uctures revealed f�f()m multi-channel seismic data (Chen, 1993) prov'ide important constraints to the starting model ot· ray tracing. The theoretical travel time gene1� ate. d from the starting model was compared with the C)bserved tra\1el time, and the tirne dift�e r ence bet\\1een then1 W(lS reduc.ed b)' adjusting either the \1elocity· structure or the layering atti tude of the model. The model was, in fact, rev1ised repeatedly until an)' further adjustment made. no signif]cant improvement, i.e., the O\/erall time differe nce is essentially remained the same . Figures 11 to 13 sho\v the fi nal models, theoretical ra)1 paths and theoretical arrival time against the observations in this stud)' . Based o . n these final models, it is fo und that: ( 1) seismic velocit)l increases eastward (Figure 13 ), (2) the base1nent with velocit)' higher than 6 km/sec is deeper in the arc-side of the trench (Figure I I \' S. 12), and (3) the depth of the high \1elocity· (8 .17 km/sec) laye1� is down to 11. 5 km below sea level ( Figure. 12 . ).

Line ORI-347
This line is the. most 1·ecent and longest (about 60 km · ) amc)ng all profiles acquired in this stud)r . The OBS was deployed near the eastern end of' the line. The record section is shown in Figure 6. Due t() '1 malt'unction c)f the air gun, there is a 20-minute, or 3-km� gap in the data set. Ho\:vever, clear t' i1·st arrivals can be observed t'o r ot�t's et less than 25 km. The ar · ri vals ot .. direct Welter-wave t'r() ffi prev'ious shot at a distance beyond 40 km are also observed near the f' ar end ot· the i� eC()rd. The fi1·st arrivals recognized in the ranges from 20 to 25 km is inter preted as signals refracted from the . Moho discontinuity, since it gives an appare . nt velocit)' of 8. 14 km/sec . The estimated one-din1ensional velc)c it)' strL1cture along this prc)file is shown in Table 2 and Figure 1 ().
The 2-D 1·a)1 tracing ine.thod was al sc) applied to confirm the \1elocity structure of the ORI-347. Since there \V as no seismic reflection prl)file <: 1(()ng this line, the �1uthors began \\1 ith a simplified fl at layer model based on a one-dimensional velocity structure \V ith a sea fl oor locc: tti(ln constrained b)1 bathytnetric data. The re sulting raypaths, models and theoretical ar rival curves generally are not t'ar f' rom the initial model ( Figure 14) \:vith some finer structura] variations. The velocity in lay·ers deeper than 7 km remained unit'o14m since no sut · t-icient constraint t' o r latet4a1 variation was av·ai lable. The re sult sugge.sts that the Moho discontinuity lies at a depth of ab()Llt I 1.5 km \v hich is compatible w· ith that derived fro m the profi le ORI-3 l 2C.

DISCUSSION
Due to ()perati()na] negligence C)n the Omega clock syste111, the14e \\1 as a time difference of less than ()ne mintite bet\:veen the Omega clock, \\lhich se14ved as the reference time, and the GPS tin1e duri ng the datcl c1c;qL1isition on both c1-uises. The computed location of each OBS pr()fi le based on the Omegct t . ime and a ship \1 e]()City of 5 knots ma)1 11ave been therefore shifted up to 150 meters from its actual positic)n . H()Wever, the relative velocity structure con\1erted f1401n the processed seismic records i11 terms of the depth of laye.r thickness and its c. orresponding velocity rem . ained essentially intact.
The velocity structures derived suggest that the Moho lies at a depth of about 1 1.5 km below sea level in the study area. This is in agreement with the crustal structure of' the South China Sea in general (Taylor, 1980). Assumi11g the sedimentary laye1-s 14 epresented by the  Allen T. Chen and Yu ng-Sen Jaw 295 sections with a " \le]ocity lo\\ler than 4 km/sec and a thickness ranging t'rom 2.7 to 3.7 km, the f'i nal models of ra)' tracing suggest that the depth to the basement is about 5.7-6.8 km, which is about 1.5 km deeper than the estimated depth of the magnetic basement ). The relatively thin crust, -w·hich ranges from 5.2 to 5.8 km in thickness, as revealed by the profiles ORI-3 12C and -347, further suggests the possibility of its having an oceanic nature. The thickness of the sedimentaf)' layer is maintained at about 3 km or more along each profile.
The overall structure is pretty uniform along the profiles ORI-3 12A and -312C which are paral lel to the trench. However, their \i' elocity models are somewhat different. While the depth of the sedimentary la)'er \"\lith a \lelocity of less than 3.7 km/sec reac.hes a depth almost 6 km on both lines, the layer with a velocity higher than 6.2 km/sec is l .5 km deeper along the eastern profile (Figures 8 and 9). The overall attitude of the profiles ORI-3 12.G and -347 also sho\\i' lay·ers dipping to the east and is consistent with the concept of an eastward subduction of the South China Sea Basin crust.

CONCLUSIONS
Some seismic refraction data from OBS experiments are presente. d so as to construct velocity models in the vicinity of the l\1 anila Trench southwestern offshore Taiwan. The apparent v·elocity of the upper mantle based on record sections of the ORI-3 J 2. C and -347 is close to 8. 1 km/sec at depth exceeding 11 km. The thickness of the sedimentary layer is 3 to 4 km and remains about the same. within 10 km on either side of the trench. The igneous crustal structure revealed b)1 the models of the profiles ORl-3 l 2A and -3 12C shows that it is thicker east of the Manila Trench. The thin crust obser\1ed along the profiles ORI-3 12C and -347 at the west side of the trench is only 5 to 6 km in thickness and is concluded to be oceanic in • • or1g1n.