Distribution of the Bottom-Simulating Reflector in the Offshore Taiwan Collision Zone

A bottom-simulating reflector (BSR) has been identified in the offshore Taiwan accretionary prism using both migrated 6-channel and 120-chan­ nel reflection profiles. The BSR is marked by a reversed polarity reflector that increases in sub-bottom depth with increasing water depth, suggesting that the BSR marks the base of the methane hydrate stability field. The BSR is located in offscraped sediments derived from the Taiwan orogen and the Chinese continental margin, which may contain high amounts of organic carbon, thereby providing a source for the methane. Gas seepage found on land north of this region also supports the presence of methane in these strata. The BSR is typically located in the crests of anticlines and mud volcanoes, implying the migration and entrapment of methane gas. Recent uplift of the seafloor in anticlines may reduce pore pressure, thereby decomposing the hydrate phase into free gas, which may enhance the acoustic impedance contrast across the BSR. BSRs are conspicuously absent be­ neath submarine canyons in the region. A "flat spot" was found under the BSR within an anticline, suggesting that the hydrate-filled pore-space may result in lower sediment permeability, therefore enhancing the entrapment of free gas. We document the areal extent and sub-bottom depth of the BSR over a 77000 km2-wide region around southern Taiwan. In particular at least 30 % of the seafloor, to as much as 60 % , along the trench ward slope of the accretionary prism is underlain by the BSR. The BSR is located at shallower sub-bottom depths on the trenchward (western) side of the ac­ cretionary prism than on the arcward (eastern) side, even though water depths are similar, implying enhanced fluid flow from depth, resulting in a higher geothermal gradient as sediments are accreted and dewatered along the Manila trench. (


INTRODUCTION
A bottom-simulating reflector, or BSR (Figure 1), is a seismic reflector subparallel to the topography of the sea floor, which in some places cuts across reflectors generated by sediment layers (Shipley et al., 1979). BSRs are commonly observed within several hundred meters of the sea-floor in continental slope sediments, particularly those associated with accretionary prisms (Hyndman and Davis, 1992). These seismic reflectors can be generated by the acoustic impedance contrast at the base of a sediment layer containing methane hydrate (Shipley et al., 1979).
Methane hydrates influence a number of important submarine processes. For example, the hydrate, and any free gas trapped below, may provide a significant hydrocarbon resource in the future (Kvenvolden and McMenamin, 1980;Kvenvolden, 1993). Development of the offshore seabed may. be influenced by seafloor instability and mass wasting, both of which are affected by hydrate formation and decomposition. The development of methane hydrate in accretionary prisms is affected by fluid flow, which also influences fault slip and potentially the seismic cycle (Moore and Vrolijk, 1992). Finally, methane hydrate in marine sediments may form a significant global sink of methane, an important "greenhouse" gas. Variations in bottom water temperature or the position of sea level may influence the stability of hydrate, and in some cases, could change the global methane storage in the atmosphere (Kvenvolden and McMenamin, 1980;Kvenvolden, 1993). One of the challenges of methane hydrate research is to estimate the global distribution of this substance and its ability to store methane gas within or below the submarine hydrate layer.
The identification of a BSR is one of the easier ways to detect the presence of methane hy drates and map their distribution, although hydrates have also been recovered in the sites without BSR (Holbrook et al., 1 996). Reed et al. (1991Reed et al. ( , 1992 first identified the presence of a BSR around southern Taiwan using 6-channel seismic reflection data. Subsequent studies have extended this work by identifying BSRs elsewhere around offshore southern Taiwan (Chi et al., 1993;Hwang, 1993). Here we present the first comprehensive regional mapping of this submarine feature around southern Taiwan using a grid of migrated 6-channel and 120channel seismic reflection profiles, the latter acquired in 1995 as part of the TAICRUST ex periment. In addition, the location of the BSR will be compared to specific structural features along the seafloor in order to examine the potential factors leading to hydrate formation. A related paper on the implications of the distribution of the BSR to regional heat flow and fluid flow within the accretionary prism is in preparation.

PREVIOUS BSR STUDIES
Methane hydrates are "ice-like" clathrate structures in which methane molecules are en closed within a solid lattice of water molecules. Most hydrates have been found in deep ocean sediments where there is a sufficient supply of methane and where pressure and temperature range between 0.2-5 MPa and 0-25 °C, respectively (Kvenvolden and McMenamin, 1980).
The increase in temperature with depth below the seafloor, associated with the geothermal gradient, causes methane hydrate to become unstable and decompose at increasing sub-bot tom depth, despite increasing pressure. As a result, the base of the methane hydrate defines a "phase boundary" separating the low-temperature stable phase of solid gas hydrate above from a field of instability below. The acoustic impedance contrast across the "phase bound ary" can be large enough to generate reflections that are easily visible on seismic reflection profiles (Shipley et al., 1979). Kvenvolden (1993) has estimated that the total amount of methane in gas hydrate within marine sediments around the world is likely over 1019 g, consequently there is significant interest in this feature as a potential energy resource. The widespread abundance of methane hydrates in marine sediments also influences other geological oceanographic processes, in cluding mass wasting and global climate pattern (Kvenvolden, 1993;Kvenvolden and McMenamin, 1980;Mciver, 1977;Reed et al., 1991;Paull et al., 1996). For example, meth ane hydrate can act as cement, mechanically strengthening the sediments above the BSR, whereas unconsolidated sediment below the BSR may be more susceptible to mass failure and downslope movement. Thus a mechanical discontinuity is produced in marine sediments that could influence the location and distribution of submarine mass wasting. Moreover, any decrease in pore fluid pressure or increase in sub-seafloor temperature could decompose the bottom of the hydrate layer, thereby decreasing the ability of the sediment to withstand shear stresses. As a result, the sediment column becomes unstable and mass wasting may be en hanced. Using 14C isotopic data from cores in the upper 7 m of the surface sediments. Paull et al. (1996) found that the glacial-age sediments in the gas hydrate field offshore the southeast-em United States are underrepresented in the core, in agreement with the prediction of in creased frequency in mass wasting when hydrate is dissolved due to a lowering of sea level.
Decomposition of significant amounts of submarine methane hydrate can release meth ane, a greenhouse gas, into the atmosphere and global climate may be influenced. The amount of methane that is trapped in gas hydrates onshore and offshore is perhaps 3000 times the amount of methane in the atmosphere (Kvenvolden, 1993). Moreover, methane has a global warming potential 20 times larger than equivalent weight ofcarbon dioxide (Shine et al., 1990). Global climate change, which can influence the position of global sea level and there fore the stabihty conditions of hydrate in the submarine environment, can impact the exchange of methane between the ocean and atmosphere. Prior to formulating models on the ocean atmosphere exchange, we first need to have a better estimate of the global storage of methane in hydrates within marine sediments and a better understanding on the formation mechanisms of the hydrate. This study adds to the growing body of information documenting the distribu tion of methane hydrate around the world.

GEOLOGICAL SETTING OF TAIWAN COLLISION
The offshore Taiwan accretionary prism ( Figure 2) is located on the boundary between the Eurasian plate and Philippine Sea plate where oceanic lithosphere of the South China Sea is subducting eastward beneath the Luzon island arc (Bowin et al., 1978). The prism is bounded on the west by the Manila Trench, and on the east by the forearc basin of the North Luzon Trough (Hayes and Lewis, 1984). The Chinese continental margin enters the subduction zone near Taiwan and the subduction-related prism is modified by arc-continent collision (Suppe, 1981). Although the timing of initial collision is still under debate, it probably began in the late middle Miocene (Teng, 1990) and is currently propagating southward (Suppe, 1984).
The submarine accretionary prism contains three distinct structural domains ( Figure 2) (Reed et al., 1992): (1) a lower slope domain composed of mostly west-vergent ramp anti clines; (2) an upper slope domain with highly discontinuous reflections, suggesting intense deformation, and, possibly, out-of-sequence thrusting and underplating; and (3) a backthrust domain located along the rear of the prism. The North Luzon Trough, a forearc basin, is truncated in the region of collision by the juxtaposition of the rear of the prism against the volcanic arc along a region of backthrusting (Lundberg et al., 1997).

SEISMIC DATA PROCESSING
This study is based on data acquired during a 6-channel seismic reflection survey con ducted in 1990 aboard the RIV Moana Wave of the University of Hawaii and 120-channel seismic reflection survey in 1995 aboard RIV Maurice Ewing of Columbia University as part of the onshore-offshore seismic experiment known as TAICRUST ( Figure 3). The 6-channel data were processed at UC Santa Cruz and San Jose State University using the SIOSEIS pro cessing software with following sequence: debias correction, resample to 4 msec, spectral analysis, quality control, near trace gather, shot gathers, trace editing, geometry, AGC, spiking deconvolution, NMO, stack(4-fold), filter, predictive deconvolution, water-bottom mute, trace weighting, filter, AGC, stack section plot, F-K migration at 1500 m/sec, filter, AGC, and finally, migrated section plot (Reed et al., 1992). The 120-channel reflection data were pro cessed at University of Hawaii using ProMAX. This dataset, with large source-receiver offset and detailed interactive velocity analysis (0.6-1.2 km spacing) after DMO, provides good ve locity information in this region, especially at shallow sub-bottom depths where BSRs are located. The velocity information is used to intially convert travel-times to subbottom depths for the BSR at several locations; more extensive conversion on a regional basis is planned for a future study.

DATA ANALYSIS
More than 8,000 kilometers of seismic reflection profiles, at an average spacing of 8 km, were examined over a region of 27,000 km2 within the offshore Taiwan accretionary prism. We have made a determined effort throughout the study to separate reflectors associated with water-bottom reverberations from BSR identifications, for example shallow reflections marked by constant sub-bottom depths beneath a dipping seafloor reflection will be interpreted as water-bottom reverberations, instead of a BSR. On these profiles, the BSR is identified as a reflector sub-parallel to the seafloor with reversed polarity. However, portions of the BSR may exhibit normal polarities where it is cross cut by bedding plane reflections. The BSR typically exhibits a single symmetrical pulse characteristic of a simple interface (Figure 1). The BSR was identified at nearly 1,500 locations along the survey lines and each pick was assigned to one of three categories according to the confidence level of the identification (Q 1, Q2 or Q3). These data were converted into methane hydrate distribution maps showing vary ing degrees of confidence in the accuracy of BSR identification ( Figure 4 ). The BSR identifi cations with the highest confidence level compose the QI category with the following reflec tor characteristics: (1) subparallel to the seafloor in shallow sub-bottom depths, (2) reverse polarity with a single symmetrical pulse; and (3) a strong coherent reflection that cuts across bedding. The Q2 category or probable BSR identifications show similar characteristics to the Q 1 picks except that the reflections are either segmented into pieces or may not cut bedding. The segment lengths of the Q2 reflectors are usually longer than the gaps between segments. The Q3 category, containing possible BSR picks, exhibits very weak and segmented reflec tions that are subparallel to the seafloor. BS Rs in the Q3 group can be traced from adjacent Q 1 or Q2 categories without abrupt changes in the subbottom depth.

STRUCTURAL RELATIONS OF BSR
BSRs are most commonly observed in the crests of anticlines ( Figure 1). As a result of fault-related folding, followed by erosion of the seafloor or turbidite deposit, the seafloor is not parallel to bedding-plane reflections in these regions, thereby making the BSR easy to identify. The BSR usually has a larger radius of curvature as compared to strata within the anticline, resulting in subbottom depths of the BSR that are greater at the crest than along the limbs of the fold. This local geometry of the BSR differs from regional BSR relation of increasing subbottom depth with increasing water depth.
The BSR was also found in many mud volcanoes and mud diapirs ( Figure 5). The chaotic reflections within these features enhance the identification of BSRs. As in the anticlines, the BSR shows a gentle convex upward curvature. Shallower subbottom depths are observed underneath the crest of several volcanoes and diapirs, rather than on the limbs.
Many of the BSRs were also found in association with faults ( Figure 5). At several loca tions in the lower slope domain, the BSR increases in amplitude in the forelimbs of fault related folds. Many BSRs can be traced into the backlimbs of these folds, but disappear in flat-lying strata away from the fold, where the water depth becomes constant. We also found several places where the subbottom depth of the BSR decreases near the fault zones. How ever, we have not observed this relationship near fault zones in the upper slope domain. BSRs are not observed where the water depth changes abruptly, for example, in the bathymetric lows in submarine canyons. .,; Plots of water depth versus sub-bottom depth of the BS R were constructed ( Figure 6). Two-way traveltime of BSR sub-bottom depth was converted into BS R sub-bottom depth in meters by using velocity-depth relations published by Hamilton (1980):    On all plots, the BSR is generally located at a sub-bottom depth of several hundred meters and increases in sub-bottom depth with increasing water depth ( Figure 6). However, there is significant scatter in these plots caused in part by local decreases in the BSR sub-bottom depth near several fault zones, on the limbs of folds, and beneath mud volcanoes ( Figure 5). The pl ot of the highest quality picks, or Q 1 category, shows the least amount of scatter along a trend of increasing sub-bottom depth with increasing water depth. The combined plot of the Ql and Q2 category data shows slightly more scatter around the general trend described above.
Finally, the plot of all BSR picks (QI, Q2, and Q3) shows the most amount of scatter.

DISCUSSION
Several lines of evidence indicate that the BSR around southern Taiwan represents the base of a methane hydrate stability field. First, the BSR in this region exhibits a reversed polarity reflection, suggesting a decrease in acoustic impedance across the BSR. This is a common observation of BSRs elsewhere in the world and is commonly attributed to the pres ence of patches of solid hydrate above the BSR and the absence of hydrate and/or presence of free gas trapped below (Shipley et al., 1979;Miller et al., 1991;Singh et al., 1993;Bangs et al., 1993;ODP Leg 146 Scientific Party, 1993;MacKay et al., 1994;Scholl et al., 1995;Dillon et al., 1995;Wood et al., 1995). Next, the BSR around southern Taiwan increases in sub-bottom depth with increasing water depth, implying that increasing hydrostatic pressures elevate the melting temperature of the hydrates. Consequently, the hydrate stability field is extended to greater subbottom depths before reaching temperatures high enough to result in decomposition. The BSR is widespread in areas of the accretionary prism composed of offscraped sediments of the Chinese continental margin and the Taiwan orogen, which may reflect high amounts of organic carbon in rapidly deposited terrigenous sediments deposited in the Manila Trench prior to accretion. Gas seepage has also been observed on land close to the study area (Central Geological Survey, 1996).
The common presence of BSRs in the crests of anticlines, mud volcanoes, and near the faults in the offshore Taiwan prism may be related to updip migration of gas and methane within folded strata. Very thick sedimentary sequences in the collision region are folded into long wavelength anticlines with the potential of tapping a very large volume of strata from which to derive methane. Gas can migrate upward within mud volcanoes, or even emit from them, as indicated by the onland mud volcanoes near this region (Central Geological Survey, 1996). The upward migration of warm fluids along faults or within mud volcanoes may have transported methane, either in a dissolved or free phase, into the crests of anticlines, and in-crease the concentration of methane at shallow subbottom depths thus enhancing the genera tion of hydrate in the region. This process could also carry heat upward from depth thereby producing a shallowing of the BSR at fault zones as seen here and in the Casacadia accretion ary prism (Zwart and Moore, 1993) The highest concentration of BSRs is located in the northwest region of the study area ( Figure 4) where offscraped terrigenous detritus derived from the Chinese continental margin and uplifted Taiwan orogen compose the lower slope domain of accretionary prism. The fine grained fraction of rapidly deposited terrigenous sediments commonly has relatively high amounts of organic carbon, compared to pelagic sediment of the ocean basin (Selley, 1983). Organic carbon contained in offshore Taiwan strata may be the source of methane in the re gion, which has contributed to the formation of the hydrate layer. High concentrations of methane, either as free gas or dissolved in pore fluids, are required before hydrate can form. High sedimentation rates in this region may also suppress the escape of the solute into the water column (Hesse et al., 1997). High concentrations of methane may ultimately become trapped in structural closures in the hinge zones of anticlines and hydrate formation itself may further assist in trapping methane in these regions.
The high amplitude and continuous nature of the BSR in the collision zone might also be affected by uplift in this area. Pecher et al. (1994) suggest that as strata are uplifted, the hydrostatic pressure on the base of methane hydrate layer will decrease, in respond to a de crease in overburden pressure associated with the overlying water column. As a result, the base of the hydrate layer will decompose into gas, which, if not dissolved, will then be trapped underneath hydrate layer, leading to an increase in the acoustic impedance contrast at the level of the BSR. This interpretation fits well with the observations in Taiwan, where the BSR is often located within the rapidly uplifting regions, such as the crests of the fault-bend folds, and on a larger scale, within the two NW-SE trending bathymetric highs. This mechanism may work in concert with the updip migration of methane gas to entrap free gas in the anticlinal ridges.
Many studies have suggested that free gas is trapped beneath the BS Rs in other regions (Miller et al., 1991;Singh et al., 1993;Bangs et al., 1993;ODP Leg 146 Scientific Party, 1993;MacKay et al., 1994;Scholl et al., 1995;Dillon et al., 1995;Wood et al., 1995). Direct hydrocarbon indicators in the form of high amplitude "flat spots" are found underneath the BSRs in several regions of the Taiwan prism (Figure 7), implying that the reduced permeabil ity in hydrate-filled pore space may trap gas underneath methane hydrate stability field. De tailed velocity analyses of the long offset 120-channel data, acquired as part of the TAI CRUST experiment, also found apparent low velocity zones below the BSR. A 50-100 m/s decrease in RMS velocity across the BSR observed in these data implies a greater decrease in interval velocity. In fact, velocity pull downs were identified in other locations, also suggesting the presence of free gas beneath the BSR. Recent vertical seismic profiling conducted during ODP Leg 184 on the Blake Ridge showed that low velocity zones may extend at 250 m be neath the BSR (Holbrook et al., 1996).
In some areas of the Taiwan prism, reflection amplitudes decrease in the layer above the BSR ( Figure 5). This "blanking" effect in the layer above the BSR has been observed in other regions around the world, including the Blake Ridge (Shipley et al., 1979;Guerin and Goldberg,  1997) and Nankai Trough (Sakai, 1997). It is commonly attributed to the small magnitude of reflection coefficients within hydrate-bound sediments. The strong reflection amplitudes of the seafloor and BSR may also act to reduce the strength of reflections in these strata due to automatic gain control (AGC). However, we do not see higher amplitude bedding reflections at an equivalent sub-bottom depth where there is no BSR, implying that an AGC "shadow" from a high amplitude BSR does not change the amplitude of stratal reflections dramatically. Detailed velocity analyses of Ewing data did not show a clear high velocity anomaly above the BSR that would be consistent with the enhanced rigidity and density in hydrate thereby producing an increase in the acoustic impedance contrast at the BSR. Also, we do not see any clear evidence of velocity pull up in those areas. These observations are similar to those of BSRs sampled at Blake Ridge (ODP Leg 146 Scientific Party, 1993), which show only small amount ( < 15 % in bulk volume) of hydrates within the sediments. As a result, the mechanism for producing low amplitude reflections above the BSR is still unclear, and re quires higher resolution geophysical/geochemical data and a better understanding of the elas tic properties of hydrated sediments to be able to solve this problem.
In several regions we have made initial attempts at estimating the geothermal gradient from the BSR sub-bottom depth, methane hydrate phase equilibria, and seafloor temperature data following the method used by Ashi and Taira (1983). These calculations suggest a geothermal gradient ranging from 17 to 160 °C/km with most values between 25-45 °C/km. The geothermal gradient decreases eastward and shows similar results to the heat flow mea surements conducted south of this region (Taylor and Hayes, 1983). Shyu et al. (this volume) have conducted in-situ heat flow measurements within this region with 3 of the 4 measured geothermal gradients exceeding those derived from BSRs in nearby areas. However, one of the measurements shows excellent fit with the BSR-based geothermal gradients. Shyu et al. (this volume) have suggested that a 90% methane+ 10% ethane gas composition in pure water yields a better estimate of the geothermal gradient in this region based on phase equilibria relations. Additional in-situ heat flow data are required to better calibrate BSR-derived esti mates to seafloor measurements on a regional basis. A better knowledge of the fine scale velocity structures, history of seafloor temperature, and sediment thermal conductivity are necessary to use the BSR to estimate absolute heat flow values (Nissen et al., 1995;Chi et al., in prep.).
We can, however, compare the relative geothermal gradients at equivalent water depths on opposite sides of the accretionary prism in order to assess the potential role of fluid flow caused by sediment dewatering. MCS data from the TAICRUST program show shallower BSR subbottom depths along the western flank of the accretionary prism (Manila Trench) than on eastern side of the prism (North Luzon Trough). Line 33 of the TAICRUST program exhibits a high amplitude BSR at a subbottom depth of 0.52 sec at a water depth of 1275 min the backthrust domain, which is located along the arcside of the accretionary prism. The BSR in the lower slope domain of the prism at the same water depth has shallower subbottom depth of 0.43 sec. Assuming that sediment velocity profiles, seafloor bottom water temperatures, and sediment thermal conductivity are similar at those two sites, the shallower BSR subbottom depth on the trench side suggests higher heat flow, which might relate to more intensive sedi ment dewatering leading to fluid flow from depth. 9. CONCLUSIONS 1. At least 20000 km2 of the Taiwan accretionary prism is covered by a BSR, which is identi fied as a reflector sub-parallel to the seafloor with reversed polarity. Reversed polarity, increasing subbottom depth with increasing water depth, and the distribution of the BSR strongly suggests that this feature marks the base of the methane hydrate stability field.
2. The BSR is most commonly found in the crests of anticlines, mud volcanoes, and near the faults, which may relate to updip migration of free gas and methane dissolved in pore fluids. Some of the gas might be trapped structurally, or beneath the hydrate layer.
3. The highest concentration of BSRs is located in the northwest region of the study area where rapidly deposited terrigenous sediments may have relatively high amounts of organic car bon, thereby providing a source for the methane.
4. Most of the Q 1 and Q2 BS Rs are located within the two NW-SE bathymetric highs in the lower slope domain. Uplift of the seafloor along these features will reduce the in-situ pres sures, causing the base of hydrates to decompose. Consequently, released gas, trapped below the hydrate layer, might enhance the reflectivity of the BSR. Bright spots located underneath the BSR elsewhere in the prism suggest fluid migration and possible gas entrap ment below the hydrate layer.

5.
Geothermal gradients, estimated from the BSR sub-bottom depth, methane hydrate phase equilibria, and seafloor temperature data, range from 17 to 160 °C/km with most values between 20-50 °C/km 6. Comparison of BSR subbottom depths on opposite sides of prism at the same water depth suggests elevated heat flow along the Manila trench, possibly due to sediment dewatering during accretion and fluid flow from depth.