Seismic velocity increase and deep-sea gas hydrate concentration above a bottom-simulating reflector on the northern Cascadia continental slope

Transcription

1 JOURNAL OF GEOPHYSICAL RESEARCH, VOL. I01, NO. B6, PAGES 13,655-13,671, JUNE 10, 1996 Seismic velocity increase and deep-sea gas hydrate concentration above a bottom-simulating reflector on the northern Cascadia continental slope T. Yuan, R.D. Hyndman, 1 G.D. Spence, and B. Desmons School of Earth and Ocean Sciences, University of Victoria, Victoria, British Columbia, Canada Abstract. The amount of gas hydrates in the accretionary wedge sediments of the northern Cascadia subduction zone off Vancouver Island has been estimated from multichannel seismic (MCS) and Ocean Drilling Program (ODP) data. Derailed semblance velocity analyses and full waveform inversion of MCS data, combined with previously published ODP Sites 889/890 sonic log and vertical seismic profile (VSP) data, show that sediment velocities increase downward more rapidly than the nohydrate/no-gas reference profile from about 1500 m/s near the seafloor to a maximum of 1900 m/s just above the bottom-simulating reflector (BSR) at a subbottom depth of 224 m. Immediately below the BSR, the MCS velocities drop to m/s. A low velocity of m/s from the VSP data probably represents a thin layer, m, containing free gas. The difference between the reference and observed velocities is used to estimate hydrate concentration, which reaches a maximum of 20-30% of the pore space above the BSR. A simple interpretation of the drill core chlorinity dilution data at the ODP site yields a similar hydrate concentration of 35 %. The estimated hydrate concentration with subbottom depth represents about 7 m 3 of hydrate per square meter of seafloor. The total methane gas at STP, including hydrate above the BSR and the small amount of free gas below, is about 800 m 3 per square meter of seafloor, or 200 TCF (trillion cubic feet) on the Vancouver Island continental slope. Application of the same method in the Blake-Bahama region of the eastern U.S. margin indicates that the velocity enhancement and inferred hydrate concentration are very similar. Introduction In the area of the Ocean Drilling Program (ODP) 889/890 sites on the lower slope of the Cascadia margin off Bottom-simulating reflectors (BSRs), associated with the Vancouver Island (Figure 1), Hyndman and Spence [1992] base of the hydrate stability field [e.g., KvenvoMen and concluded that a high-velocity layer existed above the BSR, McDonaM, 1985], are observed in the upper few hundred having a sharp base and a transitional top,. from highmeters of ocean floor sediments on many continental resolution velocity analysis, synthetic seismogra modeling, margins. BSRs are especially common in subduction zone and AVO studies. They also concluded that about 30% of accretionary prisms. In sire information from deep-sea the pore spaces were required to be fried with hydrate ff the drilling is very limited, and much of the available informa- BSR impedance contrast was the result solely of hightion as to the origin of BSRs has come from seismic reflec- velocity sediments containing hydrate over water-saturated tion data [e.g., Shipley and Didyk, 1981; Minshull and sediments. No velocity or AVO effects of free gas below White, 1989; Miller et al., 1991; Hyndman and Davis, 1992; the BSR could be detected. A thin layer containing gas was Hyndman and Spence, 1992]. Numeroustudies involving allowed by the data, provided that (1) it had a transitional forward modeling and inversion of seismic reflection data base such as not to give a reflection from the bottom of the have been carried out [Miller et al., 1991; Hyndman and gas and (2) the concentrations of gas were sufficiently low Spence, 1992; Singh et al., 1993; Minshull et al., 1994; (less than a few percent) such as not to strongly affect Katzman et al., 1994; Wood et al., 1994; Singh and Minsh- Poisson's ratio and thus AVO. Subsequent downhole ull, 1994; Andreassen et al., 1995]. These studies used BSR reflection coefficients, reflection waveform modeling, logging and vertical seismic profile (VSP) data at ODP Sites 889 and 892 showed that there were low velocities immediand amplitude-versus-offset (AVO) characteristics to con- ately below the BSR that impred the presence of free gas strain seismic velocity variations above the BSR and the [MacKay et al., 1994]. They also suggested that there was presence of low velocities associated with free gas in the only a very small velocity enhancement from the presence of sediment pore space beneath it. Also at Pacific Geoscience Centre, Geological Survey of Canada, Sidney, British Columbia, Canada. Copyright 1996 by the American Geophysical Union. Paper number 96JB /96/96JB ,655 hydrate above the BSR and that the BSR is generated primarily from the top of the free gas layer. $ingh and Minshull [1994] came to a similar conclusion of a lowvelocity gas layer beneath the BSR and httle velocity enhancement above it, based on a full waveform inversion of the multichannel data of Hyndman and Spence [1992]. It should be noted that on previous velocity determinations, there has been little direct constraint on the normal no-

2 13,656 YUAN ET AL.: SEISMIC VELOCITIES OF MARINE GAS HYDRATE 50'N 48'N 128'W 126'W 124'W Figure 1. Northern Cascadia margin off Vancouver Island showing locations of multichannel seismic lines L89-08 and L89-10 studied in this paper and locations of Ocean Drilling Program (ODP) Leg 146 Sites 888 and 889/890. hydrate/no-gas velocity-depth profile and thus little constraint on whether the BSR contrast is from high velocity over normal velocity or normal velocity over low velocity. A reference sediment velocity-depth profde unaffected by either high-velocity hydrate or low-velocity free gas is thus an important part of this study. Seismic velocities in hydrate-beating sediments overlying the BSR should be much higher than in hydrate-free sediments [e.g., Whalley, 1980; Pearson et al., 1986; Sloan, 1990]. If there is free gas beneath the BSR, a sharp reduction in compressional velocity should result. In this paper, we examine the seismic velocity structure of the BSR on the Cascadia margin offshore Vancouver Island using data from an integration of previously published ODP Tectonic and Geological Setting The coarse-clastic accretionary wedge on the northern Cascadia margin is well imaged by multichannel data [Davis and Hyndman, 1989; Spence et al., 1991a, 1991b; Hyndman et al., 1994]. The initial deformation at the frontal region of the wedge is accomplished by landward dipping thrust faults and margin-parallel folds. Most if not all off the sediment section is scraped off the incoming oceanic crust. Farther landward where the sediment section grows rapidly in thickness due to tectonic shortening, sediments undergo severe tectonic compaction and distributed small-scale deformation. In the midslope region, the rapid thickening of the wedge results in sediment elements being moved to greater depths with little porosity loss and little associated velocity downhole logging and VSP velocities with detailed change. The slow reconsolidation toward a normally multichannel seismic (MCS) velocity analysis results. The consolidated section and pore fluid expulsion in the region critical reference velocity-depth profile for the accretionary is a consequence of low sediment permeabilities and is prism sediments has been obtained from MCS velocities. associated with low velocities and high pore pressures This reference profile is also constrained by analyses of the ODP downhole density logging data. At ODP Site 889, there is a substantial velocity increase due to hydrate in a 70 [Hyndman et al., 1993]. As a consequence of the young age (5-7 Ma) of the subducted oceanic crust, the heat flow is high and the subbottom depth to the BSR at ODP Sites -100-m-thick layer overlying the BSR, and approximately 889/890 on the midslope (water depth of 1200 m) is only 2/3 of the BSR impedance contrast results from the effect of this high-velocity hydrate and the rest from the underlying 230 m (Figure 2). At ODP Site 889 (Figure 3), sediment cores were low-velocity free gas. This is a conclusion for the origin of recovered at intervals from 20 to meters below the BSR that is intermediate between the two previously pro- seafloor (mbsf) at Hole 889A and from 197 to mbsf posed extremes. For comparison with the Vancouver Island at Hole 889B [Westbrook et al., 1994]. The sediment margin data, we find that the previously published velocity data in the Blake-Bahama region of the eastern U.S. margin [Wood et al., 1994; Rowe and Gettrust, 1994; Benson et al., 1978; Sheridan et al., 1983] also show a very similar velocity enhancement and hydrate concentration. A simple interpretation of ODP core chlorinity data gives similar hydrate concentrations in both regions. section is divided into two major units. The upper unit from 0 to 128 mbsf consists of clayey silt to silty clay interbedded with silt and' fine sand layers of turbiditic origin and exhibits little deformation. This unit is interpreted to have its origin as slope basin sediments that were deposited approximately in their present location and are little deformed. The sediments of the lower unit are more indurated, through

3 YUAN ET AL.' SEISMIC VELOCITIES OF MARINE GAS HYDRATE 13,657 SW CDP ,5 L89-08 Migration X-L89-10 ODP 889/ li tjjjjjjjjjl/ /l l )l lt JJJJJJJJJJJJ/JJJJ. ljjjjjjjjjjjjj NE E 2.0 i NW CDP I... L89-10 Migration 500 m X-L89-08 SE oo I E 2.0 Figure 2. Reflection seismic sections, L89-08 and L89-10, near ODP Site 889 on the lower slope of the northern Cascadi accretionary wedge where there is a strong hydrate bottom-simulating reflector (BSR). The basic data processing included wavelet designamre, spherical divergence correction, and fmite difference migration. compaction and shght cementation. They are interpreted as accreted Cascadia Basin sediments. Studies on the Cascadia part of the site survey for ODP Leg 146 [Spence et al., 1991a, 1991b; Hyndman et al., 1994]. Detailed velocity- Basin ODP data [e.g., Davis and Villinger, 1992; Westbrook depth data have been obtained from several 1989 MCS lines et al., 1994] have suggested that the bulk sediment composi- over the accretionary prism [Yuan et al., 1994]. The air tion and average grain size (averaged over turbidite layering) do not vary significantly over the upper part of the basin gun source for the 1989 acquisition was a tuned array with a total volume of 125 L (7820 inch3). A 144-channel sediment section. streamer recorded 36-fold data to a maximum offset of 3700 m. The data processing is given in detail by Spence et al. MUltichannel Seismic Reflection Data [1991a, 1991b]. Dip moveout processing in the f-k domain was applied on the full-fold common depth point (CDP) Two multichannel seismic reflection surveys were carried gathers prior to the semblance velocity calculation in this out on the continental margin of the northern Cascadia study. subduction zone, one in 1985 [Yorath et al., 1987; Clowes Figure 2 shows portions of the two orthogonal reflection et al., 1987; Davis and Hyndman, 1989] and one in 1989 as lines L89-08 and L89-10 where the BSR is continuous and strong. L89-08 is perpendicular to the continental margin and directly across ODP Sites 889/890 (Figure 3). L E Hole Hole 500 m A,, passes about 1.5 km southwest of the drill sites. A strong BSR is imaged subparallel to the seafloor at ms below the seabed on much of the lower continental slope [Hyndman et al., 1994]. 2.00! CDP Figure 3. Detail of migrated time section from L89-08 in the immediate vicinity of Site 889. Holes 889A and 889B penetrated a small slope basin and accreted sedimento depths of and meters below seafloor (mbsf), respectively. The logged interval, however, extends only 30 m below the BSR which occurs at 224 mbsf. BSR Reflection Characteristics and Reflection Coefficients The BSR reflection is generally a single symmetrical wavelet with a reversed polarity relative to the seafloor reflection, indicating a sharp and negative impedance contrast downward across the BSR. At the higher frequencies (60-90 Hz and Hz) of a single-channel survey with small air guns of 1.97 L (120 inch ) and 0.65 L (40 inch ) near Sites 889/890 [Spence et al., 1995], no

4 , 13,658 YUAN ET AL.' SEISMIC VELOCITIES OF MARINE GAS HYDRATE reflections from the top of a hydrate layer or bottom of a gas layer have been identified (Figure 4). It has thus been inferred that the top of the hydrate accumulation and the base of the low-velocity free gas layer beneath the BSR must be gradational [e.g., Spence et al., 1995; Fink, 1995]. Similarly to Warner [1990], the seafloor reflection coefficient has been obtained from the relative amplitudes of the seafloor primary and multiple reflections. The BSR reflection coefficient has also been estimated by comparing the BSR amplitudes, compensated for the transmission loss through the seafloor, with that of the seafloor. As the multiples are reflected at different seafloor locations to the primary reflections, the reflection amplitude ratios are averaged over the entire analyzed section. This minimizes the effect of amplitude variations due to varying seafloor conditions and scattering patterns. The seafloor reflection coefficients from L89-10 are estimated as The BSR amplitudes are about 30% of that of the seafloor, and the BSR reflection coefficients are (Figure 5). Much higher BSR reflection coefficients, up to 0.15, are found along L From an analysis of a fine-grid singlechannel seismic survey near Sites 889/890, Spence et al. [1995] showed much higher BSR reflection coefficients beneath a topographic high formed by an anticlinal uplift. As hydrate density is very close to that of pore water and the amount of free gas below the BSR is assumed as very small, sediment density is changed very little by the presence of either hydrate or gas across the BSR depth. A velocity decrease of m/s is therefore required to produce the impedance contrast representing a BSR reflection coefficient of 0.1. This velocity change could come from (1) high-velocity hydrate filling sediment pore space above the BSR and sediment with normal pore water content containing little or no free gas below [Hyndman and Spence, 1992], (2) sediment with little hydrate above and consider o L89-10 Migration i p.. j i i, j I i i i ' I I I I C DP Figure 5. (a) A portion of L89-10 migration (4 km) near ODP Site 889. (b) Reflection amplitudes from seafloor primary, first multiple, and the BSR reflections from L89-10 shown in Figure 5a. The amplitudes were averaged over three near-offset traces in CDP gathers, and the amplitudes from multiples were measured at farther offsets to eliminate any amplitude variations caused by varying incidence angles. The dashed lines are actual amplitudes measurements, and the solid lines are the smoothed values. (c) The seafloor and BSR reflection coefficients estimated from reflection amplitudes in Figure 5b. (a) (b) (c) able free gas concentration below the BSR [Singh et al., 1993; Minxhull et al., 1994; MacKay et al. 1994], or some combination of 1 and Figure 4. Single-channel traces from a 120-inch 3 air gun recorded near ODP Site 889 [Spence et al., 1995]. The trace on the fight is a vertical stack of the seven traces shown to the left. The strong BSR beneath the flank of a ridge-formed anticlinal uplift is a single pulse at s with a reversed polarity relative to the seafloor reflection (SF). Velocity Data at ODP Sites 889/890 Downhole velocity measurements in ODP Sites 889A and 889B provide the first high-quality in situ velocity data extending through a BSR. At Site 889, Schlumberger logging tools, long-spaced sonic (LSS) and well-seismic tool (WST), were used to obtain sonic and vertical seismic profile (VSP) measurements of the formation velocities to a depth of 250 mbsf [Westbrook et al., 1994; MacKay et al., 1994]. The downhole data are compared with detailed MCS velocities from a 10-km-long section of L89-10 centered near the ODP sites. Sonic Velocity Logging The sonic logging data were recorded with an LSS tool that used two acoustic transmitters and two receivers to give source-receiver distances of 2.4, 3.0, and 3.6 m. Sonic log data at Sites 889A and 889B extend from 70.5 to mbsf and from 61.7 to 259 mbsf, respectively [Westbrook et al., 1994]. There are large variations in velocity of + 50 m/s

5 YUAN ET AL.' SEISMIC VELOCITIES OF MARINE GAS HYDRATE 13,659 over depth intervals of a few meters associated with the grain size changes in the turbiditc section. The layered interval, about 1800 m/s, agrees very well with the average of the sonic logs. Below the BSR, the VSP velocity drops velocity contrasts may also be enhanced by the high-velocity sharply to 1520 m/s in an interval of 15 m. The BSR thus hydrate being concentrated in the high porosity and high permeability sandier layers. Within the upper m of the slope sediments, sonic data show velocities below 1600 m/s, appropriate for recent slope sediment deposits (Figure 6b). In the 100-m interval above the BSR, the sonic logs recorded average formation velocities of 1800 m/s. We take the sonic velocities to be a minimum as the hydrate adjacent clearly represents a strong discontinuity in the VSP velocity data. MacKay et al. [1994] suggested that there was only a small velocity enhancement from the presence of hydrate above the BSR and concluded that the low velocity below the BSR is a strong indication of free gas. As neither low density nor high electrical resistivity is associated with the low velocity beneath the BSR, the concentration of free gas to the hole might have been partially dissociated or washed must be very low. However, less than a few percent of free out during the drilling. As discussed below, this velocity is gas is sufficient to produce the velocity drop and thus a above the no-hydrate reference and is interpreted to be the strong BSR. This is in agreement with the previous results result of hydrate velocity enhancement. Just below the BSR, the sonic data exhibit lower velocities but not as low as of AVO analyses which required that there be less than a few percent of free gas [Hyndrnan and Spence, 1992]. expected for free gas. The minimum velocity value for Hole 889A is 1630 m/s at 231 mbsf. On the basis of the veloc- Multichannel Seismic Velocity Analyses ities of m/s below the BSR from VSP data discussed The MCS data were collected using a 3700-m-long below, we conclude that drill fluid invasion displaced much hydrophone array that is 3 times the water depth in the study of the free gas into the formation beyond the penetration area. This large recording aperture allows unusually high depth of the sonic logs. resolution of velocity variations from CDP gathers. Two velocity studies were carried out employing conventional VSP Velocity Measurements semblance analyses. The farst was to define regional vari- The VSP at Site 889 was run from 243 mbsf, 19 m below ations in velocity-depth profiles at locations extending from the BSR, to mbsf. Surface shots fared alternately from a 4.9 L (300 inch 3) air gun and a 6.5 L (400 inch 3) water gun were recorded in the borehole by a single verticalcomponent geophone with most receiver station spaced at 5-m intervals [MacKay et al., 1994]. From 130 mbsf to the BSR at 224 mbsf, the VSP velocities range from about 1700 to 1900 m/s (Figure 6b). The average velocity in this the deep abyssal basin to the middle continental slope [Yuan et al., 1994]. In the second, detailed MCS velocities were obtained near ODP Sites 889/890 for comparison with the drill hole logging and VSP data. In a previous study, interval velocity-depth functions were determined from the Cascadia Basin seaward of the deformation front to the lower-middle slope region where the 'l's' 'e,_ sonic log B'sP X/ basin velocity trend " -::' : '. _ slope bas in sediments ' BSR depth '""='" '....' : ee e 89bsonic : *, a retionam prism.2', Figure 6. (a) L89-10 MCS velocities for a 10 km section over ODP Sites 889/890 and the downhole sonic log at the drill site. The high-velocity hydrate layer is evident to the smoothly increasing velocity with depth associated with sediment compaction. (b) Sonic logging (light dotted wiggle lines) and vertical seismic profile (VSP) data for Site 889 (open squares) are shown in comparison with the multichannel seismic (MCS) velocities taken from L89-10 (solid circles). Above the BSR, MCS velocities are in excellent agreement with the in situ measurements. The solid line fitting the velocities represents the reference velocity-depth profile in the slope basin and accreted sediments.

6 13,660 YUAN ET AL.' SEISMIC VELOCITIES OF MARINE GAS HYDRATE accretionary prism has approximately doubled in thickness VSP data, as well as our MCS velocity-depth function, to [Yuan et al., 1994]. For the latter we have concentrated in the region of ODP Sites 889/890 where a clear BSR is observed. The velocities of the continental slope sediments in the study area (Figure 6a), excluding those above the BSR depth, are well constrained and increase smoothly with provide a better-constrained starting model for the inversion. We have first applied a Monte Carlo search technique in the r-p (intercept time-slowness) domain to obtain long spatial wavelength velocity variations from the BSR and other clear reflectors. The second step was the full waveform inversion in c0-p (frequency-slowness) domain to find the fine velocity depth. The velocities are significantly lower than those in structure. Data preparation for the inversion included bandthe undeformed sequence of the Cascadia Basin at compar- pass faltering, geometrical correction, source and receiver able subbottom depths. Yuan et al. [1994] concluded that directivity correction, and transformation into c0-p domain underconsolidation compared to the basin section is the [e.g., Minshull et al., 1994]. To avoid spatial aliasing in r- primary reason for the lower average velocities in the slope p transform, four adjacent CDP gathers were combined to section. Tectonically induced fracturing that causes a give a 141-fold gather with a trace spacing of 25 m. A reduction of rigidity may be a secondary factor. source wavelet was estimated from the seafloor reflection, and the synthetic seismograms were computed using the Velocity Result From Full Waveform Inversion reflection transmission matrix method [Kennett and Kerry, 1979]. The inversion strategy consists of a sample by An automated full waveform inversion scheme for sample minimization of a misfit function between real data multichannel seismic data has proved to be well suited to and synthetic data by a nonlinear local search [Singh et al., investigate the free structure of velocity variations at the 1993; Minshull et al., 1994]. The starting model was BSR [Singh et al., 1993; Minshull et al., 1994; Singh and Minshull, 1994]. This inversion technique tries to find a progressively modified during successive iterations. The approach was implemented for L89-08 at about 2 km velocity-depth model such that the synthetic data produced seaward of the ODP Site 889. Figure 7 shows the r-p data by the model fit both the amplitude and the waveform of the real data, over ah offsets in a full-fold CDP gather. Singh et al. [1993] and Singh and Minshull [1994] applied the from the composite CDP gather with slownesses ranging from 0.01 to 0.55 s/kin. The velocity search with the Monte Carlo technique first estimated RMS velocities for waveform inversion to the data near ODP Site 889. Their four prominent reflectors by maximizing coherent energy resulting velocity structure near the BSR was confumed by along elliptical trajectories within a 60-ms time window the downhole logging and VSP data at the site but does not centered on the reflectors. The reflectors chosen were the match our MCS velocities and the downhole data in the 100- m interval just below the seafloor and the MCS data well seafloor, a reflector below the seafloor at 1.9 s, the BSR just below 2.0 s, and the reflector below the BSR at 2.3 s. below the BSR. We therefore have carried out additional Interval velocities and upper and lower velocity limits for waveform inversion using the newly available well log and each layer were obtained, and the model space defined by Slowness (s/km) SF BSR Figure 7. Slowness transform of a composite CDP gather of L89-08 for the Monte Carlo velocity search. The final velocities determined from ellipses along four reflectors are listed on the left of the figure.

7 YUAN ET AL.' SEISMIC VELOCITIES OF MARINE GAS HYDRATE 13,661 the velocity limits was searched by randomly varying each of the four layer parameters [Singh et al., 1993]. The models converged quickly to the four final velocities which are listed on the left of Figure 7 and shown in Figure 8. These velocities are in good agreement with our MCS data, the log data, and the VSP data above the BSR. The average velocity representing the sediments with no hydrate below the BSR to mbsf is 1658 m/s, much less than the velocity above the BSR. This velocity is the main difference in our result compared to that of $ingh and Minshull [1994]. A smoothed version of our semblance velocities, consistent with the model derived from the Monte Carlo search, was used as an initial model for the waveform inversion. The resulting final velocity model from the inversion (Figure 8) has velocities immediately below the seafloor that are only slightly greater than for water velocity and that increase downward. The seafloor impedance contrast mainly comes from the density change from seawater to turbiditc sediment. From mbsf to the BSR at 224 mbsf, velocities are substantially higher than velocities of the sediments below seafloor and below the BSR. At the BSR itself, a sharp decrease from a maximum velocity near 1900 m/s to a minimum of 1570 m/s is obtained in an interval of 25 m. In this inversion result, the BSR is caused mainly by the velocity increase above the BSR and less by the velocity decrease below the BSR. The low-velocity layer beneath the BSR, probably due to free gas, is too thin to be resolvable with the MCS velocity analysis. We conclude that the waveform inversion is able to resolve the velocity contrast across the BSR but unable to provide reliable velocity information below the thin gas layer and the velocity just below the BSR from the inversion is sensitive to the starting model velocity above the BSR. Thus caution must be applied in using this method alone to infer whether the BSR impedance is primarily due to high-velocity hydrate or lowvelocity free gas. reference for this section. Below 120 mbsf the ODP holes penetrated deformed and fractured sediments of the Reference Velocity of the Slope Sediments accretionary sediment prism. Upward extrapolation of the Reference Velocity From MCS Data MCS velocities from greater depths below the BSR provides the best reference for the sediments below 120 mbsf and The critical factor required to determine the velocity enhancement due to hydrate is the reference velocity-depth above the BSR. This reference velocity-depth prof'fie for the prof'de for sediments containing no hydrate. The interval slope sediments has been obtained, excluding the MCS data with logs and VSP extended only 36 m and 19 m, respectbetween 100 and 250 mbsf, using a third-order polynomial fit, ively, beneath the BSR. The logging data provide good quality velocity measurements above the BSR but do not V = z x 10-7 z x 10-9 z 3. penetrate deep enough to provide a background velocity unaffected by the presence of either hydrate or free gas. The reference velocity at the BSR depth is m/s with There is also no useful core velocity measurement from a statistical uncertainty of about +50 m/s (Figure 6a). We ODP Sites 889/890. The only velocity information from note that (1) the velocity-depth trend at the lower slope below the BSR and the gas layer is provided by the MCS region shows little variations over a wide area even at great semblance velocity analysis. We thus obtained the reference depths [Yuan et al., 1994] and thus the higher velocity velocity prof'de for most of the section above the BSR by deviation in all velocity data above the BSR from the upward extrapolation of the deeper MCS velocities, as the reference trend is a strong local anomaly and (2) there is no sediment section in this area is quite uniform in composition clear systematic sediment composition change in the upper 225 m of the ODP Sites 889/890 sediment section that could at a seismic wavelength scale [Westbrook et al., 1994]. A very similar reference is obtained by interpolation between explain such anomalously high velocities [Westbrook et al., these velocities and the near-surface velocities where there 1994]. is inferred to be little hydrate. At ODP Sites 889/890, the upper 120 m of the holes Reference Velocity Information From ODP Logs penetrated slope basin sediments [Westbrook et al., 1994]. A constraint on the reference velocity information has Although not an ideal reference, the well-def'med, deep-sea also been obtained from ODP logs at Site 889, from 70 to Cascadia Basin velocity-depth prof'de is our best no-hydrate 250 mbsf. The Schlumberger high-temperature lithodensity 50 1 oo OO 35O 4OO 45O '. MCS ref. velocity waveform inversion ø Monte I I I Carlo starting model BSR Figure 8. Comparison of velocity data near the BSR from (1) Monte Carlo search (thick dashed line), (2) starting model for the inversion (light dotted line), (3) full waveform inversion (solid heavy curve), (4) MCS data (solid dots), and (5) no-hydrate reference velocity prof'fie (light solid line).

8 13,662 YUAN ET AL.' SEISMIC VELOCITIES OF MARINE GAS HYDRATE tool (HLDT) determines formation bulk density (actually 889 are similar to the density-based porosities and have also electron density) by measuring the absorption of gamma rays emitted by a radioactive source [e.g., Westbrook et al., been converted to velocities. The neutron porosity-based velocities shown in Figure 9 are slightly higher than the 1994]. The no-hydrate reference profile has been estimated reference velocities but more than 100 m/s lower than the using a velocity-density relation assuming that the measured sonic and MCS velocities. The resistivity logs at both 889a gamma ray densities are largely unaffected by the partial replacement of the pore water with hydrate which has a and 889b have similar characters (Figure 9). The resistivitybased velocities increase with depth following the basin similar density. Anomalously low density values occurred trend from 50 to 120 mbsf. In the section below 120 mbsf, in the density log, which were considered to be artifacts of the variably enlarged hole. Jarrard et al. [1995] used a combination of caliper, photoelectric factor, and At> logs from two separate detectors of HLDT logging tool to guide the resistivity logs and resistivity-based velocities show a decrease possibly related to the hydrate dissociation. Overlying the BSR, several short intervals show higher velocities and resistivities than typical for such low-porosity the editing of the density log. Only 37 % of the recorded sediments, indicating the presence of methane hydrate. data at Site 889 were accepted as reliable. We take their acceptedata as the formation bulk density unaffected by the Two types of laboratory measurements on core recovered from the ODP Site 889 could provide reference velocity presence of hydrate. We note, however, that the edited information. Core velocities will approximate the velocities density data retained only the high values. If some of the for before original hydrate formation if the hydrate dissociarejected values are in fact valid, this editing will have biased tion upon recovery did not significantly disturb the porosity the average density to be too high and the inferred reference and sediment structure. Unfortunately, the cores were velocity to be too high. The sediment porosities have been computed from the log bulk densities using a grain density of 2680 kg/m 3, based on ODP core index measurements [Jarrard et al., 1995]. A velocity/porosity relation (Figure 11, discussed below) was used to converthe porosity log to the velocity-depth profile which is shown in Figure 9 together with the MCS velocities. The velocities from the density data average between 1600 and 1700 m/s in the logged interval, slightly higher than the reference velocities. In the vicinity of the BSR, the density based velocities follow the MCS reference velocity trend well. Neutron logs respond primarily to hydrogen content, and neutron porosity is calculated assuming that all hydrogen is highly disturbed by the slight expansion when they were moved from in situ conditions to the shipboard laboratory and the free gas,formed upon hydrate dissociation in the zone above the BSR (see photographs of Westbrook et al. [1994]), and no useful laboratory velocities were obtained. Numerous bulk density and porosity core measurements were made that could be converted to reference porosity in the same manner as described above for the density logs. The resulting reference velocities above the BSR are higher than that estimated from upward extrapolation of the MCS data, but again sample disturbance a problem and only the more indurated high-velocity samples which are nonrepresentative could be measured. We thus discounthe sample data but recognize that this data provide higher reference in the form of pore water. The neutron porosity logs at Site velocities than that from the MCS data. 15oo 17oo 19oo 15oo 17oo 19oo 15oo 17oo 19oo 15oo 17oo 19oo 15oo 17oo 19oo 50 loo BSR 25O 889a density 889a neutron 889a resistivity 889b neutron 889b resistivity Figure 9. Velocities were obtained from ODP downhole gamma ray density, neutron porosity, and resistivity logs at Site 889 (light solid lines). These data serve as a reference velocity trend in the interval where hydrate possibly exists. The MCS velocities (solid circles) and the reference velocity trend (heavy solid lines) are also plotted for comparison.

9 YUAN ET AL.: SEISMIC VELOCITIES OF MARINE GAS HYDRATE 13,663 Estimate of Gas Hydrate Concentration In the first approach, the velocity of water-saturated sediment, V,, d, is taken from the reference velocity above the Hydrate Concentration From Velocity Data BSR. The velocity of fully hydrate-saturated sediment, The most readily observable change in sediment physical V y,,d, is then determined by the simple time-average equaproperties resulting from the formation of gas hydrate is an tion which works well for combining solid matrix materials increase seismic velocity. The simplest model is for the [Wyllie et al., 1958], water in the sediment pore spaces to be partially replaced by i 1- evenly dispersed high-velocity gas hydrate "ice" that cements the sediment matrix. The recovered ODP drill cores at Sites Vhyaexl Vhyd gm ' 889/890 suggest that the hydrate is generally disseminated. where V yu, the velocity for pure methane hydrate, is 3730 Massive hydrate occurrence in the drill cores [e.g. Mathews m/s [Pearson et al., 1983; Sloan, 1990] and Vm, the matrix and yon Huene, 1985] is expected to survive recovery and velocity, is taken to be 4500 m/s [e.g., Davis and Villinger, is probably rare. The MCS velocity data discussed above, 1992]. For porosities of qb = 0.45 and qb = 0.55, the integrated with the downhole log and VSP velocity-depth calculated velocities for fully saturated sediment with hydrate data from ODP site 889, allow determination of the in situ falling all available pore space, V y,u, are 4120 and 4040 velocity enhancement from hydrate above the BSR and thus m/s, respectively. The velocity of partially hydrate-bearing estimation of the hydrate concentration. sediment with saturation S can then be approximated again The relation between velocity enhancement and hydrate by the time-averagequation combining the velocity of concentration depends upon the degree to which the hydrate hydrate saturated sediment with that of no hydrate. forms at grain boundary contacts or in the main pore spaces Figure 10 shows a plot of velocity versus hydrate satura- [e.g., Dvorkin et al., 1991]. Because we lack data on the tion in sediment for Vs, = 1600 m/s and porosity qb=0.55 gram-scale distribution of the hydrate, we have used two and for V,d= 1700 m/s and qb=0.46 (top solid line). The simple model approximations to relate the amount of hydrate porosity values are based on results from ODP logging data concentration in the pore space to sediment velocities. One and core measurements, and the velocities are appropriate is to obtain a velocity for the combination of pure hydrate for no-hydrate sediment based on the MCS velocities. The and sediment matrix, i.e., the velocity of fully hydrate- patterned area in Figure 10 indicates the estimate of the saturated sediment (no pore fluid), and then to combine that hydrate saturation in the hydrate-bearing sediments from the composite matrix with water-saturated sediment to determine enhanced velocities, ranging from 1700 to 1900 m/s. an overall model velocity [Lee et al., 1993; Wood et al., Taking more specifically the MCS velocities in the hydrate- 1994]. The other is to assume that the effect of the hydrate enriched zone and porosity values from ODP data, the solid displacing the pore water may be approximated by that of dots in Figure 10 suggest that about % of the sediment porosity reduction and to use a general velocity-porosity pore space is occupied by hydrate. relation to determine the pore space reduction due to hydrate The second approach assumes that the velocity anomaly formation. above the BSR may be approximated by a simple reduction 2000 _ > /= 1600 = I I I, I I, I,, Hydrate(% of pore space) Figure 10. Sediment velocities enhanced by hydrate concentration from two sets of parameters: (1) hydrate-free velocity is 1700 m/s; fully hydrate-saturated velocity is 4120 m/s, for the sediment whose porosity is 46 %(top solid line); (2) hydrate-free velocity is 1600 m/s; fully hydrate-saturated velocity is 4040 m/s, for the sediment whose porosity is 55 %. Measured velocities (solid dots) between 1700 and 1900 m/s suggest that % of the sediment pore spaces are filled with hydrate.

10 13,664 YUAN ET AL.' SEISMIC VELOCITIES OF MARINE GAS HYDRATE of porosity [Hyndman and Spence, 1992]. Hydrate forming in the sediment pore space effectively reduces the porosity by replacing pore fluids with high-velocity material. The mounts of hydrate required to produce the observed 70 ODP Leg 139 site 857 velocity increase can be estimated by simply calculating the difference in effective porosity between hydrate-bearing sediments and corresponding water-saturated sediments determined from the reference velocity-depth profile. 60 Several previous studies focused on the relation between sediment velocity and porosity associated with sediment compaction for the Cascadia sediments [Nobes et al., 1986; core data Davis and Villinger, 1992;.Jarrard et al., 1995]. These studies established empirical relationships between velocity Hyndman et al. (1993) and porosity using data obtained from laboratory measurements, downhole logging, and recovered drill cores. Results from ODP Legs 139 and 146 core plugs and downhole logging have suggested that the basin sediments, including those hydrothermally altered, do not show significant effects on the velocity-porosity relation from sediment cementation 40 and diagenesis [Davis and Villinger, 1992]. We thus assume that the velocity variation with depth primarily is due to porosity reduction. Using ODP downhole log data, Jarrard et al. [1995] found higher velocities for the "fractured" slope sediments compared to elsewhere on the Cascadia margin. However, their data for the fractured sediments are mainly from ODP Sites 889/890, and we conclude that these higher velocities are resulted from the presence of hydrate not from greater grain cementing. We have used the velocity-porosity relation of Hyndman Figure 11. Solid dots are velocity and porosity measurements from ODP Hole 857 of Leg 139 of the Juan de Fuca et al. [1993], which fits the measured drill core velocities Ridge region [from Davis et al, 1992]. The heavy line is and porosity well (Figure 11), to calculate inferred porosities the velocity/porosity relationship used in this study from and hydrate concentrations on three velocity-depth profiles Hyndman et al. [1993]. in the region near and at ODP Site 889: (1) downhole logging sonic velocities, (2) VSP velocities, and O) MCS velocity data (Figures 12a and 12b). As we noted above, the reference velocity-depth profile above 120 mbsf is that salts are largely excluded from the crystalline structure. The same phenomenon occurs when gas hydrates are formed for the Cascadia Basin and below 120 mbsf is that for the from pore water of deep-sea sediments. As hydrate containaccretionary prism. The small-scale variability is probably ing nearly pure water dissociates upon recovery, low pore a consequence of strongly layered grain size variations in water salinities are produced as the result of sediment pore these turbiditc sediments, rather from variations in hydrate water dilution. Because salinity variations in marine concentration. At Site 889, there is a general increase in the inferred amount of hydrate from near zero at the seafloor to an average of 20% of the pore space at the depth of BSR. The velocity of 3730 m/s for pure methane hydrate given by Pearson et al. [1983, 1986] is somewhat less than that of the sediment matrix (4300 to 5000 m/s), so the actual effect of sediments can be produced by many processes including sulfate reduction, alkalinity production, and magnesium release from interlayered positions in swelling clays, it is better to use chlorinity as an indicator for salt exclusion by hydrate formation as it is less dependent on these other processes [Hesse and Harrison, 1981]. There are many porosity reduction by hydrate filling pore space may be less other processes that can affect pore water chlorinity, such as than that predicted by the velocity-porosity relation and the dehydration of clay minerals and clay mineral filtration hydrate concentration higher. Alternatively, if the hydrate [e.g., Kastner and Ransom, 1995; Kastner et al., 1995], but is preferentially concentrated the gram contacts rather than their importance is yet to be determined. In this initial study in main pores, its effect on velocity may be greater than that we have ignored other processes and employed the simplest for simple porosity reduction. model. In this model, the in situ pore fluid prior to hydrate dissociation has near-seawater composition; chloride pro- Hydrate Concentration From Chlorinity Dilution duced by salt exclusion at the time of hydrate formation is assumed to have diffused away earlier. We estimate the Anomalies of low salinity in the pore water of recovered amount of hydrate in the pore space before recovery as deep marine sediments cores have been observed in many areas where deposits of gas hydrate are known to occur, for example, at the Blake Outer Ridge [KvenvoMen and Barn- S: 1(1- p C/p ')' sw ard, 1983], the Middle America Trench [Harrison et al., where S is the mount of hydrate in the pore space, Clpw and 1982; Hesse et al., 1985], and the Peruvian continental Cl w are the chloride concentrations of pore water and margin [Kvenvolden and Kastner, 1990]. It is well known that when ice is formed from seawater, the ions of dissolved normal bottom seawater, respectively, and Pn is the density of pure methane hydrate.

11 YUAN ET AL.: SEISMIC VELOCITIES OF MARINE GAS HYDRATE 13,665 Derived Porosity (%) Hydrate(% of pore space) ,, I... i... i... i...,,,, i... ' ).....,,, i ' ': :!':! '":7. '.'2'. ': ; -,'... Y..."7 "'.':'; --.:.:. ': :' 7...:", "..: :.::: t... / t... c P A 889 (a) (b) Figure 1. (a) Porosity-depth prof"fie calculated from velocity data near and at ODP Site 889. Also shown in the figure is the change in porosity with depth for sediment containing no hydrate (from reference velocity function). (b) Hydrate concentration in the sediment pore space. The difference between the calculated porosity and reference porosity based on velocity is related to the amount of hydrate occupying pore space. Solid circles in Figures 12a and 12b represent MCS velocities, light wiggle traces represent sonic logs, and open squares represent VSP velocities. There are several important uncertainties in this method of estimating in sire hydrate concentration: (1) the amount of residual salt trapped in the hydrate structure or between hydrate grains upon formation [see Kvenvoden and Kastner, 1990], (2) the amount of remaining hypersaline fluid excluded from hydrate into the pore water during the hydrate formation but not carded to the seafloor by advective upward fluid flow or by diffusion, (3) changing of PT conditions, and (4) other geochemical processes that affect ch!orinity as noted above. The BSR depth can migrate up or down because of several processes. Both ongoing sedimentation or seafloor warming associated with the climate warming since the Pleistocene [e.g., Westbrook et al., 1994] will cause the base of the hydrate stability field and thus the BSR to move upward. The freshwater from the dissociated hydrate below the BSR may migrate into the overlying section, reducing the pore fluid salinity where hydrate is still present. We use the constant chlorinity profile (554 mmol) at Site 888 as the reference value representing normal pore water chloride at Sites 889/890 (Figure 13a). The hydrate concentrations at these latter sites were estimated from the chlorinity anomaly in comparison with the reference chlorinity data. Figure 13b shows an increase of hydrate concentration from near zero at the seafloor to a maximum of about 35 % of the pore space at the BSR depth. Below the BSR at Sites 889/890, the chloride data show only a small decrease in concentration. This has been explained as a paleodilution process produced hy the upward migration of the hydrate stability field [Westbrook et al., 1994]. As the BSR moved upward in the sediment colurn during the Holocene, hydrate dissociation produced the observed dilution artifact which is left behind below the BSR. The hydrate concentrations using the above method have large uncertainties because of the assumptions made in the calculation. However, it is reassuring that the chlorinity data give similar hydrate concentrations and a similar concentration proœde with depth to the velocity data. Hydrate Concentration From Velocity Data in Blake-Bahama Basin Although no other BSR sites have such detailed data available as are at ODP Sites 889/890 and adjacent regions, the same analysis approach can be applied to other areas. For comparison with the Cascadia data we provide initial results analysis of published data for the Blake-Bahama region (Figure 14). Multichannel velocity data collected from the surface and near-seafloor surveys are also indicative of significant hydrate concentration in this area. The Blake Outer Ridge is a large sedimentary feature located on the east coast of the U.S. Atlantic continental margin. It was the first location where gas hydrate was concluded as the cause of the BSR [Markl et al., 1970; $toll et al., 1971; Markl and Bryan, 1983]. The sediment composition of the region is silty clay with interbedded thin layer of sandy clay, similar to that of the Cascadia margin. The prominent BSR in the region has been extensively mapped using single-channel, multichannel, and deep-tow seismic data [e.g., Tucholke et al., 1977; Shipley et al., 1979; Paull and Dillon, 1981; Dillon and Paull, 1983; Rowe and Gettrust, 1993; Dillon et al., 1994; Katzrnan et al., 1994; Lee et al., 1994]. Indications of hydrate occurrence have also been found from drilling during Deep Sea Drilling Project (DSDP) Leg 11 [Hollister et al., 1972]. DSDP Legs 44 and 76 drilled through sediments of the Bahama Basin

12 13,666 YUAN ET AL.' SEISMIC VELOCITIES OF MARINE GAS HYDRATE Chlorine Hydrate(% of pore space) d Site Site 889A 889B 88 B Site 890B loo 15o E '-' BSR o 3OO 350 ø - 0 ø (a) Figure 13. (a) Chloride concentration versus depth profries at Sites Average chloride concentration Site 888 (554 mmol) is taken to represent normal pore water chlorinity for the region on the northern Cascadia margin. (b) Concentration of hydrate in the pore spaces of sediments at Sites 889/890 estimated from chlorinity dilution. il ) o o o [Benson et al, 1978; Sheridan et al., 1983] where there is no BSR and no gas hydrate was expected. The Bahama Basin core measurements from these sites thus aid in determining the reference velocity profile for no-hydrate sediments. The data used in this study include (1) surface multichannel reflection data from Wood et al., [1994], (2).W N W Figure 14. Regional bathymetry map of the south eastern U.S. continental margin modified from Sheridan et al. [1983]. Bathymetri contours are in meters. Sites 390 and 391 from Deep Sea Drilling Project (DSDP) Leg 44 and Sites 534 from DSDP Leg 76 are located in the Blake- Bahama Basin, and Site 533 from DSDP Leg 76 is on the Blake Outer Ridge. Locations of previous studies in the region are indicated by letters: W, Wood et al. [1994], L, Lee et al. [1993, 1994], and R & G, Rowe and Gettrust [ O 28 deep-tow multichannel velocities from Rowe and Gettrust [1993], aml (3) chlorinity and core data from DSDP Legs 44 and 76 [Benson et al., 1978; Sheridan et al., 1983]. Wood et al. [1994] investigated the multichannel reflection data collected on the southeast part of the Blake Outer Ridge where water depth ranges from 1600 to 2600 m the BSR is observed at mbsf. The data were acquired using a 177-L source array and a 6-kin, 240- channel streamer. They applied an interactive velocity analysis procedure in r-p domain to determine the velocitydepth data. We have applied the same analysis approach used on the Cascadia margin to the velocity data on the Blake Outer Ridge where the sediment section is quite uniform with no major lithology variations on the scale of a seismic wavelength. The reference velocity-depth profile for hydrate-free sediment in the region was estimated from the MCS velocities of Wood et al. [1994] where there is no obvious BSR. There is some indication of high velocities in the zone of hydrate stability above the expected BSR depth. So the velocities between 220 and 650 mbsf were excluded. The remaining data exhibit a smooth velocity increase with depth associated with increasing consolidation. They were fit with a second-order regression which we take to represent sediment velocity-depth profile unaffected by hydrate (Figure 15a). The reference velocity profile was further confumed with the velocity measurement made on the core samples at DSDP Sites 390, 391, and 534 [Benson et al., 1978; Sheridan et al.,!983] where no methane hydrate was inferred. The core velocities agree well with the multichannel reference velocity-depth profile (Figure 15b). We have again used the porosity reduction method to calculate the hydrate saturation in the pore space of the sediment. The porosity-velocity relationship of Hyndman et al. [1993] proves to be very apphcable to the Blake-Bahama sediment cores (Figure 16). The method was apphed to the two sets of velocity data in the Blake-Bahama region where

13 YUAN ET AL.' SEISMIC VELOCITIES OF MARINE GAS HYDRATE 13, OO ' -' ' I... I... I... I ' ' ' ' I " '' Multichannel = Velocities I $,te 3911 Blake-Bahama Wood et al. Site Hydrate zone... BSR" Site 39C e 1 ooo Trend where no (a) reference based onx ; Wood et al. (1994) i... I... i...!... i,, x,,,,,,i, Figure 15. (a) Reference velocity for the Blake-Bahama region estimated using MCS velocities of Wood et al. [1994]. The velocities were determined from a region where no BSR is present on the seismic profile. For further accuracy, the velocities within the hydrate stability zone (open circles) have been excluded from the computation. The BSR varies from 440 to 500 mbsf. (b) Comparison between the reference velocity and the DSDP core measurements of the Bahama Basin. (b) O 50 Io!... i,ej Site 390. Site there is a strong BSR. The velocity data of Wood et al. [1994] show a high velocity above the BSR relative to the reference profile (Figure 17a). The data also show a lowvelocity zone below the BSR which could be caused by a small amount of free gas [Wood et al., 1994]. Similar to the Cascadia margin, the hydrate concentration increases with depth from zero near the seafloor to a maximum of % of pore space just above the BSR (Figure 18a). The second set of velocity data was from the deep-tow multichannel survey of Rowe and Gettrust [1993]. Both source and receiver array were towed at m above the seafloor in water depths up to 6000 m on the western flank of the Blake Outer Ridge [Gettrust et al., 1988; Gettrust and o Ross, 1990]. The high-frequency source, Hz, o 40 permits resolution of small scale (---5 m vertical, m horizontal) changes in the sediment structure within the upper m of the sediments [Bowles et al., 1991]. A strong BSR is observed at mbsf which marks the base 3o of a 400-m-thick layer of sediment inferred to contain high / ma A A A AA. hydrate concentrations [Paull and Dillon, 1981; Markl and Hyndmanetal ( 993) -- ' { -.'. Bryan, 1983; Rowe and Gettrust, 1993]. The deep-tow data 2o show a very high velocity of m/s above the BSR (Figure 17b), and the estimated hydrate concentration reaches a maximum of 50 % of the pore space (Figure 18b). The amount of hydrate in the pore space was also estimated using the pore fluid chlorinity data from DSDP Site 533 located on the crest of the Blake Outer Ridge [Sheridan et al., 1983]. Similar to the Cascadia margin, the Figure 16. Comparison between the porosity-velocity chlorinity profile shows a progressive dilution with depth relationship of Hyndman et al. [1993] with measured (Figure 19a). No pore fluid samples were taken below 400 porosities and velocities of the sediment cores from DSDP m at the site, but hydrate concentration averaging 8 % of the Sites 390, 391, and 534, in the Bahama Basin [Benson et pore space between 250 and 400 mbsf is estimated (Figure al., 1978; Sheridan et al., 1983]. 19b). Since the BSR at this location is at mbsf

14 13,668 YUAN ET AL.' SEISMIC VELOCIT] ES OF MARINE GAS HYDRATE Multichannel *., Velocities i Wood et al. [ o.o we Ro I ø X I'x e,,.. Hydrate o I - o_o e o t.- Hydrate / Free gas.,,....:_ %xe Trend where ', no BSR [,- 800 ' ' ; ' '... (a), / Ref. Velocity Figure 17. Interval velocity-depth data from Blake Outer Ridge: (a) MCS velocities of Wood et al. [1994] from the region where there is a strong BSR; (b) deep-tow data of Rowe and Gettrust [1993]. Velocities from where no and clear BSRs were observed are represented with open and solid circles, respectively. The reference velocity based on the MCS data from Wood et al. [1994] is also shown. (b) [Hollister et al., 1972], the trend of increasing concentra- which are also 100 km apart. The BSR depths are also tion with depth suggests a much higher hydrate saturation different because of the different water depth. Thus the near the BSR. It should be emphasized that the chlorinity hydrate concentrations estimated from the chlorinity and data are from a site 100 km from the two seismic surveys, velocity data cannot be directly compared. Hydrate(% of pore space) 4O 5O ' i i Hydrate(% of pore space) '! ' i i, i ' i, i 200 d co_ OO Multichannel Data from Wood et al. I,, I, I, I, I, I (a) Deep-tow Data from Rowe & Gettrust,,, I, I, t t I t I (b) Figure 18. Vertical distributions of methane hydrate in the sediment of the Blake Outer Ridge from velocity analysis: (a) from MCS velocity data of Wood et al. [1994];(b) from deep-tow velocity data of Rowe and Gettrust [1993]. The dashed lines represent polynomial fitting to the data.

15 YUAN ET AL.' SEISMIC VELOCITIES OF MARINE GAS HYDRATE 13,669 Chlorine (mm) Hydrate(% of pore space) ,, o...,,,,,,, 0 0 " / / / / / / \ \ \ \ \ \ 500 o Site 533 6OO BSR Site 533a BSR (a) (b) Figure 19. Vertical distribution of methane hydrate at DSDP Site 533 inferred from chlorinity dilution. No hydrate concentration estimate on the greater depth can be made as pore fluids were not sampled below 400 mbsf. Discussion and Conclusions thin gas layer is only resolved in the downhole VSP data. The estimated hydrate concentration-depth prof'des indicate The increase in seismic velocity associated with the a total hydrate amount of about 7 m3/m 2 of ocean floor or formation of gas hydrate in marine sediments can be used to methane amount of 800 m3/m 2 at STP. The total methane obtain semiquantitativestimate of hydrate concentration. gas is estimated to be about 200 TCF (trillion cubic feet) in An integrated analysis of multichannel seismic and ODP an area of 20 x 300 km where clear BSRs have been downhole velocity data for an area of the Cascadia midcontiobserved on the Vancouver continental margin. nental slope off Vancouver Island has allowed the velocity enhancement to be estimated. The critical reference veloc- In comparison with the Vancouver Island margin data, the previously published velocity and chlorinity data in the ity-depth prof'de for no hydrate has been obtained by upward Blake-Bahama region of the eastern U.S. margin where there extrapolation of the deeper MCS velocities and through the is a strong BSR also indicate a similar velocity enhancement downhole density log. The multichannel, downhole log, and and inferred hydrate concentration. Such hydrate concentradownhole VSP velocity-depth prof'des are in excellent tions of 20-40% of pore space or 10-20% of the sediment agreement where they overlap from the seafloor down to the volume over depth interval of several hundreds meters may BSR. The velocity enhancement increases downward from be representative of many areas where a strong BSR is near zero at the seafloor to a maximum of about 200 m/s observed. above the BSR at 224 mbsf. Just above the BSR the velocity is increased to about 1800 m/s from the no-hydrate Acknowledgments. We wish to thank M. MacKay for providreference of 1600 m/s. Two different velocity enhancement ing us her VSP data, R. Jarrard for his revised ODP logging data, versus concentration models give hydrate concentration just and M.M. Rowe and J.F. Gettrust for their deep-tow seismic velocity data. We are also very grateful to T. Minshull and S. Singh for their generous assistance in applying their inversion routines, especially during the period when TY worked at Bullard Laboratory. Our manuscript benefited greatly from the careful and constructive reviews by D.R. Hutchinson, W.P. Dillion, W.T. Wood, and E.E. Davis. The research was funded by NSERC through operating grants 0GP to GDS, 0GP to RDH, and NSERC scholarship to BD. TY is also thankful to Amoco Canada for the scholarship support during his graduate studies. This is Geological Sun, ey of Canada Contribution above the BSR of 20-30% of pore space (10-15% of the sediment volume). This amount is in approximate agreement with the maximum of % concentration in the pore space and the concentration profile with depth estimated from pore fluid chlorinity data of recovered cores at this site. On a seismic wavelength scale the velocity enhancement has a gradational top and increases smoothly with depth to the BSR, which explains why there is no seismic reflection from the top of the hydrate layer. At ODP Sites 889/890 the velocity increase due to hydrate above the BSR accounts for ---2/3 of the impedance References contrast required to produce the BSR reflection amplitudes. Andreassen, K., P.E. Hart, and A. Grantz, Seismic studies of a The remainder of the impedance contrast appears to come bottom simulating reflection related to gas hydrate beneath the from the velocity decrease associated with small concentra- continental margin of the Beaufort Sea, J. Geophys. Res., 100, tions of free gas below the BSR. The low velocity of the 12,659-12,673, 1995.

16 13,670 YUAN ET AL.: SEISMIC VELOCITIES OF MARINE GAS HYDRATE Benson, W.E., et al., Initial Reports of the Deep Sea Drilling Project, vol. 44, U.S. Govt. Print. Off., Washington, D.C., Bowles, F.A., J.F. Gettrust, and M. Rowe, Geological interpretations based on the deep-tow single channel and multichannel seismic data from the Bermuda Rise, Mar. Geol., 96, , Clowes, R.M, C.J. Yorath, and R.D. Hyndman, Reflection mapping across the convergent margin of western Canada, Geophys. J. R. Astron. Soc., 89, 79-84, Davis, E.E., and R.D. Hyndman, Accretion and recent deformation of sediments along the northern Cascadia subduction zone, Geol. Soc. Am. Bull., 101, , Davis, E.E., and H. Viilinger, Tectonics and thermal structure of the Middle Valley sedimented rift, northern Juan de Fuea Ridge, edited by E.E. Davis, et al., Ocean Drill. Program Initial Rep., 139, 9-41, Davis, E.E., et al., Proceedings of the Ocean drilling Program, Initial Reports, vol. 139, Ocean Drill. Program, College Station, Tex., Dillon, W.P., and C.K. Paull, Marine gas hydrates, II, Geophysical evidence, in Natural Gas Hydrates: Properties, Occurrences and Recovery, edited by J.L. Cox, pp , Butterworth-Heinemann, Newton, Mass., Dillon, W.P., M.W. Lee, and D.F. Coleman, Identification of marine hydrates in situ and their distribution off the Atlantic coast of the United States, in International Conference on Natural Gas Hydrates, edited by E.D. Sloan, J. Happel, and M.A. Hnatow, Ann. of the N.Y. Acad. of Sci., 715, , Dvorkin, J., G. Mavko, and A. Nur, The effect of eementation on the elastic properties of granular materia, Mech. ofmater., 12, , Fixtic, C.R., Methane hydrate distribution offshore Vancouver Island from detailed single channel seismic studies, M.Sc. thesis, Univ. of Victoria, Victoria, B.C., Canada, Gettrust, J.F. and J.H. Ross, Development of a low-frequency, deep-towed geoaeoustie system, paper presented at Oceans '90, IEEE, Washington, D.C., Gettrust, J.F., M. Grimm, S. Madosik, and M. Rowe, Results of a deep-tow multichannel survey on the Bermuda Rise, Geophys. Res. Lett., 15, , Harrison, W.E., R. Hesse, and J.M. Gieskes, Relationship between sedimentary facies and interstitial water chemistry of slope, trench, and Coco Plate sites from the Middle America Trench transeet, active margins off Guatemala, Deep Sea Drilling Project Leg 76, edited by J. Auboin, R. von Huene, et al., Initial Rep. Deep Sea Drill. Proj., 76, , Hesse, R., J. Level, and J.M. Gieskes, Interstitial water chemistry of gas-hydrate-bearing sections of the Middle America Trench slope, Deep Sea Drilling Project Leg 84, edited by R. von Huene, J. Auboin, et al., Initial Rep. Deep Sea Drill. Proj., 84, , Hesse, R., and W.E. Harrison, Gas hydrates (clathrates) causing pore-water freshening and oxygen isotope fractionation in deepwater sedimentary sections of terrigenous continental margins, Earth Planet. Sci. Lett., 55, , Hollister, C.D., J.I. Ewing, D. Habib, J.C. Hathaway, Y. Lancelot, H. Luterbacher, F.J. Paulus, C.W. Poag, J.A. Wileoxon, and P. Worstell, Sites Blake-Bahama Outer Ridge (northern end), edited by C.D. Hollister, et al., Initial Rep. Deep Sea Drill. Proj., 11, , Hyndman, R.D., and E.E. Davis, A mechanism for the formation of methane hydrate and seafloor bottom simulation reflectors by vertical fluid expulsion, J. Geophys. Res., 97, , Hyndman, R.D., and G.D. Spence, A seismic study of methane hydrate marine bottom simulating reflectors, J. Geophys. Res., 97, , Hyndman, R.D., G.F. Moore, and K. Moran, Velocity, porosity and pore-fluid loss from the Nankai subduction zone aecretionary prism, edited by I.A. Hill, et al., Proc. Ocean Drill. Program Sci. Results, 131, , Hyndman, R.D., G.D. Spence, T. Yuan, and E.E. Davis, Regional geophysics and structural framework of the Vancouver Island margin accretionary prism, edited by G.K. Westbrook, et al., Ocean Drill. Program Initial Rep., 146, , Jarrard, R.D., M.E. MacKay, G.K. Westbrook, and E.J. Screaton, Log-based porosity of ODP Sites on the Cascadi accretionary prism, edited by B. Carson, G.K. Westbrook, R.J. Musgrave, and I. Suess, Proc. Ocean Drill. Program Sci. Results, 146, , Kastner, M., and B. Ransom, Chloride concentration and chlorine stable isotope composition of fluids in subduction zones, paper presented at the Geological Association of Canada and Mineralogieal Association of Canada annual meeting, Geol. Assoc. of Can/Mineral Assoc. of Can., Victoria, Canada, Kastner, M., K.A. Kvenvolden, M.J. Whitiear, A. Camerlenghi, and T.D. Lorenson, Relation between pore fluid chemistry and gas hydrate associated with bottom-simulating reflectors at the Cascadia Margin, edited by B. Carson, G.K. Westbrook, R.J. Musgrave, and E. Suess, Proc. Ocean Drill. Program, Sci. Results, 146, , Katzman, R., W.S. Holbrook, and C.K. Paul, A combined vertical- incidence and wide-angle seismic study of a gas hydrate zone, Blake Ridge, J. Geophys. Res., 99, 17,975-17,995, Kennett, B.L.N., and N.J. Kerry, Seismic waves in a stratified half-space, Geophys. J. R. Astron. Soc., 57, , Kvenvolden, K.A., and L.A. Barnard, Gas hydrate of the Blake Outer Ridge, Site 533, Deep Sea Drilling Project Leg 76, edited by R.E. Sheriden et al., Initial Rep. Deep Sea Drill. Proj., 76, , Kvenvolden, K.A. and M. Kastner, Gas hydrate of the Peruvian outer continental margin, edited by E. Suess, et al., Proc. Ocean Drill. Program Sci. Results, 112, , Kvenvolden, K.A. and T.J. McDonald, Gas hydrate of the Middle America trench - DSDP Leg 84, edited by R. von Huene, et al., Initial Rep. Deep Sea Drill. Proj., 84, , Lee, M.W., D.R. Hutchinson, W.P. Dillon, J.J. Miller, W.F. Agena, and B.A. Swift, Method of estimating the amount of in situ gas hydrate in deep marine sediments, Mar. Pet. Geol., 10, , Lee, M.W., D.R. Hutchinson, W.F. Agena, W.P. Dillon, J.J. Miller, and B.A. Swift, Seismi character of gas hydrates on the southeastern U.S. continental margin, Mar. Geophys. Res., 16, , MacKay, M.E., R.D. Jarrard, G.K. Westbrook, R.D. Hyndman, and the Shipboard Scientific Party of ODP Leg 146, Origin of bottom simulating reflectors: Geophysical evidence from the Caseadiaeeretionary prism, Geology, 22, , Markl, R.G., and G.M. Bryan, Stratigraphic evolution of Blake Outer Ridge, AAPG Bull., 67, , Markl, R.G., G.M. Bryan, and J.I. Ewing, Structure of the Blake- Bahama Outer Ridge, J. Geophys. Res., 75, , Mathews, M.A., and von Huene, R., Site 570 methane hydrate zone, edited by von Huene, R., Aubouin, J., et al., Initial Rep. Deep Sea Drill. Proj., 84, , Miller, J.J., M.W. Lee, and R. von Huene, An analysis of a seismic reflection from the base of a gas hydrate zone, offshore Peru, AAPG Bull., 75, , Minshull, T.A., and R.S. White, Sediment compaction and fluid Inigration in the Makran accretionary prison, J. Geophys. Res., 94, , Minshull, T.A., S.C. Singh, and G.K. Westbrook, Seismic velocity structure at a gas hydrate reflector, offshore western Colombia, from full waveform inversion, J. Geophys. Res., 99, , Nobes, D.C., H. Viilinger, E.E. Davis, and L.K. Law, Estimation of marine sediment bulk physical properties at depth from seafloor geophysical measurements, J. Geophys. Res., 91, 14,033-14,043, Paull, C.K., and W.P. Dillon, Appearance and distribution of the gas-hydrate reflection in the Blake Ridge region, offshore southeastern United States, U.S. Geol. Surv. Misc. FieM Stud. Map MF-1252, Pearson, C.F., J. Murphy, and R. Hermes, Acoustic and resistivity measurements on rock samples containing hydrates: Laboratory analogues to natural gas hydrate deposits, J. Geophys. Res., 91, 14,132-14,138, 1986.

17 YUAN ET AL.: SEISMIC VELOCITIES OF MARINE GAS HYDRATE 13,671 Pearson, C.F., P.M. Halleck, P.L. McGuire, R. Hermes, and M. Mathews, Natural gas hydrate: A review of in situ properties, J. Phys. Chem., 87, , Rowe, M.M., and J.F. Gettrust, Fine structure of methane hydratebearing sediments on the Blake Outer Ridge as determined from deep-tow multichannel seismic data, J. Geophys. Res., 98, , Sheridan, R.E., et al., Initial Reports of the Deep Sea Drilling Project, vol. 76, U.S. Govt. Print. Offi., Washington, D.C., Shipley, T.H., and B.M. Didyk, Occurrences of methane hydrate offshore southern Mexico, Initial Rep. Deep Sea Drill. Proj., Carson, G.K. Westbrook, R.J. Musgrave, and E. Suess, Proc. Ocean Drill. Program $ci. Results, 146, , Stoll R.D., J. Ewing, and G.M. Bryan, Anomalous wave velocities in sediments containin gas hydrate, J. Geophys. Res., 76, , Tucholke, B.E., G.M. Bryan, and J.I. Ewing, Gas-hydrate horizons detected in seismic profile data from the western North Atlantic, AAPG Bull., 6], , Warner, M., Absolute reflections from deep seismic reflections, Tectonophysics, 173, 15-23, Whalley, E., Speed of longitudinal sound in clathrate hydrates, J. Geophys. Res., 85, , , , Westbrook, G.K., B. Carson, R.J. Musgrave, and Shipboard Shipley, T.H., M.H. Huston, R.T. Buffler, F.J. Shaub, K.J. Scientists, Proceedings of the Ocean Drilling Program, Initial MeMillen, J.W. Ladd, and J.L. Worzel, Seismic evidence for Reports, vol. 146, Ocean Drill. JProgmm, College Station, Tex., wide spread possible gas hydrate horizons on continental slopes and rises,,4apg Bull., 63, , Wood, W.T., P.L. Stoffa, and T.H. Shipley, Quantitative detection Singh, S.C., and T.A. Minshull, Velocity structure of a gas hydrate of methane hydrate through high-resolution seismic velocity reflector at Ocean Drilling Program site 889 from global seismic analysis, J. Geophys. Res., 99, , waveform inversion, J. Geophys. Res., 99, 24,221-24,233, Wyllie, M.R.J., A.R. Gregory, and G.H.F. Gardner, An experi mental investigation of factors affecting elastic wave velocities Singh, S.C., T.A. Minshull, and G.D. Spence, Velocity structure in porous media, Geophysics, 23, , of a gas hydrate reflector, Science, 260, , Yorath, C.J., et al., Marine multichannel seismic reflection, gravity Sloan, E.D., Clathrate hydrates of natural gases, 641 pp., Marcel and magnetic profiles - Vancouver Island continental margin and Dekker, New York, Juan de Fuca Ridge, Open File Rep. Geol. $urv. Can., ]66], Spence, G.D., R.D. Hyndman, E.E. Davis, and C.J. Yorath, Multichannel reflection profiles across the Vancouver Island Yuan, T, G.D. Spence, and R.D. Hyndman, Seismic velocities and continental shelf and slope, Geol. $urv. Can. Open File 2391, inferred porosities in the accretionary wedge sediments at the 1991a. Cascadia margin, J. Geophys. Res., 99, , Spence, G.D., R.D. Hyndman, E.E. Davis, and C.J. Yorath, Seismic structure of the northem Cascadia accretionary prism: Evidence from new multichannel seismic reflection data, B. Desmons, R.D. Hyndman, G.D. Spence, T. Yuan, School of Earth and Ocean Sciences, University of Victoria, P.O. Box 1700, Continental Lithosphere, Deep Seismic Reflections, Geodyn. Victoria, British Columbia, VSW 2Y2, Canada. ( $er., vol. 22, edited by R. Meissner, et al., pp , AGU, Washington, D.C., 1991 b. yuan@uvphys.phys.uvie.ea) Spence, G.D., T.A. Minshull, and C. Fink, Seismic studies of methane gas hydrate, offshore Vancouver Island, edited by B. (Received August 7, 1995; revised December 8, 1995; accepted January 4, 1996.)