Geophysical Journal International

Transcription

1 Geophysical Journal International Geophys. J. Int. (2011) 187, doi: /j X x P-wave and S-wave velocity structure of northern Cascadia margin gas hydrates Ranjan Dash and George Spence School of Earth and Ocean Sciences, University of Victoria, P.O. Box 3055, Victoria, British Columbia, V8W 3P6 Canada. Accepted 2011 September 1. Received 2011 August 30; in original form 2011 February 24 1 INTRODUCTION Off Vancouver Island at the North Cascadia margin, widespread gas hydrate has been inferred by bottom-simulating reflectors (BSRs) in seismic reflection data, by increased electrical resistivity in electrical sounding surveys, and by drilling in Ocean Drilling Program (ODP) Leg 146 (Westbrook et al. 1994; Yuan et al. 1996; Schwalenberg et al. 2005). Gas hydrate has been recovered directly from shallow piston coring and from drilling in Integrated Ocean Drilling Program (IODP) Expedition X311 (Riedel et al. 2006a, b). From the distribution of the BSR, which occurs at depths of m below the sea floor (mbsf), hydrate in the accretionary prism sediments extends over a 30 km wide band in water depths ranging from 800 to 2200 m. Seismic velocities are sensitive to the concentration and distribution of gas hydrate in the sediments. For pure methane hydrate, Now at: Chevron Energy Technology Company, 6001 Bollinger Canyon Road, San Ramon, CA 94583, USA. SUMMARY Single-channel seismic and wide-angle reflection data collected in September 2005 were analysed along a 2-D profile of 10 ocean bottom seismometers (OBSs) on the continental slope region off Vancouver Island, near ODP Site 889 and IODP Site U1327. The objectives were to determine the shallow P-wave and S-wave velocity structure associated with marine gas hydrates and to estimate the hydrate concentration and distribution in the sediment pore space. Combined traveltime inversion of single-channel and OBS data produced a P-wave velocity model down to the depth of the bottom-simulating reflector (BSR) at 230(±5) m below the seafloor (mbsf). Mean velocities, which increased from 1.50 km s 1 at the seafloor to 1.88 km s 1 at the BSR, are in good agreement with the sonic log data from Sites 889 and U1327. The increase in P-wave velocity of the hydrate-bearing sediments relative to a background no-hydrate velocity was utilized to estimate the hydrate concentration by using effective medium theory. An average concentration of 13 per cent in the interval from mbsf was estimated from the P-wave velocity model. Lateral continuity of the model data confirms that these average hydrate concentrations are also found around the drillsites out to distances of a few kilometres. Forward modelling of S-waves was carried out using the data from the OBS horizontal components. Above the BSR, S-wave velocities are higher than a background velocity profile based on a rock physics model and on global averages for unconsolidated sediments. This increase in velocity suggests that the hydrate is distributed as part of the load-bearing matrix to increase the rigidity of the sediment. Key words: Inverse theory; Gas and hydrate systems; Ocean drilling; Permeability and porosity; Seismic attenuation; Seismic tomography. P-wave velocities have been measured as 3650 m s 1, S-wave velocities are 1890 m s 1 and densities are 900 kg m 3 (Waite et al. 2000). Within the upper few hundred metres below the seafloor (mbsf), the seismic velocities of poorly consolidated sediments containing no hydrate or free gas are typically m s 1 for P-waves and m s 1 for S-waves. Hence, the measured seismic velocities of sediments containing gas hydrate are elevated, and these provide a measure of the concentration of the hydrate. The seismic velocity also depends on whether the hydrate simply fills the pore space, or whether it cements the grains or acts as part of the load-bearing frame. However, the effect of hydrate distribution is different for P-waves than for S-waves, and so the measurement of both velocities may provide the means to determine both concentration and distribution (see Helgerud et al. 1999; Chand et al. 2004; Spence et al. 2010). At the North Cascadia margin, gas hydrate was first detected in multichannel seismic surveys (MCS) in 1985 and 1989 (Davis & Hyndman 1989; Hyndman & Spence 1992). Detailed velocity analyses of the MCS data revealed substantial P-wave velocity anomalies that were interpreted to indicate relatively large hydrate GJI Marine geosciences and applied geophysics C 2011 The Authors 1363

2 1364 R. Dash and G. Spence saturations of per cent in the lower part of the stability field (Yuan et al. 1996). Sonic velocity data from Site 889 of ODP Leg 146 and from nearby Site U1327 of IODP X311 were also interpreted to indicate similar saturations (Chen 2006). However, full waveform analyses of the MCS data from 1989 were used to infer that the BSR impedance contrast is mainly due to gas below the BSR and not hydrate above the BSR (Singh et al. 1993). Problems with using seismic reflection data to determine hydrate velocities are (1) that shot-receiver offsets are limited and hence can cause large uncertainties in traveltime inversion, or (2) that long-wavelength velocity variations are poorly constrained, and so there are large uncertainties in amplitude analyses associated with BSR (Chen et al. 2007). To resolve the long-wavelength velocity variations, ocean bottom seismometers (OBSs) can record wide-angle and refracted arrivals, and horizontal OBS components can better record S-waves to provide additional constraints on velocities and hydrate distribution. OBS surveys with hydrate targets have been carried out on the Cascadia margin off Oregon (Tréhu & Flueh 2001; Kumar et al. 2006a,b) and off Vancouver Island (Hobro et al. 2005; Zykov 2006). In this paper, we present results of a detailed seismic study in 2005 near ODP Site 889 and IODP Site U1327, recorded on a 2-D profile of 10 OCBs with an instrument separation of only 100 m. A P-wave velocity model, extending several kilometres away from the drillsites, was obtained by simultaneous traveltime inversion of wide-angle arrivals from the OBSs and vertical-incidence data from a surface single-channel streamer. The seismic velocities, integrated with the sonic data from ODP and IODP drillholes, are used to interpret an average hydrate concentration of 13 per cent in the 100-m-thick layer above the BSR, and an S-wave velocity analysis from a representative OBS suggests that hydrate cements the sediment grains to increase the rigidity of the rock matrix. 2 SEISMIC DATA During a cruise on the vessel C.C.G. John P. Tully, a Generator- Injector (GI) airgun was recorded on a single-channel streamer and on 10 OBSs from Dalhousie University (Fig. 1). The GI-gun was configured with a 45 cubic inches (cu. in.) generator chamber and a 45 cu. in. injector chamber. The line of OBSs, coincident with multichannel seismic line 89 08, passed through ODP sites 889A and 889B, and IODP Site U1327 was at a distance of about 200 m from the nearest OBS. The source was recorded along 5 lines parallel to with a separation of 500 m, and 3 lines perpendicular to 89 08; the additional lines were included mainly to provide constraints in determining the accurate seafloor position of the OBSs. The normal-incidence data, recorded on the single-channel streamer at a 1 ms sample rate, show a clearly recognizable BSR at a two-way time of about 280 ms below the seafloor (Fig. 2). The data quality is good, particularly after predictive deconvolution and a filter with a bandpass of 40 Hz to 160 Hz. In the northeastern part of Line 3 and the northwestern part of Line 7, the smooth and flat seafloor is associated with a large slope basin; within the basin, the subhorizontal bedded slope sediments obscure the BSR. The uppermost sediment reflections are from clayey silts and fine sands (Riedel et al. 2006b). Below the slope sediments lie the accretionary prism sediments, which are more consolidated and highly fractured. However, the boundary between the accreted sediment layer and slope sediments is not well-defined. The accreted sediments exhibit Figure 1. OBS and single-channel streamer experiment near IODP Site U1327 on the continental slope off Vancouver Island. Magenta lines are the shot profiles and yellow circles are the deployed OBS positions. The thick black line marks the previous multichannel seismic profile and the transect for the IODP X311 drill sites (red circles). The upper inset shows the general location of the study area. Lower inset shows the enlarged area around the OBSs. Also shown are the previous ODP drill sites in the area (black diamonds).

3 P- and S-wave velocity structure 1365 Figure 2. Single-channel seismic data along shot lines 3 (a) and 7 (b) near IODP Site U1327. Line locations are shown in Fig. 1. The approximate OBS position on the seafloor is shown in (a). surprisingly few reflections probably because of pervasive deformation. The OBS data were collected out to 7 km maximum offset on either side of the deployed position. The data quality was good, but unlike the SCS data, the bubble pulse was prominent and obscured the primary reflections. A gapped-deconvolution was applied to the pressure and vertical component geophone data, and this was followed by a bandpass filter of Hz to remove the high frequency noise (Figs 3 and 4). The processed data clearly show the reflections including the BSR, but refractions are almost absent in the OBS data. Where present, their amplitudes are very weak and their apparent velocities are very small ( 1.65 km s 1 ). The absence of any strong refraction indicates that no layer with a velocity higher than 1.65 km s 1 is reached within the offset range that the small airgun source can be recorded. Although the horizontal components have varying data quality, the presence of shearwaves(s-waves) in the horizontal components is evident on most OBSs. In particular, arrivals can be seen below the BSR, which occurs at a time of about 1150 ms at near offset (Fig. 3). In pressure and vertical geophone components (Figs 3a and b), events below the BSR are virtually absent which may be explained as strong attenuation of P-waves below BSR due to the presence of free gas. However, S-waves are insensitive to gas and are very useful in areas where P-waves are not recorded. 3 OBS RELOCATION With ship s navigation from the differential Global Positioning System (GPS), the deployment position of an OBS on the sea surface is known with high accuracy, but its actual rest point on the seafloor is not well known. While sinking to the seafloor, the OBS can drift by several hundred metres from the point of deployment, depending on the water depth and the current speed. Although the depth of the seafloor is constrained by bathymetry data, an accurate water depth at the final position of the OBS is not known. The shot positions recorded from the ship have an uncertainty of the order of tens of metres. Also, the time drift of the clock is assumed to be very small and linear with time, which may not be a valid assumption and may contribute to the error in the seismic data. Thus the OBS position, clock drift and the shot locations need to be known more precisely to get a better model of the subsurface. The OBS relocation problem is similar to the determination of hypocentre location and source time in earthquake studies. It is an

4 1366 R. Dash and G. Spence Figure 3. Pressure (a) and vertical geophone component (b) of OBS-A recorded from shots along Line 3. The BSR is clearly seen at about 1.18 s at near offsets. The lower two panels show the presence of shear waves on the horizontal components of OBS-A (c) and OBS N (d). inverse problem in which the objective is to find the seafloor location of the OBS and a time correction that minimize the error between the observed and calculated traveltimes of the seismic signal through the water column. Several OBS relocation procedures exist that utilize the traveltimes through the water column to estimate the true OBS location. For this study, the source receiver localization (SRL) scheme of Zykov (2006) was used. It is similar to the regularized inversion method of Hobro (1999), but it also solves for the GPS clock drift. Zykov (2006) argued that the SRL problem is ill-conditioned since OBSs are concentrated in a very small area compared to the shot geometry. The azimuths from far offset shots to the OBSs are concentrated in a narrow sector and this causes instability in the solution. Therefore, the method uses a regularized inversion approach by incorporating aprioriinformation into the solution. The model parameters (such as the shot and receiver positions) were regularized with a prior estimate and uncertainty. Also, smoothness constraints were imposed on the shot positions by assuming small curvature of the shot lines. Direct arrival traveltimes were used for the SRL. Traveltimes through the water column were obtained from all the shots by picking the first zero-crossing. An uncertainty value of 2 ms was assigned to all the picks. The effect of clock drift was then examined by inverting for parameters b and k in a linear drift function, t drift = b+kt, wheret is the shot instant time at the beginning of the survey, b is the drift at the beginning and k is the drift rate. Convergence was achieved after four iterations with a normalized χ 2 value of Traveltime residual values were generally less than 2 ms, which were comparable to the picking error of the direct arrival traveltimes. The average drift in position of the OBSs, in water depths near 1300 m, was about 100 m to the northwest. 4 P -WAVE VELOCITY MODELLING Seismic reflection data provide a very good structural image of the subsurface, but the accuracy of velocities is limited by the source receiver offset, particularly for deeper horizons. Along a vertical-incidence ray path, an increase in both the interface depth and the velocity above the interface will produce the same traveltime in the seismic time section. On the other hand, wide-angle data from OBSs provide accurate velocity information but a poorly resolved structural image (although migration of seafloor multiple arrivals recorded on the OBSs provides a structural image nearly equivalent

5 Figure 4. Examples of OBS (pressure) data showing traveltime picks for the reflection events used in velocity modelling. to the vertical-incidence image; see Dash et al. 2009). Thus, combining both the normal-incidence and the OBS wide-angle data sets in a traveltime inversion can provide improved constraints on both the velocity and the interface depth. 4.1 Event identification The 2005 seismic survey near IODP Site U1327 provided singlechannel reflection data coincident with wide-angle data recorded on anarrayof10closelyspacedobss.reflectioneventsthatoriginated from the same interface in both single-channel and OBS data were identified and picked. The selection of the events was based on their consistency throughout the modelling area. Single-channel seismic data from line 3 (location in Fig. 1) coincident with the line of OBS deployments were used for the velocity analysis. Three distinct events (two events from the upper slope sediments and the BSR) P- and S-wave velocity structure 1367 were identified below the seafloor (Fig. 5). Based on the signalto-noise ratio, the seafloor arrivals were assigned an uncertainty of 2 ms. For shallow events, E1 and E2, pick uncertainties were 5 ms and 6 ms, respectively. For the BSR, a pick uncertainty of 4 ms was assigned. Wide-angle data from all ten OBSs were used in the velocity analysis. The OBS data in the mid-slope region contain only reflections and almost no refraction events, consistent with other OBS surveys in this area (Hobro et al. 2005; Zykov 2006). Hydrophone (pressure) components were chosen for picking. To facilitate the event identification, the OBS and the SCS data were aligned side by side to match the seafloor and the BSR. Then the OBS reflection events at zero offset corresponding to the single-channel events were selected (Fig. 5). Picks were obtained up to 2.5 km on either side of the OBS locations. Beyond this distance, OBS arrivals are difficult to identify due to weak signal and interference with other arrivals. Pick uncertainties, similar to the uncertainties of the single-channel events, were assigned to the OBS arrivals. 4.2 Modelling The velocity model was obtained using the traveltime inversion algorithm of Zelt & Smith (1992), which employs a forward ray tracing step and a damped least-square inversion step to modify the model parameters (velocity and depth) to minimize the difference between the observed and the predicted traveltimes. Since traveltime inversion is a non-linear problem, the algorithm first determines the ray paths in the initial model and then updates the velocity model assuming stationary rays. The initial P-wave velocity model was constructed from the single-channel seismic data. The traveltime picks were converted to depth using the velocity information from previous studies in the area. The initial velocity gradients were 0.50 s 1 for layer 1, 0.80 s 1 for layer 2 and 1.3 s 1 for layer 3. After initial forward modelling, layer velocities and interface depths were adjusted in a linearized inversion scheme that minimized the difference between the observed and calculated traveltimes. All the OBSs were included in the inversion (Fig. 6). A layer stripping approach was followed where the top layer was modelled first and then held fixed. The ray tracing and inversion steps were repeated until the solution converged, that is, the root-mean-square (RMS) traveltime misfit fell to the order of the pick uncertainties and the normalized χ 2 misfit approached 1. For the final model, an RMS traveltime residual of 6 ms was achieved with a normalized χ 2 value of 1.6. The velocity gradients for the final model were 0.50 s 1 for layer 1, 0.88 s 1 for layer 2 and 1.62 s 1 for layer 3. Overall, the perturbation from the starting velocity model was very small. The uncertainties in the velocity model from the traveltime inversion were estimated using the procedure of Katzman et al. (1994). Uncertainty of a particular model parameter was estimated by perturbing its value from that of the final model and holding it fixed while allowing the other parameters to vary during inversion. The parameters were perturbed until the traveltime residual and the normalized χ 2 increased significantly from the value of the final model. In particular, the BSR depth and the velocity above the BSR were examined. The velocity gradient for this layer was fixed at the final value while allowing lateral variations in velocity. From Fig. 7(b), the velocity value for the hydrate layer can probably be accepted within 3 per cent uncertainty. For an average velocity of 1800 m s 1 in the 160-m-thick layer above the BSR, this results in an average velocity error of about ±50 m s 1 andanaverage depth error of less than 5 m.

6 1368 R. Dash and G. Spence Figure 5. Event identification by aligning OBS and single-channel data. The middle panel is single-channel data. Following Zelt & Smith (1992), resolution of the model parameters was estimated by analysing the diagonal elements of the resolution matrix. Ideally, a value of 1 implies perfect resolution and values less than 1 indicate a spatial averaging of the true earth structure by a linear combination of model parameters. According to Zelt & Smith (1992), a resolution value of greater than 0.5 indicates that the model parameter is well resolved. It should be noted that this method of estimating model resolution does not take into account the improvement in constraint due to crossing of ray paths. Fig. 7(c) shows the resolution values of the depth and velocity nodes for the final velocity model. The depth nodes show a better resolution than the velocity nodes. Between 5 km and 6.5 km model distance (below the OBS locations), the depth nodes are very well resolved. 4.3 P-wave velocity model The final P-wave velocity model includes three layers above the BSR (Fig. 8). The model is constrained over a distance range of about 4 km. The model provides higher resolution, especially at shallower depths, compared to the results from previous OBS studies in this area (e.g. Hobro 1999, Hobro et al. 2005; see Fig. 9). The maximum thickness of the uppermost layer is about 40 m. It has very low velocities of km s 1 representing very unconsolidated marine sediments (Fig. 8). The middle layer, also containing slope sediments, has a varying thickness with an average value of about 50 m and has slightly higher velocities ( km s 1 ) resulting from progressive sediment compaction with increasing depth. The bottommost layer has an average thickness of about 160 m. Most of the layer constitutes consolidated sediments of the accretionary prism and has considerably higher velocities varying from 1.62 km s 1 at the top to 1.88 km s 1 at the BSR. The BSR depth is about 230 m. This value, which is well-constrained with an uncertainty of less than 5 m, agrees well with the BSR depth inferred from the VSP data at Site 889B (MacKay et al. 1994) and from the sonic data at Site U1327 (Riedel et al. 2006b; Fig. 9). However, the VSP data at Site U1327 implies that the BSR depth is about 250 m; it seems likely that the disagreement is due to measurement error associated with the VSP data. The boundary between the slope and accreted sediments is not well defined in the seismic data. Based on core and downhole log data from ODP Site 889 and IODP Site U1327, Riedel et al. (2006b) interpreted the boundary at about 90 m below seafloor. However, the single-channel reflection data along line 3 suggest that the boundary is slightly deeper. In Fig. 8, it is interpreted to lie m below the second layer. The layer below the BSR is believed to have a very low velocity due to the presence of free gas, but is unexplored by the P-wave data. 5 S -WAVE VELOCITY MODELLING Although difficult to study because they are all secondary arrivals, S-waves provide information that is complementary to P-waves. S-wave propagation is more dependent on the sediment rigidity provided by grain-to-grain contacts. P-waves are sensitive to the presence of gas within the sediment pore spaces, and so structures beneath the BSR are virtually absent in the single-channel reflection and OBS vertical-component data because of strong attenuation due to the presence of free gas. However, the horizontal components of the OBS data show strong arrivals at greater depths (Figs 3 and 10), which are assumed to be produced by conversion of the downward going P-waves at reflection interfaces. These arrivals can potentially provide information on the structures below the BSR. Also, the S- wave velocity (V s ) contrast between pure methane hydrate (1890 ms 1 ) and hydrate-free shallow marine sediments ( m s 1 ) is very high (Waite et al. 2000). Thus, any increase in V s above the BSR can be attributed to the presence of gas hydrates.

7 P- and S-wave velocity structure 1369 Figure 6. The final ray tracing and traveltime fit for the P-wave velocity model. All 10 OBSs were used simultaneously with the single-channel seismic reflection data. For clarity, only four OBS are shown and the OBS (middle panel) and the single-channel (bottom panel) times are plotted separately. The OBS times are reduced with a velocity of 2.6 km s 1. The small vertical bars (coloured) indicate the picks and their uncertainty range. The final RMS residual was 6 ms with a normalized χ 2 value of Modelling procedure Seven S-wave arrivals were identified on the OBS radial components based on their consistency across different OBSs (Fig. 10). Three PS arrivals (Es1, Es2 and Es4) were identified at about 1.2, 1.38 and 1.7 s at minimum offset in the horizontal component data (Figs 10 and 11). Event Es4 was identified as the PS arrival corresponding to the BSR. A fourth PS arrival (Es3), whose corresponding P-wave arrival was unclear in the vertical component data, was identified at about 1.55 s at near offset. Below the BSR, three more PS arrivals were identified at minimum-offset times of 1.85, 2.0 and 2.2 s. Since the exact arrival times of the PS arrivals were uncertain, they were assigned large picking errors. For the first two events (Es1, Es2), an uncertainty of 20 ms was assigned. Events 3, 4 and 5 (Es3, Es4 and Es5) were assigned a pick uncertainty of 30 ms each, while events 6 and 7 (Es6, Es7) were assigned 40 ms uncertainty. S-wave velocities were determined by forward modelling using the ray-trace algorithm of Zelt & Smith (1992). Interface depths and P-wave velocities were held fixed, and the PS-traveltime residuals were minimized by varying the Poisson s ratio in each layer, beginning at the shallowest layer and progressing downwards. A similar Figure 7. Normalized χ 2 misfit (line) and RMS traveltime residual (dots) with respect to change of BSR depth (a) and velocity above the BSR (b). For velocity, a conservative estimate of uncertainty is ±3 per cent, as indicated by the shaded box in (b). Resolution of depth (circle) and velocity (square) nodes for the final model are shown in (c). Dashed lines represent the model interfaces. Velocity values for selected nodes are given in km s 1. approach to estimate S-wave velocities from multicomponent data has been followed in several other studies, including Shillington et al. (2008); Westbrook et al. (2008); Eccles et al. (2009); Peacock et al. (2010) and Exley et al. (2010). To correlate the S-wave arrivals on the OBS radial component with their corresponding P-wave arrivals on the hydrophone/vertical component, several events were tested for each layer to find the event with the best traveltime fit. A preliminary Poisson s ratio model was obtained from the P- and S-wave sonic log data from IODP Site 1327 using the following relation: ( / ) 2 Vp Vs 2 ν = [ (Vp / ) ] 2, 2 Vs 1 where ν is the Poisson s ratio. For very shallow layers (less than 75 m depth), no drillhole data were available. Thus, Poisson s ra-

8 1370 R. Dash and G. Spence Figure 8. Time-migrated single-channel seismic data plotted with the final velocity model. The red dashed lines are the interfaces (from the slope sediments and the BSR) used in the modelling. The white dashed line indicates the approximate interface between the slope and accreted sediments. Figure 9. The velocity-depth profiles at the drillhole locations extracted from the final velocity model (blue) plotted with the drillhole sonic data (red) and VSP data (thick black). Also shown are the velocity-depth profile from a previous study by Hobro (1999) (green) and the reference velocity profile for 100 per cent water saturation (grey) based on the rock physics model of Dvorkin et al. (1999). A BSR depth of 230 m is inferred from the seismic velocity model, the VSP data at Site 889 and the sonic data at Site U1327; however, the BSR from the VSP data at Site U1327 is anomalously deep.

9 Figure 10. S-wave picks (white dots) from radial components of OBS-C (a) and OBS-I (b) plotted with pick uncertainties (white vertical bars on the left most pick of each arrival). Also plotted are the P-wave picks from the respective vertical geophone components. P- and S-wave velocity structure 1371 tio values of and 0.48 were assigned to the top two layers that correspond to unconsolidated water-saturated sediments. The third layer in the P-wave velocity model, which has a large velocity gradient, was approximated by a number of thin layers. The layer boundary just above the BSR was assumed to produce event Es3. During ray tracing, the depths and Poisson s ratios for the thin layers above the BSR were perturbed while keeping the original P-wave velocity model fixed until an acceptable fit was obtained with the observed arrivals in the horizontal component. For modelling below the BSR, both the layer depths and the Poisson s ratios were perturbed until a satisfactory fit was obtained. PS ray-trace modelling was done for OBS C and I. Only the results from OBS C are presented here (Fig. 11). The combined RMS traveltime residual for OBS C is 17 ms, with a normalized χ 2 of Ray-trace modelling for OBS I using the same model as above produced an equally satisfactory fit. In Fig. 11, we note that there are still some systematic variations in traveltime for individual events, but these are likely due to local heterogeneity and/or anisotropy (Exley et al. 2010). To determine the uncertainty in the S-wave velocity, we followed the same procedure as used in the error estimate for P-wave velocity and layer depths, and similar to the approach used in the S-wave study of Exley et al. (2010). For a particular layer, RMS traveltime residuals and χ 2 values were calculated for a range of Poisson s ratios varying from 95 to 105 per cent of the final model value while keeping V p and the Poisson s ratio for other layers fixed at their final calculated value. Uncertainties were estimated for events Es1 to Es5. From Fig. 12, the shallower events, especially Es1 and Es2, are highly sensitive to change in Poisson s ratio, where both traveltime residual and increase rapidly after 1 per cent change in Poisson s ratio. This implies that the S-wave velocity values are well-resolved for those layers. For deeper layers, the resolution decreases gradually, which may be related to the large uncertainties in the traveltime picks. In general, the overall error in the Poisson s ratio is between 2 and 3 per cent. 5.2 S-wave velocity model As expected, the S-wave velocity model provides a good match with the S-wave drillhole data at Site U1327 (Fig. 13). It increases more or less continuously with depth and reaches a value of about 530 m s 1 at the BSR. Compared to the P-wave velocity, which decreases sharply below the BSR (observed from the drillhole data), there is only a small decrease in S-wave velocity below BSR. In the shallow slope sediments, the V p and V s values match very well with the sediment velocity values predicted by Hamilton (1979), but the deviation from the Hamilton (1979) model is evident in the zone of mbsf. The V p /V s ratio decreases continuously from the seafloor downwards, from a value of 6.5 to about 3.2 at the BSR. The final model from the S-wave ray tracing provides information to about 200 m below the BSR a region not sampled by P-wave arrivals. However, due to large pick uncertainties, the velocity values below the BSR have relatively large errors of up to ± 100 m s 1. A crossplot of V p versus V s can often provide classification information related to rock types. In Fig. 14, the velocity model of Hamilton (1979) and the mudrock line of Castagna et al. (1985) are also shown. The mudrock line is an empirically derived relationship between V p and V s for water-saturated sediments (mostly silicates). However, in gas-charged sediments the mudrock equation underestimates V s due to the decrease in V p. The crossplot of V p versus V s can easily distinguish the hydrate-bearing sediments above the BSR from the gas-charged sediments below. As evident in the figure, the V p V s trends for both the hydrate- and gas-bearing sediments differ significantly from the mudrock line. The hydratebearing layer has a similar trend as the Hamilton line at intermediate depths, but the trend deviates near the base of the layer. A firstorder linear trend line was calculated for the layers above the BSR, producing V p = 1.08 V s , where V p and V s areinms 1.

10 1372 R. Dash and G. Spence Figure 11. PS-wave and PP-wave ray tracing on the radial component of OBS-C showing P (solid) and S (dashed) ray paths. The PS arrivals are mode-converted after reflection from the layer boundaries. The dashed layer interfaces above the BSR were inserted to approximate a gradient in Poisson s ratio. The dashed layer immediately above the BSR was assumed to produce event Es3. Layer boundaries below the BSR are interfaces interpreted from PS waves. 6 HYDRATE CONCENTRATION AND DISTRIBUTION IN PORES 6.1 Concentration estimates based on P-wave velocities Hydrate concentration was estimated from the final P-wave velocity model obtained from the traveltime inversion. Since the presence of hydrate in the sediment pore spaces increases the sediment velocity, the amount of hydrate can be estimated from the velocity difference relative to the background no-hydrate velocity. Numerous models have been proposed for estimating hydrate concentration from velocity data (Chand et al. 2004; Chen 2006; Spence et al. 2010), including the simple effective porosity-reduction model (Hyndman et al. 1993), time-averaging models (Lee et al. 1993) and rockphysics models (Helgerud et al. 1999; Carcione & Tinivella 2000; Jakobsen et al. 2000). Estimates of hydrate concentration using the above methods can be very different from each other. Although some of the variation is due to different assumptions in the methodology, an even greater effect may be the determination of the background or reference velocity, for sediments containing no hydrate and no gas. Several reference velocity-depth profiles exist for this area. Yuan et al. (1996) and Chen (2006) computed reference velocity profiles from normal-move-out analysis of multichannel seismic data. Riedel et al. (2005) calculated the background velocity profile from neutron-porosity logs at ODP Site 889, which they smoothed and converted to velocity using the Hyndman velocity porosity relation and the Lee time-average equation. Chen (2006) also used porosity logs from ODP Site 889 and from several sites of IODP Expedition 311; he converted the smoothed porosity-depth profiles at each site to velocity-depth profiles, using either the Hyndman velocity porosity relation, the Lee time-average equation or the rock-physics approach of Helgerud (2001). In this study, hydrate concentration was estimated with the rockphysics model of Helgerud et al. (1999) based on effective medium theory. This method calculates a reference velocity model for fully water-saturated sediments by following procedures developed by Dvorkin et al. (1999). A source of uncertainty in this method is that lithology must be estimated; following Riedel et al. (2006a) based on results from IODP Expedition 311, we used an average mineralogy of 85 per cent clay and 15 per cent quartz. The elastic parameters of the sediment constituents are summarized in Table 1. In Fig. 15(a), the baseline P-wave velocity model from the rocks physics model compares well with the reference velocity

11 P- and S-wave velocity structure 1373 Figure 12. RMS traveltime residual versus change in Poisson s ratio for different S-wave events. Approximate values of normalized χ 2 are plotted on the right axis. profile from MCS velocity analyses (Chen 2006), although the MCS model is m s 1 lower at all depths. The rock physics velocity profile is also similar to that reported by Hobro et al. (2005), who used a combination of the self-consistent approach and differential effective medium theory (SCA/DEM). Gas hydrate concentration was modelled by assuming hydrate is part of the load-bearing sediment matrix (eqs 4, 7 and 8 from Helgerud et al (1999)). In this gas hydrate in-frame model, both Figure 14. Crossplot of P- and S-wave velocities derived from the raytrace modelling plotted with Castagna et al. (1985) s mudrock line and the Hamilton velocity model. Crosses and dots represent the velocity values above and below the BSR, respectively. P-wave and S-wave velocities increase with hydrate concentration as hydrate adds stiffness to the sediment frame. At a given depth, porosity is taken from a smoothed velocity-depth profile based on downhole log measurements at nearby IODP Site 1027 (Chen 2006; Riedel et al. 2006a), in which porosity is assumed to decrease exponentially with depth with an e-folding depth factor of 1.0 km. Using the elastic moduli for the sediment components (Table 1), we calculate the elastic wave velocities for different values of hydrate Figure 13. Results from S-wave ray-trace modelling (blue lines), (a) V s values and (b) V p /V s ratios. The drillhole sonic data from IODP Site U1327E are shown in red. The black line indicates the reference velocity based on the rock physics model of Dvorkin et al. (1999) Ham. : velocities from Hamilton (1979. Velocity profiles below about 300 mbsf are uncertain and hence are plotted as dotted lines. The thick dashed line represents the approximate depth of the BSR.

12 1374 R. Dash and G. Spence Table 1. Elastic properties of sediment constituents (after Helgerud 2001). Sediment Bulk Shear Density P-wave S-wave constituent Modulus (GPa) Modulus (GPa) (g cm 3 ) Velocity (km s 1 ) Velocity (km s 1 ) Clay Quartz Pore Water Methane Hydrate Methane Gas saturation at each depth. The observed velocity at each depth is then used to interpolate the corresponding concentration value (Fig. 15b). This procedure is the same as that used by Chen (2006) to determine hydrate concentration based on sonic velocity data at the sites of the IODP Expedition 311 transect. At the location of Site U1327, the P-wave velocity model yielded hydrate concentrations that increase linearly from 0 to 26 per cent over the range from about 60 mbsf to the BSR at a depth of 230 mbsf. The average hydrate concentration of about 13 per cent matches well with the estimate from sonic data as calculated by Chen (2006), who obtained an average of about per cent in the interval from 120 to 225 mbsf that corresponds to the section of accreted sediments. Earlier estimates from ODP Site 889 typically yielded higher values. From velocity data, Yuan et al. (1996) obtained saturations of about 20 per cent in the 100-m interval above the BSR. In a downhole resistivity study, Hyndman et al. (1999) estimated per cent pore space hydrate saturation in this same interval. However, recent interpretations of log resistivity data from IODP X311 by Chen et al. (2008) yielded saturations of 11(±7) per cent over this range of gas hydrate occurrence. In the OBS seismic study of Hobro et al. (2005), who used a rock-physics modelling approach, saturations of 3 12 per cent were found in this interval. Similarly, interpretations by Malinverno et al. (2008) of both chlorinity and log resistivity data from Expedition 311 yielded average saturations of 4 10 per cent for the gas hydrate occurrence zone at each of the four sites in the transect, although the thickness of this zone decreased systematically landward of the deformation front. TheP-wave velocity model provides a well-constrained estimate of hydrate concentration beneath a 4-km long profile centred near the IODP Site U1327 drillholes. Velocities increase uniformly with depth over the entire region. That is, the seismic velocity model (Fig. 8) suggests that gas hydrate is widely distributed within the sediment column in the area around the drillhole sites, with averagesaturations of about per cent consistent with estimates from the drillholes. However, no evidence was found from the P-wave modelling for hydrate layers with highly elevated concentrations (>50 per cent) at depths between 120 and 138 mbsf, as detected in Figure 15. P-wave velocity-depth profiles (a) and gas hydrate saturations from P-andS-waves estimated using the effective medium theory of Helgerud et al (1999) (b and c). In (a), the model extracted 1-D P-wave velocity profile is plotted in blue. The green line indicates the background velocity based on the rock physics model of Dvorkin et al. (1999), the red line is the reference velocity estimated with the rock physics approach of Hobro et al. (2005), and the thick black line is the reference velocity from MCS data. The shaded region bounded by the orange dashed lines indicates approximate error in the estimation of P-wave velocity. In (b) and (c), the blue lines are the mean saturations from V p and V s, respectively and the shaded regions are the approximate errors in concentration estimation.

13 Hole U1327A. Thus, the occurrence detected in this hole must be very localized, since it was not detected in the nearby Hole U1327E (at 70 m distance) or in the P-wave OBS models. 6.2 Hydrate distribution in pores from S-wave velocities TheS-wave velocity model, based on the OBS converted phases, is generally consistent with the directly measured downhole S-wave velocities at IODP Site U1327E (Fig. 13a). The OBS velocities, which likely have measurement errors near ±100ms 1, appear to be slightly higher than the average trend of the downhole measurements in the interval mbsf, where the downhole values also vary by about ±100 m s 1. Over this interval, both sets of values are consistently higher than estimates for a reference S-wave velocity profile in Fig. 13(a), a reference profile based on the rock-physics modelling approach of Dvorkin et al. (1999) is very similar to the S-wave velocity-depth profile corresponding to global averages for marine sediments (Hamilton 1979). While the presence of hydrate in the sediment pore space increases the P-wave velocity, its effect on S-wave velocity can be very different depending upon the type of hydrate distribution. If hydrate is present in the sediment pore volume but without grain contact, then the S-wave velocity is not expected to increase. Since we do observe an increase in S-wave velocity relative to a background profile, this suggests that the hydrate adds stiffness to the sediment frame by acting as part of the load-bearing matrix or even as cement between sediment grains (Helgerud et al. 2000). For S- wave velocities that are approximately m s 1 larger than a background velocity profile, effective medium theory can be used to estimate the S-wave velocity for a given hydrate concentration (Helgerud et al. 1999). Hence, where the observed OBS S-wave velocities are larger than the reference velocities, a hydrate saturation profile can be interpolated (Fig. 15c). This shows that the S-wave velocities are consistent with saturations of per cent, although the errors are equivalent to the estimated values. A notable increase in S-wave velocity with gas hydrate saturation was also observed (1) in a PS-converted BSR phase study at Hydrate Ridge (Petersen et al. 2007), where S-wave velocities yielded hydrate concentrations of 16 per cent and (2) in sonic log measurements at the Mallik gas hydrate well (Chand et al. 2004; Chand & Minshull 2004), where gas hydrate saturations based on log resistivities and on rock-physics modelling are very high, averaging 50 per cent with localized increases to 80 per cent. In most other shear wave studies in gas hydrate regions, velocities show little or no change at the depth of the BSR, and little difference above the BSR relative to the best estimate of a reference velocity profile. Such measurements have been made in the Storegga Slide area and off Svalbard (Bünz et al. 2005; Westbrook et al. 2008), at Hydrate Ridge on the Cascadia margin off Oregon (Kumar et al. 2006a,b), and at the Blake Ridge (Guerin et al. 1999; Chand et al. 2004). In these regions the gas hydrate concentrations are thought to be relatively low (Storegga: 6 12 per cent; Hydrate Ridge: maximum 12 per cent; Blake Ridge: 5 10 per cent based on resistivity, per cent based on rock-physics modelling of seismic velocities). For most authors, the interpretation of the lack of correlation between S-wave velocity and the presence of hydrate is that the concentrations are too low to affect the stiffness of the sediment and so hydrate is disseminated within the pore spaces of the sediment. Gas hydrate concentration at the North Cascadia margin may be just high enough to produce a measurable correlation between S-wave velocity and hydrate concentration. P- and S-wave velocity structure 1375 S-waves are also observed from horizons as deep as 200 m beneath the BSR, whereas no equivalent P-waves are seen from these horizons. This is thought to be due to the presence of free gas beneath the BSR, which has a strong effect in attenuating P-waves but little effect in attenuating S-waves. S-wave velocities decrease by only a small amount beneath the BSR, where they return to values near the background velocity profile (Fig. 13a). The elevated S-wave velocities above the BSR are due to hydrate acting as part of the sediment frame, whereas free gas beneath the BSR has little effect on the stiffness of the sediment frame. 7 CONCLUSIONS On the continental slope off Vancouver Island, the P-wave velocity structure associated with marine gas hydrates was estimated by simultaneous traveltime inversion of vertical-incidence and wideangle arrivals recorded on a 2-D profile of 10 closely spaced OCBs. The P-wave velocities increase fairly uniformly from 1.5 km s 1 at the seafloor to 1.88 km s 1 at the BSR, which is well-constrained at a depth of 230(±5) mbsf. The measured velocities agree well with the average trend of the P-wave sonic velocity-depth profile at IODP Site U1327. This suggests that a 25-m-mismatch in depth between the sonic velocities and the Vertical Seismic Profile (VSP) data is most likely due to a problem with the VSP data set. Over the depth range from 120 to 230 mbsf, the OBS velocities are about m s 1 faster than a reference profile determined for sediments containing no hydrate and no gas. The higher velocities are produced by an average gas hydrate saturation of 13 per cent of the pore space, calculated using a rock-physics method for both the reference velocity profile and the conversion to hydrate saturation. The OBS velocity model is fairly uniform laterally beneath a 4- km-long profile passing through the OBSs, which suggests that gas hydrate is widely distributed in the region around the U1327 drillholes. Using P-to-S conversions observed on the OBS horizontal components, an estimate was made for the S-wave velocity structure. Above the BSR, S-wave velocities are about m s 1 higher than the global estimates of V s in poorly consolidated sediments as determined by Hamilton (1979), although the deviation is comparable to the estimated errors. The increase in V s relative to a background velocity implies that the hydrate acts as part of the load-bearing matrix to increase the rigidity of the sediment. But unlike V p, the presence of free gas below the BSR does not produce a significant decrease in V s since S-wave velocities are insensitive to free gas. ACKNOWLEDGMENTS We thank Bob Iulucci, Chris LeBlanc and Keith Louden of Dalhousie University for the use and operation of the OCBs, and Bob Macdonald and Ivan Frydecky for their technical expertise in ship operations. This work was supported by grants from the Natural Science and Engineering Research Council, and by technical support and ship time from the Geological Survey of Canada and the Department of Fisheries and Oceans. REFERENCES Bünz, S., Mienert, J., Vanneste, M. & Andreassen, K., Gas hydrates at the Storegga Slide: constraints from an analysis of multicomponent, wide-angle seismic data, Geophysics, 70(5), B19 B34.

14 1376 R. Dash and G. Spence Carcione, J.M. & Tinivella, U., Bottom-simulating reflectors: seismic velocities and AVO effects, Geophysics, 65(1), Castagna, J.P., Batzle, M.L. & Eastwood, R.L., Relationships between compressional-wave and shear-wave velocities in clastic silicate rocks, Geophysics, 50(4), Chand, S. & Minshull, T.A., The effect of hydrate content on seismic attenuation: case study for Mallik 2L-38 well data, Mackenzie delta, Canada, Geophys. Res. Lett., 31(14), L14609, doi: /2004gl Chand, S., Minshull, T.A., Gei, D. & Carcione, J., Elastic velocity models for gas-hydrate-bearing sediments a comparison, Geophys. J. Int., 159, Chen, M., Northern Cascadia marine gas hydrate: constraints from resistivity, velocity, and AVO, Master s thesis, University of Victoria, Victoria, BC. Chen, M.-A., Riedel, M., Hyndman, R. & Dosso, S., AVO inversion of BSRs in marine gas hydrate studies Geophysics, 73(2), C31 C43. Chen, M.-A. P., Riedel, M., Spence, G. & Hyndman, R., A downhole electrical resistivity study of northern Cascadia marine gas hydrate, Technical report, Proc. Integr. Ocean Drill. Program, 311, doi: /iodp.proc Dash, R., Spence, G., Hyndman, R., Grion, S., Wang Y. & Ronen, S., Wide-area imaging from OBS multiples, Geophysics, 74(6), Q41 Q47. Davis, E.E. & Hyndman, R.D., Accretion and recent deformation of sediments along the northern Cascadia subduction zone, Bull. geol. Soc. Am., 101, Dvorkin, J., Prasad, M., Sakai, A. & Lavoie, D., Elasticity of marine sediments; rock physics modeling, Geophy. Res. Lett., 26, Eccles, J.D., White, R.S. & Christie, P.A.F., Identification and inversion of converted shear waves: case studies from the European North Atlantic continental margins. Geophys. J. Int., 179, , doi: /j x x. Exley, R.J.K., Westbrook, G.K., Haacke, R.R. & Peacock, S., Detection of seismic anisotropy using ocean bottom seismometers: a case study from the northern headwall of the Storegga Slide. Geophys. J. Int., 183, , doi: /j x x. Guerin, G., Goldberg, D., & Meltser, A., Characterization of in situ elastic properties of gas hydrate-bearing sediments on the Blake Ridge, J. geophys. Res., 104, Hamilton, E.L., Vp/Vs and Poisson s ratios in marine sediments and rocks, J. acoust. Soc. Am., 66(4), Helgerud, M.B., Wave speeds in gas hydrate and sediments containing gas hydrate: a laboratory and modeling study, PhD thesis, Stanford University, CA. Helgerud, M.B., Dvorkin, J., Nur, A., Sakai, A. & Collett, T., Elastic-wave velocity in marine sediments with gas hydrates: effective medium modeling, Geophys. Res. Lett., 26(13), , doi: /1999gl Helgerud, M.B., Dvorkin, J. & Nur, A., Rock physics characterization for gas hydrate reservoirs. Ann. New York Acad. Sci., 912, Hobro, J.W.D., Three-dimensional tomographic inversion of combined reflection and refraction seismic travel-time data, PhD thesis, University of Cambridge. Hobro, J.W.D., Minshull, T.A., Singh, S.C. & Chand, S., A threedimensional seismic tomographic study of the gas hydrate stability zone, offshore Vancouver Island, J. geophys. Res., 110(B09102), doi: /2004jb Hyndman, R.D. & Spence, G.D., A seismic study of methane hydrate marine bottom simulating reflectors, J. geophys. Res., 97(B5), Hyndman, R.D., Moore, G.F. & Moran, K., Velocity, porosity, and pore-fluid loss from the Nankai subduction zone accretionary prism, Proc. Ocean Drill. Program, Sci. Results, 131, Hyndman, R.D., Moran, K. & Yuan, T., The concentration of deep sea gas hydrates from downhole resistivity logs and laboratory data, Earth planet. Sci. Lett., 172(1 2), , doi: /s x(99) Jakobsen, M., Hudson, J.A., Minshull, T.A. & Singh, S.C., Elastic properties of hydrate-bearing sediments using effective medium theory, J. geophys. Res., 105(B1), Katzman, R., Holbrook, W.S. & Paull, C.K., Combined verticalincidence and wide-angle seismic study of a gas hydrate zone, Blake Ridge, J. geophys. Res., 99(B9), , doi: / 94JB Kumar, D., Sen, M.K. & Bangs, N.L., 2006a. Seismic characteristics of gas hydrates at Hydrate Ridge, offshore Oregon, Leading Edge, 25, Kumar, D., Sen, M.K., Bangs, N.L., Wang, C. & Pecher, I., 2006b. Seismic anisotropy at Hydrate Ridge, Geophys. Res. Lett., 33, L01306, doi: /2005gl Lee, M.W., Hutchinson, D.R., Dillon, W.P., Miller, J.J., Agena, W.F. & Swift, B.A., Method of estimating the amount of in situ gas hydrates in deep marine sediments, Mar. Pet. Geol., 10, MacKay, M.E., Jarrard, R.D., Westbrook, G.K. & Hyndman, R.D.; the Shipboard Party of ODP Leg 146, Origin of bottom-simulating reflectors: geophysical evidence from the Cascadia accretionary wedge, Geology, 22, Malinverno, A., Kastner, M., Torres, M.E. & Wortmann, U.G., Gas hydrate occurrence from pore water chlorinity and downhole logs in a transect across the northern Cascadia margin (Integrated Ocean Drilling Program Expedition 311), J. geophys. Res., 113, B08103, doi: /2008jb Peacock, S., Westbrook, G.K. & Bais, G., S-wave velocities and anisotropy in sediments entering the Nankai subduction zone, offshore Japan. Geophys. J. Int., 180, doi: /j x x. Petersen, C.J., Papenberg, C. & Klaeschen, D., Local seismic quantification of gas hydrates and BSR characterization from multi-frequency OBS data at northern Hydrate Ridge, Earth planet. Sci. Lett., 255, Riedel, M., Collett, T.S. & Hyndman, R.D., Gas hydrate concentration estimates from chlorinity, electrical resistivity, and seismic velocity, Geol. Surv. Can. Open-File Rep., Riedel, M., Collett, T.S., Malone, M.J. & the Expedition 311 Scientists, 2006a. Proceedings of the Integrated Ocean Drilling Program, 311, doi: /iodp.proc Riedel, M., Novosel, I., Spence, G.D., Hyndman, R.D., Chapman, N.R., Solem, R.C. & Lewis, T., 2006b. Geophysical and geochemical signatures associated with gas hydrate related venting in the northern Cascadia margin, Bull. geol. Soc. Am., 118(1 2), Schwalenberg K., Willoughby E.C., Mir R. & Edwards R.N., Marine gas hydrate signatures in Cascadia and their correlation with seismic blank zones, First Break, 23, Shillington, D.J., Minshull, T.A., Peirce, C. & O Sullivan, J.M., P- and S-wave velocities of consolidated sediments from a seafloor seismic survey in the North Celtic Sea Basin, offshore Ireland. Geophys. Prospect., 56, , doi: /j x. Singh, S.C., Minshull, T.A. & Spence, G.D., Velocity structure of a gas hydrate reflector, Science, 260, Spence, G.D., Haacke, R.R. & Hyndman, R.D., Seismic indicators of gas hydrate and underlying free gas, in Geophysical Characterization of Gas Hydrates, eds Riedel, M., Willoughby, E.C. & Chopra, S., Soc. Exploration Geophysicists, Tulsa, OK. Tréhu, A.M. & Flueh, E.R., Estimating the thickness of the free gas zone beneath Hydrate Ridge, Oregon continental margin, from seismic velocities and attenuation, J. geophys. Res., 106, , doi: /2000jb Waite, W.F., Helgerud, M.B., Nur, A., Pinkston, J., Stern, L.A., Kirby, S.H. & Durham, W.B., First laboratory measurements of compressional and shear wave speeds through pure methane hydrate at in situ conditions, Gas Hydrates: Challenges for the Future, Ann. N.Y. Acad. Sci.Vol. 912, pp , eds Holder G.D. & Bishnoi P.R. Blackwell Publishing, Malden, MA.