Acoustic imaging of gas hydrate and free gas at the Storegga Slide

Transcription

1 JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 109,, doi: /2003jb002863, 2004 Acoustic imaging of gas hydrate and free gas at the Storegga Slide Stefan Bünz and Jürgen Mienert Department of Geology, University of Tromsø, Tromsø, Norway Received 22 October 2003; revised 2 February 2004; accepted 12 February 2004; published 10 April [1] Analysis of P wave velocities of an ocean-bottom cable (OBC) data set demonstrates the existence of gas hydrates and free gas on the northern flank of the Storegga Slide. The distribution and concentration of gas hydrates and free gas show significant variation over the length of the OBC. Average gas hydrate saturation of pore space is 5%, when modeled by effective-medium theory with hydrates as a sediment-frame component. Average gas saturation is 0.45% assuming homogeneous distribution. The bottomsimulating reflector (BSR) is generally identified as termination of enhanced reflection but in some places appears as a reflection proper. Amplitude anomalies related to the BSR are primarily caused by the gas and not the hydrates. The gas hydrates at the Storegga Slide develop from gas-rich fluids migrating into the gas-hydrate stability zone (GHSZ) from the sediments below. Development of polygonal faults and the related expulsion of formation water might drive the fluid flow in the area. The physical and geological properties of the sediments control the distribution and concentration of gas, which migrates predominantly along strata rather than along the base of the GHSZ. This migration mechanism controls the flux of gas-rich fluids into the GHSZ and hence the distribution of gas hydrates and the variation in gas hydrate concentration. INDEX TERMS: 0935 Exploration Geophysics: Seismic methods (3025); 3022 Marine Geology and Geophysics: Marine sediments processes and transport; 3025 Marine Geology and Geophysics: Marine seismics (0935); 3094 Marine Geology and Geophysics: Instruments and techniques; KEYWORDS: gas hydrates, free gas, Storegga Slide Citation: Bünz, S., and J. Mienert (2004), Acoustic imaging of gas hydrate and free gas at the Storegga Slide, J. Geophys. Res., 109,, doi: /2003jb Introduction [2] Gas hydrates and free gas occur worldwide in sediments of continental margins. They react on a dynamic interplay between ocean temperature, sea level and climate and might play a key role in affecting the fluid flow activity and stability of continental slopes. Furthermore, if methane escapes from the hydrate/free gas system through the ocean into the atmosphere it may have an impact on global climate [Kvenvolden, 1993; Dickens, 1999; Paulletal., 2000a; Kennett et al., 2003]. Gas hydrate evidence on the mid- Norwegian margin exists in an area of 4000 km 2 on the northern flank of the km 2 -large Storegga Slide [Bünz et al., 2003]. A link between hydrates and the sliding has been suggested [Mienert et al., 1998]. Theoretical modeling suggests that the gas hydrate stability zone (GHSZ) is highly sensitive to bottom water temperature changes [Mienert et al., 2001; Vogt and Jung, 2002] and pressure changes due to mass movements [Berndt et al., 2002]. Numerous fluid-escape features and widespread polygonal faults underneath the hydrate system provide evidence for an active fluid flow system on the mid-norwegian margin [Berndt et al., 2003]. Here, we focus on the understanding of the distribution and dynamics of hydrate and free gas at the northern flank of the Storegga Slide scar, a key area for an assessment of slope stability from the past to the future. Copyright 2004 by the American Geophysical Union /04/2003JB002863$09.00 [3] Gas hydrates are an ice-like crystalline solid in which cages of hydrogen-bonded water molecules entrap individual gas molecules [Sloan, 1998]. They require specific pressure-temperature conditions and an adequate supply of methane and water, conditions commonly met on continental margins in water depths exceeding 500 m. Existence of gas hydrates is frequently inferred from the observation of bottom-simulating reflectors (BSR) in reflection seismic data. The BSR generally corresponds to the base of the gas hydrate stability zone (BGHS) and is the result of an acoustic impedance contrast between hydrate-bearing sediments and free gas trapped in the sediments underneath gas hydrates [Holbrook et al., 1996; Pecher et al., 1996]. As a result of its pressure-temperature dependence the BSR often mimics the seafloor thereby crosscutting stratigraphic horizons. Due to the strong negative seismic impedance contrast, the BSR often shows enhanced seismic amplitudes and a phase reversal relative to the seafloor reflection. Gas hydrates can exist without a BSR, because this reflection is primarily caused by free gas [Holbrook, 2000]. Only very high hydrate concentrations produce perceptible amplitude anomalies in seismic data if they are not accompanied by free gas [Hornbach et al., 2003]. [4] Detailed information on the distribution of gas hydrates and gas within the sediments can be found through analyses of seismic velocities, obtained from well-log, vertical seismic profiling, multichannel or ocean-bottom seismic data [Singh et al., 1993; Katzman et al., 1994; 1of15

2 Holbrook et al., 1996; Korenaga et al., 1997; Ecker et al., 2000; Hornbach et al., 2003]. Hydrate s high intrinsic P wave velocity ( km/s) [Sloan, 1998] can raise the P wave velocity in hydrate-bearing sediments relative to hydrate-free sediments. Conversely, the presence of gas lowers the measured P wave velocity in the sediments relative to gas-free sediments [Domenico, 1977]. A number of models for hydrate- and gas-bearing sediments relate seismic velocity changes to estimates on hydrate and gas concentration [Lee et al., 1996; Helgerud et al., 1999; Jakobsen et al., 2000; Gei and Carcione, 2003]. [5] The objective of this paper is to image the distribution of gas hydrates and gas on the northern flank of the Storegga Slide. More specifically we aim (1) to estimate their concentration, (2) to investigate hydrate formation mechanisms and (3) to study the dynamics of the free gas. We use an ocean bottom cable (OBC) data set from an area on the northern flank of the Storegga Slide, where a strong BSR has been identified [Bünz et al., 2003]. An earlier interpretation of the OBC data has been given by Berteussen et al. [1999] and Andreassen et al. [2003]. Here we conduct a detailed analysis of P wave velocities on the vertical component of the OBC line while the inline component represents an image obtained from P to S converted (PS) waves. The PS-wave data is useful to investigate the micro-scale distribution of gas hydrates and gas. Seismic velocity changes are subsequently converted into gas hydrate and free gas concentrations. The resulting image provides new and important information on the interaction of gas hydrate and free gas accumulation with stratigraphy and subsurface structure. Moreover, a geotechnical borehole provides important constraints on hydrate and gas concentration estimates. 2. Geologic Setting [6] A BSR as an indicator for the presence of gas hydrates on the mid-norwegian margin has been mapped at the transition from the Vøring to the Møre basins (Figure 1) [Bünz et al., 2003]. Those basins developed as a result of several rifting episodes leading to Late Paleocene/Early Eocene continental break-up and subsequent thermal subsidence [Skogseid and Eldholm, 1989; Brekke, 2000]. The basin fill in some places is up to 10 km thick [Brekke, 2000]. A number of contractional structures of assumed Late Eocene and Mid Miocene age occur on the mid-norwegian margin and are known to be potential hydrocarbon reservoirs, e.g., the Ormen Lange gas reservoir [Brekke and Riis, 1987; Doré and Lundin, 1996; Vågnes et al., 1998; Bryn et al., 1998]. [7] The two sedimentary successions important to the gas hydrate and fluid flow systems are the Miocene/earliest Pliocene Kai Formation with predominantly fine-grained hemipelagic sediments [Dalland et al., 1988; Blystad et al., 1995; Rokoengen et al., 1995] and the overlying Naust formation that encompasses a thick wedge of clastic sediments of the Plio-Pleistocene glacial-interglacial cycles [Stuevold and Eldholm, 1996; Hjelstuen et al., 1999]. The Naust formation is characterized by two types of deposits. First, contourite drift sediments that were deposited by S-Ndirected contour currents maintaining a high sedimentation rate on the margin [Laberg et al., 2001]. Second, glacigenic debris flows that were released at the shelf break during glacial periods, transported large amounts of sediments downslope [Vorren et al., 1998; Ottesen et al., 2001]. A giant retrogressive submarine slope failure, the Storegga Slide is the last of a series of slide events coupled to Pleistocene climatic fluctuations [Bryn et al., 2003]. This 8.2 ka B.P. multiphase Storegga Slide covers large parts of the Møre and the southern Vøring basins [Haflidason et al., 2002]. The eastern headwall runs in N-S direction and reaches a height of up to 300 m. The northern sidewall runs along the border of the Møre and the Vøring basin, and passes through the area where a gas-hydrate related BSR occurs [Bünz et al., 2003]. Mienert et al. [1998] suggested that gas hydrate dissociation contributed to slope instability in the area. [8] The BSR occurs outside and inside the slide area, runs parallel to the seafloor and lies at its theoretically predicted depth [Bouriak et al., 2000; Mienert et al., 2001; Bünz et al., 2003]. The BSR can be identified as the envelope of enhanced reflection terminations and exhibits only in few places as a reflection proper with reversed phase, when compared to the seafloor reflection [Bünz et al., 2003]. The regional distribution of the BSR is limited by the shoaling of the BGHS toward the continental shelf and on the slope by sediments that are less conducive for gas hydrate growth [Bünz et al., 2003]. The BSR only occurs within the contouritic and hemipelagic deposits of the Naust formation. The northern flank of the Storegga Slide is the only place on the margin where these types of sediments have been built up continuously. [9] Gas hydrates in the pore space of sediments focus the fluid flow upslope [Bouriak et al., 2000; Bünz et al., 2003] and trap fluids at the top of the slope between the Møre and Vøring basins. Fluids are then released possibly due to overpressure. Bouriak et al. [2000] and Bünz et al. [2003] suggested that the Storegga gas hydrates develop primarily from fluids that originate from deeper sedimentary strata distinctly beneath the GHSZ. Bünz et al. [2003] hypothesized that some fluids originate from the reservoir within the contractional dome structures. A polygonal fault system within the Kai formation sediments (Figure 1) [Berndt et al., 2003] might be another major source of fluids, because the development of polygonal faults is related to sediment contraction and pore fluid expulsion [Cartwright and Lonergan, 1996]. The processes leading to contraction and water expulsion are still debated. Possible processes involved in their development include syneresis of colloidal sediments and Rayleigh-Taylor instabilities due to density inversions. Henriet et al. [1991] speculated that polygonal faults can control fluid flow on a regional scale. A belt of fluid-escape features (Figure 1) is associated with the gas hydrates and polygonal fault systems on the mid-norwegian margin. Berndt et al. [2003] reported that some of the features originate at the BGHS, others at the top of the polygonal fault system. They suggested that fluid expulsion is related to the development of polygonal faults, which is an ongoing process since the Early Miocene. 3. Material and Methods 3.1. Data Seismic Data [10] Multichannel seismic line NH (Figures 1 and 2) was provided by Norsk Hydro, Oslo. This data set 2of15

3 Figure 1. Distribution of the BSR and pockmarks/pipes on the mid-norwegian margin after Bünz et al. [2003]. Extent of polygonal faults after Berndt et al. [2003]. The location of the OBC line lies within the BSR area on the northern flank of the Storegga Slide. was recorded using a 140 in 3 sleeve gun array and a 1200-m-long 96-channel streamer. Adjacent traces were summed, giving a total of 48 channels per shot. The resulting CDP spacing is 12.5 m. The data contains frequencies of up to 300 Hz with a main frequency between Hz. The line runs approximately in SW-NE direction across the northern sidewall of the Storegga Slide between water depths of about m. [11] The OBC seismic data (Figures 1 3) was acquired in a two-ship operation by PGS, Oslo. One ship was shooting, and another ship was draping the OBC on the ocean floor. The acquisition of shear waves in the marine environment using a surface-towed source assumes that waves propagate downward as P waves, convert upon reflection at a lithologic boundary whenever the angle of incidence is not zero and propagate upward as an S-wave, which then can be recorded by horizontal geophones at the seafloor. These wave types are called converted waves or PS-waves. It is generally assumed that only these first-order converted waves contain enough energy to be successfully recorded [Stewart et al., 2002]. The PS-waves are recorded on the inline component of the OBC, whereas the vertical component of the OBC records reflected P waves. [12] The OBC was deployed five times for a total length of 4 km. While the cable was lying at the seafloor, data were recorded with offsets up to 3500 m from both sides of the cable in a split-spread bi-directional design. The cable has altogether 128 channels, 32 for each component: hydrophone, vertical geophone, horizontal inline geophone and horizontal crossline geophone. The spacing between receivers of the same type is 25 m. The shot spacing was also 25 m, giving a CMP spacing of 12.5 m. The source was towed at a depth of 6 m and had a volume of 3180 in 3. The data contains frequencies up to 200 Hz with a main frequency between Hz. The location of the OBC lies at the central part of line NH (Figures 1 and 2) in water depths between m Geotechnical Data [13] The geotechnical data (Figure 4a) were provided by Norsk Hydro, Oslo, who initiated the drilling to investigate soil conditions and gas hydrates at the northern flank of the Storegga Slide. The geotechnical borehole is situated in the center of the OBC line (Figures 1 3) and has a total penetration of 310 m. It is located on the slope between the Møre and the Vøring basins. Although it penetrated into sediments well below the BSR, no hydrates could be recovered due to technical limitations or the nonexistence of hydrates at this location [Mienert and Bryn, 1997]. The geological data analyses were mainly conducted by the Norwegian Geotechnical Institute (NGI) and published in a report [NGI, 1997]. The porosity has been determined from the total unit weight and the unit weight of the solid 3of15

4 Figure 2. Time-migrated seismic section from the northern flank of the Storegga Slide. The BSR on the mid-norwegian margin is mainly identified as the termination of enhanced reflections. A polygonal fault system occurs in the sediments of the Kai formation. The base of the Naust formation is characterized by high reflection amplitudes. Pipes can be identified on the upper part of the slope between the Vøring and the Møre basins. Underneath them is a seismically transparent zone. The dashed box marks the location of the OBC line; the rectangle marks the geotechnical borehole. fraction. The unit weights have been determined from standard density measurements multiplied by the standard acceleration of gravity. Uncertainties are mostly due to the fact that the solid unit weight was only determined at eleven depths (indicated by crosses in Figure 4a). We interpolated in between those samples. However, the given water content is very closely related to porosity and shows a very similar trend with depth as the porosity curve. We therefore assume that our calculated porosity values represent a reasonable floating average for the sediments at the northern flank of the Storegga Slide. Five different mineral constituents were analyzed using X-ray diffractometry: Mica, clay, quartz, feldspar and calcite. The clay fraction encompasses the minerals smectite, chlorite and kaolinite. In addition the content of organic carbon is shown (Figure 4a). The presented data provides important constraints on the modeling of hydrate and gas concentrations Data Processing and Velocity Analysis [14] The multichannel seismic line (Figure 2) was processed including deconvolution, bandpass filtering, FX-filtering and wave equation migration. The seismic processing of the vertical component of the OBC line included prestack bandpass filter of Hz, true amplitude recovery, predictive deconvolution. Semblance velocity analysis, normal moveout (NMO) correction, stacking, post-stack scaling and filtering was applied to small offsets of ±250 m, because this gave the best resolution for shallow depths [Andreassen et al., 2003]. The seismic image of the vertical component (Figure 3) shows the result of our processing sequence. [15] The geologic structure at the northern flank of the Storegga Slide shows an almost horizontal stratigraphic layering (Figure 3), only the seafloor and the horizons immediately underneath show a slight dip to the southwest that reaches a maximum of 1.2. The horizontal layering allows an analysis of stacking velocities on the prestack data to obtain root-mean square (RMS) velocities, and convert RMS velocities into physical interval velocities using Dix s [1955] equation. [16] In order to properly analyze velocities, the shots and receivers have to be at the same datum. Therefore we edited the traces for spike and noise burst and applied prestack wave equation datuming to the data. In a first run, we 4of15

5 Figure 3. Stacked sections of vertical and inline component of the OBC line, which has a length of approximately 4 km. Note the difference in time axis. The inline component has been linearly stretched in time to match the P wave data based on horizon number 4, which occurs just below a sedimentary layer with chaotic internal texture. The rectangle marks the location of the geotechnical borehole. The center part of the vertical component shows the BSR as a reflection proper. The inline component does not show any amplitude anomalies at the level of the BSR. Because the inline component results from the acquisition of shear waves and is therefore mainly affected by the shear properties of the sediments, the amplitude anomalies on the vertical component must be related to the pore fill of the sediments. datumed receivers, and in a second run we datumed shots to an elevation of 900 m below sea level. Afterward, we applied residual statics, F-K filtering and predictive deconvolution. The velocity analysis included dip-moveout (DMO) correction and common-offset prestack time migration and followed the most popular sequence given by Yilmaz [2001, pp ]. Velocities were picked at sparse intervals and NMO correction was applied using this velocity. Data were sorted into common-offset gathers, DMO corrected and prestack time migrated using an F-K migration operator [Stolt, 1978]. Subsequently, we sorted the data back to common midpoint gathers and applied inverse NMO correction using the sparsely picked velocity field. The velocity was then analyzed at frequent locations of every 5 10 CMP stations in supergathers of 3 9 CMPs. The final stacking velocity field was converted into an interval velocity field using Dix s equation. Finally, we smoothed the interval velocity field to intervals of every second CMP and every 4 ms using smoothing operators of 10 CMPs and 8 ms, respectively. [17] The uncertainty of the interval velocities is directly related to picking errors of the stacking velocities. A maximum picking error of ±5 m/s is estimated on the basis of the width of the semblance peak. This error in RMS velocity can cause an error of m/s in the interval velocity. As it is the case with stacking velocities, there is a trade-off between adjacent layers separated by a horizon that is defined by the velocity pick. If velocity increases in one of the two layers, then velocity decreases in the other. In case of the reflection at the BGHS, it can either enhance velocity anomalies in the hydrate- and gas-bearing zone or it can suppress them. However, because velocities were determined frequently along the line and picking errors have been canceled out by the smoothing operation, we anticipate an error in the interval velocity of well below 50 m/s. Determined velocities might be slightly higher than they normally would be due to an anisotropy effect inherent to wide-angle seismic data in such environments [Holbrook, 2000]. [18] The processing of the inline component is discussed by Andreassen et al. [2003] and included asymptotic conversion point binning, bandpass filtering Hz, prestack wave equation datuming, true amplitude recovery, F-K filtering, predictive deconvolution, NMO correction, time-variant scaling, converted wave stacking and poststack scaling and filtering. [19] Complex trace attributes [Taner et al., 1979] may offer additional information to geologic interpretation, es- 5of15

6 Figure 4. (a) Data from the geotechnical borehole. The porosity has been determined from total unit weight and solid unit weight. The solid unit weight has been analyzed only at certain depths marked by the crosses. (b) Stratigraphic units at the northern flank of the Storegga Slide. The small indicated horizons separate individual layers for the rock physical modeling. U, M, and L denote upper, middle, and lower. pecially in hydrocarbon settings [Taner and Sheriff, 1977]. Thus those attributes may also offer insights in gas hydrate settings, where free gas exists underneath the hydrates. We used the hydrophone component of the OBC data to determine complex trace attributes, because this component is independent of the direction of the displacement of the wave field. The complex-trace attributes envelope of the amplitude (reflection strength) and apparent polarity 6of15

7 Figure 5. Reflection coefficients at the seafloor and the BSR of the OBC line determined using the multiple method of Warner [1990]. were calculated from a near-zero offset (±250 m) stack. Therefore the prestack data were only corrected for spherical divergence and inelastic attenuation using a t 2 -scaling function, as well as bandpass filtering Hz for the apparent polarity only. 4. Results 4.1. Seismic Observations [20] Four stratigraphic units of the Naust formation can be identified in the seismic data using the stratigraphy given by Evans et al. [2002]. Those units are separated by major reflections, marked with numbers 1 4 in Figures 3 and 4b. Naust units A C encompass approximately 600 ms, corresponding to about 500 m of the sediments of the Naust formation. Naust unit A C sediments are significantly younger (<0.7 Ma B.P.) in comparison to sediments of Naust unit E ( Ma B.P.), which only is ms thick. Naust unit E lies just above the Kai formation, in which polygonal faults have been identified [Berndt et al., 2003]. [21] Naust units B and C are crosscut by a reflection that mimics the seafloor in a depth of approximately 350 ms below seafloor (Figures 2 4). This BSR occurs at the theoretically predicted depth of the GHSZ [Mienert et al., 2001]. Generally it can be recognized as the upper termination of enhanced reflections. Only over a small distance of about m downslope from the geotechnical borehole, the BSR shows as a reflection proper (Figure 3). Whether the BSR is a true reflection in its own right on the seismic data is mainly the result of the frequency bandwidth of the acquisition system [Wood et al., 2002]. In this paper, we use the term BSR proper in cases in which the BSR is a true reflection. The amplitude of the enhanced reflections varies slightly as does the length over which they continue underneath the BSR. Further upslope, these reflections vanish. Here, the multichannel seismic line (Figure 2) shows a seismically transparent zone underneath a layer of high reflection amplitudes. Only the horizon that separates the Naust and Kai formations is clearly visible (for comparison, see also Bünz et al. [2003, Figure 4a]). [22] Using the multiple method of Warner [1990] on the zero-offset trace of each CMP of the hydrophone component of the OBC, we obtain a mean seafloor reflection coefficient of 0.29 ± 0.04 (Figure 5). This is in good agreement with seafloor reflection coefficients determined from piston core data [Berndt et al., 2002]. The BSR reflection coefficient shows strong variation between almost 0to 0.11 that correlate with the termination of enhanced reflections (Figure 5). In the area, where the BSR is identified as a coherent event (labeled A in Figure 6a), the BSR reflection coefficient is 0.06 ± Further downslope it is significantly lower ( 0.015), but here no clear BSR event could be identified and reflection coefficients are calculated at the predicted position of the BSR. [23] The fact that the BSR is primarily evident as a termination of enhanced reflections impairs the investigation of BSR reflection coefficients along the whole OBC line. Therefore the reflection strength (Figure 6c) may provide more insight into amplitudes at BSR depth. High reflection strength correlates with enhanced reflections beneath the BSR. The small zone of a BSR proper (labeled A in Figure 6a) on the vertical component (Figure 3) is not a continuous event of high reflection strength. This zone is instead characterized by individual spots of high reflection strength that align along the BSR. On the apparent polarity however (Figure 6d), this zone shows up as a clear bottomsimulating event with reversed polarity. Also, enhanced reflections further upslope underneath the BSR have negative apparent polarity, whereas the enhanced reflections further downslope shows positive polarity at the depth, in which event 3 crosscuts the BSR. [24] The seismic data (Figure 2) show vertical narrow zones of acoustic wipe-out. These features have been interpreted as pipes and are suggested to be the result of fluid expulsion related to a polygonal fault system [Berndt et al., 2003]. Pockmarks are found on the seafloor above some of the pipes. Most of the fluid-escape features occur on the top of the slope between the Møre and Vøring basins above the layer of high reflection amplitudes. [25] Another high amplitude reflection occurs at a subbottom depth of approximately 600 ms at the top of Naust unit E. Medium to high reflection strengths are associated with the unit and the apparent polarity is negative in some places (Figures 6c and 6d). Unit E overlies the Kai formation, where polygonal faults are widespread. [26] All major stratigraphic horizons can be identified on the inline component of the OBC line (Figure 3) on the basis of their lateral continuity. S-wave reflections mainly result from stratigraphic boundaries, where an S-wave impedance contrast exists, and thus the mode-converted waves yield direct and more accurate information about the shear properties of the subsurface. Due to the fact that the S-wave velocity is much lower than the P wave velocity, the seismic resolution increases and the inline component of the OBC provides a much more detailed image of the subsurface. A BSR or high reflection amplitudes associated with the BGHS on the vertical component are not observed on the inline component. A strong reflection originates from the top of an unconformity (Figure 3). This unconformity is hardly visible on the vertical component, where it is situated directly underneath the enhanced reflections Seismic Velocities [27] The seismic velocity analysis resulted in the P wave interval velocities shown in Figure 6a, where it is overlain 7of15

8 Figure 6. 8of15

9 with a wiggle trace plot of the vertical component. The velocity increases from about 1500 m/s at the seafloor to m/s just above the BSR. The BSR can be clearly identified by a velocity inversion, separating higher velocities above from lower velocities beneath. Underneath the BSR, the velocities are as low as 1280 m/s. The velocities increase slowly to 1.1 s, then increase significantly to about 2000 m/s in a seismically chaotic layer that is related to an unconformity (Figure 3). Below that layer is a second low-velocity zone with velocities between m/s. This zone corresponds to Naust unit E, which shows strong reflection amplitudes on the seismic data (Figure 3) and occurs just above a polygonal fault system. Here, high reflection strength and negative apparent polarity coincide with zones of low P wave velocity (Figures 6a, 6c, and 6d). [28] A considerable increase in velocity just above the BSR is distinguishable. This velocity increase follows the BSR and thereby crosses lithologic horizons. Velocity measurements at a depth of about ms within base Naust B vary between 1650 m/s 1850 m/s. In the lowvelocity zone underneath the BSR, velocity variations within stratigraphic units are even higher ( m/s). [29] The velocity above the BSR is quite variable. The highest velocities occur in the northern half of the line, with a maximum (1868 m/s) at a location where two strong enhanced reflections terminate at BSR depth (labeled B in Figure 6a). The velocities at the southern end are significantly lower ( m/s). Velocity variations underneath the BSR are also observable. Lowest velocities (1280 m/s) occur at about the center of a line, just underneath a reflection that has been identified as a proper BSR (labeled A in Figure 6a). The location of low P wave velocities are generally in good agreement with high reflection amplitudes observed on the seismic data. For example, the two enhanced reflections at the northern end (labeled B in Figure 6a) coincide with distinctly lower P wave velocities ( m/s). In the small zone between those two reflections and the BSR, velocities are moderately low at about 1550 m/s. The thickness of the low-velocity zone underneath the BSR also seems to change slightly between ms, corresponding to approximately m. The thickness of the zone of increased velocities above the BSR is difficult to determine with accuracy but it is in the range from approximately ms, corresponding to m. 5. Discussion 5.1. Distribution of Gas Hydrates and Gas [30] The P wave velocity data (Figure 6a) provides evidence of large velocity variations on the northern flank of the Storegga Slide. These variations occur at the depth of the BSR, which delineates the BGHS, and at the base of the Naust formation. The vertical component of the OBC shows high reflection amplitudes (Figure 3), high reflection strength and a polarity reversal (Figures 6c and 6d) that coincide with the P wave velocity variations. The inline component of the OBC data does not suggest strong lateral changes in reflectivity along stratigraphic horizons, nor does it show a BSR that is crosscutting the sedimentary bedding. Because the inline component is obtained from the acquisition of shear waves on the seafloor, it is predominantly affected by the shear properties of the sediment matrix. Therefore the amplitude anomalies on the vertical component must be related to the pore fill of the porous matrix. [31] Above the BSR, the magnitudes of lateral velocity variations within stratigraphic units (Figure 6a) are significant (up to 12%). These variations are difficult to explain by changes in the lithologic composition of stratigraphic units, because the inline component does not suggest lateral reflectivity changes along individual horizons (Figure 3). The distinct increase in P wave velocity follows the BSR (Figure 6a) and suggests that it is related to this reflector. Therefore we conclude that higher P wave velocities above the BSR document the presence of an appreciable amount of gas hydrates within the pore space of the sediments. Lateral P wave velocity variations underneath the BSR are even higher than above (up to 36%). Here, the low P wave velocities are typical for a gas hydrate-related BSR [Holbrook, 2000]. Consequently, low P wave velocities are interpreted to be caused by the presence of gas in the pore space of the sediments. [32] The observations presented here are in agreement with earlier studies, which interpreted the BSR as the result of a negative impedance contrast between hydrate and-gasbearing sediments [Mienert et al., 1998]. Determined OBC interval velocities are consistent with previous studies of the seismic velocity of hydrate- and gas-bearing sediments at individual locations on the northern flank of the Storegga Slide. Modeling of ocean-bottom seismometer (OBS) data [Posewang and Mienert, 1999], full-waveform inversion of multichannel seismic data and a vertical seismic profiling interval velocity profile [Andreassen et al., 2000] show velocities between 1800 and 1850 m/s for the hydratebearing sediments and velocities as low as 1350 m/s for the gas-bearing sediments. [33] P wave velocities at the base of the Naust formation are as low as 1550 m/s. This 450 m/s-velocity inversion at the base of the Plio-Pleistocene glacial-interglacial wedge has been earlier identified throughout the Vøring Basin and on the continental shelf [Reemst et al., 1996; Hjelstuen et al., 1999]. Reemst et al. [1996] suggested that it is related to overpressure caused by the trapping of formation water under a layer of shale. This water could Figure 6. (a) Wiggle trace plot of the vertical component overlain with color-coded P wave velocities. The rectangle marks the location of the geotechnical borehole [NGI, 1997]. The BSR can be clearly identified by a velocity inversion. (b) Wiggle trace plot of the vertical component with hydrate and gas saturations of pore space calculated from the hydrateas-part-of-frame and the homogeneous-gas-distribution rock physics model of Helgerud et al. [1999]. Hydrate and gas saturations vary considerably along the line. (c) Reflection strength of the hydrophone component. (d) Apparent polarity of the hydrophone component. The coincidence of free gas zones with high amplitude reflections on the seismic data (see also Figures 3 and 5) suggests a model in which the BSR is predominantly caused by free gas that is trapped underneath the hydrates. 9of15

10 have been expelled from the polygonal fault system in the Kai formation underneath [Berndt et al., 2003] (Figure 2). Overpressure formation would also be supported by the fact that fluid expulsion rates of up to 60% have been reported for polygonal faults elsewhere [Verschuren, 1992]. Hjelstuen et al. [1999] related the velocity inversion to significant lithologic changes. However, they also recognized that cyclic ice sheet loading may have resulted in an overpressure within the sediments of the preglacial stages, and that this overpressure led to widespread diapirism, that is observed in the Vøring Basin [Hjelstuen et al., 1997; Judd and Hovland, 1992]. A third explanation for the velocity inversion might be the presence of gas, which also is suggested to be part of the diapirism [Hjelstuen et al., 1999]. [34] Velocity deviations around the BSR can be translated into estimates of gas hydrate and free gas saturations of pore space. These estimates vary depending on the rock physics model that applies. Three gas-hydrate models have been suggested [Ecker et al., 1998; Helgerud et al., 1999]: (1) Hydrates fill the pore space and are modeled as part of the pore fluid; (2) hydrates act as sediment grains and are modeled as part of the sediment frame; (3) hydrates act as intergranular cement and are modeled as contact cement [Dvorkin et al., 1994] between sediment grains. Hydrate saturation estimates are highest for the first model and lowest for the third. For the third model, already very low saturations of hydrates would result in a strong increase in s-wave velocity [Ecker et al., 1998] and hence in increase in shear resistance. The fact that a BSR cannot be observed on the inline component of the OBC rejects an increase in shear resistance (Figure 3) [Andreassen et al., 2003]. Therefore we suggest that hydrates at the northern flank of the Storegga Slide do not cement the sediments. The two remaining models can not be unambiguously distinguished because hydrate concentrations are probably too small to be able to discriminate between them [Andreassen et al., 2003]. Only accurate velocity information could yield additional insights into this distinction. Here, we calculate the more conservative estimate of hydrate saturation of pore space using the hydrate-as-part-of-frame model (Figure 6b). If hydrate saturations had been calculated for the pore fluidcomponent model, a rough estimate would be double the saturations of the hydrate-in-frame model. [35] Using Helgerud et al. s [1999] effective-medium theory, we calculate gas saturations assuming a homogeneous distribution of gas. In this case the composite bulk modulus of the pore fluid is the isostress average of the brine and gas bulk moduli. For a given velocity anomaly this results in significantly lower estimates than assuming a patchy distribution, in which gas is supposed to occur in patches that are larger than the scale of individual pores of the sediment. Also, we tentatively estimate gas saturations for the low-velocity layer at the base of the Naust formation. [36] The parameters for the modeling are taken from the geotechnical borehole (Figure 4a) [NGI, 1997]. The porosity and mineral constituents were averaged over individual layers separated by horizons as indicated in Figure 4b. The subbottom temperature was calculated using a temperature gradient of 52 C/km and a seafloor water temperature of 1 C [Mienert et al., 2001]. The salinity was kept at 32%. We assume the average number of grain contacts is 9 and Table 1. Elastic Properties of Sediment Solid Phase Components a Sediment Constituent Density, g/cm 3 Bulk Modulus, GPa Shear Modulus, GPa Mica Clay Quartz Feldspar Calcite Methane hydrate a Mavko et al. [1998], Dvorkin et al. [1999], Waite et al. [2000], and Batzle and Wang [1992]. the critical porosity is 40% [Mavko et al., 1998]. Density, bulk and shear moduli of mineral constituents and hydrate are taken from Mavko et al. [1998], Dvorkin et al. [1999], and Waite et al. [2000] and are listed in Table 1. The density and bulk modulus of brine and gas vary with depth and are calculated after Batzle and Wang [1992]. Because geotechnical data is not available at the depth of the base of the Naust formation, we assume 48% porosity, 25% mica, 25% clay, 25% calcite, 15% feldspar and 10% quartz. Hydrate and gas saturations of pore space are then estimated for every velocity point at every second CMP and 4 ms. [37] The errors of the estimates of hydrate and gas saturations are mainly related to the velocity errors, but also porosity errors and uncertainties in the mineral constituents might have a small effect. Considering that the error in interval velocity might be as much as 150 m/s, the saturations errors are significant, e.g., a velocity deviation of 30 m/s causes a change in hydrate saturation of approximately 4%. Furthermore, the rock physics modeling assumes a packing of spheres [Helgerud et al., 1999], but the fraction of sheet silicates (mica and clay minerals) is high. Because we assume velocity errors much below 50 m/s and errors from the modeling are systematic, total saturation errors for gas hydrates might be up to 5%. Due to the fact that only little amount of free gas significantly reduces the P wave velocity, and that subsequently the P wave velocity varies little with the amount of gas present, errors for the free gas saturation are estimated to a total of up to 0.4%. However, relative saturation changes of both hydrates and gas should be well preserved. Accurate velocity determination is crucial for accurate hydrate and gas saturation estimates, but necessary accuracies can only be obtained from sonic-log data, which so far is not available on the northern flank of the Storegga Slide. [38] Figure 6b shows the distribution and saturations of gas hydrate and free gas obtained from the rock physics modeling. Hydrate saturation of pore space varies significantly along the line. Hydrates generally seem to concentrate just above the BGHS. Highest hydrate saturations of up to 11% of pore space occur in the northeastern half of the OBC line with one concentrated occurrence above two terminating enhanced reflections (labeled B in Figure 6a). In the south-western half, saturations are up to 7%. The average gas hydrate saturation along the whole OBC line is 5%. The inferred zone of hydrate existence is generally 50 ms (40 50 m) thick with one exception: a location, where two strong enhanced reflections terminate at the BSR at approximately ms depth (labeled B in Figure 6a). Here, hydrates exist more extensive in a zone that reaches a thickness of about 100 ms (90 m). 10 of 15

11 [39] The geotechnical borehole penetrated through the BSR (Figure 6b), and should have retrieved gas hydrates in small concentrations just above the BSR. The fact that gas hydrates could not be recovered [Mienert and Bryn, 1997] suggests that hydrates must have been dissociated due to the coring activity. Ocean Drilling Program Leg 204 demonstrated the need for pressure coring to recover gas hydrates from subbottom sediments [Tréhu et al., 2002]. [40] Free gas saturations are highest at about 1.2% of pore space in the area underneath the BSR proper. Toward the downslope end of the line, gas saturations decrease, reaching maximum values of approximately 0.7%. Upslope, a distinct zone of higher gas saturations (about %) follows the BSR and two enhanced reflections (labeled B in Figure 6a). Above those two reflections and directly underneath the hydrates, gas saturations are lower ( %). For the whole OBC line, the average gas saturation of pore space in the gas-bearing layer underneath the BSR is 0.45%. Free gas saturations at the base of the Naust formation generally decrease upslope. They are highest (up to 1%) at the downslope termination of the line Nature of the BSR [41] Whether the BSR at the BGHS is the result of free gas underneath the hydrates, or the result of hydrate-bearing sediments overlying normally fluid-saturated sediments, has been a controversial issue [Hyndman and Spence, 1992; Singh et al., 1993; Holbrook et al., 1996]. ODP legs 146 and 164 indicated that free gas is present underneath hydrate-bearing sediments [Carson et al., 1995; Paull et al., 2000b]. Recent insights from the most well-known hydrate system at the Blake Ridge off the US east coast shows that hydrates can exist without a BSR [Holbrook, 2000]. However, at locations where gas is present, only 15% of the impedance contrast is attributed to the presence of hydrates [Holbrook, 2000]. [42] The fact that seismic amplitudes (Figure 3), BSR reflection coefficients (Figure 5), reflection strength (Figure 6c) and negative apparent polarity (Figure 6d) coincide with the zone where free gas is predicted to exist (Figure 6b), supports a model for the Storegga gas hydrates in which the BSR is predominantly the result of the free gas trapped underneath the hydrate-bearing sediments. This conclusion is further corroborated by the fact that higher gas saturations correlate with enhanced reflections underneath the BSR, and that highest hydrate saturations do not coincide with the area where a proper BSR is observed (Figures 3 and 6b, labeled A in Figure 6a). On the other hand, this area coincides with highest gas saturations, which suggest a relationship between the amount of free gas trapped and BSR formation. However, the appearance of a BSR is strongly related to the resolution of the seismic system [Chapman et al., 2002; Wood et al., 2002], and tuning [Widess, 1973] is supposed to affect amplitudes at BSR depth [Katzman et al., 1994; Andreassen et al., 1995]. It is nevertheless remarkable that the BSR proper is not a coherent event on the reflection strength plot (Figure 6c), but rather consists of individual spots of high reflection strength. These spots align at BSR depth and belong to enhanced reflections that stretch underneath and away from the BSR. This observation might indicate that the BSR proper is the result of tuning. The tuning might be caused by thin gas-bearing layers, whereby the thickness of these layers might vary at locations where the BSR crosses the stratigraphic bedding at small angles Dynamics of the Hydrate/Free Gas System [43] Gas hydrates and free gas within the sediments on the northern flank of the Storegga Slide show variations in distribution and concentration that should be the consequence of the origin and migration pathways of gas. Consequently, there should be geologic factors that control the origin and migration of gas and hence the distribution and concentration of hydrates and gas. [44] The Storegga gas-hydrate system is a relatively young system. Because gas hydrates are only observed in the Naust formation sediments [Bünz et al., 2003], the system must have started its development shortly after the start of the glacial cycles. The northern flank of the Storegga Slide is the only place on the margin which is unaffected by glacigenic debris flows, such that the gas hydrate system could develop continuously within the contourite drift sediments. [45] At the location of the OBC, today s GHSZ comprises mainly Naust units A and B, dating back to Mid-Pleistocene. This implies that the BGHS must have been moved upward approximately 280 m in a geologically very short period of time. Gas below the BGHS might result from recent methanogenesis within Naust formation sediments, from hydrate recycling [Paull et al., 1994], or from advection from deeper strata. The available data cannot conclusively explain the origin of gas, and it is likely to be a combination of all three processes. Bünz et al. [2003] suggested that hydrate recycling and methanogenesis within the Naust formation do not contribute significantly, because of an average organic carbon content of approximately 0.5% (Figure 4a). [46] The low-velocity layer at the base of the Naust formation that is related to expulsion of fluids from a polygonal fault system within the Kai formation (Figures 2 and 6a) might be a source of gas-rich fluids that ascend into the GHSZ. However, it is not completely clear yet if the gas in this layer originates from methanogenesis within the Kai formation sediments or if it originates from deep-seated hydrocarbon reservoirs within the compressional dome structures as hypothesized by Bünz et al. [2003]. Water that is expelled during the development of polygonal faults might be involved in the formation of gas hydrates. Due to the fact that the origin of the low-velocity layer at the base of the Naust formation cannot be conclusively answered, also the latter question remains open for further investigation. [47] The strongest control on gas concentrations is likely to be porosity, as highest gas saturations underneath the BSR are observed in strata with highest porosity (Figures 4a and 6b) within Naust unit C (Figure 4b). However, permeability contrast between strata might also play an important role, as suggested by Bouriak et al. [2000]. Enhanced reflections stretch laterally underneath the BSR and coincide with stratigraphic layers. This suggests that gas is present over some distance within layers. Strata associated with two strong enhanced reflections (labeled B in Figure 6a) show higher gas saturations than exist above or beneath them. This further corroborates that physical and 11 of 15

12 Figure 7. Schematic model summarizing the processes of the dynamic hydrate/free gas system on the northern flank of the Storegga Slide. geological properties of strata (e.g., porosity) control gas saturation. Because enhanced reflections are associated with the presence of free gas, their varying reflection amplitudes are the result of contrasting gas saturations due to different porosities in adjacent layers. [48] Within individual strata underneath the BSR, gas saturation generally increases toward the BSR. Together with the lithologic control on gas saturation, this suggests that gas predominantly migrate along strata. The stratigraphic bedding is slightly updip at the depth of the BSR, which suggests that gas migrate updip along strata until it is trapped by the low-permeability hydrate-bearing sediments. Hydrate formation sufficiently reduces the permeability to act as a barrier for the gas [Nimblett and Ruppel, 2003]. Moreover, sediments of the Naust formation on the northern flank of the Storegga Slide lack faults that may provide major pathways for gas-rich fluids. [49] Generally, hydrates are distributed in a mthick zone just above the BSR, which seems to be consistent with the relatively young age of the gas-hydrate system and the idea that hydrates form from advected fluids. However, varying gas-hydrate saturations seem to be related to the availability of free gas, i.e., the vertical and lateral flow of gas-laden fluids. Hydrates occur most extensively and in highest saturations in an area where two strong enhanced reflections terminate (labeled B in Figure 6a). This suggests that the stratigraphic layer that gives rise to those two enhanced reflections is a major pathway for gas into the GHSZ. Consequently, hydrate saturations are controlled by gas migration mechanism in this area. [50] At the downslope termination of the OBC line both, hydrate and gas saturations are generally lower than elsewhere along the line. Moreover, strata with higher porosity lie above the BSR, but unlike with gas saturations, hydrate saturations are low in this high-porosity layer. These lower hydrate and gas saturations might be explained by decreased availability of gas in this area. For the most part, the availability of gas seems to increase upslope along the OBC line. [51] Gas-rich fluids migrate either vertically or laterally into the zone around the BSR. The seismically chaotic unit related to the unconformity (Figure 3) does not show evidence for fluid migration. This suggests that fluids migrate mainly laterally into the area covered by the OBC. Further upslope a seismically transparent zone might provide evidence for gas-rich fluids (Figure 2; compare also Bünz et al. [2003, Figure 4a]). The only evidence for focused fluid flow is seen in acoustic pipes that originate in the low-velocity zone at the base of the Naust formation [see, e.g., Berndt et al., 2003, Figure 3]. Except for these pipes, the flow of fluids is supposed to be diffuse [Berndt et al., 2003]. At certain depth, the diffuse flow might change into a more focused flow along strata and consequently migrate laterally following updip the strata into the gascharged zone underneath the BSR. Finally, it enters into the GHSZ. [52] The fact that the BGHS has moved significantly upward since Mid-Pleistocene and the fact that gas today occurs in an only m-thick layer underneath the BSR suggests that fluid migration is an efficient process and happens on a geological timescale of a few 100 ka. The polygonal fault system in the Kai formation might be an explanation for rapid fluid migration. Fluids that are expelled from a polygonal fault system seem to create an overpressured layer just above the polygonal faults. This overpressure drives the fluid flow on the northern flank of the Storegga Slide, whereas in other areas in the Vøring Basin it is responsible for widespread diapirism [Hjelstuen et al., 1997; Hovland et al., 1998]. Pipes and pockmarks are a consequence of fluid expulsion [e.g., Hovland and Judd, 1988] in the study area [Berndt et al., 2003]. Because most of the pipes originate at the BGHS, gas hydrates might act 12 of 15