The relatively lower reflection amplitudes of some sediments within the Blake Ridge GHSZ compared to sediments below the GHSZ have been previously interpreted as reduction of impedance contrast and reflectivity due to cementation of sediment by gas hydrate (Lee et al., 1993). Holbrook et al. (1996) did not find the anomalously high velocities one would expect for frozen sediment, and they attributed the lack of impedance contrasts within the GHSZ to the homogeneity of the sediments at this location, which may only be reflective below the GHSZ because of the presence of free gas within preferentially porous strata. Although the velocities determined from the VSPs suggest minimal cementation on a scale of tens to hundreds of meters, the relative amplitude reduction (or blanking) may involve factors other than merely sediment homogeneity.
In some highly localized zones, particularly near faults, the Leg 164 sampling and logging results imply that gas hydrate concentrations may be high enough to fill a significant portion of pore space and to affect seismic velocity at submeter- to meter-length scales. Dillon et al. (1994) describe "fingers" of blanking that extend upward into the GHSZ from faults below the BSR. Such blanking zones may continue to even shallower depths in the section along unresolved faults. Cementation from high concentrations of gas hydrate are more likely in fault zones than in adjacent, unfaulted sediments characterized by presumably lower fluid flux. However, these potentially cemented fault zones are not oriented horizontally, may be quite irregular in shape and impedance, and may thus be more efficient in scattering energy outward rather than directly upward to the receivers. Thus, a "blank" zone in seismic data may imply a paucity of nearly horizontal reflectors, not necessarily a lack of impedance contrasts. The kind of scattering described here typically exhibits strong frequency dependence (e.g. Clay and Medwin, 1977) and would likely yield different results when studied with seismic systems of different frequencies.
Another phenomenon that may be responsible for blanking at larger scales is the increasing throw with depth on the syndepositional faults imaged in the DTAGS seismic data (Fig. 5). A spherical, spreading seismic wave has a finite lateral resolution that decreases with decreasing frequency and increasing distance from the reflector. Inhomogeneities that are below the resolution are averaged, and, due to scattering, generally result in lower reflectivity being recorded by the receivers (Clay and Medwin, 1977). For wavelengths greater than ~30 m, (less than ~60 Hz), a 15-m displacement in a reflector creates destructive interference, and the observed reflectivity is diminished. For data of this frequency, we would therefore expect a decrease in observed reflectivity with depth, because the lateral heterogeneity increases due to increasing fault throw. Destructive interference will increase with the distance from the source as more heterogeneity is incorporated into the first Fresnel zone. This may explain why the high-frequency, deep-tow data shown in Figure 5 (250-650 Hz, at 750 m above the BSR) does not show as great a discrepancy between reflectors just above and just below the top of gas as seen in the surface-tow data of Figure 1 (10-240 Hz, 3250 m above the BSR). The lateral heterogeneity of the faulted sediments, combined with the nonhorizontal distribution of hydrate, leads us to believe that the reduced reflectivity is more likely due to scattering and destructive interference than to reduced impedance contrasts from sediment cementation.
Theoretical modeling of the steady-state methane gas hydrate system in porous marine sediments (Xu and Ruppel, 1999) reveals several misconceptions about the significance of the BSR relative to the base of the GHSZ and GHZ. Although the intersection of the temperature profile with the phase boundary does mark the lower limit of gas hydrate stability (base of GHSZ) and the potential upper limit for the top of free gas, several factors control where gas hydrate (GHZ) and free gas actually occur. Most notably, the GHZ can only extend as deep as the base of the stability zone/top of the free-gas zone if the rate of methane supply exceeds a critical rate whose value depends on fluid-flux rate, energy flux, and other parameters. Furthermore, under certain conditions, the top of the free-gas zone may be separated from the base of the GHSZ by a layer of sediments lacking both free gas and gas hydrate.
Xu and Ruppel (1999) note that these results have particular significance for interpreting seismic data and other observations made on ODP Leg 164. At Site 994, where a strong BSR continuous with those at the adjacent sites is lacking, Holbrook et al. (1996) inferred the presence of free gas at ~560 mbsf. Chloride anomalies between 196 and 456 mbsf at this site (Paull, Matsumoto, Wallace, et al., 1996) are consistent with pore-water freshening (Hesse and Harrison, 1981) and the in situ occurrence of gas hydrate. Thus, the top of the free-gas zone at Site 994 appears to lie significantly below the base of the actual zone of gas hydrate occurrence (GHZ), consistent with models in which the methane supply rate is less than the critical value. At Sites 995 and 997 (strong BSR), the top of the free-gas zone and the base of the GHZ are effectively coincident, which implies that the methane supply rate at these sites exceeds the critical value. Thus, if all other physical parameters at Sites 994 and 995 are approximately equivalent, the 3-km distance that separates the two sites may mark a transition from precritical to critical methane supply rates (C. Ruppel, unpubl. data). This transition in methane supply rates is accompanied by a transition from noncoincidence to coincidence of the top of free gas and the base of the GHZ/GHSZ and a consequent change from no BSR (Site 994) to a strong BSR (Site 995). If all methane were supplied by biogenic processes occurring below the GHSZ, then the observed coincidence between the top of free gas and the base of the GHZ can be used to determine the critical rate of methane supply and the potential size of the methanogenic bacterial population (Xu and Ruppel, 1999).
Figure 6 shows the predicted depths of different boundaries within the Blake Ridge gas hydrate system and the relationship between methane supply rate and the depth to the top of the free-gas zone, determined using the Xu and Ruppel (1999) analytical model. Note that the base of the GHSZ in these plots lies closer to the actual observed depth of the BSR at Site 995 than would be predicted based on traditional stability curves and the measured geotherm (Ruppel, 1997). Two factors combine to explain the close agreement of the observed BSR depth (~440 mbsf) and the predicted depth to the top of free gas and base of the GHSZ/GHZ in Figure 6. First, the adopted value for constant energy flux (combination of advective and conductive components) is 40 mW/m2, which is at the high end of the range of values (36.2-39.9 mW/m2) determined at Site 995 (C. Ruppel, unpubl. data). Second, the phase equilibria used in the Xu and Ruppel (1999) model are based on state-of-the-art statistical thermodynamics calculations by Tohidi et al. (1995) for methane hydrate and 3.5% seawater. These predicted phase equilibria yield somewhat lower temperatures at the base of the GHZ and a predicted BSR (where the top of free gas is in coincidence with the base of the GHZ/GHSZ) at a depth closer to that observed from seismic methods.
Large amounts of visible hydrate were not recovered at Sites 994, 995, and 997, except for the single 20-cm-long massive piece recovered at 330 mbsf in Hole 997A, probably at the intersection of a fault and the borehole. The overall lack of visible hydrate deposits, coupled with catwalk temperature measurements made immediately upon core recovery (Paull, Matsumoto, Wallace, et al., 1996), suggests that much of the gas hydrate in the section probably is finely disseminated. Independent analyses based on pore-water chloride anomalies, (Paull, Matsumoto, Wallace, et al., 1996), resistivity logs (Paull, Matsumoto, Wallace, et al., 1996), and seismic data (Holbrook et al., 1996) yield an estimate of 2%-3% gas hydrate throughout the GHZ.
If disseminated gas hydrate forms at grain contacts, it may act as a cement, effectively freezing the sediment. Dvorkin and Nur (1993) have shown that for cementation at an idealized spherical grain contact, both Vp and Vs reach ~90% of their saturation values when the cement occupies only ~3% of the total volume. However, the VSP data reveal only modest increases in velocity and do not support the cementation hypothesis on the scale of tens to hundreds of meters. Conversely, theoretical considerations and laboratory experiments (e.g., Handa and Stupin, 1992; Clennell et al., in press) suggest that capillary forces arising in fine-grained sediments like those on the Blake Ridge may inhibit gas hydrate stability, leading to an inference of initial formation in pore spaces away from grain boundaries (Clennell et al., in press). Ecker et al. (1998) used a model of gas hydrate formation away from grain contacts to explain MCS amplitude vs. offset results from the Blake Ridge. For seismic waves on the scale of meters or larger, the medium is a continuum, and gas hydrate in pore spaces would likely affect the Vp of the sediment in a time-averaged sense, based on the fractional volume of gas hydrate present (Wyllie et al. 1958; Lee et al., 1996). A small (2%-3%) concentration of hydrate would thus have only a small effect on elastic properties, a result consistent with the observed ZVSP velocities.
WVSP data are also affected by the microscale distribution of hydrate. The converted S-waves that arise within the GHZ occur at horizons where shear modulus changes abruptly, possibly due to gas hydrate forming at grain contacts over zones that are hundreds of meters thick. However, it is difficult to unambiguously ascribe changes in shear moduli to the presence of gas hydrate or the cementation of sediments by gas hydrate. Whether the contrasts in shear moduli are more significant than would be expected for compacting sediment lacking gas hydrate remains to be determined. Although Vp, Vs, and Q measurements from Leg 164 do not clearly resolve the microscale distribution of gas hydrate, current analyses suggest that the elastic effects from 2% to 3% gas hydrate by volume are small.
The Leg 164 geophysical data are consistent with increased concentration of gas hydrate near faults of the type shown in the seismic images of Figure 1 and Figure 5. Such features were previously suspected to be possible conduits for advective flux (Wood et al., 1997) and the escape of methane gas (Rowe et al., 1995). As shown quantitatively by Xu and Ruppel (1999), large concentrations of gas hydrate or high accumulation rates are almost unequivocally linked to high rates of fluid flux and methane supply to porous sediments within the stability zone. At present, we lack a detailed high-resolution image that can precisely constrain the angle at which the boreholes may have intersected a fault. Furthermore, the degree of variability in gas hydrate concentration (Paull, Matsumoto, Wallace, et al., 1996) between Holes 997A (up to 100% at 330 mbsf) and 997B (~23% at 365 mbsf) and sparse lateral sampling render simple interpretations impossible, particularly if the holes intersect the same fault. Quantitative calculations on a one-dimensional system indicate that gas hydrate accumulations in homogeneous porous media should be either relatively uniform with depth or more concentrated near the base of the zone (~450 mbsf), depending on the rate of mass flux through advection-dominated systems (Xu and Ruppel, 1999). If a fault represents a zone of effectively 100% porosity and Holes 997A and 997B intersect the same fault, then the inferred pattern of gas hydrate concentrations in both the vertical and lateral directions does not conform to the pattern predicted by simple models. More extensive modeling using more realistic parameters based on better field constraints on the lateral and vertical distribution of gas hydrate will be necessary to address the issue of gas hydrate distribution along conduits.
The thermal data do not reveal a significant advective component at any of the sites, but the slightly elevated thermal gradient at Site 997 is consistent with the intuitive notion that gas hydrate near a fluid conduit (fault) lies closer to the stability boundary than gas hydrate in the surrounding sediments and at adjacent sites. The fault zone provides an obvious escape pathway for warmer than in situ methane-laden fluids produced through dissociation of gas hydrate. Gas hydrate in and near such fault zones may represent a less stable, more mobile subreservoir whose dissociation might be triggered by relatively small bottom-water temperature increases or sea-level falls (possibly accompanying climate change; Paull et al., 1991; Dickens et al., 1995, 1997) or even episodes of rapid sedimentation (e.g., Pecher et al., 1998).
The potential high concentration of gas hydrate along faults may also affect seafloor stability. Dissociation of gas hydrates has long been linked to large mass-wasting phenomena produced when sediments presumably liquefy during gas hydrate dissociation (Kayen and Lee, 1991). If the more volatile (more mobile and less stable) gas hydrate deposits are concentrated along sites of previous strain (faults), then initial dissociation zones should be more localized. Thus, there is an even greater potential for the initiation of mass-wasting events at existing faults than would be expected in areas with no gas hydrate.