2H
ValuesThe Cl- pore-water concentration profiles for Holes 996A, 996D, and 996E (Fig. 3) are characterized by a general increasing trend with increasing depth, and large fluctuations. Usually the concentration of Cl- in pore waters from deep-sea sediments only vary by a few percent. At Site 996 the concentrations range from seawater values (560 mM) near the sediment/water interface to 970 mM at 54.26 mbsf in Hole 996E. The large variations are believed to be caused by varying amounts of hydrate decomposition during pore-water recovery and the influence of saline water from below. Many (possibly all) the pore-water samples are mixtures of hydrate meltwater and in situ pore water.
Meltwater samples from pure hydrates should contain zero Cl- concentrations because hydrates exclude ions during their growth. The concentration of Cl- in the hydrate meltwaters range from 21 to 352 mM (Table 1); thus, all samples contain significant amounts of interstitial water. This is a common feature for hydrate samples from oceanic sediments (e.g., Kastner et al., 1990).
Numerous processes may produce subsurface brines in sedimentary sequences (e.g., Martin et al., 1995; Egeberg, 1992). In the present setting, the two most probable mechanisms are evaporite dissolution and ion exclusion by hydrate formation. The pronounced increase of the Na/Cl value with increasing depth (Paull, Matsumoto, Wallace, et al., 1996) strongly suggests that the interstitial water has acquired its high salinity by dissolution of evaporites and that the Blake Ridge Diapir is of evaporitic nature.
2H
ValuesThe 2H
values of the pore-water samples range from -2 to 3.3
(Table 1) and show wide
scattering when plotted vs. depth (Fig.
4).
Most pore-water profiles from deep-sea sediments show a progressive depletion in deuterium with depth below the seafloor (Friedman and Hardcastle, 1988). This has been attributed to glacial effects (Friedman and Hardcastle, 1988), alteration of volcanic ash or basalts (Cole et al., 1987), dewatering of shales (Yeh, 1980), formation of hydrate (Kvenvolden and Kastner, 1990), and oxidation of mantle-derived methane and hydrogen (Lawrence and Taviani, 1988).
The hydrate meltwater
samples are enriched in deuterium (range 8.1-17.1).
The strongest heavy isotope enrichment is observed for Sample 164-996E-7H-CC,
which contains only 21 mM Cl-. For comparison, Kastner et al. (1990)
reported 2H
values of 19.3
and 17.6 for
two hydrate samples containing 51.4 and 90.6 mM Cl-, respectively.
The first subsample of Core 164-996B-1H (2.1 mbsf) contains about 32% more Cl-
than the second subsample and is proportionally less enriched in deuterium (Table
1). The differences between the two subsamples of Core 164-996C-1H (2.3
mbsf) is about 16% in Cl- and -12% in deuterium.
Prior to Leg 164, Paull et al. (1995) had presented several evidences for pore-water advection at Site 996. The presence of a deep-sea chemosynthetic biological community suggests unusual conditions. These authors also observed high concentrations of CH4 and H2S in the shallow sediment cores and in the water column above, and acoustically defined water-column plumes that apparently emanate from a fault connecting the seafloor to the base of the BSR.
Direct sampling of hydrates during Leg 164 confirms that pore waters at Site 996 contain concentrations of CH4 sufficient to stabilize hydrates. This is unexpected for sediments with moderate organic carbon concentrations (0.41%, n = 14), and suggest influx of CH4 from below. Gas hydrate-filled veins that were 1-3 mm thick and could be traced 30-40 cm along cores represent possible flow conduits. Numerous veins similar to these are probably closely spaced in the sediments because two parallel veins were observed in an individual core and because veins were found in most of the cores below 30 mbsf.
Egeberg and Dickens (1999) showed that the amount of hydrate and rate of fluid migration may be determined by solving the transport equations for two conservative pore-water constituents, provided the concentrations of these species in pore waters squeezed from hydrate-free sediments (sediments from below the BSR) have been determined. At Site 996 the BSR was not penetrated; hence an alternative method is used here. The method used for estimating the amount of hydrate is based on the assumption that the rate of exchange of pore water by fluid migration is so rapid that formation of hydrates and other diagenetic reactions do not influence the concentration of pore-water Cl- and deuterium. This assumption will be discussed below when discussing the source of water and gas.
In situ concentration
profiles of Cl- and deuterium are generated by the model included in Appendix
A. The concentration profiles are very sensitive to the assumed rate of
pore-water advection (Fig. 5A).
However, when the simulated 2H
values are plotted vs. the simulated Cl- concentrations, this
flow-rate dependency vanishes (Fig. 5B).
The fact that the flow-rate dependency vanishes when the concentrations of one conservative dissolved species are plotted vs. the concentrations of a second conservative species is the basis for the method of estimating the hydrate amount used here. The flow-rate independent curve (Fig. 5B, Fig. 6) will be referred to as the FRI-curve.
Actual pore-water compositions should plot at the FRI-curve if the sediment from which the pore-water sample was obtained did not contain hydrate. Pore-water samples diluted by meltwater from hydrates should plot above the FRI-curve, because hydrates are enriched in deuterium (Fig. 6).
The distance from the
FRI-curve provides a measure of the amount of hydrate meltwater in a pore-water
sample and may be used for quantification. The difficult part is to determine
the input parameters for constructing the FRI-curve, which is the lower boundary
conditions for modeling of the in situ Cl- and 2H
profiles. Seawater values for Cl- (560 mM) and 2H
(0 SMOW) are
assumed for the upper boundary conditions. Because all of the recovered
pore-water samples may be diluted by hydrate meltwater, the lower boundary Cl-
concentration must be higher or equal to the highest measured Cl-
concentration (970 mM, Table 1),
and the lower boundary 2H
value must be more negative or equal to the most negative measured 2H
value (-2.0
SMOW,Table 1). However, there
is no lower constraint on the FRI-curve (Fig.
6). Minimum estimates of the amount of hydrate may be obtained by
choosing lower boundary conditions, which results in a FRI-curve that intersects
the most negative 2H
value. One set of lower boundary conditions that fulfill this condition when the
lower boundary is set at 250 mbsf is 1400 mM Cl-, -5.6 SMOW
(curve 1, Fig. 6). It is
possible to construct similar curves with other lower boundary conditions, for
example 1600 mM Cl-, -6.4
SMOW. However this uncertainty has no effect on hydrate amount estimates because
the distances between the measurements and the FRI-curve remain the same.
The procedure for estimating hydrate amount is described in Appendix B. The hydrate-amount estimates (Fig. 7) suggest the presence of hydrates immediately below the seafloor in all holes.
Shallow occurrence of hydrate is consistent with the observation of massive pieces of hydrates in the first core from Holes 996B, 996C, and 996D. The first two cores from Hole 996A were left for over an hour to degas H2S before splitting, and no visual observations of their hydrate content could be made (Paull, Matsumoto, Wallace, et al., 1996). In Hole 996E, the first visual detection of gas hydrate was made in Core 164-996E-4H (23.2-32.7 mbsf). Recovery was extremely low (3.8%) for Core 164-996E-3X. In Cores 164-996A-9H and 164-996D-6H, which contain the highest theoretical estimates of hydrate amount (29% and 31% respectively), hydrates were observed as wavy, approximately vertical veins as much as 0.5 cm thick and 3-4 cm wide that could be traced about 30-40 cm along the cores. The mean hydrate concentration in the sampled sequence is 10% of pore space. As explained above, this is a minimum estimate because there is no lower constraint on the FRI-curve (Fig. 6). However, the fact that several samples plot close to the selected FRI-curve (i.e., several samples do not contain hydrate) does suggest that the selected curve is the most probable, because visual core observations did reveal a patchy distribution of hydrate. Hence, it seems unlikely that all samples taken for pore-water squeezing contained hydrate.
All samples of hydrate meltwaters contain appreciable concentrations of Cl- and are less enriched in deuterium than one would expect for marine hydrates (e.g., Kastner et al. 1990), because the hydrate samples are contaminated by pore water. Estimates of the amounts of hydrate in these samples range from 44% to 99% (Table 1). Note that because the vectors used to estimate the amounts of hydrate in the hydrate samples intersects at low Cl- concentrations (Fig. 6), the solutions are not unequivocal. However, it is possible to choose the correct solution if one also considers the sample depths, because the in situ Cl- concentration must increase with increasing depth. Hence, the composition of the shallow (2.1-3.5 mbsf) hydrate samples (filled squares) was determined by means of vector Va, and the composition of the deep hydrate samples (32.1 and 58.63 mbsf, Table 1), as determined by vector Vb (Fig. 6).
Estimation of rates of fluid migration is based on the estimates of in situ Cl- from above (Table 1) by adjusting the advection term in the transport equation (Appendix A) until the simulated Cl- concentrations match the data.
It is evident that all data cannot be fitted with a single flow rate (Fig. 8). This is not surprising because the permeability field is probably quite heterogeneous because of the presence of fractures and carbonate-cemented layers (Paull, Matsumoto, Wallace, et al., 1996). Most of the data can be fitted with upward flow rates between 0.125 and 0.5 m/k.y. These flow rates are one to two orders of magnitude greater than pore-water flow rates generated by sediment compaction. Nevertheless, they seem too low to sustain the benthic biological community and are difficult to reconcile with the observations of water-column plumes (Paull et al., 1995). However, the derived flow rates must be regarded as averages for the whole diapir cross section, and the flow rates may be considerably larger on a local scale (i.e., along faults; see "Influence of the Fracture System" section, this paper).
Quantification of fluid flow is important, because it may help explain the shallow occurrence of hydrate at Site 996. An elegant model to account for formation of hydrate in organic-poor sediments involves recycling of CH4 by gas bubbles rising across the BSR (e.g., Paull et al., 1994). According to this model, most of the hydrate growth should occur close to the BSR. At Site 996 the BSR is located at ~250 mbsf (estimated from Paull et al. (1995), fig. 1). It is evident from direct observations and hydrate amount estimates presented here that prevailing conditions at Site 996 allow formation of hydrate at much shallower depth, despite the low content of organic matter.
At Site 996 vein-filling hydrates were recovered from several cores (Paull, Matsumoto, Wallace, et al., 1996), supporting the hypothesis that formation of massive hydrates is associated with fluid advection. Other models for the formation of natural gas hydrate differ from Paull et al. (1994) in that they involve active fluid flow (e.g., Soloviev and Ginsburg, 1994; Hyndman and Davis, 1992; Egeberg and Dickens, 1999) rather than rising CH4 bubbles.
The rate of pore-water flow at the nearby ODP Site 997 has been estimated to be about 0.2 m/k.y. (Egeberg and Dickens, 1999). The sediments at Site 997 contain a mean hydrate concentration of 2.3% (Egeberg and Dickens, 1999). At Site 996 the mean flow rate is 0.35 m/k.y., and the mean hydrate concentration is about 10% of pore space. Based on chlorinity data, Hyndman and Davis (1992) estimated a mean hydrate concentration at Deep Sea Drilling Project (DSDP) Sites 496 and 497 (lower and mid-slope of the Pacific active margin off Guatemala respectively) of about 25%, in association with a upward rate of fluid migration of 1 m/k.y. Similar calculations for the northern Cascadian accretionary ridge showed that a hydrate layer of up to 100 m thickness may form in response to a rate of fluid expulsion of 3 m/k.y.
Vertical advection of fluids promote precipitation of hydrates at shallow depth by two effects. First, it provides a supply of CH4 that might otherwise not be sufficient to saturate the pore water with respect to hydrates in organic-poor sediments. Second, it provides a mechanism to reduce the concentration of sulfate and thereby allow concentrations of CH4 to build up, both by increasing the rate of sulfate reduction via methane-oxidizing bacteria (e.g., Borowski et al., 1996) and by counteracting diffusion of seawater sulfate into the sequence. At Site 996 significant concentrations of sulfate (>1 mM) were not observed below 7 mbsf in any core (Paull, Matsumoto, Wallace, et al., 1996), similar to Site 892 at Cascadia where there is also seafloor seepage and a bioherm present (Kastner et al., 1995).
It has been suggested
previously that decomposition of hydrates below the BSR may be the source of
advecting fluids and gas at the Blake Ridge Diapir (Paull et al., 1995). Using a
rate of sediment accumulation of 0.066 m/k.y., a concentration of hydrate of
10%, and a porosity of 50% result in an estimated rate of burial of
hydrate-bound H2O of about 3.1 kg/m2/k.y. When the
hydrate-bearing sediments are buried across the BSR and melt, this meltwater may
contribute to the upward advective flux of pore fluids. However, the flow rates
derived by modeling of the Cl- data gives a total advective
pore-water flux of 60-250 kg/m2/k.y. Hence, the proportion of hydrate
meltwater in the ascending pore water is probably not more than a few percent;
most of the ascending pore water must come from another source than melting of
hydrate. The best-fit lower boundary conditions (1400 mM Cl-, -5.6
2H)
are consistent with this finding. If hydrate meltwater constitutes a significant
proportion of the ascending pore water, one would expect heavy 2H
values and low Cl- concentration. We believe that the source of the
ascending water is pore water from the sedimentary sequences surrounding the
Blake Ridge Diapir. By using the rate of sedimentation for Site 996 as an
estimate for the regional rate of sedimentation, a rate of burial of pore water
of about 30 kg/m2/k.y. in regions without externally induced
pore-fluid flow is estimated. Hence, the water budget for the diapir may be
balanced by assuming that pore water is focused towards the diapir from a
surrounding area that is 2-8 times greater than the area of the diapir itself.
On the other hand, if one assumes that hydrate meltwater is the main source of
the ascending water (Paull et al., 1995), then the drainage area would have to
be 20-80 times greater than the area of the diapir, which seems less realistic.
We suggest that the local high rate of pore-fluid flow at Site 996 is caused by
focusing of pore water from a larger area by the structures created by the
rising diapir. Paull et al. (1995) proposed that gas and fluid flow may be
focused laterally towards the diapir along a permeable conduit beneath the base
of a gas hydrate seal. However, if free gas does exist below the BSR, as
suggested by Dickens et al. (1997) for other Leg 164 sites, then this zone will
be characterized by low water permeability because of the reduced cross section
available for pore-water flow.
Parts of the CH4 incorporated into hydrates may come from CH4 released during decomposition of hydrates below the BSR and advected upward (recycling). The CH4 flux associated with the burial of hydrate (28 mol/m2/k.y.) is of the same order of magnitude as the advective flux of dissolved CH4 (11-43 mol/m2/k.y., assuming a saturation concentration of 170 mM). However, an unknown amount of dissolved CH4 is discharged across the sediment/water interface and is not available for formation of hydrate. Hence, it seems that simple recycling of CH4 from the BSR is not sufficient to balance the CH4 budget. Furthermore, an additional amount of CH4 is required to also balance the sulfate budget, because CH4 is consumed by sulfate reducing bacteria with a 1:1 stoichiometry (Borowski et al., 1996). Due to large spatial variations in the sulfate gradient, it is difficult to estimate the size of the sulfate flux. The steepest sulfate gradient (Core 164-996D-1H, -266 mMm-1) gives a sulfate flux of approximately 1600 mol/m2/k.y., whereas the lowest sulfate gradient (Hole 996E, -4 mMm-1) gives a sulfate flux of 20 mol/m2/k.y. Although we can not accurately estimate the apparent CH4 deficiency, it may amount to several tens to hundreds of mol/m2/k.y. Methanogenesis of dissolved organic carbon supplied by the ascending pore water and internal CH4 generation from sedimentary organic matter are the most likely sources of the missing CH4. Egeberg and Barth (1998) reported concentrations of nonvolatile dissolved organic carbon (DOC) of up to 95 mM in the ascending pore water from the nearby Site 997. Conversion of this amount of carbon would add another 6-24 mol/m2/k.y. to the supply of CH4. Pore waters from Site 996 have not been analyzed for DOC. The flux of sedimentary organic carbon is about 18 mol/m2/k.y. (rate of sedimentation 66 m/m.y.; porosity: 70%; concentration of organic carbon: 0.4%, Paull, Matsumoto, Wallace, et al., 1996). Hence, internal generation of CH4 from sedimentary organic carbon is restricted to a few mol/m2/k.y., because only a fraction of the sedimentary organic matter can be microbiologically converted to CH4 (Henrichs and Reeburgh, 1987).
Mass balance calculations may also be used to justify the fundamental assumption made in this study that the system is transport dominated, rather than reaction dominated with respect to Cl- and deuterium. A system is transport dominated if the flux of dissolved species through the system is several times larger than the rates of diagenetic reactions that may influence the concentration of the species. The distinction is important because modeling of transport-dominated systems does not require incorporation of reaction terms into the diagenetic equation. The rate of exclusion of Cl- by formation of hydrate (3 mol/m2/k.y.) derived from the rate of burial of hydrate (3.4 kg/m2/k.y.) and a pore-water Cl- concentration of 1000 mM constitutes only a small fraction of the flux of Cl- through the sequence (60-250 mol/m2/k.y.). Similar calculations for deuterium give similar results; the fluxes of these components through the system are one to two orders of magnitude higher than the integrated rates of hydrate formation. Hence, the assumption of a transport dominated system seems to be valid because no other diagenetic reaction is known that could possibly affect the concentration of these species at this level.
The model used to estimate the pore-water advection rates (Appendix A) assumes diffusion and advection through a homogeneous porous matrix and does not explicitly handle flow through fractures, although the presence of vein-like hydrates in cores indicates that most of the flow takes place through the closely spaced fracture system. The validity of using this model under the present conditions may be judged from the following considerations. The time required for molecular diffusion to eliminate concentration gradients between the highly permeable fracture system and the blocks of sediments between the fractures is probably small compared to the vertical component of the pore-water advection velocity. For example, for a fracture spacing of 10 cm, 90% of the concentration difference is eliminated by molecular diffusion in less than 0.13 yr (Fig. 9).
Note that the time required to homogenize the concentration field is independent of the magnitude of the concentration gradient, because large concentration gradients cause large mass fluxes per unit time. The time required to eliminate the difference in concentration between fractures and a sediment block increases by the square of the block dimensions, but it only takes about 13 yr to eliminate 90% of a concentration difference for a fracture spacing of 100 cm. Even this is short compared to the residence time of water in the sequence (>150 k.y). Based on the core observations, the fracture spacing is probably on a centimeter rather than a meter scale; hence it seems likely that the interstitial water is always close to equilibrium with the water flowing through the fractures. Therefore, the model probably generates valid results, even if it does not include fracture flow.
