4.1. Flow Rates
Freshening due to mixing with buried brackish water from the Marmara Lake is observed at all sites except at one location of the Western High were fluids having a salinity higher than seawater are expelled. Even at locations with no fluid flow through the seafloor, the change in water salinity after the reconnection with the Mediterranean Sea 14.7 kyrs BP [Vidal et al., 2010] is expected to cause a freshening gradient within the sediment. For a chloride diffusion coefficient in the sediment of 10−9 m2/s, the chlorinity gradient at the seafloor in a pure diffusion model is 13.5 mM/m [Zitter et al., 2008]. Taking into account uncertainties in diffusion coefficients, on the timing of the salinization, on true depth, and on the salinity of the end-member fluid, any gradient deviating by more than±5 mM/m from this value in the first 10 m below the seafloor almost certainly indicates fluid movement with respect to the seafloor. Furthermore, profile curvature at the 10 m scale indicates vigorous advection at rates in the mm/yr range or more. It follows that coring sites can be classified between sites with no net water outflow or downflow (KC-2, KC-5, KC-13, KC-17, KC-26, VT-2737, VT-2741, VT-2742), sites with local outflow on the order a mm/yr (KC-7, KC-19, KC-20, KC-26, KC-31, VT-2740) where the gradient of chlorinity is increased, and one site where upward fluid velocity can be estimated from the shape of the profile (KC-30). Fluid velocity is determined to be 10 to 15 mm/yr for this core by fitting data below 2 m depth with a steady state advection-diffusion model [e.g., Martin et al., 1996]. Fluids expelled though the carbonate mound have higher chlorinity than seawater. In core KC-27, the curvature of the chlorinity profile above the depth at which gas hydrates were sampled indicates upward fluid flow at a velocity of about 20 mm/yr. In core KC-14, advection rate is too low to cause significant curvature, and no hydrate was found. Subsurface fluid flow is a complex process and flow rates are both spatially and temporally heterogeneous [e.g., Tryon et al., 2002] so these rates should not be extrapolated to the scale of the mounds but may be indicative of the order of magnitude of flow in these locations. Overall, these results are comparable with typical rates in areas of faulting in marine sedimentary environments such as convergent and passive margins [e.g., Luff et al., 2004; Torres et al., 2002; Tryon et al., 2002; Tryon and Brown, 2004].
4.2. Fluid Sources
Sources of pore fluids in the Sea of Marmara are significantly different than in typical ocean basin or margin environments. The continental setting replaces igneous oceanic basement with primarily metamorphic rocks of sedimentary origin and its recent period as a lacustrine environment adds fresh water to the sources. The mix of bottom water, Lake Marmara brackish water, and deep sources associated with the fault zone and/or gas reservoirs can be overprinted with silicate and carbonate diagenetic processes and, at the mound site, removal of water during gas hydrate formation. These multiple sources and processes cause the genesis of some of these fluids and their initial compositions to be not readily determined. Some general observations and interpretations can, however, be made which can bracket the sources and processes involved.
Major ion concentrations at most sites in the Sea of Marmara can be modeled from the mixing of two components. The basin core pore fluids are dominated by simple mixing of a present-day bottom water upper end-member with a brackish, low-density Lake Marmara end-member that is advecting buoyantly and/or diffusing from a relatively shallow depth. This mix is overprinted by shallow redox reactions and carbonate precipitation that consumes methane, sulfate, Ca, Mg, and Sr, and produces carbonates and sulfides. Some traces of low-temperature silicate diagenesis are also present as well as traces of thermogenic gas but, in general, no deep source fluid is apparent in the basin and basin margin cores.
The most altered fluids sampled at core KC-20 on the Central High are compatible with mixing of 60% seawater and 40% brackish water with an overprinting of local low-temperature silicate diagenesis. The profiles of K, B, Li, Mg, and Na fall significantly below the mixing line between seawater and Lake Marmara water, consistent with low-temperature silicate mineral diagenesis (Figure 5). Ca-bearing silicate minerals such as plagioclase feldspar are common components of basin sediment in continental environments and these weather to form a variety of authigenic clay minerals such as kaolinite. A large number of different weathering reactions are possible depending on the abundance of different silicate minerals and available cations but all ultimately consume cations such as K, B, Li, Mg, and Na and release Ca [e.g., Gieskes and Lawrence, 1981; Martin et al., 1996]. Ca, Sr, and Ba increase with depth (Figure 2d). Ba and Sr are common elements in sedimentary rocks and would be expected to be released. The weathering of K feldspar and mica contribute Ba and Sr to pore fluids and the dissolution of biobarite and recrystallization of calcite below the SMI may also be involved in the increase in these elements. Gas collected on dive 1664 near this site was analyzed and found to be primarily methane with a thermogenic origin [Bourry et al., 2009] suggesting the possibility that there may be an additional deep source of fluids. Silicate diagenetic processes at this source or as a result of fluid-rock interaction during gouge formation in the MMF zone may contribute to the observed composition.
The three mound cores exhibit the most overwhelming evidence for deep-sourced fluids and deep processes including, most notably, thermogenic gas and oil. One method for estimating the source temperature and depth is the use of geothermometers. In a recent review of available geothermometers, those based on Na-K and silica appeared as the most reliable, at least for crustal hydrothermal systems [Verma et al., 2008]. Li-Mg and Li-Na geothermometers have been proposed for sedimentary basin fluids, notably these associated with oil [Kharaka and Mariner, 1989]. Application of these geothermometers to fluid expelled through deep-sea mud volcanoes also yielded consistent results [e.g., Martin et al., 1996]. When applied to fluids from KC-27 and KC-14, geothermometers based on Na-K [Fournier, 1979] yield temperatures of 118°C and 138°C, respectively, and on Na-K-Ca [Fournier and Truesdell, 1973] yield temperatures of 136°C and 154°C, respectively. When applied to the Li-rich fluids sampled at the oil seep site, the Li-Na geothermometer of Kharaka and Mariner [Kharaka and Mariner, 1989] yields temperatures of 122°C and 132°C, almost identical to the Na-K geothermometer; however, the Li-Mg geothermometer [Kharaka and Mariner, 1989] results in 76°C–78°C for both cores. Silica geothermometers merely indicate near equilibrium with chalcedony at in situ temperature. The results for KC-27 are consistently 10°C–20°C lower than KC-14 suggesting a greater overprinting of low-temperature silicate diagenetic reactions at the former site. Overall the geothermometer analysis indicates that the fluid expelled at the hydrocarbon seeps reacted with the sediment within the 75°C–150°C temperature range. This maximum source temperature inferred for the fluids also corresponds to the upper limit of the seismogenic zone, assumed to occur 100°C–150°C [Hyndman et al., 1997].
In contrast to the other Marmara sites, the Western High mound fluids are brines of up to twice seawater concentration. The formation of hydrates surely drives this brine formation to some extent. For example, all components of the short core, KC-33, fall on a mixing line between pure water and a highly altered end-member nearly identical to the KC-14 end-member. Since this core had large amounts of gas hydrate, the composition is surely a result of water removal during hydrate formation and/or dissociation during core recovery. Cores KC-14 and KC-27 have nearly identical concentrations of K, Mg, and Li at their lowest depth, but differ by a factor of two in Ca and Sr concentration and by a factor of 1.3–1.4 in Cl and Na, precluding a sole gas hydrate cause for their concentrations. Halite dissolution from a buried evaporite can also be ruled out as the brine source as the Na/Cl ratio would approach 1 rather than the observed trend lower to 0.6. A potential deep source for fluids at this site is the water associated with the reservoir from which the gas and oil is seeping. Oilfield waters are known to have strong enrichments of a variety of elements and often form brines [Collins, 1975]. Analysis of the gas and hydrate at the mound site suggests that the gas source is the same as that of the Kuzey Marmara gas field [Bourry et al., 2009] which is located on a NE trending anticline that appears to be a continuation of the Sea of Marmara Western High. The source rock for this field is the Eocene Hamitabat Formation which consists of sandstone, shale, and conglomerate and the reservoir rock is the limestone Sogucak Formation [Hosgormez and Yalcin, 2005]. At the Kuzey Marmara field the Hamitabat formation is absent and the reservoir lies unconformably over the metamorphic basement which consists primarily of metamorphosed sedimentary rocks. Overlying formations consist primarily of shale, sandstone, siltstone, claystone, and conglomerates, and minor amounts of tuffite and is typically 1200–1500 m thick at the anticline crest [Hosgormez and Yalcin, 2005].
While downcore increases in Ca and Sr and decreases in B, K, and Na are compatible with low-temperature silicate diagenetic processes, the unusually high Li concentrations are incompatible with a low-temperature diagenetic process. Li concentrations in the mound cores reach 1100 μM, a factor of nearly 40 greater than bottom water, while all other Marmara cores exhibit lower Li concentrations downcore. As a general rule, low-temperature diagenetic processes tend to take up Li while high-temperature processes release Li [You and Gieskes, 2001]. For example high Li is a common occurrence in fluids from oceanic high-temperature hydrothermal vents (300°C–400°C) in which seawater reacts with basaltic crust [e.g., von Damm, 1990]. High Li concentrations have also been observed in the deepest oil reservoirs of the Gulf of Mexico Basin [Macpherson, 1989]. Macpherson points out that such concentrations cannot be explained by mineral diagenesis or salt dissolution and must result from metamorphic processes occurring much deeper. High Li concentrations are also a common occurrence associated with the dehydration of smectite clays; however, these are nearly always also associated with low chlorinities and high B concentrations [e.g., You et al., 1993] while, in our case, we have high chlorinities and lessening B concentrations downcore and an end-member B/Li molar ratio of 0.1–0.4. As pointed out by Tryon et al. , B/Li molar ratios of less than 1 are extremely rare in marine sedimentary environments and are almost always associated with interaction with igneous rocks. In the continental environment, fluid-rock interactions with pegmatite veins in the metamorphic basement could also produce such fluid compositions; however, these are unknown in our Sea of Marmara study area.
One possibility for a brine producing process that also produces fluids both high in Li and low in B is serpentinization. The serpentinization of ultramafic rocks takes up water and thus leaves a residual brine. Boron is also strongly taken up and Li leached from the rock during serpentinization [Agranier et al., 2007; Lee et al., 2008; Snyder et al., 2005; Vils et al., 2008]. This is indeed what is observed in deep boreholes in oceanic crust [Expedition 309 and 312 Scientists, 2006]. A potential source formation in the Marmara area is the IntraPontide suture zone. This accretionary complex is made up of serpentinized peridotites and igneous rocks, blueschist, radiolarian chert, and limestone [Okay and Tüysüz, 1999]. The Thrace Basin, presumably a fore-arc basin, was deposited over this complex and, in its northern part, on continental crust [Görür and Okay, 1996]. The western NAF roughly follows this suture zone; however, it is not known how this relationship continues beneath the Sea of Marmara. It is generally thought to intersect the Sea of Marmara in the east at the southern extension of the NAF, south of the Armutlu Peninsula, and in the west at the Ganos Fault [Okay and Tüysüz, 1999]. The suture may be offset from the southern NAF to the northern NAF (MMF) along the N–S trending West Black Sea Fault, a Cretaceous transform plate boundary, that intersects the Sea of Marmara in the north at the Central High. Thus, the MMF may cut this suture zone at depth at the Western High. We envision a deep hydrothermal system within the NAF and IntraPontide suture zone that feeds residual brines along fault zone conduits to overlying formations where it is further modified by mixing and diagenetic reactions to ultimately exit at the mounds.
An intriguing additional related possibility exists for a source rock for the suggested serpentinization: an upper mantle sliver at depth in the transform fault, (e.g., as has also been suggested for the San Andreas Fault [Ozacar and Zandt, 2009]). Preliminary reports of He isotopic compositions of >1.0 Ra at Marmara fluid seeps [Burnard and Bourlange, 2008] also suggest that fluids originating from subcrustal depths are expelled at the Marmara seafloor.
Regardless of its source, a third fluid is, therefore, required to model the major ion composition of the fluids at the mound sites, KC-14, KC-27, and KC-33. There was no evidence of hydrate in core KC-14 and its end-member composition intersects the KC-33 hydrate water removal line, so this intersection is our best model for a brine end-member (Figure 8). Chloride is the only element that is assumed conservative in this analysis. There is no evidence for Lake Marmara brackish water at theses sites; however, its presence would be difficult to resolve in a hydrate environment given its similarity to pure water. The composition of the fluid sampled at the southern mound hydrate site (KC-27) would correspond approximately to the composition of the brine source if up to 25%–30% of the water flowing through is taken up by local hydrate formation. In this model, however, the initial K, Mg, and Li concentrations at the southern mound would be about 45% lower and Ca, Sr, and Ba about 25% higher than at the northern mound prior to hydrate formation. The proximity of the two sites virtually assures that they have a common deep source; therefore, differences in composition must be attributable to shallower, local effects such as low-temperature diagenetic reactions. Preliminary high-resolution 3-D seismic images from the 2009 Marmara-DM cruise suggest that the fluid conduits for the two mounds are separate to at least the depth of our resolution, about 250–300 m. Even at flow rates of up to 10 times that modeled from the KC-27 profile, fluid transit times in the conduits would be at least a thousand years, allowing time for significantly different fluid evolutions. These results, combined with the geothermometer results, suggest a much greater level of low-temperature silicate diagenesis at the southern mound.
Figure 8. Plot of Cl versus Na for the mound site cores. KC-14 exhibits a simple mixing line between seawater and a brine end-member; KC-27 composition indicates a similar mixing of seawater and brine, overprinted by water removal during hydrate formation; and KC-33 falls on a mixing line between pure water and the brine indicating both removal of water by hydrate formation and addition of water during hydrate dissociation during core recovery and processing.
Download figure to PowerPoint
The sites exhibiting the clearest evidence for a deep aqueous fluid source are also sites of thermogenic gas expulsion [see Bourry et al., 2009] and are located on anticlinal ridges, which can presumably act as fluid traps. Furthermore, the two carbonate and hydrate mounds are located some distance form the fault zone (350 and 700 m away along the ridge axis) and thus are probably outside of the damage zone of the main strike-slip fault. Our hypothesis to explain this is that the bulk of the fluids originate from the anticlinal traps rather than the fault zone itself. These traps probably extend several kilometers along the anticlines but may not be continuous. Where the anticline is breached by the fault, the interaction of MMF induced fractures and extensional fractures associated with the anticline allow fluids to escape. Much of the driving force of the fluid expulsion is gas and oil buoyancy and this will tend to exit at local highs, while the MMF on the Western High lies in a valley. Ultimately fluid outlet sites are controlled by a combination of local topography, fracture locations, and driving forces.
In summary, the fluid samples collected during Marnaut suggest that part of the water expelled along the MMF at the ridges comes from at least thermogenic oil and gas generation depths. Anomalous He isotopic compositions [Burnard and Bourlange, 2008] and the low B/Li molar ratio brines are, so far, the main evidence that fluids originating from seismogenic or greater depths are expelled at the seafloor. Our samples also confirm that the fluids expelled along seafloor fault ruptures in the basins mostly originate at a shallow level in Pleistocene lacustrine sediments [Zitter et al., 2008].
4.3. Factors Influencing the SMI
Three cores taken at various locations away from fluid emission sites in the Çınarck and Central Basin display nearly ideal linear sulfate profiles down to the SMI at 2.5 m depth, consistent with a simple diffusion-reaction model [e.g., Halbach et al., 2004]. Sulfate concentration gradient is 11 mM/m in these cores. Considering that the tracer diffusion of sulfate in sediment is about half that of chloride [Iversen and Jørgensen, 1993] at about 1.5 m2/yr for a sediment porosity of 70% and a temperature of 14°C, sulfate consumption, and hence methane flux, is estimated to 120 mmol.m−2.yr−1, in agreement with earlier estimates from the Sea of Marmara [Çağatay et al., 2004],
It is a remarkable paradox that, at several sites where the presence of free gas in the sediment is inferred from geophysical and/or visual observations (e.g., KC-17, VT-2737 and VT-2740 in Tekirdağ Basin, KC-19 and VT-2742 in the Central Basin, KC-26 on the western high), the SMI lies at a deeper level than in the reference cores. Overall, a relationship can be drawn between a sharp downward decrease of sulfate concentration above the SMI (gradient generally exceeds the 11 mM/m background value) and the depth at which chloride decreases below the 595 mM threshold. This relationship between sulfate concentration and chloride, a presumably conservative element, is puzzling. In fact, most of the chloride profiles near fluid emission sites are nearly constant to some depth within the sediment (typically 1 to 5 m) and start decreasing (or increasing in the case of cores KC-27 and KC-14) at the SMI or just above it. These profiles cannot be modeled with pore water advection-diffusion models unless some penetration of seawater in the sediment is considered. Various processes can, in principle, account for this observation: salinity driven convection [Henry et al., 1992], bioirrigation [Wallmann et al., 1997], episodic methane emission [Tryon et al., 1999], eddy mixing in the wake of gas bubbles [Haeckel et al., 2007]. All these processes require effective permeabilities in the upper few meters of sediment that would be considered atypical for clay rich sediments. However, one can hypothesize that wherever streams of small (1–5 mm diameter) bubbles are seen escaping through the seafloor, mm size conduits are present to some unknown depth in the sediment and influence water and chemical fluxes. As the primary objective of this manuscript is to identify fluid sources and diagenetic reactions, we save further discussion of fluxes through the seafloor for future work. We here merely point out that variations of the sulfate gradient with depth in the sediment may be caused by a steady state pore water irrigation process, rather than by migration of the SMI after an earthquake, as hypothesized by Halbach et al. .