Solute sources and geochemical processes in Subglacial Lake Whillans, West Antarctica

National Science Foundation (NSF, grants OPP-0838933, 1346250, and 1439774); Italian National Antarctic Program (Barbante); graduate fellowships from the NSF Integrative Graduate Education and Research Traineeship program (0654336) and NSF Center for Dark Energy Biosphere Investigations; dissertation grant from the American Association of University Women


INTRODUCTION
Research over the past two decades has demonstrated abundant water beneath the Antarctic ice sheets, with subglacial water volume estimates of 10 4 -10 5 km 3 (Priscu et al., 2008). Water-rock interactions in glaciated catchments provide abundant reaction sites for chemical and biological weathering, which generates dissolved aqueous species (Tranter, 2004). Water beneath the West Antarctic Ice Sheet (WAIS) resides and flows in hydrologic networks of streams, lakes, and saturated sediment, leading to prolonged water-rock interactions and solute generation, with the waters ultimately draining to the ocean (Fricker et al., 2007;Carter and Fricker, 2012). These subglacial aqueous environments represent a source of suspended and dissolved biotic and abiotic components and fresh water to the Southern Ocean with potential impacts on ocean productivity, geochemistry, and circulation (Priscu et al., 2008;Statham et al., 2008).
Active subglacial lakes, such as those along the Siple Coast of West Antarctica, are an important component of the subglacial hydrologic system because they form reservoirs along the hydraulic drainage path to the Ross Sea. Subglacial Lake Whillans (SLW) and other lakes in the Siple Coast region are characterized as "active lakes" (Smith et al., 2009). The filling phase of the lake hydrological cycle occurs over ~2-4 yr, slowly increasing the depth and volume of the lake (Siegfried et al., 2014). Drainage events punctuate the filling cycle, moving large (~0.15 km 3 , in the case of SLW) volumes of water to the grounded margin of the Whillans Ice Stream and the Ross Sea (Carter and Fricker, 2012) on sub-decadal time frames. However, sedimentwater interaction times beneath the Whillans Ice Stream can be thousands of years (Christoffersen et al., 2014). This contrasts with the hydrologic conditions beneath valley glaciers and marginal regions of the Greenland Ice Sheet where chemical weathering processes have been more widely studied (Tranter et al., 2002;Tranter, 2004;Wadham et al., 2010). In these (non-Antarctic) subglacial systems, the main water source is surficial melt water, and subglacial water rock-contact times are typically much shorter, hours to months (Skidmore and Sharp, 1999;Tranter et al., 2002;Wadham et al., 2010). Another important distinction between the non-Antarctic subglacial systems and SLW is that during interglacial periods, portions of the WAIS basin, including the present-day location of SLW, were periodically inundated by ocean water, and marine sediments accumulated Studinger et al., 2001). There may be a biogeochemical legacy from these deeper marine sediments that has yet to be explored, given the potential for chemical and redox gradients to exist, for example, with more anoxic and solute-rich sediment pore waters at depth, underlying shallow waters that are more dilute and oxic. Here we present the first detailed description of solute sources and geochemical processes for the water column and sediment pore waters from an Antarctic subglacial lake.

METHODS
Samples of water and sediment were collected from SLW (84.240°S, 153.694°W), 801 m beneath the WAIS, which was accessed using a clean, hot-water drill (Priscu et al., 2013). Water samples were collected from the middle of the ~2.2 m water column of SLW using a 10 L Niskin sampling bottle, and a 38 cm sediment core was retrieved with a multicorer (60 cm long × 6 cm diameter). Lake water samples for dissolved major ions and trace elements were filtered through 0.4 µm polycarbonate filters and 0.2 µm PTFE (polytetrafluoroethylene) filters, respectively. Samples of sediment pore water were extracted at 2 cm intervals using Rhizons (www.rhizosphere.com/rhizons) equipped with 0.2 µm pore-size filters, and distributed into cleaned vials. Samples for major ion and trace metal determinations were stored frozen before analysis via ion chromatography and inductively coupled plasma-sector field mass spectrometry (ICP-SFMS), respectively (Turetta et al., 2004;Montross et al., 2013). Stable isotopes (H and O) of the water were measured on unfrozen samples via cavity ring-down spectrometry. Silicon was determined colorimetrically (Iler, 1979 GSA Data Repository 1 for full description of the methods.

SLW Seawater Component
The lake water had a d 18 O (H 2 O) value of -38. 0‰ (Christner et al., 2014) with pore-water values increasing down core to -36.5‰ at 37 cm depth, the core bottom (Fig. 1A). dD (H 2 O) values showed a similar trend toward more positive values at depth (Fig. 1A). Lake water dD (H 2 O) values were comparable to those of meteoric ice from other Antarctic outlet glaciers (Souchez et al., 2004). A two-component mixing model of d 18 O (H 2 O) was used to determine the percentage of glacial melt water and seawater in the SLW lake and pore waters (Phillips and Gregg, 2001), with d 18 O (H 2 O) of -39‰, measured from deep glacial ice upstream of SLW (Vogel, 2004), as the glacial ice end member and 0‰ as the seawater end member. This analysis indicated that the lake water was composed of 97.4% glacial melt and 2.6% seawater, with the proportion of seawater increasing down core to a maximum of 6.4% at the core bottom, and the balance, 93.6% from glacial melt. Chloride concentrations also increased down core from 3.5 meq L -1 in the lake to 22.5 meq L -1 at the core bottom, equivalent to 4.1% of seawater Clconcentration (Fig. 1B). The Clconcentration profile is consistent with the stable isotope data indicating that the pore waters and solutes are partially derived from seawater. Chloride is a conservative tracer, and the linearity of the porewater Clprofile suggests that this system is in steady state, relative to the upward diffusion of Clfrom a more concentrated source at depth (>38 cm). The seawater influence is unlikely to be contemporary, as tidally driven pumping of modern, ice-shelf cavity seawater is limited to a distance of ~10 km upstream from the grounding line (Walker et al., 2013), whereas SLW is ~100 km upstream. The seawater signal measured at SLW is likely due to the lasting influence of past seawater inundation and deposition of marine sediments. The other major ions (Na + , K + , Mg 2+ , Ca 2+ , SO 4 2-, and HCO 3 -) also show trends similar to that of Clwith increased concentrations at depth and profiles that indicate upward diffusion from a more concentrated source (>38 cm depth) ( Fig. 1B; Table DR1 in the Data Repository). The Br -:Clin the lake waters was close to that of seawater, and we assumed all Clcame from a seawater source (Christner et al., 2014). The seawater component of the lake and pore waters was calculated using the Clconcentration and the ratio of each ion to Clin seawater (Holland, 1978). The geochemistry of SLW and its sediment pore waters are atypical for subglacial systems as there is a significant component of seawater-sourced solute (Fig. 2). The seawater component of total solute in SLW sediment pore water increases from 57% in the lake to 77% at 37 cm depth, where total solute concentrations are ~13 meq L -1 and ~67 meq L -1 , respectively.
The seawater component has a notable influence on SLW geochemistry and distinguishes SLW from subglacial waters of other Antarctic systems, and non-Antarctic valley and outlet glaciers ( Fig. 2A), with the exception being the subglacial outburst waters near Casey Station, which have 34% seawater solute (Goodwin, 1988). The SLW waters exhibit a compositional shift after subtraction of the seawater component (Fig. 2). Other subglacial pore-water samples from the Siple Coast (i.e., Kamb Ice Stream and Bind schadler Ice Stream) are not influenced by the seawater correction as only 3% of the solute is of marine origin Fig. 2).

SLW Solute from Crustal Weathering
We define whole water as the total solute load measured in the sample. The crustally derived, non-seawater component is defined as the total solute load with the seawater component subtracted. Following subtraction of the seawater component, significant crustally derived K + and Na + were evident in the SLW lake water and throughout the pore-water profile, likely from silicate mineral weathering (Fig. 1C), when compared to other subglacial waters ( Fig. 2A). This is consistent with the mineralogical data on the sediments; e.g., the 63-125 µm size fraction was primarily composed of feldspar (35% of grains analyzed; n = 206; Table DR2). Aqueous Ca 2+ and Mg 2+ are typically sourced mainly from silicate or carbonate weathering, with a greater proportion of Ca 2+ from carbonates than for Mg 2+ (Meybeck, 1987). The concentrations for Mg 2+ are negative after applying the seawater correction, indicating a Mg 2+ sink from solution exists in the sediment. Cation exchange reactions on clay minerals are the most likely sink given the high proportion (50% by dry weight; Table DR3) of silt-and clay-sized particles in SLW sediments, which is similar to the till composition found upstream of SLW .
Clay minerals are prevalent in SLW sediments and provide cation exchange reaction sites that can modify Na + , K + , Mg 2+ , and Ca 2+ ion concentrations derived from mineral dissolution and seawater sources. Ion exchange experiments on SLW sediments using a NaCl exchange solution with a similar concentration (20 meq L -1 ) to that at the bottom of the sediment core resulted in 0.9 meq g -1 Ca 2+ and 0.6 meq g -1 Mg 2+ desorbing from clay minerals. The exchange experiments demonstrate that clay minerals are a sink for Ca 2+ and Mg 2+ from SLW pore waters.

Weathering Regime in Subglacial Lake Whillans
Calcium is the dominant cation in alpine subglacial waters irrespective of lithology due to the rapid dissolution kinetics of carbonates (Anderson et al., 1997). By contrast, crustally derived Ca 2+ is not a dominant ion in SLW, as demonstrated by the low Ca 2+ :Scation molar ratio (0.04-0.13) compared to the world mean for glacial and non-glacial catchments (~0.45; Anderson et al., 1997). Na + is the dominant crustally derived cation in SLW lake and pore waters. This is also true for samples from Antarctic subglacial waters beneath the Kamb Ice Stream, Bindschadler Ice Stream, and Casey Station which also have low crustally derived Ca 2+ :Scation ratios: 0.22, 0.17, and 0.09, respectively . The low Ca 2+ :Scation ratios from Antarctic subglacial aqueous samples are generated through a different weathering regime from Ca 2+ -dominated Alpine glacial systems. These lower Ca 2+ :Scation ratios are most likely due to increased silicate weathering and ion exchange reactions during longer water-rock interaction times in Antarctic subglacial environments. Long water residence time produces proportionately greater amounts of crustally sourced Na + and K + to the subglacial waters (Wadham et al., 2010). Clay minerals in SLW sediments are also a product of longterm silicate weathering and are not observed in valley glacier tills characterized by shorter residence times (e.g., Griggs, 2013). The Mg 2+ and Ca 2+ concentrations in SLW waters are lower than one would expect from such long residence-time waters, and the deficit is assumed to be due to cation exchange on clay minerals, leading in part to the low Ca 2+ :Scation ratio.
Silicate weathering can occur abiotically, but given the presence of an active microbial community in SLW (Christner et al., 2014) and previous research demonstrating the key role of microbes in weathering of glacial sediments (Montross et al., 2013), biotic weathering may also occur. CO 2 released from heterotrophic activity in the SLW water column (Christner et al., 2014) will dissolve and produce carbonic acid, which is responsible for indirect weathering of minerals (Montross et al., 2013) and contributes in part to the HCO 3 in the lake and sediment pore waters. The balance of HCO 3 may come from carbonate dissolution and possibly hydrothermal sources given the high geothermal heat flux in the region (Fisher et al., 2015). Direct weathering of sulfur-containing minerals (i.e., pyrite), which are present in SLW sediments (Fig. DR1 in the Data Repository), through sulfide oxidation produces protons and dissolved SO 4 2- (Fig. 1B). Microbial sulfide oxidation is inferred from the excess of crustally derived SO 4 2with respect to Clin the top 15 cm of the core (Fig. 1D) and has been demonstrated in other glacial sediments (Mitchell et al., 2013;Boyd et al., 2014). Microbial community and functional gene analysis of SLW and Kamb Ice Stream sediment supports the possibility of microbial pyrite oxidation, as microbial phylotypes and sulfide-oxidizing functional genes related to sulfide-oxidizing bacteria are present and abundant in these environments (Lanoil et al., 2009;Christner et al., 2014;Purcell et al., 2014). These microbially driven sulfide oxidizing processes are likely the source of crustally derived, non-seawater sulfate in SLW.
Microbial sulfide oxidation is most energetically favorable in the presence of O 2 . To estimate the depth below which sulfide oxidation is unlikely to occur, we estimated the reductionoxidation (redox) state of the sediment using the charge of the element vanadium (V) and thus its aqueous concentration. Redox-sensitive elements, such as V, are useful indicators of the redox state of the sediments (Morford and Emerson, 1999). As dissolved O 2 concentrations decrease, vanadate (H 2 VO 4 -) is reduced to vanadyl [V(IV)O 2+ ] and V is removed via precipitation (Morford and Emerson, 1999). Therefore, the decrease in V concentrations to below detection at depths lower than 15 cm in the SLW sediments implies decreasing O 2 concentrations with depth (Fig. 1E). The O 2 -consuming processes are likely microbial, involving either heterotrophic or lithotrophic activity such as sulfide and ammonia oxidation, processes which are thought to be important in SLW (Christner et al., 2014). The silicon profile is similar to that of V and shows that silicon concentrations are highest in the surficial (0-2 cm) sediments (Fig.  1E). Increased microbial activity in the surficial sediments may be due to the presence of oxidizing conditions, unlike those at depth. Microbial activity in the sediments is therefore likely to be important in liberating solutes via both direct and indirect biotic weathering processes.

CONCLUSIONS
Crustally derived solute in SLW is generated primarily from the weathering of silicate minerals, a process that is likely microbially mediated. Ion exchange reactions with clay minerals also modify the proportions of cations in solution. Seawater is also an important component of SLW solute, in contrast to Kamb Ice Stream and Bindschadler Ice Stream pore waters and subglacial waters from other non-Antarctic systems. The major ion profiles all indicate a more concentrated solute source in pore water of sediments below the base of our 38 cm core, which we were unable to sample. These deeper pore waters likely have an increased seawater component, resulting from marine inundation of the SLW region when the WAIS was less extensive. SLW represents a subglacial system where www.gsapubs.org | Volume 44 | Number 5 | GEOLOGY contemporary chemical and biological weathering processes are active but seawater solute is a component of the water that is ultimately cycled back to the Southern Ocean during lake drainage.

ACKNOWLEDGMENTS
The Whillans Ice Stream Subglacial Access Research Drilling (WISSARD) project was funded by the National Science Foundation (NSF, grants OPP-0838933, 1346250, and 1439774). We thank the WISSARD science team for assistance in sample collection; WISSARD science team members can be found at wissard.org. Partial support was provided by the Italian National Antarctic Program (Barbante); graduate fellowships from the NSF Integrative Graduate Education and Research Traineeship program (0654336) and NSF Center for Dark Energy Biosphere Investigations (Michaud); and a dissertation grant from the American Association of University Women (Vick-Majors). The University of Nebraska-Lincoln provided hot-water drill support. A. Schauer, B. vanHeuvel, and E. Steig helped with stable isotopic analyses, and C. Allen assisted with major ion analysis. M. Tranter provided valuable discussion and input. We thank the three reviewers for their constructive criticism which improved the manuscript. This is C-DEBI contribution #321.