Little is known about the habitability of cold liquid environments sealed off from the atmosphere, be it the subglacial lakes of Antarctica or beneath the icy shell of Europa . Located in the McMurdo dry valleys of Antarctica, Lake Vida has the thickest ice of any subaerial lake on Earth and is one of the few “ice-sealed” ecosystems known to support a diverse and active microbial population in a cold, anoxic, aphotic brine . The brine was first discovered 16 m below the surface of Lake Vida and is hypothesized to have been sealed from the atmosphere for several millennia . At present, the lake level (i.e., the ice surface) of Lake Vida is rising, which implies that the brine is progressively getting farther from the surface. In this paper, we present evidence for how and when Lake Vida formed to further understand the structure and evolution of the existing brine system beneath Lake Vida.

On most lakes in the dry valleys, the thickness of ice ranges from 3–6 m . This thickness is maintained by energy loss at the surface (conduction and ablation) and energy gained at the bottom of the ice cover (freezing) . Constant ablation at the ice surface and freezing at the bottom of these floating ice covers limit the maximum age of the surface ice to ∼ 5 years . However, the Lake Vida ice cover is at least partially grounded , so the ice does not turn over in the same way. Water that is flowing to the lake is trapped on the surface of the ice where it freezes and is later ablated or buried by subsequent ice buildup. In this way, the thick ice on Lake Vida may record past hydrological changes similar to a glacier; however, unlike a glacier, intermittent accumulation and ablation may lead to large discontinuities in the ice cover during prolonged cold/dry periods.

In Victoria Valley, it is theorized that the Lake Vida basin was occupied by a 200 m deep glacial lake 8600 14C yr BP , after which lake levels began to decline. This lake is inferred to have had a water column (likely with an ice cover) that permitted sedimentation and led to the lacustrine deposits seen on the landscape today. A deep water column would preclude the presence of bottom ice, and therefore it is improbable that any of the observed ice existed during this time. Therefore, the 27 m of ice currently on Lake Vida was formed subsequent to 8600 yr BP.

During mid- to late Holocene, it is possible that Lake Vida was influenced by events similar to the repeated lake level drawdowns and complete desiccation events recorded in lacustrine sediment cores and geochemical diffusion profiles of the large lakes of Taylor and Wright valleys. For instance, it is speculated that Taylor Valley underwent a valley-wide desiccation event at 1000–1200 yr BP , and Lake Fryxell had lowstands at 6400, 4700, 3800, and around 1600 yr BP . Lake Vanda, in Wright Valley, underwent a lowstand at 1200 yr BP , or prior to 2000 yr BP . In these closed basin lakes, desiccation is thought to be the result of climatic changes and not a result of large drainage events. If Lake Vida was influenced by similar climatic patterns, some or all of these events may be recorded in the ice cover.

The aim of this study is to reconstruct the history of the ice cover on Lake Vida. We examine the isotopic and ion geochemistry, sediment characteristics, and diatom composition of a 27 m ice core, as well as ground penetrating radar (GPR) profiles, to extrapolate the strata in single cores. Both radiocarbon dating and optically stimulated luminescence (OSL) dating are employed to establish the time of deposition of sediment layers, and ice core stratigraphy is used as a means of establishing periods of lake level drawdown.

Lake Vida (77∘23′ S, 161∘56′ E), situated in Victoria Valley, Antarctica, is one of the largest (6.8 km2) and highest (350 m a.s.l. – above sea level) lakes in the McMurdo dry valleys (Fig. 1). The lake is endorheic (closed basin) and receives inflow via streams originating from the Victoria Upper, Victoria Lower, and Clark glaciers. Annual precipitation is < 35 mm . Lake Vida occupies a unique climatological niche where summer temperatures can rise slightly above 0 ∘C to generate stream flow, yet unusually cold winters (compared to the other major valleys in the region) maintain a thick ice cover on the lake. From 1995 to 2000, the mean annual air temperature at Lake Vida (-27.4 ∘C) was 7 to 10 ∘C lower than valley bottom temperatures in Taylor and Wright Valleys, but mean summer temperatures were similar .

From drilling in 2010, it is known that the ice on Lake Vida extends to at least 27 m . A unique feature of Lake Vida is the presence of liquid brine within the ice cover, which infiltrates the drill hole at approximately 16 m and rises to 10.5 m below the surface. The brine is anoxic, with salinity of 195 g L-1 and temperature of 13.4 ∘C . It is hypothesized that the brine is contained within small fractures or channels in the ice and rises to 10.5 m when the confining layer of freshwater ice in the upper 16 m is breached.

An electric 15 cm diameter SideWinder drill was used to retrieve a 27 m and a 20 m long ice core located 6 m apart in the center of Lake Vida in November 2010. The 27 m ice core was split in half for archival purposes and subsampled into 5 cm lengths. Where recovery was incomplete for the 27 m core (between 16 and 20 m), the 20 m ice core was subsampled. Longitudinal thick-sections (∼ 0.5 cm thick) were cut from the ice core face and viewed under cross-polarized light for ice crystal fabric analyses. Subsamples were washed with deionized MilliQ water to remove possible brine contamination and allowed to completely melt for processing on a Dionex 1500 ion chromatograph for major ion analysis and a Los Gatos Research liquid water isotope analyzer for isotopic composition of 2H and 18O (analytical errors provided in Table 1). When necessary, samples were diluted to near seawater salinity prior to analysis. Isotopic values are reported with respect to the VSMOW international standard. Salinity is reported as the sum of concentration of total ions.

Sediment layers in the 27 m ice core were subsampled in duplicate 1 cm segments, which were freeze-dried or allowed to melt in order to extract pore water by centrifugation. For grain size analyses, 2 g samples were sieved through a 1000 µm sieve and pretreated with 30 % H2O2 for 18 h in a 50 ∘C water bath. Following pretreatment, samples were shaken following the addition of 1 mL of 30 mg L-1 Graham's salt (Na4P2O7) as a dispersant and analyzed on a Micromeritics Saturn DigiSizer 5200 particle size analyzer (detection limit 0.1–1000 µm). Sand/silt classifications are based on the Udden–Wentworth scale . Sediment layers were also subsampled to evaluate diatom assemblages and absolute abundance (valves/g dry weight) via light microscopy. Preparation methods followed standard techniques . A known mass of freeze-dried sediment was reacted with 10 % HCl and 10 % H2O2 to remove carbonates and organics. Abundance per gram was extrapolated from diatom counts on coverslips in a beaker of known area.

Freeze-dried samples were analyzed for total carbon (TC) and total inorganic carbon (TIC) with an elemental analyzer (Dimatec Co.). Total organic carbon was calculated from the difference in TC and TIC. Six samples with the highest organic carbon content (0.4–2.1 %) were chosen for radiocarbon dating of the organic fraction. Samples were prepared by removing carbonates and humic acids by acid–alkali–acid extraction prior to graphite conversion in an automated graphitization system. This required an overnight treatment with 1 % HCl to remove carbonates, followed by 4-hour humic acid extraction with 1 % NaOH and a secondary overnight treatment of 1 % HCl to eliminate any CO2 that may have been absorbed during the NaOH treatment . Two samples at 23.90 and 26.43 m with 1.3–2.3 % TIC were selected for radiocarbon dating of carbonates and were treated with dilute sulphuric acid. Two 5 cm long sediment sections at 21.51 and 25.54 m were cut by band saw under red light for OSL dating (UIC Luminescence Dating Research Laboratory) and remained frozen until analysis. The advantage of frozen sediment for OSL dating is that the water content is expected to have remained constant since freezing took place . OSL dating of quartz grains from sediment layers was performed using single aliquot regeneration protocols .

To confirm the continuity of horizons and sediment layers noted in the ice cores, 55 km of GPR transects was recorded over the surface of Lake Vida in 2010 (Fig. 1) using a GSSI SIR-3000 acquisition unit equipped with a 400 MHz antenna. Transects were recorded at 400 ns time range and 2048 16-bit samples per trace, with five manual gain points at -20, 0, 25, 30, and 50. A dielectric constant of 3.15 was initially chosen for depth calibration, but this was altered based on known characteristics of the ice cores. In post-processing, radar profiles were triple stacked and passed through a 200 and 500 MHz triangle FIR filter to remove high- and low-frequency noise. Lake levels were annually surveyed from benchmarks tied into historical optical survey transects conducted by New Zealand Antarctic Program and were recorded in m a.s.l.

A water column of brine was not encountered 20 m below the surface of Lake Vida as previously hypothesized . Rather, wet ice and sediment continued below this depth. After four thick (> 10 cm) sediment layers were encountered below 21 m, the drill became lodged in what was almost certainly a sediment-rich layer (based on the slow progress of drilling) at 27.01 m. The last sample obtained was an ice layer from 26.62 to 26.81 m.

There was an overall trend of increasing salinity in the ice core with depth as well as a shift at 23 m from Cl- to SO42- as the dominant anion and an accompanying decrease in Mg2+ (Fig. 2). Below 21 m, the ice salinity was variable from < 1 to 34 g L-1. During extraction, all cores below 16 m depth were in contact with the brine. If samples were contaminated by brine in the drill hole, we would expect the percentage of major anions/cations and the ratio of Na+ : Cl- and Na+ : SO42- in the ice to be similar to the brine (Fig. 2). As the ice chemistry is distinct from the brine, we conclude the ice was not substantially contaminated by brine in the drill hole. The fabric of the ice also changed with depth, from large individual ice crystals with c-axes oriented upwards, which is typical of freshwater ice, to ice composed of randomly oriented small crystals that appear to have recrystallized over time (Fig. 3).

Throughout the cores, there are many small pockets of sediments and thin sediment layers (Fig. 4). All sediments sampled from the ice core were predominately sand (grain size: 62.5 to 2000 µm), with only 4 out of 79 samples having a mean grain size < 100 µm or a percentage of silt and clay (< 62.5 µm) > 6 % of the total volume. In the 27 m core, the four thick sediment layers below 21 m will be referred to by their depth in the core: SL21.62, SL22.88, SL25.59, and SL26.28 (Fig. 3). These layers had water contents < 10 % and total thicknesses of 19, 15, 11, and 19 cm, respectively. At the base of SL26.28, mean grain size began to decrease with a concomitant increase in TC (Fig. 5). Microscopy of the lower sediment layers and ice revealed abundant diatom frustules of Luticola gaussi and the genera Pinnularia, specifically P. deltaica and P. quaternaria, only in SL26.28. These species are commonplace in the sediments and water column of Taylor Valley lakes, streams, and cryoconite holes . L. gaussi is considered a freshwater species and cosmopolitan to the Antarctic diatom flora , as are P. deltaica and P. quaternaria. However, no studies have been done to specifically test the salinity tolerance of these species.

The upper sediment layers noted in two ice cores retrieved in 2010 correspond to those cored in 1996 and 2005 (Fig. 4). The radiocarbon dates vary significantly, especially in the upper ice where there is no correlation between age and depth. The dates from SL26.28 do show increasing age with depth (4909 ± 46 to 6300 ± 49 14C yr BP). It is noted that δ13C values are unreliable and not reported due to high fractionation during accelerator mass spectrometry measurements.

OSL samples showed no evidence of exposure to sunlight during collection and returned dates of 320 ± 40 and 1200 ± 100 yr BP for SL21.62 and SL25.59, respectively, dates which are younger than the radiocarbon ages from the respective horizons.

Stable isotope (δ18O and δ2H) values of both the ice and sediment pore water fell on or below the local meteoric water line but on a slope consistent with the sublimation of ice and isotopic values reported around the dry valleys (Fig. 6). Most of the ice and sediment pore water samples have δ18O between -25 and -32 ‰. There are, however, two notable exceptions. The first is the sample at 12.75 m that is significantly depleted in 18O versus all other ice samples. The second is the pore water of the four thick sediment layers below 21 m, as well as some thinner sediment layers above, that are relatively enriched in 18O in comparison to the ice.

GPR profiles distinctly map the edge of the lake basin until an impenetrable basal reflector around 21 m (Fig. 7). Along the basin edges there are features resembling ancient terraces, especially at 8 m (Fig. 7). From 8 to 12 m, synchronous, wavy reflectors are spaced approximately 1 m apart. In all profiles, ice and sediment layers appear to be continuous across the lake.

The low ion concentration, absence of large sediment layers, and clear GPR returns in the top 8 m of ice suggest that the upper ice has formed recently under a positive water balance. The level of Lake Vida has risen 3.5 m in the last 40 years and has a hydrologic history similar to Lake Bonney (Fig. 8), which has been documented to have risen ∼ 16 m from 1903 to 2010 . If a linear extrapolation is applied to the Vida record based on the correlation of volumetric change to Lake Bonney, the surface of Lake Vida would have risen ∼ 7.7 m from 1903 to 2010 when our cores were collected. Therefore, the 8 m contour in Fig. 1 may be an approximate representation of the lake shore in 1903. This is an indication of the rapidity with which the level of Lake Vida can change over time.

Our analyses point toward the formation of sediment layers in the Lake Vida ice cover from the accumulation of sediment entrained in the ice cover accumulating during periods of ice ablation and lake level drawdown. The capacity of Lake Vida to integrate watershed processes presents a fundamental framework for understanding hydrological and climatological shifts over time.

The inconsistency in radiocarbon dates makes a full reconstruction of the history of the Lake Vida ice cover challenging. However, several conclusions are gained from this ice/sediment record:

A hydrologically variable climate is not unique to recent times. Lake Vida has experienced major drawdowns that led to the accumulation of four thick sediment layers in the lower ice cover. These drawdowns may have occurred as early as 6300 14C yr BP, but OSL ages and a presumed reservoir effect in radiocarbon ages suggest these events were likely constrained to the last 1–3 millennia.

Lake Vida represents a unique lacustrine system that has recorded a hydrologic history in the growing ice cover. As climate is projected to change , Lake Vida may provide an ideal environment for tracking the influence of climate on hydrology in the dry valleys.