The life and death of a subglacial lake in West Antarctica

More than 600 subglacial lakes have been identified beneath the Antarctic ice sheet, including 140 “active” lakes that fill and drain, modulating subglacial meltwater transport from the continental interior to coastal ocean (Livingstone et al., 2022). Geophysical evidence shows that sediment deposition occurs at beds of subglacial lakes (e.g., Peters et al., 2008; Smith et al., 2018); this sediment record has been hypothesized to provide a record of subglacial processes and ice-sheet history (Bentley et al., 2011; Yan et al., 2022). However, Antarctic subglacial lakes are among the least accessible features on Earth, leaving this hypothesis largely untested: the only documented subglacial lake sediment cores were retrieved from Whillans Subglacial Lake (SLW) and consisted of structureless diamict indistinguishable from those retrieved from non-lacustrine subglacial environments (Kamb, 2001).

We present unequivocal evidence that lake sediment deposition occurs in Mercer Subglacial Lake (SLM), located 63 km south of SLW beneath the West Antarctic Ice Sheet. We demonstrate a tight coupling between the initiation of deposition in SLM and regional ice-stream variability, wherein stagnation of Kamb Ice Stream reorganized ice flow of adjacent ice streams (Catania et al., 2012) and led to SLM formation 180 ± 20 a before core recovery (BCR) based on our most likely sedimentation rate (SR) estimate. Subglacial lakes modulate ice-flow velocity (e.g., Stearns et al., 2008), impact location and timing of freshwater and nutrient-rich discharge into the Southern Ocean (Carter and Fricker, 2012), and interact with deeper, more expansive subglacial groundwater systems (Gustafson et al., 2022). Therefore, establishing the life cycle of these features remains crucial for quantifying transport of carbon (Vick-Majors et al., 2020), nutrients (Hawkings et al., 2020), and dissolved gases (Michaud et al., 2017) within and beyond the subglacial system.

The downstream confluence of Mercer and Whillans Ice Streams hosts a widespread, active subglacial hydrologic network (Fricker et al., 2007). Interconnected chains of subglacial lakes, including SLM and SLW (Fig. 1A), temporarily store water for months to decades before episodic drainage events release water downstream (e.g., Smith et al., 2009). Based on updated lake surface-elevation time series, SLM completes a fill-drain cycle every 4–6 a (Fig. 1B) with a clear connection to the activity of two lakes upstream: Conway Subglacial Lake (SLC) and Upper Conway Subglacial Lake (USLC) (Fig. 1C; Carter et al., 2013).

Ice flow in this area has switched between active streaming (>10² m a^–1) and stagnation (<10¹ m a^–1) throughout the last millennium (e.g., Catania et al., 2012), hypothesized to be driven by changes in subglacial water routing (Alley et al., 1994). Inferences of coupling between large-scale dynamics and subglacial water systems have been largely limited to analysis of advected surface features (e.g., Marsh et al., 2016). As WIS continues to slow at 1.9%– 2.7% a^–1 (Siegfried et al., 2016), resulting ice thickness perturbations are expected to alter the hydrologic system (e.g., Winberry et al., 2014). The time scales over which these changes occur are unknown because our observational record is limited to modern satellite missions since 2003 (Siegfried and Fricker, 2021).

We cleanly (i.e., minimizing all aspects of contamination and disturbance of the subglacial environment) accessed SLM in December 2018 by melting an ~0.4-meter-diameter borehole through 1087 m of overlying ice at 84.64029°S, 149.50134°W (Priscu et al., 2021). Our updated lake activity time series (2003–2021) based on multi-mission satellite altimetry and GPS data (Fig. 1B) indicated SLM had just entered a draining phase at time of access with surface height 14.9 ± 0.3 m above lowstand.

SLM borehole science operations included in situ observations, sediment coring, and installation of a fiber-optic mooring for long-term temperature observation. We focus on a 0.49-meter-long sediment core collected with a multicoring device modified for use in a narrow borehole (Michaud et al., 2016). See the Supplemental Material¹ for complete operations and analysis descriptions.

Camera imagery and temperature and conductivity profiling were combined to estimate a water-column thickness at SLM of 15 m (Priscu et al., 2021); this in situ measurement is consistent with our estimate of the drill site’s surface height above lowstand at time of access. SLM surface-elevation anomalies (±4 m) for the subsequent year were an order of magnitude larger than ice thickness change (+0.54 m) as indicated by autonomous phase-sensitive radio-echo sounding (ApRES) (Fig. 1D), suggesting our surface-elevation time series (Fig. 1B) can be interpreted as water-column thickness changes between the ice base and underlying sediments.

Two profiles of water-column temperature and conductivity collected during borehole science operations revealed it to be well mixed, with uniform temperature (-0.74 ± 0.01°C), uniform conductivity (287.2 ± 1.1 μS cm^–1), and negligible water velocity (<1 cm s^–1) (Priscu et al., 2021). Although we were unable to calibrate our fiber-optic distributed temperature sensing for precise temperature retrievals (see the Supplemental Material for methods), analysis of differential attenuation within the optical fiber indicated uniform water-column temperature through April 2019 and non-uniform water column temperature beginning May 2019 (Fig. 1E). This timing coincides with the switch from SLM draining to filling (Fig. 1B), suggesting a significant change in the water column as SLC drainage temporarily increased SLM’s water volume during the SLM drainage phase (Carter and Fricker, 2012; Fig. 1C).

We collected multicores from SLM that captured the sediment-water interface and recovered the first laminated sedimentary sequence from contemporary subglacial Antarctica. Core handling and transit disturbed fine-scale structure in the top 45 mm, but preservation of laminations in the lower 75 mm suggests the total laminated lake sediment thickness is 120 ± 2 mm. This low-density (1.24 ± 0.02 g cm^–3), clast-poor (0.13 ± 0.22 > 2 mm clasts per 25 cm³ sediment) mud (>80 vol%) overlays a massive clastrich muddy diamict (Fig. S1 in the Supplemental Material) characteristic of other Siple Coast subglacial sediment cores (Kamb, 2001).

Using computed tomography imaging, we identified planar interlaminated clay and silt couplets (N = 6–7 by visual inspection) as observed in grayscale values reflecting grainsize change (Reilly et al., 2017) (Fig. 2). Spectral analysis of grayscale values along vertical traces through the undisturbed portion identified significant (≥90% probability) cycles with greatest power at periods of 12 mm and 4.1 mm (Fig. 3A; Fig. S2; Table S1). We interpret the rhythmically laminated mud in our core (“rhythmites” hereafter) as sediment deposition in contemporary SLM.

We considered various subglacial processes, which act over time scales spanning several orders of magnitude, that could produce rhythmite deposition in SLM: (1) large-scale stick-slip motion of WIS (hourly); (2) subglacial lake fill-drain cycles (sub-annual to decadal); and (3) ice-stream stagnation-reactivation cycles (centennial). Sediments recovered from SLW (Hodson et al., 2016) and elsewhere across the region (Kamb, 2001) did not recover rhythmites, ruling out subglacial processes 1 and 3 given that these processes occur across the Siple Coast. We propose that the lake depositional environment changes significantly during SLM fill-drain cycles, including large changes in water-column thickness (Fig. 1B) and structure (Fig. 1E); these contrasts may provide the conditions needed to produce the observed rhythmites. Therefore, we interpret the rhythmites as an archive of past multi-year fill-drain events of SLM.

Physical interpretation of the rhythmite sequence requires understanding the sediment sources and depositional processes driving observed variability in SLM sedimentation. We considered three separate, interrelated processes affected by fill-drain dynamics (Fig. 4):

1. Source-sediment differences due to upstream subglacial lake dynamics, which would suggest SLM rhythmites represent dynamics of and pre-sorting in (Schroeder et al., 2019) SLC (Fig. 4A).

2. Drainage-channel erosion between lakes (Carter et al., 2017), which would suggest SLM rhythmites include remobilized subglacial diamict from between SLC and SLM (Fig. 4B).

3. Changes in SLM water-column thickness driving changes in water velocity, which would suggest SLM rhythmites represent local lake dynamics that cause removal of fine sediments near lowstand or a change in relative position of the sediment source (Fig. 4C).

We propose that coupled dynamics between SLC and SLM (i.e., sedimentation processes 1 and 2) were a key factor that drove rhythmite deposition because this interpretation also accounts for the lack of rhythmites in nearby SLW (Hodson et al., 2016), a more isolated active subglacial lake with no lake immediately upstream (Fig. 1A).

We estimated the time at which SLM formed by linking the satellite altimetry record (Fig. 1B) and sediment core laminae frequency (Fig. 3A). The contact at 120 ± 2 mm in the sediment core delineates a sharp contrast between lake sediments and underlying diamict (Fig. 2), marking the geologically rapid transition from a sub-ice-stream environment to the modern depositional lake. To estimate a long-term average SR for the lake sediment package, we employed statistical techniques developed for geological records that contain cyclicity associated with depositional processes (in this case, fill-drain cycles) but lack adequate chronological constraints (Meyers and Sageman, 2007). Our resulting distribution of possible (i.e., able to reject our null hypothesis) SRs, obtained through Monte Carlo simulation (Fig. 3B) and the observed 4–6 a fill-drain cycle frequency (see the Supplemental Material for methods), range from 0.49 ± 0.12 mm a^–1 to 2.3 ± 0.2 mm a^–1. This SR range suggests that SLM formed between 53 and 260 a BCR. Based on the most consistently identified long-term average SR of 0.68 ± 0.08 mm a^–1 (Fig. 3B), we estimate SLM formation occurred 180 ± 20 a BCR.

Timing of SLM formation coincides with a known large-scale ice-stream reorganization: Kamb Ice Stream stagnated and upper WIS narrowed through inward shear-margin migration 182 a BCR (Catania et al., 2012). Ongoing WIS thickening (e.g., Smith et al., 2020) and deceleration (e.g., Beem et al., 2014) imply that complete stagnation may occur in the coming decades due to basal freezing (e.g., Joughin et al., 2005). If Kamb Ice Stream thickening and deceleration enabled SLM initiation 180 a BCR, then continued WIS stagnation will likely force the demise of SLM as a subglacial lake: the dynamic, high-volume water system beneath WIS will devolve toward a low-volume lake system along ice-stream margins similar to that beneath the currently stagnated Kamb Ice Stream trunk (Kim et al., 2016). The entire life cycle of SLM—from its formation 180 a BCR to its future disappearance—would then span ~200–250 a. Quantifying this active subglacial lake time scale allows us to contextualize the periodicity of carbon and nutrient transfer between the subglacial and marine environments and provides us a metric by which subglacial hydrologic processes may be included in future models. Our work fingerprinting the genesis of subglacial rhythmites provides a means to connect modern subglacial processes to similarly laminated subglacial lake deposits found in the geologic record (e.g., Kuhn et al., 2017; Remmert et al., 2022).

Recovery and statistical analysis of lake sediments from contemporary subglacial Antarctica provided a constraint on the time scale of Antarctic subglacial lake activity prior to the satellite observational record. This result allowed us to infer a fundamental paleoglaciological connection between regional ice-stream variability and subglacial hydrology 180 ± 20 a BCR and provides strong evidence for a tight coupling between ice-stream stagnation processes and the onset of dynamic subglacial water systems in Antarctica. Future use of surface geophysical data combined with direct sediment sampling of additional subglacial lakes will further refine our process-scale understanding of critical biogeochemical cycles associated with the Antarctic ice sheet and surrounding Southern Ocean.

This work was supported by the U.S. National Science Foundation (grants 0636970, 0838885, 1327315, 1543347, 1543405, 1543441, 1543537, 1832109, 1832170, and 1836328) and NASA (grants 80NSSC20K1734 and 80NSSC21K0912). We thank the United States Antarctic Program; Kenn Borek Air; the New York Air National Guard; UNAVCO; the Center for Transformative Environmental Monitoring Programs (jointly operated by Oregon State University and the University of Nevada, Reno); Sue Cook; the University of Nebraska–Lincoln hot-water drill team; the European Space Agency; the U.S. National Snow and Ice Data Center; the Oregon State University Marine and Geology Repository, and Val Stanley; pre-WISSARD (Whillans Ice Stream Subglacial Access Research Drilling project), WISSARD, and SALSA (Subglacial Antarctic Lakes Scientific Access project) field crews; and two anonymous reviewers and the scientific editor for thoughtful comments that improved our manuscript.