Dynamic subglacial meltwater history archived in Antarctic subglacial lake sediments Open Access

Direct observations and geologic records from beneath Antarctic ice streams are critical to understanding the development of subglacial hydrologic systems and their influence on ice stream dynamics but archives remain spatially and temporally limited. At least 675 subglacial lakes have been identified beneath the Antarctic ice sheet (Livingstone et al., 2022), many of which are “active,” filling and draining on sub-decadal time scales and transporting substantial volumes of water through a connected hydrologic system (Fricker et al., 2007; Siegfried and Fricker, 2018, 2021). Sediments recovered from lakes beneath ice streams can provide critical information on subglacial processes and ice sheet history (Bentley et al., 2011; Yan et al., 2022; Siegfried et al., 2023).

Ross Ice Shelf, the largest in Antarctica, is fed by several fast-flowing ice streams that flow over many subglacial lakes (Fig. 1). Water and soft, deformable, saturated sediments underlie these ice streams and lubricate the bed to facilitate ice flow despite low driving stresses (e.g., Alley et al., 1986, 1987; Engelhardt and Kamb, 1997; Kamb, 2001). Over decadal to millennial time scales, these ice streams exhibit stagnation-reactivation cycles (Retzlaff and Bentley, 1993; Catania et al., 2006), characterized by large velocity changes (Beem et al., 2014; Hulbe et al., 2016; Siegfried et al., 2016) and changes in their flow direction (Conway et al., 2002; Hulbe and Fahnestock, 2007; Catania et al., 2012; Verboncoeur et al., 2025), all of which can influence ice sheet mass balance (Seroussi et al., 2023). Almost half of the net ice loss from West Antarctic Ice Sheet is offset by the positive mass balance along the Siple Coast (Pritchard et al., 2009; Rignot et al., 2019), which is partially attributed to the stagnation of Kamb Ice Stream ~180 years ago and the ongoing deceleration of Whillans Ice Stream (Joughin and Tulaczyk, 2002; Beem et al., 2014). Because ice-flow variability is affected by the generation and routing of subglacial meltwater (e.g., Alley, 1989; Anandakrishnan and Alley, 1997; Christoffersen et al., 2010; Carter et al., 2013; Elsworth and Suckale, 2016; Siegfried et al., 2016), it is predicted that changes in the distribution of basal meltwater could trigger future ice stream reorganization and reduce the positive mass balance of the Siple Coast sector within the near future (tens to hundreds of years; Bougamont et al., 2015; Seroussi et al., 2023; Siegfried et al., 2023). Observations on sub-decadal scales have advanced our understanding of contemporary Antarctic subglacial hydrology (Livingstone et al., 2022), but few records on longer time scales (decadal to millennial scales) exist from the interior of the West Antarctic Ice Sheet.

More evidence for paleo-subglacial lake and relict meltwater deposits has been observed from the deglaciated continental shelf in the Ross Sea (Simkins et al., 2017; Prothro et al., 2018) as well as offshore from Pine Island Glacier (Witus et al., 2014; Kuhn et al., 2017; Smith et al., 2017; Lepp et al., 2022), than directly from current, active subglacial lakes. Sediments from Whillans Subglacial Lake (SLW; Fig. 1), one of only two (Tulaczyk et al., 2014; Priscu et al., 2021) subglacial lakes accessed to date, were composed of massive diamict and interpreted as subglacial till deposited by Whillans Ice Stream through direct ice-bed interaction (Hodson et al., 2016). Thus, no record of in situ subglacial lake sedimentation or fluvial erosional surfaces was found in the SLW sediments (Hodson et al., 2016), in contrast to the laminated sediments from Mercer Subglacial Lake (SLM), described in Siegfried et al. (2023). Here we describe the complete stratigraphic distribution of lithofacies (hereinafter referred to as facies) of a 2.06 m composite sediment sequence recovered from SLM, located beneath Mercer Ice Stream, West Antarctica (Fig. 1), to expand and contextualize the description of the laminated sediments capping the sequence and to reconstruct the evolution of basal conditions and subglacial meltwater activity.

During the 2018–2019 field season, SLM was cleanly accessed using a hot water drill to melt an ~0.4 m diameter borehole through overlying 1087 m ice into the 15 m deep lake water cavity (Priscu et al., 2013; Michaud et al., 2020; Priscu et al., 2021). Short sediment cores (0.06 m diameter) were collected using a three-barrel UWITEC multicoring device (Rosenheim et al., 2023; Table S1 in the Supplemental Material¹). Cores from the first multicorer deployment are described here (01UW-A, 01UW-B, and 01UW-C). Two longer cores (1 m and 1.7 m long; 0.10 m diameter) were collected using a free-fall gravity corer (Rosenheim et al., 2023) and are hereafter referred to as 01FF and 02FF (Table S1). Downstream ice movement was 0.64 m day⁻¹ and created an ~5 m of lateral offset from the first to the last cores collected, which limited disturbance between coring locations (Priscu et al., 2021). Magnetic susceptibility was measured on all sediment cores using a Bartington MS3 magnetic susceptibility meter. Cores 01UW-A, 01UW-C, 01FF, and 02FF were stored and shipped upright at 4 °C to the Marine and Geology Core Repository at Oregon State University, Corvallis, Oregon, USA, following protocols described in Venturelli et al. (2021).

X-ray computed tomography (CT) scans were performed on unsplit cores using a Toshiba Aquilon 64 Slide Medical CT Scanner. The resulting high-resolution CT images were processed using SedCT MATLAB to produce a coronal slice through the center of each core at an effective pixel resolution of 0.5 × 0.5 mm and a vertical interval of 0.25 mm (Reilly et al., 2017). The quantitative data were stored in pixels as relative grayscale values or Hounsfield units (HU; Hounsfield, 1973). HU are defined as the attenuation coefficient of the imaged material. By this definition, the HU value of water is 0 and the HU value of air is −1000. HU values are dependent on sediment density (Hounsfield, 1973; Reilly et al., 2017). CT HU values were converted to wet bulk density (WBD) following methods of Reilly et al. (2017) for fine-grained glacial sediments using the equation:

The coarse clastic material was quantified for each sediment core using the CT scans. An automated image analysis code counted the number of clasts greater than 2 mm (but less than core barrel diameters of 60 mm and 100 mm; the clast counts are hereafter reported as >2 mm) per coronal slice using the algorithm from Reilly et al. (2019). Clast counts were normalized to 10 cm³ volumes based on core diameter and coronal slice thickness of 2 cm. For example, a 2-cm-thick slice of the 6-cm-diameter core would have a 5.66 × 10 cm³ volume. This provided an automated and objective method but should be considered as a clast index rather than an absolute count of grains >2 mm (Reilly et al., 2019). Clast counts and visual confirmation with CT images were used to evaluate clast volume percent.

Following CT scanning, cores were split, visually described using the Moncrieff classification scheme for poorly sorted terrigenous clastic sediments containing gravel (Moncrieff, 1989) and the Munsell color system, and then photographed using a GEOTEK Line Scan Camera at Oregon State University. Gamma ray attenuation bulk density (GD) was logged for split cores using a GEOTEK Multi-Sensor Core Logger (MSCL) at 1 cm intervals. GD was calculated using the second-order polynomial equation:

where H is core thickness, GA is MSCL-measured gamma attenuation, and A, B, and C are coefficients. Sediment porosity values were calculated using the equation:

where FP is fractional porosity, MGD is mineral grain density (g cm⁻³; used 2.75), GD is gamma density as determined by the gamma density processing panel, and WD is fluid phase density (g cm⁻³; assumed to be 1.00). We have assumed that (1) The sediment is fully saturated with water, (2) mineral grain density is constant, and (3) fluid density is constant. Absolute porosity values presented here are likely higher than in situ conditions due to the production of air-filled voids in the cores created by degassing as the cores equilibrated at the surface upon collection. Thus, changes in MSCL porosity downcore should be treated relatively, assuming degassing was constant throughout the cores.

X-ray fluorescence (XRF) core scanning was carried out on the split core surface of archive halves at 0.5 mm resolution using an ITRAX XRF Core Scanner at the Oregon State University Marine and Geology Repository with generator settings at 10 kV, 30 kV, and 50 kV. Here we use elements with robust signal quality and typical environmental indicators in Antarctic subglacial sediments (e.g., Ca, Fe, K, Ti; Monien et al., 2012). Sediment shear strength was measured with a handheld Torvane shear tester on split cores at discrete intervals down the core.

A composite 2.06 m record was obtained by correlating all cores collected within ~5 m of each other using multiple physical parameters and standard splicing methods (Hagelberg et al., 1992; Lisiecki and Herbert, 2007; Venturelli et al., 2023; Rosenheim et al., 2023). The core top of 01UW-A from our first deployment of the multicorer represents the sediment-water interface. The composite depth record was created by correlating cores using magnetic susceptibility field measurements and similarities in lithology, with lengths of 109 cm and 28 cm added to the measured core depths of 01FF and 02FF, respectively (Fig. S1). The degree of overlap between 02FF and the shorter multicores was determined by the absence of the upper 28 cm of sediment from the multicores in the 02FF core. The 28 cm of softer sediment, which includes 11.5 cm (LU1) and the uppermost 16.5 cm of LU2; (see Results section for a description of sediment properties) was lost during core collection due to overpenetration of the core barrel (Rosenheim et al., 2023). Core 01FF was tied to 02FF using common lithologies (Fig. S1). No indication of stretching or compaction was observed between cores, and offset lengths were used to correlate them.

Particle size distribution of the sediment matrix (<2 mm, i.e., sand-silt-clay fraction) was measured at 10 cm intervals over the entire core. Approximately 1 g of sediment was transferred into 50 ml plastic centrifuge tubes containing 30 ml of 2.5 g L⁻¹ sodium hexametaphosphate solution as a dispersant agent modified from Sperazza et al. (2004). Samples were subjected to gyratory shaking at 200 rpm for at least 10 h and then wet-sieved at 2 mm to remove gravel, before analysis on a Malvern Mastersizer 3000 (Sperazza et al., 2004). The Mastersizer 3000 instrument offers accuracy within 0.6% and precision within 0.5%. Particles were assumed to be non-spherical in shape. Individual aliquots (~10 mL of sediment slurry) were removed from each sample and analyzed on the Mastersizer 3000. The instrument made five measurements for each aliquot, with duplicate aliquots analyzed for each sample, providing a minimum of 10 measurements per sample to ensure results were reproducible and uncertainty minimized. Third and fourth aliquots were analyzed where abundant sample permitted, providing additional measurements for those samples (Table S2). Average volume % and standard deviation of grain sizes within clay (<4 µm), silt (>4 µm, <63 µm), and sand (>63 µm, <2 mm) were calculated using the total number of measurements collected (minimum n = 10; Table S2).

Bulk mineral identification of the mud (<63 µm) fraction of the sediments was determined using X-ray powder diffraction (XRD) following Moore and Reynolds (1989). Samples were dried at 95 °C for 24 h and lightly disaggregated using a mortar and pestle to homogenize material, sieved at <63 µm to remove coarser grains, and then loaded on XRD slide holders to ensure full crystallite randomization. XRD analyses were conducted using a Scintag X1 Diffraction System with Cu K alpha X-rays generated with a 40 kV beam voltage and 45 mA beam in the range 2°–72° 2θ, with a step size of 0.02° 2θ and a measuring time of 2 s per step. All mineral identifications and semiquantitative estimates were completed using MDI JADE 6 software.

Samples of sediment pore water were extracted at 2 cm depth intervals through pre-drilled holes from core 01UW-B using syringes attached to MicroRhizon samplers (0.15 μm pore size; Michaud et al., 2016). The filtrate was distributed into acid-washed vials. Sediment samples with volumes of 25 cm³ and 35 cm³ were collected with cut-off plastic syringes (2.7 cm diameter) from whole rounds of core 01FF at two depths (1.54–1.59 m and 1.89–1.94 m). The difference between sample volumes is due to different lengths of sediment extracted with the syringes. Pore waters from these samples were similarly extracted using MicroRhizons. Specific conductance was determined in the field using an Oakton cup-style sensor with ±1%; the two deeper samples were diluted with 18.2 MΩ water prior to analysis. Porewater vials for major ions (including chloride) were frozen in the field, thawed in the laboratory, and diluted with 18.2 MΩ water before analysis on a Metrohm Peak Ion Chromatograph as described in Hawkings et al. (2020).

Measurements were taken from two different multicores and each of the two gravity cores to ensure reproducibility and a representative stratigraphy. The lateral offset caused by ice flow ensures that the sediment lithologies in the collected cores accurately reflect the actual depositional conditions. Summary statistics (mean, standard deviation, median, interquartile range, min, and max) were calculated to show sample variability, and Kruskal-Wallis tests were used to assess the probability of statistically significant differences in the clast counts, bulk density, porosity, magnetic susceptibility, and Ca/Fe ratio between lithostratigraphic units (1–4) (Table S3; McKight and Najab, 2010). Statistically significant differences provided a quantitative method to support boundaries drawn based on visual appearance between the lithostratigraphic units (LU1–LU4) and to aid in grouping of units into facies associations (facies 1–3).

The SLM stratigraphy comprises four main lithologic units observed in CT imagery and characterized by differences in physical properties, grain size distributions, clast abundance, and XRF geochemistry (Figs. 2 and 3). All lithostratigraphic units are observed in two or more sediment cores. Tables 1 and S3 summarize the distinctions between each of the four units and show the distribution of three statistically significant (p-values <0.05) facies associations that are used to interpret the paleoenvironmental history of Mercer Ice Stream.

The lowermost lithologic unit (LU4) is a 0.31-m-thick (182–206 cm below surface) massive matrix-supported muddy diamict (following Moncrieff, 1989) and was recovered only in the bottom of cores 01FF and 02FF (Figs. 2 and 3). LU4 is a greenish-gray color (Gley 5/10Y). This unit contained abundant granule (2–4 mm) and pebble-sized (4–64 mm) clasts that account for a high mean wet bulk density (1.94 ± 0.04 g cm⁻³; Figs. 2B and 2D). This clast-rich (5–30 vol%) diamict was very poorly sorted (defined as SD = 2.7) following Folk and Ward (1957). Grain size distribution is composed of predominantly a 6 µm, silt mode (52 vol%), with contributions of sand (25 vol%) and clay (23 vol%), warranting the mud-rich designation (Fig. 4D). The pebble fraction included sub-rounded to rounded clasts, whereas angular- and subangular-shaped clasts were uncommon (Fig. 3D). This unit was macroscopically structureless and homogeneous in composition based on XRF geochemistry (Fig. 2F), and MSCL porosity varied from 35% to 55% (Fig. 2E). The single shear strength value in the unit was 90 kPa, the highest measured in the entire sedimentary sequence (Fig. 2G). The mean Ca/Fe and K/Ti were 0.05 ± 0.003 and 1.5 ± 0.07, relatively higher compared to the mud beds and laminae of the overlying LU3 (Figs. 2F and 5B–5E; Table S3).

LU3, contained in cores 01FF and 02FF, is a matrix-supported, stratified diamict with interbedded parallel to sub-parallel laminae <1 cm thick (LU3A) and a massive, 5-cm-thick mud bed (LU3B). This unit extends from the top of LU4—142 cm below the surface—and is greenish-gray (Gley 5/10Y). Mud and laminae are best observed in CT images rather than visually on the split core face (n = 6; Figs. 3D, 3E, and 6). The darker coloring in the CT images compared to the diamict matrix is due to the enrichment of clay-sized particles (<4 µm; 38.5 vol%) over sands (17.1 vol%) in the sediment matrix (lighter coloring = coarser grains = higher bulk density; Figs. 2C, 3D, 3E, and 6). The diamict is very poorly sorted and separated by interbedded well sorted mud layers. There was an overall decrease in wet bulk density (1.79 ± 0.04 g cm⁻³) and Ca/Fe, K/Ti, as well as an increase in porosity compared to the underlying LU4 because coarser sediments tend to have lower porosity and higher bulk density (Figs. 2D, 2E, 5B, and 5C). Compositional analysis of the silt and clay fraction (<63 µm) showed an increase in the abundance of clay minerals, which is likely due to an increase in the particles <4 µm in diameter (Table S4). The greater thickness of the mud bed from 174 cm to 179 cm below the surface than the surrounding laminae supports subdividing LU3 into LU3A and LU3B members (Table 1).

LU2 extends from 11.5 cm to 142 cm and is a second matrix-supported, massive diamict (~130 cm thick) containing large granule to pebble-sized clasts (5%–30%; Figs. 2 and 3). All four cores (01UW-A, 01UW-C, 01FF, and 02FF) contain sediments of this lithology. This unit is a slightly darker greenish-gray (Gley 4/10Y). LU2 is characterized by a mean bulk density of 1.87 ± 0.07 g cm⁻³, a mean Ca/Fe of 0.05, and abundant clasts (Fig. 2). LU2 displays textural and compositional similarities to LU4 (Figs. 2–5; Table S3). The top 16.5 cm of LU2 has increased porosity relative to the lower part of the unit (Fig. 2E). This sediment also shows a lower bulk density, which reflects an increase in the void ratio (Fig. 2D). The shear strength of this upper 15 cm is lower than the underlying sediments of the same unit (Fig. 2G).

LU1, originally described in Siegfried et al. (2023), extends from 11.5 cm below the surface to the top of the cored sequence. It consists of planar laminated muds, with interbedded coarse beds of a greenish-gray color (Gley 5/10Y). LU1 was only recovered in multicores (01UW-A, 01UW-B, and 01UW-C) that captured undisturbed sediment-water interfaces; core 01UW-B was sacrificially sampled for pore waters in the field. The mm-scale sorted silt and clay laminae are a key characteristic of this unit and exhibit normal as well as reverse grading (light and dark transition in CT scans; Figs. 3C and 7) and are repeated rhythmically at an ~1 cm interval (Fig. 3C). The matrix contains sand-sized grains (<17.4 vol%) in addition to the high silt and clay fraction (45.5 vol% and 37.2 vol%, respectively; Figs. 4 and 5A). The unit has distinctive coarse laminae, the most prominent at 8 cm and 9 cm (Fig. 7). These contain numerous granules and very coarse sand grains (Figs. 3A, 3B, and 7). Additional laminae are observed at depths of 6 and 7 cm, containing fewer grains of very coarse sand (1–2 mm diameter). Layers of very coarse sand to granules were also evident at depths of 2.5 and 4 cm; however, this portion of the sequence showed disturbance due to coring and handling; hence, further analysis of these laminae was not conducted. An ~1 cm diameter clast is associated with the coarse laminae at 9.3 cm, which deformed the underlying sediments (Figs. 3C and 7). LU1 is soft, unconsolidated, and water-saturated. The MSCL porosity values (88% ± 7%) account for the low bulk density (1.24 ± 0.03 g cm⁻³), which highlights the transition between this unit and the underlying diamict of LU2 (Figs. 2D, 2E, and 3B). The XRF Ca/Fe and K/Ti ratios also show the LU1-LU2 contact as this unit has a lower average Ca/Fe value of 0.02 ± 0.003 relative to LU2 (Figs. 2F and 5E).

The fine-grained SLM sediments (<63 µm) were primarily composed of common silicate minerals such as quartz, feldspars (albite and orthoclase), and clay minerals (smectite, chlorite, illite, kaolinite; Table S4). The major mineral species were common to the four units but did show some differences in relative abundances between units. Notably, LU1 and LU3 are composed of higher amounts of clay minerals relative to quartz and feldspars (Table S4). This variability is likely due to an increase in the clay size fraction (<4 µm) observed in units LU1 and LU3 compared to LU2 and LU4.

To complement the sedimentological analysis, chloride ion concentration and specific conductance were used as proxies for the salinity of the sediment pore waters. Chloride concentrations in the sediment pore water increased down the core from 1.2 mM at the sediment-water interface (core top) to 41.2 mM at the bottom (~190 cm; Fig. 2H). Specific conductance (corrected to 25 °C) also increased with depth, with values of the water column, sediment-water interface, and bottom of core record of 224 µS cm⁻¹, 281 µS cm⁻¹, and 3975 µS cm⁻¹, respectively. We used a two-component mixing model of Cl⁻ concentration to determine the percentage of glacial meltwater and seawater in the SLM lake and pore water (Michaud et al., 2016; Phillips and Gregg, 2001), with Cl⁻ concentration of 547 mM as the seawater endmember and 0.010 mM as the freshwater endmember, based on the highest snowpack Cl⁻ concentrations for interior West Antarctica (Kreutz and Mayewski, 1999). The maximum likely seawater component of the porewaters increased from 0.2% at the sediment-water interface to 7.3% at the bottom of the record.

Kruskal-Wallis tests were used to assess statistical differences in the bulk density, porosity, and XRF data between the lithostratigraphic units (McKight and Najab, 2010). P-values from Kruskal-Wallis multiple pairwise-comparison tests show that all units are statistically different for all properties (p-values <0.05) and provide an independent measure to support the boundaries designated between the lithostratigraphic units. Similarities in clast abundances, grain size distributions, and bulk densities between LU2 and LU4 support grouping these lithologies into a common facies association (facies 3: massive, clast-rich matrix supported diamict). Although LU1 and LU3 share similar grain size distributions, the differences in structure, WBD, and porosity (Fig. 3) support differentiating these two lithostratigraphic units into separate facies (facies 2: stratified diamict with interbedded mud laminae/bed; facies 1: rhythmically interlaminated muds with coarse thin granule beds and dispersed clasts; Table 1).

Chloride (Cl⁻) is a conservative tracer, with seawater as the primary Cl⁻ source for the sediment porewaters (e.g., Michaud et al., 2016; Gustafson et al., 2022). The relatively low Cl⁻ concentrations in the porewaters indicate that most of the sediment porewater is from glacial ice melt (Michaud et al., 2016) and sediments were deposited in freshwater conditions. The porewaters show an ~7% increase in chloride content over the entire sediment record, consistent with the upward diffusion of ions from a deeper fossil marine water source based on the geophysical observations of Gustafson et al. (2022).

The lowermost sediments recovered in cores belong to the massive clast-rich muddy diamict facies (LU4 and LU2: facies 3; Table 1). The lack of sorting and overall homogeneity of these diamicts are consistent with a subglacial till deposited beneath a grounded ice sheet due to a combination of lodgement and deformation (Evans et al., 2006; Smith et al., 2019). Such sediments have been previously described from beneath Siple Coast ice streams (Tulaczyk et al., 1998; Kamb, 2001; Licht et al., 2005; Hodson et al., 2016) as well as on the deglaciated continental shelf (Domack et al., 1999; Licht et al., 1999; Licht et al., 2005; McKay et al., 2008, 2016; Prothro et al., 2018). Similar bimodal clay and silt peaks as shown in particle size data from adjacent ice streams and downstream sites in the Ross Sea (Licht et al., 2005). The range in shear strength values (Fig. 2G) shows that both lodgement and deformation are likely key sedimentary processes since deformation tills are characterized by low values (<20–40 kPa) and lodgement tills have higher shear strengths (>20–40 kPa; Evans et al., 2005). The upper part of LU2 has physical properties (low density, increased porosity, low shear strength) consistent with a volume increase or till dilation that arises from sediment deformation when pore water pressures are high (Damsgaard et al., 2016). We interpret the top 15 cm of this facies (12–27 cm below the surface) as the dilated active zone, following Evans et al. (2006), based on elevated sediment porosity and lower bulk density.

Overlying facies 3 are stratified diamicts with thin mud laminae (LU3A) and a mud bed (LU3B) that archive different depositional conditions (facies 2; Table 1). The thin (<1 cm thick) mud laminae interbedded in the stratified diamict (LU3A) indicate deposition that oscillated between suspension settling from ponded or slowly flowing meltwater at the ice-bed interface, depositing fine-grained sediments (Ó Cofaigh et al., 2007; Witus et al., 2014; Buechi et al., 2017; Prothro et al., 2018; Lepp et al., 2022) and a grounded ice sheet in contact with the sediment, forming glacial till (Evans et al., 2006). The thinness of the laminae suggests that the meltwater events were relatively short. The transition between meltwater deposits and till deposition archives the switching of basal conditions from coupled (grounded) to decoupled (lift-off) states.

Notably, a 5-cm-thick mud bed at 174–179 cm below the surface lacks clasts (>2 mm) and sedimentary structures, such as laminae or coarse lags. The matrix of this mud bed is dominated by silts and clays with prominent modes at 6 µm and 0.5 µm in the grain size distribution (Figs. 4 and 5A). The density of these sediments is similar to the thinner mud laminae (Fig. 6). The lack of sorted sand or gravel beds suggests that if there was flowing water, the water velocity must have remained below 20 cm s⁻¹ or else coarser-grained beds and/or erosion surfaces would be present. The thickness of this bed is ~5× greater than the other fine-grained laminae, suggesting an additional depositional process. This fine-grained and well-sorted sediment package is interpreted to originate from the settling of suspended particles within a water column. Assuming this bed was deposited with a sedimentation rate close to those of the laminae, this package reflects a longer and more stable basal condition. Thus, we interpret the thicker mud bed to archive deposition in a paleo-subglacial lake or a water-filled cavity that decoupled the ice-bed interface for a longer duration than the laminae.

The subglacial muds of facies 2 have not previously been observed in sediment cores from other contemporary settings beneath the Siple Coast ice streams. Similar cm thick mud interbeds in tills have been reported in terrestrial glacial sedimentary sequences from dynamic sectors of the Laurentide (Boyce and Eyles, 2000), Cordilleran (Brown et al., 1987), and Fennoscandian (Lesemann et al., 2010) ice sheets and interpreted to represent ice-bed separation and low energy sedimentation in stagnant or slowing flowing subglacial water (e.g., Boyce and Eyles, 2000). Recently presented sedimentological evidence from SW Finland reveals till-buried subglacial erosional channels filled with fine-grained laminated sediments, deposited by slowly flowing yet intermittently changing water flow at the bed of the Fennoscandian Ice Sheet (Ojala et al., 2022). We note though that some of these studies report laminated muds rather than the massive muds we observe. We suggest that at the SLM location, the shearing of sediments via loading from the overlying flowing ice would destroy sedimentary laminae from the time of deposition (e.g., Clerc et al., 2012; Smith et al., 2018). Collectively, this body of evidence from various glaciated landscapes highlights the dynamic and widespread role of subglacial meltwater drainage systems and offers valuable insight for interpreting both recent (Quaternary) and ancient (Paleozoic) glacial deposits.

Terrigenous silts with similar fine-silt grain size modes (Fig. 4) have been observed within glaciomarine sediments recovered from the deglaciated continental ice shelf in the Ross Sea as well as from offshore Thwaites Glacier and interpreted as sediments transported via meltwater plumes sourced from subglacial environments (Witus et al., 2014; Prothro et al., 2018, 2020; Simkins et al., 2017; Lepp et al., 2022).

The fine-grained deposits of facies 3 demonstrate that ice-bed separation has occurred on at least six occasions given the number of mud layers (Fig. 6). The thicknesses of these units are variable, suggesting varying sedimentation rates and/or duration of deposition. The thinner (<1 cm) mud layers could reflect punctuated and/or low sedimentation rates, whereas the thicker (5 cm) mud provides evidence of longer meltwater deposition and/or higher sedimentation rates, following Lepp et al. (2022).

Facies 1 is composed of laminated silts and clays, described as rhythmites (Siegfried et al., 2023), with thin granule-pebble beds, and caps the SLM sedimentary sequence (Figs. 3B–3C and 7). This sediment facies archives deposition in the contemporary SLM (Siegfried et al., 2023), rather than from sediment fallout released by drilling operations (Priscu et al., 2021). The water-saturated and unconsolidated properties of the sediments and lack of deformation indicate an absence of ice grounding during this period of deposition. The soft, water-saturated nature of these sediments highlights that the ice stream does not fully ground during lake drainage events at the coring location, notably the deepest portion of the lake selected for this reason. Facies 1 was not present in the sediments recovered from Whillans Subglacial Lake (SLW), where evidence for ice grounding in the surface sediments was reported (Hodson et al., 2016). Expanding upon the depositional model reported by Siegfried et al. (2023), we elaborate on two sediment depositional pathways to explain the accumulation of the SLM lake sediment facies: (1) sediment transported in suspension by subglacial meltwater into the lake (Siegfried et al., 2023) and (2) sediment fallout from basal ice melting (Table 1; Fig. 8).

The diagnostic features of facies 1 are the rhythmically laminated silts and clays, which were the focus of Siegfried et al. (2023). High-resolution images from CT-scan grayscale measurements, used as a density indicator to infer grain size (e.g., Reilly et al., 2017, 2019), demonstrate both normal and reverse grading of the fine-grained laminae (light overlain by dark laminae and vice versa; Figs. 3C and 7). These sediments reflect transport and suspension settling through the lake water column under oscillating energy environments (Table 1). The appearance of both normal and reverse grading suggests that changes in water velocities are gradational and that a single rhythmite sequence includes both a fining- and coarsening-upward pattern. Siegfried et al. (2023) proposed three processes contributing to SLM rhythmite deposition: (1) sediments pre-sorted in Conway Subglacial Lake (SLC) and eroded and transported downstream to SLM; (2) channels are eroded into the underlying sediments between SLC and SLM; and (3) SLM water velocity changes are caused by water column thickness changes, which prevents fine-grained deposition during lake low-stand conditions. These processes are discussed in the following two paragraphs to incorporate our sedimentological data.

Because rhythmites were not observed at SLW, which is more hydrologically isolated than SLM (Fig. 1), Siegfried et al. (2023) argued that a combination of mechanisms 1 and 2 must be required for their deposition. Mechanism 1 requires transport, deposition, and then increased flow velocities to induce erosion of presorted silty sediments in an upstream basin such as SLC (must exceed 10 cm s⁻¹ to cause erosion following Schroeder et al., 2019). Similar velocities would be required to erode channels down into the underlying till of meltwater draining from SLC (Carter et al., 2017). Once eroded, the silt and clay would be transported in suspension unless the flow velocity fell below the grain-dependent settling velocity in an environment such as SLM (Fredsøe and Deigaard, 1992). Both facies 1 and 3 share similar 0.5 µm and 6 µm grain peaks (Figs. 4 and 5) and are composed of similar silicate minerals (Table S4), suggesting that facies 1 is derived from a similar upstream subglacial till source and that sediment of both facies are eroded from till locally, that has spatially homogeneous fine particle size distribution.

The sediment record may reflect internal lake dynamics in addition to external processes. Satellite ice-surface observations show that SLM completes a fill-drain cycle every 4–6 years and experiences a >15 m change in water column thickness at the center of the lake between low- and high-stands (Siegfried et al., 2023). Siegfried et al. (2023) proposed that flow velocities through the lake are altered following changes in water-column thickness. Fine-grained particles (clays and fine silts; <10 µm) will remain in suspension if the water velocity remains equal to or greater than the particle settling velocity. Clay cohesion could result in the formation of clay flocs, leading to clay-rich particles falling out of suspension as a secondary mechanism. Lake high-stand conditions would have minimal flow (negligible water velocity during coring at <1 cm s⁻¹), favoring net fine-grain (<10 µm) deposition. As draining commences, mixing within the water column and flow velocities would increase, carrying coarser grains in suspension with particles >10 µm deposited. As drainage continues and water velocity increases, deposition would be restricted to fine sands. As the lake begins to fill, flow velocities would decrease but mixing may restrict deposition to silts and sands, or until the energy is sufficiently lowered to allow fines to settle out.

In contrast to the fine-grained sediments (<125 µm), the coarse pebble to granule planar laminae represent event-based deposition from rainout processes from the base of the ice sheet when compared to the background sedimentation from suspension settling (Clerc et al., 2012; Livingstone et al., 2012; Siegfried et al., 2023). The event-based deposition is highlighted in the coarse laminae, which contain a large (1 cm) diameter clast, the top of which is aligned with the pebble to granule layer, but that deforms the underlying sediments. This indicates the large clast and the granule layer were deposited synchronously. We interpret the ~1-cm-diameter clast as a dropstone given the contortion and displacement of the underlying laminae (Thomas and Connell, 1985; Figs. 3B–3C and 7). The most plausible source for the granule-to-pebble material is suspension settling of material melting out of basal debris-rich ice from the lake ceiling, which contains clasts of that size (Priscu et al., 2021) and is analogous to ice-rafted debris in sub-ice shelf proximal settings (Livingstone et al., 2012). Water velocities of 50 cm s⁻¹ and >100 cm s⁻¹, respectively, would be required for fluvial transport of granule- to pebble-sized material (Hjulström, 1939). These high flow rates are not consistent with the preservation of the fine-grained sediments and depressed laminae associated with the dropstones, nor with the low current velocities (~1 cm s⁻¹ or less) observed when sampling the lake during a drain cycle (Priscu et al., 2021). Because the coarse material (>1 mm) is confined to discrete laminae and beds, sediment rainout from this source must have been an episodic event-based deposition. The prominent coarse sand–dominated laminae reflect sediment release into the lake from basal ice melt and are marker beds for this process. The SLM water temperature was −0.74 °C during field operations, consistent with the pressure melting point of ice with a thickness of 1087 m, so basal ice and lake water were at approximate thermal equilibrium (Priscu et al., 2021). Water flowing into or out of SLM (Siegfried et al., 2016; Siegfried and Fricker, 2018, 2021) could generate an increase in frictional heat at the ice-water interface in addition to geothermal heat and/or water pressure transients, causing basal melt and thereby releasing sediment. The presence of sedimentary structures in facies 1 contrasts with the absence of structures in facies 2 despite similar depositional conditions within a subglacial lacustrine environment. It is plausible that conditions in the paleo-subglacial lake were different from those of the current SLM; i.e., periodic roof melting of debris-rich basal ice, the source of the coarse sand and larger size fractions in facies 1, was not a significant sediment delivery mechanism. Further, the higher density and lower porosity of the muds in facies 2, relative to facies 1, indicate sediment compaction and deformation consistent with the overlying subglacial till (facies 3), which would have been deposited by grounded ice. Thus, the shearing of sediments, via loading from the overlying flowing ice would result in the destruction of sedimentary laminae from the time of deposition (e.g., Clerc et al., 2012; Smith et al., 2018).

Recent work has shown that at least some portions of West Antarctic Ice Sheet grounding line in the Ross Sea retreated inland during the Holocene before readvancing to its modern position (Kingslake et al., 2018; Venturelli et al., 2020, 2023; Neuhaus et al., 2021). The ¹⁴C-bearing organic matter at SLM allowed a more precise estimate of the timing (6.3 ± 1.0 ka) of the most recent marine incursion into and upstream of the SLM environment (Venturelli et al., 2023). This conclusion illustrates that Holocene changes in grounding line position enable ¹⁴C-bearing (marine) organic matter from the latest incursion to become part of the subglacial sedimentary mixture, with components of marine Oligocene and Miocene sediments (Coenen et al., 2019; Venturelli et al., 2020, 2023).

The Cl⁻ and specific conductance of SLM sediment porewaters indicate the reworked glaciomarine sediments were deposited under freshwater conditions following a marine incursion, but the timing of the readvance at the SLM location and deposition of the lowest till recovered (LU4) remains unknown. The model of Neuhaus et al. (2021) indicates that the readvance of the West Antarctic Ice Sheet grounding line at sites proximal to SLM beneath Whillans Ice Stream could have occurred as recently as 1100 years ago at SLW. These timing constraints from Neuhaus et al. (2021) for readvance are consistent with previously published ice stream reorganization events (Catania et al., 2012) as well as higher than present accumulation rates during the late Holocene (Bodart et al., 2023). Catania et al. (2012) proposed changes in meltwater discharge and/or basal freeze on as two mechanisms for driving the reconfiguration events during these periods. Changes in hydraulic gradient due to ice thickness changes or grounding line position may also have been responsible for the reorganization of subglacial meltwater (Lowe and Anderson, 2002). The meltwater-generated muds of facies 2 may be related to hydrological changes associated with ice stream scale reorganization processes in the late Holocene associated possibly with high mid-Holocene accumulation rates and grounding line readvance. In contrast to the lower meltwater muds, the laminated sediments of the contemporary SLM have the best-constrained age of the four lithostratigraphic units reflecting deposition over a period ranging from 53 years to 260 years (Siegfried et al., 2023) when laminae can be related to subglacial lake fill-drain cycles, as we have shown.

Collectively, the sediments described in our study archive numerous lines of evidence of paleo-subglacial conditions that have varied temporally and likely spatially, corroborating previous hypotheses suggesting subglacial lake sediments as potential repositories of subglacial environmental change (Bentley et al., 2011; Siegfried et al., 2023). The SLM sedimentary record of subglacial processes includes (1) emplacement of subglacial tills during stages of ice grounding; (2) deposition of sorted, meltwater deposits indicative of subglacial meltwater drainage; and (3) rhythmic sedimentation within a subglacial lake caused by alternating flow conditions and variable sediment delivery (Fig. 8). The sedimentary sequence in the SLM cored interval first records a period of grounded ice (facies 3) that is interspersed with periods of ice-bed separation associated with meltwater flowing through a distributed subglacial hydrologic system (facies 2; Fig. 8). There is then a return to grounded ice conditions (facies 3) before the transition into the contemporary subglacial lake (facies 1). Notably, the upper portion of LU2 shows evidence for till dilation, likely reflecting increased subglacial water pressures immediately before the development of the contemporary subglacial lake.

The stiffest and most compact sediments, according to shear strength and bulk density/porosity data, are found at the bottom of the cored interval (Figs. 2D, 2E, and 2G). The presence of this stiff till prevented the collection of deeper sediments (Rosenheim et al., 2023) and likely formed through a combination of lodgement and deformation, with sediment properties suggesting lodgement-favored formation (Ó Cofaigh et al., 2007; Halberstadt et al., 2018). The stiff till requires a period of strong ice-bed coupling, low sediment porewater pressure, and high basal effective stresses. The stiff till grades into a softer till, capturing a shift toward deformation and shearing as the dominant sedimentary processes and lodgement as a secondary mechanism. The simplest explanation for this transition is an increase in sediment porewater pressure, facilitating the reworking of the underlying stiff till through deformation and shearing (Ó Cofaigh et al., 2007). If these properties extend spatially, it is plausible that this transition reflects a local increase in ice flow velocity due to the reduction in basal effective stress caused by an increase in porewater pressure and reduced resistive stresses at the ice-bed interface (Christianson et al., 2014).

The fine-grained interbeds (facies 2) indicate that the ice-bed interface alternated between coupled and decoupled states during a minimum of six events (Figs. 2, 3, and 6). The thickest of these mud layers is interpreted as a potential paleo-subglacial lake. Ice-bed separation was induced by pressurized meltwater that exceeded the ice stream overburden pressure. Fine-grained sediments were carried in suspension and then settled out of slow-flowing water or ponded water (e.g., Greenwood et al., 2016). The presence of the mud layers argues for meltwater production and hydraulic conditions causing directional flow through the area presently occupied by SLM.

The shear strength in the middle section of facies 2 is characteristic of deformation till (Evans et al., 2005; Halberstadt et al., 2018). This transitions into a dilated till that is evident in the upper 15 cm, indicating the presence of water at the ice-bed interface. Evans et al. (2006) described this process as a “hydrogeological jack” that causes the lifting of the overlying ice sheet and reduces friction at the ice-bed interface. This active layer represents the area of maximum downstream displacement in the sediments and increased movement of the ice stream across its bed. The absence of sorted, thinly bedded meltwater deposits of facies 2 indicates that porewater pressures were not high enough to cause extensive or complete basal decoupling (Piotrowski and Tulaczyk, 1999; Ó Cofaigh et al., 2007). Although the availability of meltwater was limited to the sediment porewaters, the hydrogeologic jacking may have been a precursor to subglacial lake formation, which is reported to have occurred ~180 ± 20 years before core recovery (Siegfried et al., 2023).

The presence of rhythmically laminated silts and clays (facies 1) reflects deposition in a suspension load-dominated environment in a subglacial lake with alternating drainage conditions including water velocity and sediment content (Fig. 8). The observed sedimentary structures require relatively low flow velocities repetitively changing through time (Ojala et al., 2022; Siegfried et al., 2023). The coarser silts, inferred from CT imaging, indicate low-energy underflow currents, whereas the finer clays are evidence of suspension settling in stagnant waters (Clerc et al., 2012). Changes in sedimentation and sorting could be due to water velocity changes within the lake while sediment delivery remains constant. Water column height changes associated with internal fill-drain processes or broader catchment-scale variability could modulate water velocities (Siegfried et al., 2023). Alternatively, the laminated sequence could be the product of punctuated or pulsed sediment delivery in which sedimentation rates may vary dramatically through time. This continuum between processes influencing sorting and melt-out most likely has persisted because of the changing subglacial hydrology and lake level changes (Siegfried and Fricker, 2021; Siegfried et al., 2023).

The temporally dynamic hydrologic basal conditions recorded in SLM sediments provide the first example of multiple ice-bed separation events due to flowing meltwater beneath an active ice stream in West Antarctica. These findings have important implications for our understanding of the dynamics of Antarctic ice streams (e.g., Flowers, 2015) and for the biogeochemistry of the subglacial hydrologic system (e.g., Vick-Majors et al., 2020; Michaud and Priscu, 2023). The presence of meltwater at the ice stream bed interspersed with ice grounding events may be associated with regional ice stream reorganization events that were caused by ice thickness changes (Catania et al., 2012). The lake sediments and their structure provide a record of dynamic subglacial hydrologic conditions on more recent time scales, of decades to centuries (Siegfried et al., 2023). The active hydrologic regime evident from the sediment record before the contemporary SLM indicates that a different (distributed) drainage configuration may have existed at the SLM location earlier in the Holocene.

We employed a suite of sedimentological and geochemical analyses to reconstruct past basal conditions and meltwater activity at Mercer Subglacial Lake, West Antarctica extending our observational record. A composite 2.06 m sediment record comprised massive-to-stratified diamicts, massive muds, and laminated muds with drop stones. We interpreted the lithostratigraphic variability to reflect the emplacement of glacial tills interbedded with meltwater drainage deposits and capped by rhythmically laminated subglacial lake sediments. The meltwater sediments were deposited by suspension settling in a slowly flowing or ponded setting, likely a distributed subglacial drainage system The rhythmically laminated lake sediments were produced by changes in the sedimentation rate and sorting of suspended sediment transported into the lake and fallout of material from melting basal ice. Collectively, the sedimentary record presented here provides a complex archive of subglacial hydrologic conditions and thermal histories that are linked to past ice stream variability within the late Holocene. More broadly, this archive can be used as an analog for interpreting ancient glaciogenic sedimentary rocks.

The Subglacial Antarctic Lakes Scientific Access (SALSA) project was funded by the National Science Foundation (NSF-OPP, grants 1543537, 1543396, 1543405, 1543441, and 1543347). We thank the United States Antarctic Program for logistical support that enabled our field season, the New York Air National Guard and Kenn Borek Air for providing air support, V. Stanley and the Oregon State University Marine and Geology Repository, C. Dean as SALSA project manager, R. Ricards as SALSA project coordinator at McMurdo Station, B. Reilly for the development of SedCT and sharing script for automated clast counts, and D. Mogk for XRD assistance. We thank B. Reilly for his useful conversations and sharing of scripts for data analysis related to CT scan images. We are grateful to the University of Nebraska–Lincoln hot water drill team for subglacial lake access. Partial support was provided to T. Campbell by the Geological Society of America and by the Clay Minerals Society. We thank science editor T. Rasbury and reviewers K. Licht and G. Dunbar for their helpful comments that improved the manuscript.