Heterogeneous melting near the Thwaites Glacier grounding line

Abstract

Thwaites Glacier represents 15% of the ice discharge from the West Antarctic Ice Sheet and influences a wider catchment^1-3. Because it is grounded below sea level^4,5, Thwaites Glacier is thought to be susceptible to runaway retreat triggered at the grounding line (GL) at which the glacier reaches the ocean^6,7. Recent ice-flow acceleration^2,8 and retreat of the ice front^8-10 and GL^11,12 indicate that ice loss will continue. The relative impacts of mechanisms underlying recent retreat are however uncertain. Here we show sustained GL retreat from at least 2011 to 2020 and resolve mechanisms of ice-shelf melt at the submetre scale. Our conclusions are based on observations of the Thwaites Eastern Ice Shelf (TEIS) from an underwater vehicle, extending from the GL to 3 km oceanward and from the ice-ocean interface to the sea floor. These observations show a rough ice base above a sea floor sloping upward towards the GL and an ocean cavity in which the warmest water exceeds 2 °C above freezing. Data closest to the ice base show that enhanced melting occurs along sloped surfaces that initiate near the GL and evolve into steep-sided terraces. This pronounced melting along steep ice faces, including in crevasses, produces stratification that suppresses melt along flat interfaces. These data imply that slope-dependent melting sculpts the ice base and acts as an important response to ocean warming.

Fig. 1. Warm water reaches near the ice base and retreating GL of the TEIS.

a, Historical GL positions (coloured lines/zones after ref. ) demonstrate notable GL retreat over the past two decades (QGIS map: Landsat 8, 15 m pixel⁻¹, band 8 image LC08_L1GT_003113_20200131_20200211_01_T2_B8, 31 January 2020; the red box denotes the study region). b,c, Warm water is delivered close to the ice base (upper grey regions), shown by contours of thermal driving (degrees above in situ freezing point). The ice (black line) and seabed (brown regions) elevation profiles are measured by up and down altimetry from Icefin, which compare with bathymetry from mapping and forward sonar (Fig. 2). The small circles denote the Icefin track, along two transects approaching the GL, T1 (red) and T2 (blue) shown in the lower inset (red box from a). The yellow circle in the inset and vertical line through the ice denote the location of the borehole. The T1 track is oriented 5–10° oblique to the flow direction of the glacier and T2 approximately 50° oblique to flow; Icefin reached the grounded point of the glacier at the end of T2. Triangles in b and c mark historic GL locations estimated from satellite interferometry for 2011 (white), and the furthest downstream estimate in 2016 (blue). In b, the yellow triangle denotes the potential GL wedge detected by Icefin (Fig. 2). Nearest to the GL, although temperatures are colder than the deep water, the ocean water holds more than one degree of thermal driving. The ice base transitions from rough near the GL to terraced (progressively steeper-sided step-like features) near and downstream of the borehole, suggestive of progressive melting. Crevasses also contain terraces, especially clear in c.

Fig. 2. Sea-floor bathymetry shows smooth retreat and ice–bed interactions near the TEIS GL.

The sea-floor bathymetry near the TEIS GL is characterized by along-flow ridged bedforms having several different wavelengths, as well as evidence for two possible former GL positions (white and red boxes) and channelized subglacial outflow (black box). Data in a–d are from downward-facing bathymetric sonar and e from forward sonar on Icefin. Reworked sediments (white box) are observed near the borehole (yellow circle). b, A single sinuous 2–3-m-tall slope consistent with a GL sediment wedge is found about 200 m north of the 2016 GL estimated from remote sensing (red boxed region from a; red arrows denote wedge). The wedge crosscuts the 2–5-m wavelength along-flow bedforms (Extended Data Figs. 5 and 6). c, An isolated 4-m-deep channel cut into the sea floor makes two sharp turns and includes a segment that cuts perpendicular to most bedforms, suggesting that this feature formed from rerouting of subglacial water as the GL retreated (Extended Data Fig. 5). d, The bedform topography near the GL of T2 shows evidence of linear ridges striking north (Extended Data Fig. 4). e, Forward-looking sonar data of the ice base near the GL shows that the ice has the same 2–5-m-wavelength ridges as the shortest-wavelength features on the sea floor. These data together suggest that GL retreat has been largely continuous over the observable period, since at least 2011 based on remote sensing. Moreover, the similarities between the bed and ice morphology at the GL suggest that the ice–bed interactions set up slopes that are then progressively melted by the intruding seawater. Bathymetric sonar, a–d, was processed in Qimera and projected using QGIS. Forward sonar are projected using the Oculus ViewPoint software.

Fig. 3. Ocean conditions influence ice base morphology, which varies with distance from the GL.

The Icefin vehicle track is shaded by relative along-track distance from downstream (white) to upstream (black). Light-blue data denote regions with cooling and freshening in terraces and dark blue denotes the coldest/freshest data observed. a, Conditions in the near-GL water cavity show the influence of melting (freshening) close to the GL along T2 (left). Coloured stars denote close passes to the ice that also have distinct signatures of mixing and melting. Vertical profiles of thermal driving (Θ − Θ_f), absolute salinity (S_A) and dissolved oxygen (DO) binned with distance from the ice base show complex signatures that vary with location (Extended Data Fig. 3), suggesting the influence of both melting and SGW outflow (centre). Imagery near the GL (red box) shows ridged ice topography and sediment-laden clear basal ice at the GL (yellow star) (right). Scale bar, approximately 0.5 m. b, Ocean conditions in a large terrace formed in the ice base imply melting near the sidewalls (red boxes, 800 m from the GL along T2) (left). Warm, salty water (black, grey) is found along the sidewalls, whereas much fresher and more oxygenated water with low thermal driving (cold relative to in situ freezing) collects in the terrace roof (centre). Imagery of terrace sidewalls across the TEIS uniformly show scalloped surfaces reflecting turbulent melting (Extended Data Fig. 8 and Supplementary Video 1) (right). Scale bar, approximately 0.5 m. c, As in b but for a small terrace at 2,400 m downstream along T1 that contains cold, fresh and oxygen-rich water along its roof. Here the water becomes supercooled, with ice crystals forming laterally (right) across the heavily stratified interface (red box) between this 0.1 m upper boundary layer and the warm, saline and more oxygen-poor lower ocean waters. Scale bar, approximately 0.1 m.

Fig. 4. Ocean currents and ice topography contribute to variable melting in terraces and crevasses.

Here the Icefin vehicle track is shaded by relative along-track distance from downstream (white) to upstream (black) and current velocities are shaded from slowest (white) to fastest (purple). a, Horizontal and vertical trends near a corner of a wide terrace (1,900 m downstream in T1 near the borehole) show freshening and cooling water inside the terrace and slowing currents as the water feels the influence of the ice interface. The grey lines denote the bottom of the terrace. Vertical profiles of ocean-current speed (U), thermal driving (Θ − Θ_f), absolute salinity (S_A) and dissolved oxygen (DO) binned with distance from the ice base show that, although the water is warm close to the interface, the current velocity slows in the boundary layer, suggesting breaking from friction at the interface. b,c, As in a for the furthest crevasse from the GL, observed along both T1 (b) and T2 (c). The panels on the right are binned with distance from the top of a step in the crevasse sidewall along T1 marked with the upper grey line. The lower grey line indicates the elevation of the bottom of the crevasse in T1. Stars in b relate to the location in the left panel. These panels show warm water with thermal driving of nearly 1.8 °C (Θ − Θ_f) reaching the crevasse walls accompanied by very slight freshening and oxygen increase that indicate melting (S_A and DO) that would then rise into the crevasse.

Fig. 5. Highly variable melt rates are found beneath the TEIS.

a,b, Estimates of the spatially varying ice-shelf melt rate are shown for each of four subregions along T1 (a) and T2 (b) (r1–r4 are the same regions as in Extended Data Table 2). The ice surface is coloured by melt rate calculated along each slope (top panels) from the three-equation parameterization (Methods) under regionally averaged ocean conditions, demonstrating the increased melt rate along steep slopes. Horizontal coloured lines (bottom panels) correspond to the mean melt rates in each region. For regions r2 in T1 and r3 in T2, two means are presented, as conditions were observed to change with height in the crevasses, in which the water higher in the crevasses was colder and fresher than the water lower in these features. The lower bar indicates the melt rate determined by variable ocean forcing in the upper crevasse above the dashed lines in the top panels; the upper bar represents the mean melt rate below the dashed line in the crevasses. The means for each of these regions are as follows: T1: r1: 3.07 m year⁻¹; r2: 16.16 m year⁻¹ (below dashes), 9.72 m year⁻¹ (above dashes); r3: 3.48 m year⁻¹; r4: 4.11 m year⁻¹; T2: r1: 1.47 m year⁻¹; r2: 4.18 m year⁻¹; r3: 9.12 m year⁻¹ (below dashes), 6.82 m year⁻¹ (above dashes); r4: 5.76 m year⁻¹.

Fig. 6. The ice-shelf melt rate is strongly slope-dependent and steep slopes contribute up to 27% of the ice loss under the TEIS along only 9% of the ice base.

a, Estimated spatially varying ice-shelf melt rates along T1 and T2 show the strong influence of local slope. Here each curve consists of individual melt-rate data points that have been calculated using the regionally averaged ocean conditions (Methods) corresponding to the regions labelled in Fig. 5. Red curves are from T1 and blue curves are from T2. b, Sideways melting along slopes greater than 30° contributes an estimated 27% of the melting under the TEIS, whereas these slopes account for only 9% of the ice base. Upward melting along low slopes is still the most notable source of melting, in which slopes less than 30° account for 73% of melting, while representing 91% of the ice.

Extended Data Fig. 1. The Icefin vehicle is a modular underwater vehicle designed to deploy through boreholes in the ice (diagram produced using SolidWorks).

The vehicle consists of a total of seven modules: forward science including oceanographic sensors (CTD, DO), forward-looking sonar, cameras and lighting; forward directional thrusters; customizable science payload, shown here as configured for the Thwaites missions with a bathymetric sonar; electronics module; aft science and navigation module with DVL/ADCP, altimeter, high-definition camera and light; aft directional thruster; and rear thruster with rear camera. The fibre-optic tether mounts to a bridle at the tail of the vehicle and connects through the rear bulkhead of the electronics module, delivering live data feeds to the surface.

Extended Data Fig. 2. Salinity and dissolved oxygen generally track with temperature near the retreating GL of the TEIS.

a, As in Fig. 1, image of the TEIS, with historical GL positions in coloured lines showing notable retreat over the past two decades (QGIS map: Landsat 8, 15 m pixel⁻¹, band 8 image LC08_L1GT_003113_20200131_20200211_01_T2_B8, 31 January 2020; the red box denotes the study region); inset denotes the geographic location of the TEIS in relation to Antarctica. b–g, Hydrography of the ocean under the ice shows that absolute salinity (d,e) and dissolved oxygen (f,g) track with temperature (Fig. 1b,c) under the TEIS. The inset in b provides a focused view of the study region: the yellow circle denotes the location of the hot-water-drilled access hole, the red line represents T1 (5–10° oblique to the flow direction of the glacier) and the blue line represents T2 (50° oblique to flow). Triangles in b–g mark historic GL locations estimated from satellite data (white, 2011; blue, 2016–2017) and shown by the Icefin bathymetric sonar data (yellow).

Extended Data Fig. 3. Salinity, temperature and dissolved oxygen profiles show signatures of melting and mixing below the TEIS.

T–S (a) and DO–S (b) diagrams compare hydrographic data from T1 and T2. Data are colourized after the data shown in Figs. 3 and 4 (with distance along track and for which blue colours denote extremely fresh section) and stars denote locations also called out with stars in Figs. 3 and 4. The warm, salty and oxygen-poor data in red are not shown in Figs. 3 or 4 but come from the outermost data from Extended Data Fig. 2 at distances greater than 10 m from the ice base. This water does not interact with the ice base in the surveyed region. Heavy lines show linear mixing lines between the source water mass responsible for melting the ice base locally (red star) and a pure mixture of GMW or SGW under fully turbulent mixing conditions^,,. These may not fully describe sheltered environments along terrace roofs, in which diffusive processes may dominate^,.

Extended Data Fig. 4. The bathymetry near the GL of the TEIS along the T2 survey is characterized by ridged bedforms of varying wavelengths.

a, Sonar-derived bathymetry from nearest the GL, coloured by depth, showing examples of linear ridges (after Fig. 2, bathymetry and profiles produced with Qimera and projected in QGIS). White lines denote the position of profiles found in c and d. b, Forward-looking sonar shows that the ice-shelf basal topography near the GL (after Fig. 2) is characterized by similar ridges, having an approximately 2.5 m ridge-crest spacing and sloped faces. Forward-looking sonar is projected in Oculus ViewPoint. c,d, Linear profiles of the bed topography across the region show evidence for ridges with approximately 1-m, 2–2.5-m and 5-m wavelengths. These data show that the shape of the ice surface at the GL is inherited from scraping over the bedforms and is later modified into terraces.

Extended Data Fig. 5. Ridged bedforms with pronounced ridge-crest amplitudes are found along the T1 survey.

a, Sonar-derived bathymetry from the borehole to the nearest approach made to the GL, coloured by depth (after Fig. 2, bathymetry and profiles produced with Qimera and projected in QGIS), in which boxes denote the sections shown in b and c. b,c, Close-up views showing details of two regions of the survey. Red (b) and white (c) lines denote the position of profiles found in d and e. d,e, Linear profiles of the bed topography showing that small-scale ridges formed across larger, longer-wavelength topography. f,g, Nearest the GL, 1-m, 5-m and 10-m wavelength along-flow ridges are observed.

Extended Data Fig. 6. Sea-floor topography suggests past dynamics influencing the retreat of Thwaites Glacier.

a, Sonar-derived bathymetry from the borehole to the nearest approach made to the GL, coloured by depth, with callouts for panels b and c (after Fig. 2, bathymetry and profiles produced with Qimera and projected in QGIS). b,c, Close-up views showing details of two regions of the survey, after Fig. 2. Red and white lines indicate the position of profiles found in d–g. d,e, Linear profiles of a crosscutting approximately 3-m-tall sinuous ridge are consistent with a past grounding event allowing enhanced sediment deposition in this area. This is the only such feature in the survey data. f,g, Linear profiles across a possible former subglacial channel near the GL. In f, a 5-m-deep U-shaped trough in the sediment begins parallel to ice flow but then cuts perpendicular to ice flow across lineated bed features and then turns sharply, which could be consistent with a channel incised by subglacial outflow into the sediment. In g, the upstream extent of the trough is less conspicuous, suggesting either modification as the ice ungrounded or that the SGW was not routed discretely through this area.

Extended Data Fig. 7. Bed elevation and ice speed for the TEIS show a changing system susceptible to GL retreat.

a, Colourized bed elevation from BedMachine v3 (ref. ) for the TEIS is overlaid on a Landsat 8 image with the historical GL positions (coloured lines/zones after ref.  are identical to Fig. 1a and Extended Data Fig. 2a, in which green is 2000, white is 2011 and blue is 2016–2017). QGIS map: Landsat 8, 15 m pixel⁻¹, band 8 image LC08_L1GT_003113_20200131_20200211_01_T2_B8, 31 January 2020; the red box denotes the study region. Note that the regions upstream of the present GL are grounded more than 800 m below sea level. b, As in a except coloured by the average 2019 seaward ice-flow speeds for this region from the 120 m pixel⁻¹ ITS_LIVE remote-sensing product. Note that the flow speeds increase as the ice crosses the GL and decrease near the pinning point.

Extended Data Fig. 8. Topography of terraces under the TEIS demonstrates strongly asymmetric melting and melt processes.

a–d, Images of a terrace upstream of the borehole along T1 show a steep, curved sidewall (a), flat roof (b), sharp transition from the wall to the flat base (c) and close-up of scallops in the wall show differences between modes of melting upward along the roof and base with sideways turbulent melting along the sidewall. Upward-looking views of the base (e) and sidewall (f) of another terrace downstream of the borehole show similar features. Small terraces near the GL along T1 show the initiation of scallops along smaller features in the basal ice (g), in which asymmetric ice melting is clear from the shape and streams of particulates (h,i).