Heterogeneous melting near the Thwaites Glacier grounding line

The study investigates how ocean-driven processes control melting and grounding line (GL) retreat at Thwaites Glacier’s Eastern Ice Shelf (TEIS). Thwaites Glacier, grounded below sea level and fed by warm Circumpolar Deep Water (CDW), has undergone accelerated changes over the past two decades, with GL retreat rates of 0.6–1.2 km per year, widespread thinning, and structural weakening. Despite its critical role in potential sea-level rise, the mechanisms by which ocean conditions and small-scale ice–ocean interactions drive heterogeneous basal melting—particularly near the GL—remain poorly constrained due to limited direct observations beneath ice shelves and coarse model representations that often impose zero melt at the GL. This work aims to directly observe and quantify the spatial variability of melt processes near the GL at submetre scales, assess the role of ice basal topography and crevasses, and evaluate how stratification and circulation deliver heat and salt to the ice base.

Prior research documents rapid mass loss and GL retreat across West Antarctic glaciers, including Thwaites, driven by intrusions of warm CDW onto the Amundsen Sea continental shelf. Remote sensing indicates persistent thinning of the TEIS (about 25 m per decade, with airborne radar suggesting up to 45 m per decade). However, models typically simplify ice-shelf geometry and frequently impose zero melt at the GL, inconsistent with observed thinning and retreat. Earlier in situ studies under Antarctic ice shelves have characterized basal channels, terraces, and turbulent boundary layers, and have highlighted the importance of topography, turbulence, and thermohaline structure for melt rates (for example, observations and parameterizations at Pine Island, Ronne, and Petermann). Yet, before this work, there were no in situ measurements at the GL of a substantial Antarctic ice shelf. Laboratory and numerical studies indicate that sloping and vertical interfaces can experience enhanced melting via turbulent compositional convection, and that topography can be generated by melting/freezing feedbacks. These insights motivate high-resolution observations to resolve how slope, stratification, and flow govern heterogeneous melt near the GL.

The team deployed the Icefin autonomous underwater vehicle (AUV) through a borehole to survey the TEIS from the GL to roughly 3 km downstream and from the ice–ocean interface to the seabed. Icefin carried: (1) downward- and forward-looking sonars to map sea-floor bathymetry and ice-base morphology at high resolution; (2) oceanographic sensors to measure potential temperature, salinity (absolute salinity), dissolved oxygen, and currents (including ADCP-derived velocities) within the near-ice boundary layer, in terraces, and within crevasses; and (3) imaging systems documenting basal ice textures and debris. To minimize borehole freshwater contamination, transects were conducted along and into the boundary layer, up to within approximately 5 cm of the interface, including perpendicular approaches to vertical sidewalls. Bathymetric and forward-sonar data were processed (e.g., in Qimera, QGIS, and Oculus ViewPoint) to resolve bedforms and ice-base features. Mixing behavior was assessed via temperature–salinity–oxygen relationships relative to glacial meltwater (GMW) mixing lines. Evidence of subglacial water (SGW) influence near the GL was evaluated using T–S slopes and dissolved oxygen–salinity behavior. Melt rates were estimated using the standard three-equation parameterization for ice–ocean heat and salt exchange, assuming shear-driven turbulent mixing. Local ice-base slope was derived from sonar-resolved topography, and ocean forcing (currents, temperature, salinity, oxygen) was regionally averaged in subdomains with similar conditions along two main transects (T1 and T2). Spatially varying melt was then computed along slopes (including flat roofs and steep sidewalls). Results were compared with co-located autonomous phase-sensitive radio echo sounders (ApRES) and an oceanographic mooring at the borehole to assess consistency with long-term observations.

• Sustained GL retreat: Bathymetry and ice-base morphology indicate largely continuous GL retreat since at least 2011 over a prograde bed, with only one clear former stable GL position evident near the 2016–2017 locations.
• Warm ocean cavity: The warmest water in the cavity exceeds 2 °C above the local freezing point, providing ample thermal driving to melt the ice base.
• Rough ice near GL and inherited topography: The ice base near the GL is rough, with short-wavelength (2–5 m) ridges that mirror small-amplitude seabed ridges; roughly 30% of surfaces within ~1 km of the GL have high slopes. Basal ice is debris-laden (sand to mud, angular clasts) and laminated, consistent with interactions over the bed upstream.
• Terrace formation and slope-dependent erosion: Small terraces and scalloped textures appear within ~200 m of the GL and evolve into steep-sided, flat-roofed terraces (heights from tens of centimeters to >6 m) farther downstream. Scalloped textures on steep walls indicate vigorous turbulent melting, whereas flat downstream areas have very low slopes (<5°).
• Stratification suppresses upward melt on flats: Strong stratification develops near flat roofs of terraces; extremely fresh boundary layers (salinity down to ~20 g kg−1 and 36–42% freshwater) occur in shallow terrace roofs, where diffusive processes dominate over fully turbulent mixing.
• Boundary-layer dynamics and currents: Thermal driving is ~1.75 °C within 1 m of the ice base. Currents weaken from ~3 cm s−1 in the background to near-zero within ~5 m of flat interfaces, but increase in crevasses up to ~5.9–6 cm s−1.
• Subglacial water influence near GL: Water closest to the GL is cooler and fresher (excluding terrace roof layers) with a T–S slope of ~2.05 °C (g kg−1)−1 and decreasing dissolved oxygen with freshening, indicating admixture of SGW. Estimated SGW concentrations reach maxima of ~7 ml l−1 (from T–S) and ~24 ml l−1 (from DO–S), consistent with oxygen-poor debris-laden basal ice accretion upstream.
• Heterogeneous melt rates and strong slope dependence: • Average upward melt on flat surfaces: ~5 m yr−1, consistent with ApRES and mooring records and historical radar-based estimates. • Near-GL average melt: ~2 m yr−1, ranging ~1–10 m yr−1. • Vertical/steep faces: estimated melt rates approach ~30 m yr−1 under regionally averaged forcing; crevasse sidewalls can reach up to ~43 m yr−1 at observed locations. • Areal contribution: Slopes >30° represent ~9% of the ice base but account for ~27% of total melting; slopes <30° represent ~91% of area and ~73% of melting.
• Crevasse processes: Warm water with high thermal driving (~1.8 °C above freezing) reaches within ~1 m of vertical crevasse walls; water cools, freshens, and becomes more oxygenated with height inside crevasses, indicating accumulation of meltwater and efficient lateral heat/salt transfer that enhances sidewall erosion.

The observations demonstrate that basal melting beneath TEIS is strongly controlled by ice-base slope and small-scale morphology, especially near the GL. Stratification in the boundary layer suppresses upward melt over flat interfaces, whereas turbulent lateral exchange and buoyant GMW enable warm water to access and vigorously erode sloped and vertical faces. This creates and maintains terraces and scalloped sidewalls, and concentrates melt in crevasses where currents are stronger, thereby widening crevasses and basal rifts. The inherited rough topography at the GL—formed during ice–bed interactions upstream—provides numerous sloped surfaces that are preferentially melted as the ice transitions into flotation, helping to sustain melting even where ambient currents are modest. The results reconcile heterogeneous melt patterns with sustained GL retreat and emphasize that current ice–ocean models, which often assume flat interfaces and zero GL melt, likely underestimate localized but dynamically significant melt on steep slopes. Incorporating slope-dependent melt physics and stratification effects is essential to accurately predict ice-shelf thinning, crevasse evolution, calving propensity, and GL migration at Thwaites and similar systems.

This study provides the first in situ, submetre-scale characterization of ice–ocean interactions at and downstream of the Thwaites Glacier grounding line. It reveals that melting is highly heterogeneous and strongly slope-dependent: flat interfaces experience moderate, stratification-limited upward melt (~5 m yr−1), while steep faces and crevasse walls undergo much higher lateral melt (up to ~30–43 m yr−1). Though steep slopes cover a small fraction of area (~9%), they account for a disproportionately large share (~27%) of total basal mass loss, with implications for rapid crevasse widening, rift propagation, and calving. The team also documents continuous GL retreat since at least 2011 over a prograde bed, warm ocean conditions (>2 °C above freezing), strong boundary-layer stratification, and SGW influence near the GL. These findings call for high-resolution observational campaigns and ice–ocean models that explicitly resolve or parameterize slope-dependent melting, stratification, and crevasse processes. Future work should map larger areas near GLs, quantify temporal variability (including tides and seasonal forcing), trace SGW pathways and oxygen dynamics, and integrate these processes into prognostic models to improve sea-level rise projections.

• Spatial scope: Observations are limited to two main transects (T1, T2) and the vicinity of a single borehole on the TEIS, which may not capture full spatial variability across Thwaites Glacier or other ice shelves.
• Temporal coverage: Surveys provide a snapshot relative to multi-annual variability; processes such as tidal modulation or seasonal changes are not comprehensively resolved in these specific transects.
• Parameterization assumptions: Melt-rate estimates rely on the three-equation parameterization with regionally averaged ocean forcing and assumed shear-driven turbulence; local departures from these assumptions and unresolved microstructure could affect absolute rates.
• Borehole effects and proximity: Although Icefin minimized contamination by operating outside the immediate borehole plume, residual impacts cannot be entirely excluded.
• Indirect SGW inference: Subglacial water influence is inferred from T–S and DO–S signatures and nearby bathymetric features; direct observation of SGW sources was not achieved.
• Generalizability: The strong role of slope and stratification observed here likely applies to other warm-based ice shelves with low to moderate currents, but specific magnitudes and spatial patterns will vary with local geometry and forcing.