Seafloor Roughness Reduces Melting of East Antarctic Ice Shelves

Global mean sea level projections have large uncertainty due to the poorly constrained future contribution of the Antarctic ice sheet. Around Antarctica, basal melting of ice shelves is driven by ocean thermal forcing associated with intrusions of warm Circumpolar Deep Water (CDW) onto the continental shelf and into ice shelf cavities. Climate models project warming of shelf waters but exhibit warm biases and circulation biases near Antarctica, likely due to unresolved processes at small spatial scales. The Denman Glacier region in East Antarctica, despite relative stability of East Antarctica overall, has shown rapid changes and is sensitive to CDW access. Transport of heat toward ice shelves is controlled by barriers such as the Antarctic Slope Front/Current (ASF/ASC) at the slope and the Antarctic Coastal Current (AACC) near the coast, with eddies and tides mediating cross-front transports. Bathymetry strongly modulates these circulations, yet small-scale seafloor roughness on the continental shelf is poorly resolved in commonly used bathymetry products and in climate models. This study tests the hypothesis that small-scale seafloor roughness over the continental shelf damps shelf circulation, limits CDW access, and thereby reduces ice shelf basal melting in the Denman region.

Prior work highlights that: (1) climate models often exhibit warm biases and circulation biases in the Antarctic margin due to coarse resolution and unresolved processes; (2) ocean access of CDW to shelves and cavities is governed by the ASF/ASC and AACC, with eddies and tides facilitating cross-front heat transport; (3) bathymetric features such as canyons, ridges, and troughs can enable or redirect warm inflows and modulate heat transport; and (4) small-scale seafloor roughness (abyssal hills) imparts topographic form stress on currents and eddies and generates internal waves, affecting momentum and mixing in the Southern Ocean. However, the effects of small-scale seafloor roughness on continental shelf circulation, shelf water mass properties, and ice shelf melt in East Antarctica have not been systematically explored. Satellite-derived bathymetry (e.g., SRTM15) suggests ubiquitous roughness at 10–60 km scales over shelves that is absent from smoother products (e.g., BedMachine/IBCSO-based), indicating a potential missing mechanism in global ocean and climate models.

A high-resolution regional ocean-sea ice-ice shelf model based on MITgcm was configured for the Shackleton Ice Shelf/Denman Glacier region (90–114°E, 60–67°S). Horizontal grid resolution is ~1.4 km (1/40° longitude, 1/80° latitude) with 160 vertical levels (3 m near surface to 100 m at depth; 100 levels in the top 1 km). The model includes dynamic and thermodynamic sea ice, and a thermodynamic ice shelf component using the three-equation ice-ocean thermodynamics. Bottom drag is quadratic with coefficient 2×10−3; K-profile parameterization is used for vertical mixing; Smagorinsky viscosity is applied horizontally. Lateral sponge layers (4° wide) at eastern, western, and northern boundaries restore ocean and sea ice fields to monthly means from the 1990–91 repeat-year ACCESS-OM2-01 run, with restoring timescales decreasing from 3 months to 10 days across the sponge. Surface atmospheric forcing uses JRA55-do (v1.3) applied 3-hourly. Barotropic tides (10 constituents) are imposed at open boundaries from TPXO9v4. Ice shelf geometry is from BedMachine Antarctica. Two main simulations differ only in ocean bathymetry outside a control box around the ice shelf (94.5–104°E, 64.8–67°S where bathymetry is identical): (1) BedMachine-based bathymetry (smoother shelf), and (2) SRTM15 bathymetry (with pronounced small-scale roughness). Both were integrated for 10 years; equilibrium is reached after ~5 years; analysis uses the final 5 years. To isolate the effect of small-scale roughness from large-scale bathymetric differences, two additional simulations added synthetic roughness (10–60 km scales) to BedMachine while preserving large-scale (>60 km) bathymetry; one matched SRTM15 variance and one doubled the variance per multibeam observations. Diagnostics: Kinetic energy (KE) was decomposed into mean (MKE), seasonal (SKE), and eddy (EKE) components. MKE = (ρ0/2)(um^2+vm^2), with um, vm 5-year means; SKE = (ρ0/2)(us^2+vs^2), us, vs monthly anomalies relative to um, vm, averaged over 5 years; EKE = KE − MKE − SKE. Heat fluxes were similarly decomposed into mean, seasonal, and eddy components. Heat flux divergence was computed and integrated over domains: whole shelf (onshore of 1000 m isobath; 90–114°E) and under-ice cavity (coastline to ice front). Water mass properties over a control region with identical bathymetry between simulations were characterized using volumetric temperature-salinity (TS) histograms and density classes: Antarctic Surface Water (AASW, σθ < 27.4), Winter Water (WW, 27.4–27.6), and modified CDW (mCDW, σθ > 27.6). Model validation against available observations in the region is summarized in Supplementary materials.

- Seafloor roughness substantially alters shelf circulation: In the SRTM15 run, the AACC weakens markedly (mean kinetic energy less than half that in BedMachine) and shifts offshore, while the ASC along the slope intensifies. Shelf EKE, especially in front of the ice shelf, is strongly suppressed with roughness. Shelf-wide zonal transport onshore of the 1000 m isobath reduces from ~2 Sv (BedMachine) to ~1.5 Sv (SRTM15). - Water mass changes: In the control region, the fraction of mCDW volume decreases from 7% (BedMachine) to 1% (SRTM15), largely compensated by an increase in WW volume (+7%). AASW volumes change negligibly (~1%). Shelf temperatures decrease by ~0.1–0.2 °C across much of the shelf and beneath the ice shelf, and by up to ~1 °C at parts of the shelf break. - Heat transport and redistribution: Integrated convergence of ocean heat flux over the entire shelf decreases from 3.62 TW (BedMachine) to 3.45 TW (SRTM15), a −0.17 TW (−5%) change. Partitioning shifts: mean +0.14 TW (0.87→1.01 TW), seasonal +0.11 TW (1.07→1.18 TW), eddy −0.42 TW (1.68→1.26 TW). Over the ice cavity, heat convergence drops by 0.04 TW (−12%), from 0.33 to 0.29 TW; mean contribution becomes negligible, seasonal ~0.13–0.14 TW, eddy ~0.16 TW unchanged. Spatially, heat convergence shifts toward the outer shelf and upstream, and weakens near the coast and ice front, consistent with a weaker AACC and reduced eddy activity. - Melt rate impacts: Modeled basal melt rates range from ~0–1 m yr−1 in the interior to ~4–5 m yr−1 near the edge (BedMachine), broadly consistent with previous estimates. With roughness (SRTM15), local melt rates reduce by up to ~1 m yr−1, especially west of Denman around Shackleton Ice Shelf. Integrated basal meltwater discharge reduces by 4 Gt yr−1 relative to BedMachine. Synthetic roughness experiments show a reduction of ~2.5 Gt yr−1 (SRTM15-like variance) and up to ~8.3–8.4 Gt yr−1 when roughness variance is doubled. - Mechanism: Small-scale roughness applies form stress and enhances internal wave generation that damps deep-reaching currents and eddies on the shelf. This reduces AACC strength and eddy transports, limiting mCDW presence and heat delivery to the cavity, thereby reducing basal melting.

The study demonstrates that small-scale seafloor roughness over the East Antarctic continental shelf acts as a missing control on shelf circulation and basal melting. By damping the AACC and eddies while intensifying the slope current, roughness reduces on-shelf CDW presence and the delivery of ocean heat to ice shelf cavities. Consequently, basal melt rates and total meltwater discharge decrease (by ~4 Gt yr−1 in the SRTM15 case). These results directly address the research question by revealing a physical mechanism unresolved in typical climate models that can lower ocean thermal forcing at the ice base. Because melt sensitivities to temperature are nonlinear, including roughness effects could reduce the sensitivity of modeled basal melting to projected warming, potentially narrowing the spread in Antarctic contributions to sea level rise. The spatial redistribution of heat convergence toward the outer and upstream shelf further suggests that local melt patterns are sensitive to shelf roughness. The findings underscore the need to represent small-scale bathymetric roughness—either resolved or parameterized—in global climate and high-resolution regional models to improve projections of ice-ocean interactions.

This work provides evidence that ubiquitous small-scale seafloor roughness over the Antarctic continental shelf reduces on-shelf CDW presence, suppresses shelf circulation (especially the AACC and eddies), diminishes heat convergence beneath ice shelves, and lowers basal melting in the Denman/Shackleton system. Integrated meltwater discharge decreases by ~4 Gt yr−1 with observed-like roughness and up to ~8.3–8.4 Gt yr−1 with stronger roughness, highlighting the potential magnitude of this effect. Incorporating seafloor roughness effects into ocean and climate models—through higher resolution or parameterizations of topographic form stress and internal wave drag/mixing—could reduce biases and the spread in sea level rise projections. Future work should: (1) obtain improved bathymetric observations beneath and around Antarctic ice shelves to constrain roughness; (2) develop and test parameterizations for roughness-induced form stress and internal wave/tidal processes suitable for climate models; (3) further disentangle the relative roles of lee waves, internal tides, and diapycnal mixing over shelves; and (4) assess the robustness of these effects across other Antarctic regions and under future warming scenarios.

- Bathymetry products differ at large scales as well as small scales; while synthetic roughness experiments attribute impacts to 10–60 km roughness, some ASC changes near the shelf break may reflect large-scale bathymetric differences. - Limited direct bathymetric observations over the shelf and beneath ice shelves lead to uncertainties; SRTM15 likely underestimates true roughness, so impacts may be underestimated. - The model domain includes only the Shackleton/Denman ice shelf system; other nearby cavities are excluded, potentially affecting regional circulation pathways. - Subglacial freshwater discharge and some mixing processes are not explicitly represented; diapycnal mixing may need additional parameterization. - Results depend on model resolution and forcing choices; while the resolved scales (∼1.4 km) capture key roughness effects, some processes remain parameterized. - Heat flux integrals depend on lateral boundary placement (though differences between simulations are robust). - The study does not fully separate contributions from lee waves vs internal tides to form stress and mixing.