Sub-ice-shelf bathymetry is a major boundary condition for models of Antarctic ice sheets and the ocean cavities beneath them. The shape of the seabed helps determine where relatively warm, salty ocean water can enter a cavity, how currents are steered across the basin, and where deep troughs may permit dense water to penetrate toward grounded ice. Bathymetric roughness also influences turbulence in the cavity and the transfer of heat to the ice. Direct observations remain rare because the cavities are difficult for ships and other platforms to access, leaving gravity inversion as an important tool for estimating hidden terrain between scattered measurements.

The approach works because the large density contrast between rock and seawater leaves a measurable imprint on Earth’s gravitational field. Airborne surveys can recover broad features of a sub-ice-shelf basin, typically at wavelengths of about 5 kilometres and above. Geological complications make the calculation less straightforward. Lateral variations in crustal thickness, rocks with different densities, and other subsurface structures can distort the measured anomaly. Earlier inversions addressed some of these issues by tying reconstructed topography to known elevations, but the new analysis indicates that this can be insufficient when a large, deep basin lacks centrally positioned tie points.

New observations collected beneath the Dotson and Crosson ice shelves revealed the scale of the problem. Autonomous underwater vehicles Ran and Autosub Long Range surveyed the ocean-facing margin during the RV Nathaniel B. Palmer expedition in January and March 2022. Ran collected multibeam profiles with a Kongsberg EM2040 system, supported by Doppler velocity log and inertial navigation; comparisons with shipborne swaths indicated an absolute horizontal positional error below 40 metres and a relative error below 10 metres. Autosub Long Range followed three routes extending 2, 16 and 39 kilometres inboard from the ice-shelf margin with a single-beam depth sounder whose local depth accuracy was below 1 metre. Three profiling floats and seismic measurements from 27 active shot points added observations farther across the cavity.

Comparison of these observations with the earlier gravity-derived model showed that parts of the Dotson basin had been estimated as too shallow by at least 400 metres. The discrepancies were most systematic at wavelengths longer than 30 kilometres, whereas the previous inversion reproduced much of the observed variation at intermediate scales of 10 to 30 kilometres. Differences below 10 kilometres were also expected because the airborne gravity data had been filtered at a 5-kilometre wavelength and collected with line spacing of 7.5 to 10 kilometres. The long-wavelength mismatch pointed to a systematic regional bias rather than merely unresolved small-scale terrain or scattered measurement errors.

The study’s explanation is that glacial excavation removes mass from the crust and prompts an upward adjustment of the Mohorovičić discontinuity, the boundary between less-dense crust and denser mantle. This isostatic compensation brings dense mantle material closer to the surface beneath a deep basin. Its positive gravitational contribution can offset part of the negative anomaly produced when excavated rock is replaced by water, causing the basin to appear shallower than it is. The method therefore calculates a model of the compensating gravity field and subtracts it from the original free-air anomaly before the bathymetric inversion. The researchers describe it as a pseudo-isostatic correction because its initial topography is inferred from gravity rather than directly observed.

For the calculation, the team used an Airy isostatic model, which assumes that the crust has no lateral strength. They adopted a reference Mohorovičić depth of 30 kilometres, crustal density of 2,670 kilograms per cubic metre, and mantle density of 3,330 kilograms per cubic metre. The chosen reference depth was deeper than some local seismic estimates, but the authors considered it a reasonable global average given uncertainty in the crustal and upper-mantle density structure. The Airy model generates the maximum local Moho deflection and therefore the largest correction among models with different crustal strengths. The researchers also calculated the compensation while treating excavated material as having displaced air rather than water, amplifying the modelled Moho topography by 1.6 times. This selected treatment produced a good fit with the direct observations beneath the ice shelf.

After subtracting the modelled isostatic gravity effect, the adjusted anomaly was converted into an initial topographic surface using a rearranged Bouguer-slab equation. The researchers then modelled the gravity field of that surface using prisms, calculated the residual against the adjusted observations, and iteratively updated the topography. Two further iterations brought the residual below about 1 milligal across most of the survey area. Where direct observations were available, the difference between the reconstructed surface and measured elevation was interpolated into a correction field. Subtracting that field from the gravity-derived estimate produced a model that followed the observations while retaining the regional shape supplied by gravity. The inversion was also run without the sub-ice-shelf observations as an independent test of the isostatic correction.

Against the longest Autosub Long Range profile, the isostatically corrected inversion reduced the mean error from 195 to 3.2 metres and the standard deviation from 105 to 70 metres. Comparison with float observations produced a smaller improvement in the mean error, from 313 to 135 metres, while the standard deviation remained 187 metres. The researchers attributed the difference partly to the floats following the steep coast of Bear Island, where an approximately one-kilometre topographic step falls from coastal terrain of about 300 metres to adjacent depths near 700 metres below sea level. That feature is not recoverable using gravity data filtered at wavelengths above 5 kilometres. Float navigation also had horizontal uncertainty of about 1 kilometre or more, which could make measured depths appear displaced from the model. The vehicle profile lay closer to the centre of the basin and away from the steep fjord wall, making it a more favourable test of the method.

The revised method changed the regional picture most strongly beneath Dotson and Crosson, where the basin is deep and direct constraints are sparse. The final model combined the corrected inversion with the new observations and was slightly deeper in places than the unconstrained, isostatically corrected result. The authors said this version should provide the best bathymetric estimate currently available for the area given the available data. They also cautioned that the 70-metre standard deviation along the main vehicle profile is location-specific: it incorporates unresolved short-wavelength topography, errors of about 1.67 milligals in the gravity observations, local geological variability, and navigation uncertainty. Unknown geology remains difficult to quantify, while thick sediment layers could create a different, negative gravity effect that the model does not explicitly resolve.

Beyond the study area, the results show that gravity-derived bathymetry can contain regional errors of several hundred metres when isostatic compensation is ignored. Such errors may affect simulations of ocean circulation beneath ice shelves and assessments of grounded-ice vulnerability to warming oceans, because a modelled seabed raised above a sill can artificially restrict the entry of warm water. The correction is likely to matter most for small or poorly explored ice shelves containing broad erosional basins and few direct measurements, rather than the much larger Ross or Ronne-Filchner ice shelves. Its influence was smaller near Thwaites Glacier, where the predicted compensation field had a low-amplitude, approximately uniform gradient, so interpolation between observations on either side of the data gap could provide a reasonable result. The study therefore indicates that direct observations remain essential, while a physically informed isostatic correction can make gravity inversion substantially more reliable for recovering regional bathymetry beneath Antarctic ice shelves.

Subject of Research: Gravity-based recovery of bathymetry beneath the Dotson and Crosson ice shelves using an isostatic correction

Article Title: Improved recovery of sub-ice shelf bathymetry from gravity data using an isostatic correction: a case study from the Dotson and Crosson ice shelves, West Antarctica

Article References: Jordan, T. A., Heywood, K. J., Wåhlin, A., Hall, R. A., Muto, A., Dutrieux, P., Hogan, K., Girton, J., Alley, K. E., & Pettit, E. (2026). Improved recovery of sub-ice shelf bathymetry from gravity data using an isostatic correction: a case study from the Dotson and Crosson ice shelves, West Antarctica. Solid Earth, 17(10), 1187-1199. https://doi.org/10.5194/se-17-1187-2026

Image Credits: AI Generated

DOI: 10.5194/se-17-1187-2026

Keywords: Antarctica, ice-shelf bathymetry, gravity inversion, isostatic compensation, Dotson Ice Shelf, Crosson Ice Shelf, glacial erosion, seabed mapping, autonomous underwater vehicles, West Antarctica, Improved, recovery