Methodology:

Free air gravity data was first converted into equivalent topography using the Bouguer slab formula. This initial topographic surface was used to calculate the deflection of a theoretical Moho (crust/mantle boundary) at 30 km depth and the associated isostatic gravity anomaly, following an Airy isostatic assumption (the crust has no lateral strength) and assuming densities for crust and mantle of 2670 and 3330 kgm-3 respectively, and that the initial topographic surface was in air. This calculation was done using the GMT module gravfft. The calculated isostatic gravity correction was subtracted from the initial Free air gravity anomaly.

The subsequent isostatically corrected gravity anomaly was then used in place of the free air anomaly as input in the inversion procedure outlined in Jordan et al. 2020. The corrected gravity anomaly was converted to a topographic surface, using the Bouguer slab formula. The resulting gravity field was modelled, and the topographic surface updated to improve the fit between model and isostatically corrected data anomaly. This modelling and adjustment procedure was iterated three times.

The output adjusted topography was compared with known bathymetric observations including offshore swath bathymetry, airborne radar over grounded ice, and four distinct sets of direct bathymetric observations beneath the Crosson and Dotson ice shelves, including from Autosub Long Range and Ran Autonomous Underwater Vehicles (AUV), from drifting depth profiling floats, and from seismic observations from the ice shelf surface. Errors between direct observations and the adjusted topography were interpolated across the region and subtracted from the adjusted topography to give the final gravity derived bathymetry for the region.

Data collection:

Gravity and topographic data outside the Crosson and Dotson ice shelves is the same as in Jordan et al 2020 (https://doi.org/10.5285/7803de8b-8a74-466b-888e-e8c737bf21ce). Additional bathymetric data beneath the Crosson and Dotson ice shelves was collected with Autosub Long Range and Ran Autonomous Underwater Vehicles (AUV), from drifting ORBIS EM-APEX profiling floats (Girton et al., 2019), and from 27 active seismic shot points from the ice shelf surface (Muto et al., 2024).

Data quality:

We estimate the recovered bathymetric data to have a standard deviation of ~70 m, based on comparison with direct observations. However, the minimum resolvable wavelength of the data is ~5 km, so it is likely that very steep, or short wavelength topography is not well represented. In such areas significantly larger errors may be expected.