Airborne radio-echo sounding (RES) surveys are currently the primary method of measuring ice sheet thickness to derive subglacial topography, which underpins modelling efforts to improve projections of sea-level rise. However, the scientific impact of these campaigns is hampered by the time required to process the resulting radargrams. Here, we provide an overview of recent advances in machine learning (ML) relevant to ice sheet RES research to show that ML can enhance the value of RES data collected during past campaigns. We highlight two key areas where ML is already being used to enhance RES data analysis: (i) denoising and automated picking of radar returns; (ii) improved spatial interpolation and uncertainty quantification of flightline data. In addition, we suggest two areas where ML may also have a role when planning future surveys. We present examples from Antarctic and Greenland Ice Sheets that demonstrate how ML-driven approaches can outperform traditional methods for interpolation of basal topography and show that advances have been made in ML-based automated extraction of reflecting horizons from radargrams. As numerical ice sheet models become increasingly sophisticated, integrating ML throughout the workflow may help maximize observational value and guide future strategic efforts in the Polar Regions. This article is part of the Theo Murphy meeting issue 'Next generation ice-sheet bed measurements'.
Palmer et al. (Thu,) studied this question.