Random forest parameterization of Antarctic subglacial hydrology

As a step towards hydrologically coupled, centennial-scale ice-sheet simulations, we developed a random forest parameterization of subglacial hydrology. The parameterization is trained using a library of subglacial drainage model simulations. When used as part of Antarctic ice-flow solutions, the new parameterization outperforms other simple models such as the perfect ocean connection assumption. Being 100,000x faster than running subglacial drainage model simulations, the random forest parameterization shows promise for coupling to ice-sheet models for the purpose of making sea-level-change projections.

Preprint

Hill, T, Hoffman, M. J., Flowers, G. E., & Bingham, D. (2026) Random forest parameterization of Antarctic subglacial hydrology for coupled ice-flow modelling. EGUsphere preprint [submitted 2026-01-29]. https://doi.org/10.5194/egusphere-2026-343.