Predicting ocean-induced ice-shelf melt rates using deep learning

• Published in: The Cryosphere
• Main Authors: S. H. R. Rosier, C. Y. S. Bull, W. L. Woo, G. H. Gudmundsson
• Format: Article in Journal/Newspaper
• Language: English
• Published: Copernicus Publications 2023
• Subjects: Environmental sciences; GE1-350; Geology; QE1-996.5; Antarctic; The Antarctic; Antarc*; Ice Sheet; Ice Shelf; Ice Shelves; The Cryosphere
• Online Access: https://doi.org/10.5194/tc-17-499-2023; https://doaj.org/article/45546bd5577642bf8f5078491020b924

Through their role in buttressing upstream ice flow, Antarctic ice shelves play an important part in regulating future sea-level change. Reduction in ice-shelf buttressing caused by increased ocean-induced melt along their undersides is now understood to be one of the key drivers of ice loss from the Antarctic ice sheet. However, despite the importance of this forcing mechanism, most ice-sheet simulations currently rely on simple melt parameterisations of this ocean-driven process since a fully coupled ice–ocean modelling framework is prohibitively computationally expensive. Here, we provide an alternative approach that is able to capture the greatly improved physical description of this process provided by large-scale ocean-circulation models over currently employed melt parameterisations but with trivial computational expense. This new method brings together deep learning and physical modelling to develop a deep neural network framework, MELTNET, that can emulate ocean model predictions of sub-ice-shelf melt rates. We train MELTNET on synthetic geometries, using the NEMO ocean model as a ground truth in lieu of observations to provide melt rates both for training and for evaluation of the performance of the trained network. We show that MELTNET can accurately predict melt rates for a wide range of complex synthetic geometries, with a normalised root mean squared error of 0.11 m yr −1 compared to the ocean model. MELTNET calculates melt rates several orders of magnitude faster than the ocean model and outperforms more traditional parameterisations for > 96 % of geometries tested. Furthermore, we find MELTNET's melt rate estimates show sensitivity to established physical relationships such as changes in thermal forcing and ice-shelf slope. This study demonstrates the potential for a deep learning framework to calculate melt rates with almost no computational expense, which could in the future be used in conjunction with an ice sheet model to provide predictions for large-scale ice sheet models.