AIAI: Artificial Intelligence to Improve the Coupling Between the Antarctic Ice Sheet and the Ocean/Atmosphere System

AIAI (2023–2027) aims to improve the integration of ice sheets into Earth System Models through neural network emulators at the interfaces between the Antarctic ice sheet and the global atmosphere, and between the Antarctic ice sheet and the global ocean. The project develops convolutional neural networks (U-Net) trained on high-resolution, polar-oriented regional models — the MAR atmospheric model (surface mass balance) and the NEMO-SI3 ocean model (ice-shelf basal melt) — to emulate their behaviour within the coarser-resolution IPSL-CM6 climate model, coupled to the Elmer/Ice ice-sheet model. The goal is to reduce uncertainty in projections of the Antarctic contribution to sea level rise by better capturing the two opposing drivers of ice mass change: increased surface accumulation/melt and increased ocean-driven dynamical mass loss. At ULiège, I supervise Achille Gelens, a research engineer based at LSCE, who is developing the neural network emulator of the MAR atmospheric model. I also contribute to discussions on the ocean (NEMO-SI3) emulator and its integration into the IPSL climate model. This work was supported by the French National Research Agency through the AIAI project (ANR-22-CE01-0014). It is also directly connected to my FNRS postdoctoral project — a link between the two will be added here once that project page is online.