Graph Neural Networks for Emulation of Finite-Element Ice Dynamics in Greenland and Antarctic Ice Sheets
Younghyun Koo, Maryam Rahnemoonfar
TL;DR
The paper introduces an equivariant graph convolutional network (EGCN) as a fast, mesh-compatible emulator for finite-element ice dynamics modeled by ISSM. By comparing EGCN to a traditional GCN and a CNN baseline on Helheim Glacier and Pine Island Glacier, the authors demonstrate superior accuracy and substantial speedups, with R-values exceeding 0.99 and RMSE improvements in velocity and thickness. The work demonstrates that EGCN preserves equivariance to graph rotations and translations, enabling better generalization on unstructured meshes and boundary-sensitive regions such as fast ice streams. These results suggest that EGCN-enabled emulation can dramatically accelerate parameter studies and improve projections of ice-sheet mass loss and sea-level rise.
Abstract
Although numerical models provide accurate solutions for ice sheet dynamics based on physics laws, they accompany intensified computational demands to solve partial differential equations. In recent years, convolutional neural networks (CNNs) have been widely used as statistical emulators for those numerical models. However, since CNNs operate on regular grids, they cannot represent the refined meshes and computational efficiency of finite-element numerical models. Therefore, instead of CNNs, this study adopts an equivariant graph convolutional network (EGCN) as an emulator for the ice sheet dynamics modeling. EGCN reproduces ice thickness and velocity changes in the Helheim Glacier, Greenland, and Pine Island Glacier, Antarctica, with 260 times and 44 times faster computation time, respectively. Compared to the traditional CNN and graph convolutional network, EGCN shows outstanding accuracy in thickness prediction near fast ice streams by preserving the equivariance to the translation and rotation of graphs.