Graph Neural Networks for Emulation of Finite-Element Ice Dynamics in Greenland and Antarctic Ice Sheets

TL;DR

This paper introduces an equivariant graph convolutional network (EGCN) to emulate ice sheet dynamics, achieving faster computation and higher accuracy than traditional CNNs and GCNs, especially near fast ice streams.

Contribution

The study develops an EGCN model that accurately emulates ice sheet dynamics with significantly improved efficiency over CNNs and GCNs, preserving equivariance to translation and rotation.

Findings

EGCN reproduces ice thickness and velocity changes with 260x and 44x faster computation.

EGCN outperforms CNN and GCN in accuracy near fast ice streams.

EGCN preserves equivariance to translation and rotation, enhancing modeling fidelity.

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.