Learning Spatio-Temporal Patterns of Polar Ice Layers With Physics-Informed Graph Neural Network
Zesheng Liu, Maryam Rahnemoonfar
TL;DR
This work tackles the challenge of predicting deep ice-layer thickness from shallow layers in the Greenland ice sheet, where echogram noise hinders image-based methods. It introduces PSAGE-LSTM, a physics-informed graph neural network that fuses GraphSAGE feature learning with an LSTM to capture spatio-temporal dynamics, augmented by MAR weather-derived physical node features. Through comprehensive experiments on 2012 Greenland echogram data, PSAGE-LSTM consistently outperforms non-physical and non-inductive baselines, demonstrating the value of incorporating physical context into graph-based thickness inference. The approach offers a robust, scalable framework for monitoring ice-sheet dynamics and could enhance predictions of deep ice-layer evolution under climate change.
Abstract
Learning spatio-temporal patterns of polar ice layers is crucial for monitoring the change in ice sheet balance and evaluating ice dynamic processes. While a few researchers focus on learning ice layer patterns from echogram images captured by airborne snow radar sensors via different convolutional neural networks, the noise in the echogram images proves to be a major obstacle. Instead, we focus on geometric deep learning based on graph neural networks to learn the spatio-temporal patterns from thickness information of shallow ice layers and make predictions for deep layers. In this paper, we propose a physics-informed hybrid graph neural network that combines the GraphSAGE framework for graph feature learning with the long short-term memory (LSTM) structure for learning temporal changes, and introduce measurements of physical ice properties from Model Atmospheric Regional (MAR) weather model as physical node features. We found that our proposed network can consistently outperform the current non-inductive or non-physical model in predicting deep ice layer thickness.