Bryan Riel from School of Earth Sciences Publishes a Perspective Article in Science

Recently, Associate Professor Bryan Riel from the Institute of Geophysics at the School of Earth Sciences of Zhejiang University was invited to publish a PERSPECTIVE article titled "How does Antarctic ice deform? —— A deep-learning model infers large-scale dynamics of Antarctic ice shelves" in Science, commenting on the research results of Wang et al. published in the same issue.

Wang et al. reconstructed the viscosity distribution and stress state of Antarctic ice shelves using a physics-informed neural network (PINN) model, revealing the complexity of ice flow dynamics and pointing out significant differences in the applicability of the traditional Glen flow law in different regions. This study provides a new perspective for understanding ice shelf deformation and helps improve the accuracy of ice sheet dynamics simulations.

According to Riel, ice shelves play a key role in supporting upstream ice flow and maintaining the stability of ice sheets. The rheological properties of ice are affected by factors such as temperature, grain size, and crystal orientation, leading to significant spatial variability in the relationship between effective stress and strain rate. The study by Wang et al. found that in the compression zones near the grounding line of the ice shelves, the stress index is generally lower than the traditionally assumed value, while in the extension zones, the relationship between ice deformation rate and effective stress deviates significantly from the traditional power-law model. These findings indicate that new modeling methods are needed to accurately assess the mechanical stability of ice shelves and predict the risk of future sea-level rise.

Riel further emphasized that the application of deep learning in glaciology holds great potential, especially in cases where data is scarce or traditional methods are limited. PINN can combine remote sensing observations, physical theories, and deep neural networks to achieve more refined simulations of ice flow behavior. However, he also pointed out that more direct measurement data (such as the internal crystal structure and grain size of ice shelves) is needed to optimize model inputs and combine traditional ice flow models to improve prediction accuracy. This study shows that the integration of artificial intelligence and geophysics will be an important direction for future ice sheet dynamics research, which is crucial for understanding the stability of Antarctic ice shelves and global sea-level changes.

References:

Bryan Riel, How does Antarctic ice deform? 2025. Science, 387, 1150-1151. DOI: 10.1126/science.adw3158

Wang et al., Deep learning the flow law of Antarctic ice shelves. 2025. Science 387,1219-1224. DOI: 10.1126/science.adp3300