Physics Informed Deep Learning for Modeling Coseismic Crustal Deformation

Crustal deformation, which can be modeled by dislocation models, provides critical insights into the evolution of earthquake processes and future earthquake potentials. In this presentation, we introduce our recent work on a novel physics-informed deep learning approach for modeling coseismic crustal deformation (Okazaki et al. 2022). Physics-informed neural networks were proposed to solve both the forward and inverse problems by incorporating partial differential equations into loss functions (Raissi et al. 2019). The use of neural networks enables to represent continuous displacement fields in arbitrary geometrical structures and mechanical properties of rocks without discretization. To accurately model the displacement discontinuity on a fault, which cannot be directly approximated by neural networks composed of continuous functions, the polar coordinate system is introduced. We illustrate the validity and usefulness of the proposed approach through forward modeling of antiplane dislocations, which are used to model strike-slip faults. This approach would have considerable potential for extension to high-dimensional, anelastic, nonlinear, and inverse problems in a straightforward way.

Reference

Okazaki T, Ito T, Hirahara K, Ueda N, Physics-informed deep learning approach for modeling crustal deformation. Nature Communications, 13, 7092 (2022). https://doi.org/10.1038/s41467-022-34922-1