FLOOD SIMULATION WITH PHYSICS-INFORMED MESSAGE PASSING
Primary Area: applications to physical sciences (physics, chemistry, biology, etc.)
Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.
Keywords: Physics-informed GNN, flood simulation, PDEs
Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2024/AuthorGuide.
Abstract: Flood modeling is an important tool for supporting preventive and emergency
measures to mitigate flood risks. Recently, there has been an increasing interest
in exploring machine learning-based models as an alternative to traditional hydrodynamic models for flood simulation to address challenges such as scalability and accuracy. However, current ML approaches are ineffective at modeling early stages of flooding events, limiting their ability to simulate the entire evolution of the flood. Another key challenge is how to incorporate physics domain-knowledge into these data-driven models. In this paper, we address these challenges by introducing a physics-inspired graph neural network for flood simulation. Given a (geographical) region and precipitation data, our model predicts water depths in an autoregressive fashion. We propose a message-passing framework inspired by the conservation of momentum and mass expressed in the shallow-water equations, which describe the physical process of a flooding event. Empirical results on a dataset covering 9 regions and 7 historical precipitation events demonstrate that our model outperforms the best baseline, and is able to capture the propagation of water flow better, especially at the very early stage of the flooding event.
Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors' identity.
No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.
Submission Number: 8838
Loading