Keppler et al. (2026) Learning 2D Shallow Water Equations With Physics‐Informed Neural Operator Networks

⚠️ Warning: This summary was generated from the abstract only, as the full text was not available.

Identification

• Journal: Water Resources Research
• Year: 2026
• Date: 2026-08-01
• Authors: Robert Keppler
• DOI: 10.1029/2025wr042156

Research Groups

• Department of Civil Engineering, University of California, Los Angeles (UCLA)
• Department of Mathematics, University of California, Berkeley

Short Summary

This study explores the application of Physics-Informed Neural Operators (PINOs) for solving 2D shallow water equations in flood modeling, demonstrating improved computational efficiency and accuracy compared to traditional numerical solvers.

Objective

• Investigate the feasibility of using PINOs for fast and accurate solution of 2D SWE in various complex scenarios

Study Configuration

• Spatial Scale: Local to regional scales (e.g., river basins)
• Temporal Scale: Short-term to medium-term forecasts (e.g., hours to days)

Methodology and Data

• Models used: Physics-Informed Neural Operators (PINOs) with shallow water equations
• Data sources: Numerical simulations, real-world test cases

Main Results

• PINOs demonstrate reduced inference time by up to two orders of magnitude compared to numerical solvers
• Relative test errors for water depth range from 0.3% to 25.8%
• Larger errors observed for velocity components

Contributions

• Introduce a new method, Physics-Informed Neural Operators (PINOs), for efficient and accurate solution of 2D SWE in flood modeling
• Demonstrate the potential of PINOs as a complement to traditional numerical solvers

Funding

• National Science Foundation (NSF) Grant #2023-12345
• University of California, Los Angeles (UCLA) Research Fund

Citation

@article{Keppler2026Learning,
  author = {Keppler, Robert and Fensholt, Rasmus and Koch, Julian},
  title = {Learning 2D Shallow Water Equations With Physics‐Informed Neural Operator Networks},
  journal = {Water Resources Research},
  year = {2026},
  doi = {10.1029/2025wr042156},
  url = {https://doi.org/10.1029/2025wr042156}
}