Deep learning of flood forecasting by considering interpretability and physical constraints

Zhang, Ting; Zhang, Ran; Li, Jianzhu; Feng, Ping

doi:https://doi.org/10.5194/hess-29-5955-2025

Articles | Volume 29, issue 21

https://doi.org/10.5194/hess-29-5955-2025

© Author(s) 2025. This work is distributed under
the Creative Commons Attribution 4.0 License.

https://doi.org/10.5194/hess-29-5955-2025

© Author(s) 2025. This work is distributed under
the Creative Commons Attribution 4.0 License.

Articles | Volume 29, issue 21

Research article

|

04 Nov 2025

Research article |

| 04 Nov 2025

Deep learning of flood forecasting by considering interpretability and physical constraints

Ting Zhang, Ran Zhang, Jianzhu Li, and Ping Feng

Supplement

https://doi.org/10.5194/hess-29-5955-2025-supplement

Model code and software

Interpretable physics-guided DL for flood forecasting (v1.0.0) Ran Zhang https://doi.org/10.5281/zenodo.17337746

Short summary

This study presents a model integrating attention mechanisms and physical constraints to improve flood prediction. It forecasts floods up to 6 h in advance. The model enhances accuracy by focusing on critical input features and historical patterns. Results demonstrate its superior performance compared to other models, offering improved flood prediction with greater interpretability and alignment with physical laws.