Liu et al. (2025) Spatiotemporal Flood Prediction From Single Frame Input With a Post‐Processing Method

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

Identification

• Journal: Water Resources Research
• Year: 2025
• Date: 2025-09-01
• Authors: Dejun Zhu, Danxun Li
• DOI: 10.1029/2025wr040975

Research Groups

Not specified in the provided text.

Short Summary

The study introduces a "single frame prediction" framework using a U-Net architecture and physics-based post-processing to predict spatiotemporal flood maps based on boundary conditions and the previous time step's state.

Objective

• To develop a hydrodynamic prediction framework that generates flood maps using single-frame inputs and boundary conditions, relying on hydrodynamic principles rather than historical trends or traditional hydrodynamic models.

Study Configuration

• Spatial Scale: Grid-based flood maps (specific area not mentioned).
• Temporal Scale: Long-term spatiotemporal flood event prediction.

Methodology and Data

• Models used: U-Net (Convolutional Neural Network) and a novel physics-based post-processing method.
• Data sources: Not specified (utilizes boundary conditions and flood maps containing water depth and unit discharge).

Main Results

• The single frame prediction framework is feasible and produces accurate flood maps.
• The physics-based post-processing method effectively mitigates accumulated errors during long-term predictions.
• Quantitative performance achieved an average root-mean-square error (RMSE) of 0.041 m for water depth and 0.003 m²/s for unit discharge.

Contributions

• Proposes a novel spatiotemporal prediction approach that eliminates the need for traditional hydrodynamic models or historical trend analysis.
• Introduces a physics-based refinement technique to reduce error accumulation in deep learning-based long-term flood forecasting.
• Provides a more interpretable depiction of flood processes by incorporating essential hydrodynamic variables (water depth and unit discharge).

Funding

Not specified in the provided text.

Citation

@article{Liu2025Spatiotemporal,
  author = {Liu, Ziqi and Zhu, Dejun and Li, Danxun},
  title = {Spatiotemporal Flood Prediction From Single Frame Input With a Post‐Processing Method},
  journal = {Water Resources Research},
  year = {2025},
  doi = {10.1029/2025wr040975},
  url = {https://doi.org/10.1029/2025wr040975}
}