Advanced Flood Prediction with Physics-Guided Deep Learning: Combining UNet, FNO, and SAR/Optical Imagery

2026-06-08 · Source: cs.CV updates on arXiv.org · Field: Science & Research — Artificial Intelligence & Machine Learning, Environmental Science & Earth Systems · Depth: Expert, long

Summary

A new physics-guided deep learning framework significantly enhances flood prediction by integrating multi-modal remote sensing data with hydrodynamic principles. Developed by Tewodros Syum Gebre, Jagrati Talreja, and Leila Hashemi-Beni, this hybrid architecture combines a UNet for capturing fine-scale spatial details with a Fourier Neural Operator (FNO) to model basin-scale hydraulic interactions. It leverages Sentinel-1 SAR, Sentinel-2 optical imagery, and DEM-derived terrain features, while enforcing mass and momentum consistency through physics-informed residual losses from depth-averaged shallow water equations. The model achieved an Intersection over Union (IoU) of 0.82 and an F1 score of 0.90 for flood extent prediction, outperforming UNet-only (IoU: 0.75, F1: 0.85) and FNO-only (IoU: 0.77, F1: 0.87) baselines. It also demonstrated an RMSE of 0.21 m for water depth and 0.15 m/s for flow velocity, maintaining mass imbalance below 2.1%. This work has been accepted for publication in the Proceedings of the IEEE Radar Conference (RadarConf 2026).

Key takeaway

For Machine Learning Engineers developing flood forecasting systems, integrating physics-guided deep learning with multi-modal remote sensing data is crucial. You should consider hybrid UNet-FNO architectures constrained by shallow water equations to achieve higher accuracy and physical consistency in predictions. This approach yields more reliable water depth and flow velocity estimates, improving operational monitoring and emergency response capabilities.

Key insights

Embedding hydrodynamic principles into deep learning models significantly improves flood prediction accuracy and physical consistency.

Principles

• Multi-modal data fusion enhances robustness.
• Hybrid architectures capture multi-scale dynamics.
• Physics-guided regularization improves generalization.

Method

The framework fuses UNet and FNO outputs, then uses an MLP head to predict water depth, surface elevation, and velocity. Physics residuals from shallow water equations are incorporated as penalties during training.

In practice

• Combine UNet and FNO for multi-scale feature learning.
• Integrate SAR, optical, and DEM data for robust inputs.
• Use SWE residuals to enforce physical consistency.

Topics

• Flood Prediction
• Physics-Guided Deep Learning
• UNet-FNO Architecture
• Multi-modal Remote Sensing
• Shallow Water Equations
• Hydrodynamic Modeling

Best for: AI Scientist, Machine Learning Engineer, Research Scientist

Related on AIssential

• Advanced Flood Prediction with Physics-Guided Deep Learning: Combining UNet, FNO, and SAR/Optical Imagery — Physics-guided deep learning improves flood prediction accuracy and consistency by integrating multi-modal remote sensing with hydrodynamic equations.
• Outperforming Self-Attention Mechanisms in Solar Irradiance Forecasting via Physics-Guided Neural Networks — Physics-guided hybrid neural networks outperform complex self-attention models in noisy solar irradiance forecasting.
• Evaluating Transformer and LSTM Frameworks for Prediction in Ungauged Basins — LSTMs demonstrate superior performance over encoder-only Transformers for upstream streamflow prediction, especially when augmented with downstream hydrologic context.
• Physics-informed convolutional neural networks for fluid flow through porous media — Physics-informed CNNs accurately predict fluid flow in porous media, significantly accelerating traditional LBM simulations.
• Learning the coupled dynamics of global climate modes — UniCM unifies global climate mode forecasting by learning coupled ocean-atmosphere dynamics and inter-mode interactions, revealing emergent predictability.
• Kaggle Winners Walkthroughs: Yale/UNC-CH - Geophysical Waveform Inversion with Team greySnow — Combining physics-based models with machine learning enhances Full Waveform Inversion for subsurface imaging.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by cs.CV updates on arXiv.org.