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

A new physics-guided deep learning framework that fuses satellite radar, optical imagery, and hydrodynamic equations has demonstrated high accuracy in predicting flood extent, water depth, and flow velocity, according to a preprint posted to arXiv on June 2, 2026 [1]. The hybrid architecture combines a UNet to capture fine-scale spatial details with a Fourier Neural Operator (FNO) to model basin-scale hydraulic interactions, while physics-informed residual losses enforce mass and momentum consistency [1]. The model ingests Sentinel-1 synthetic aperture radar, Sentinel-2 optical imagery, and terrain features derived from digital elevation models, then constrains its predictions with the depth-averaged shallow water equations [1]. Across diverse floodplain settings, the framework achieved an Intersection over Union of 0.82 and an F1 score of 0.90 for flood extent prediction, outperforming UNet-only and FNO-only baselines [1]. Using hydrodynamic simulations as reference data, the model recorded a root-mean-square error of 0.21 m for water depth and 0.15 m/s for flow velocity [1]. Mass imbalance remained below 2.1%, and ablation studies confirmed that removing physics-based regularization significantly degraded performance [1]. The authors argue that embedding hydrodynamic principles into deep learning yields more physically coherent flood predictions and offers potential for operational monitoring and large-scale deployment [1]. The preprint appeared on arXiv, an open-access repository that hosts electronic preprints across physics, computer science, and related fields and has grown to receive roughly 24,000 submissions per month as of late 2024 [6]. The work was posted under the Electrical Engineering and Systems Science category, specifically Image and Video Processing, and was accompanied by experimental HTML rendering and links to code and data tools through the arXivLabs framework [1][4]. arXivLabs, launched in 2020, provides a formalized space for community collaborators to build tools that enhance the reading and discovery experience while adhering to arXiv’s values of openness and user data privacy [5].