Toward an Operational GNN-Based Multimesh Surrogate for Fast Flood Forecasting

2026-04-03

AI Summary

This study addresses the computational inefficiency of traditional high-fidelity two-dimensional hydrodynamic models, which hinders their applicability in rapid flood forecasting across large urban areas. To overcome this limitation, the authors construct a synthetic flood database using the high-resolution unstructured-mesh Telemac2D model and propose a graph neural network (GNN) surrogate model that integrates a projected grid with a multi-grid connectivity mechanism. This architecture expands the spatial receptive field without increasing network depth and incorporates explicit discharge inputs Q(t) alongside a teacher-forcing training strategy to significantly enhance the stability and accuracy of long-term predictions. Experimental results demonstrate that the proposed model completes a 6-hour flood forecast in just 0.4 seconds on a single NVIDIA A100 GPU—over 27,000 times faster than the original Telemac2D simulation running on a 56-core CPU—while maintaining high fidelity to the reference solution.

Abstract

Operational flood forecasting still relies on high-fidelity two-dimensional hydraulic solvers, but their runtime can be prohibitive for rapid decision support on large urban floodplains. In parallel, AI-based surrogate models have shown strong potential in several areas of computational physics for accelerating otherwise expensive high-fidelity simulations. We address this issue on the lower Têt River (France), starting from a production-grade Telemac2D model defined on a high-resolution unstructured finite-element mesh with more than $4\times 10^5$ nodes. From this setup, we build a learning-ready database of synthetic but operationally grounded flood events covering several representative hydrograph families and peak discharges. On top of this database, we develop a graph-neural surrogate based on projected meshes and multimesh connectivity. The projected-mesh strategy keeps training tractable while preserving high-fidelity supervision from the original Telemac simulations, and the multimesh construction enlarges the effective spatial receptive field without increasing network depth. We further study the effect of an explicit discharge feature $Q(t)$ and of pushforward training for long autoregressive rollouts. The experiments show that conditioning on $Q(t)$ is essential in this boundary-driven setting, that multimesh connectivity brings additional gains once the model is properly conditioned, and that pushforward further improves rollout stability. Among the tested configurations, the combination of $Q(t)$, multimesh connectivity, and pushforward provides the best overall results. These gains are observed both on hydraulic variables over the surrogate mesh and on inundation maps interpolated onto a common $25\,\mathrm{m}$ regular grid and compared against the original high-resolution Telemac solution. On the studied case, the learned surrogate produces 6-hour predictions in about $0.4\,\mathrm{s}$ on a single NVIDIA A100 GPU, compared with about $180\,\mathrm{min}$ on 56 CPU cores for the reference simulation. These results support graph-based surrogates as practical complements to industrial hydraulic solvers for operational flood mapping.

Problem

• flood forecasting
• hydraulic simulation
• computational efficiency
• surrogate modeling
• operational decision support

Innovation

• graph neural network
• multimesh connectivity
• projected mesh
• pushforward training
• surrogate modeling