Accelerating Flood Warnings by 10 Hours: The Power of River Network Topology in AI-enhanced Flood Forecasting

TL;DR

This paper enhances flood forecasting by transforming river network graphs to improve GNN performance, enabling 24-hour water level predictions that outperform traditional models by 71% in predictive horizon.

Contribution

Introduces a reachability-based graph transformation that densifies river network topology, significantly improving GNN-based flood prediction accuracy and horizon.

Findings

Transformed GNNs outperform EA-LSTM in flood prediction accuracy.

Achieves 24-hour water level forecast accuracy equivalent to 14-hour EA-LSTM forecasts.

Extends predictive horizon by 71% using topological densification.

Abstract

Climate change-driven floods demand advanced forecasting models, yet Graph Neural Networks (GNNs) underutilize river network topology due to tree-like structures causing over-squashing from high node resistance distances. This study identifies this limitation and introduces a reachability-based graph transformation to densify topological connections, reducing resistance distances. Empirical tests show transformed-GNNs outperform EA-LSTM in extreme flood prediction, achieving 24-h water level accuracy equivalent to EA-LSTM's 14-h forecasts - a 71% improvement in long-term predictive horizon. The dense graph retains flow dynamics across hierarchical river branches, enabling GNNs to capture distal node interactions critical for rare flood events. This topological innovation bridges the gap between river network structure and GNN modeling, offering a scalable framework for early warning systems.