Graph learning forecasts road disruptions from rainfall across regions
From Rainfall to Road-Network Functionality: Transferable Graph Forecasting for Flood-Disrupted Transportation Networks
Machine Learning
Summary
Heavy rain can quickly damage roads and make travel difficult, especially in emergencies. The authors developed a new machine learning method called G-R2R that predicts which roads will be affected by floods using data about the landscape and recent rain, without needing special adjustments for each new area. They tested it on computer simulations from several regions in China and found it predicts road problems accurately, even in places the system hasn’t seen before. This helps planners and responders quickly understand and prepare for flood impacts on transportation.
What this means in practice
- •For emergency response planners: Predict how flooding will disrupt roads in new regions to improve evacuation and aid delivery plans without retraining models for each area.
- •For infrastructure maintenance teams: Use rapid forecasts of flood-affected road nodes to prioritize inspection and repair efforts across multiple regions with diverse terrain.
Authors
Junyan Wang, Naiyu Wang
Abstract
Extreme rainfall can rapidly degrade road-network functionality (RNF), disrupting emergency mobility, accessibility, evacuation, logistics, and infrastructure restoration. Existing flood-transportation impact models are often difficult to deploy across regions because they rely on local calibration, intensive simulation workflows or target-region retraining. This study proposes G-R2R, a transferable spatiotemporal graph-learning framework that reframes rainfall-induced road-network disruption as a region-conditioned, topology-flexible graph forecasting problem for unseen transportation networks. The framework integrates ego-centered subgraph sampling, terrain-conditioned graph representation, region-conditioned dual-level static encoding, and dynamic rainfall-network feature learning. Static geographic and infrastructure attributes are fused with rainfall forcing and recent RNF states to forecast node-level RNF across heterogeneous road networks. G-R2R is evaluated using simulation-derived RNF sequences from ten flood-prone regions in Wenzhou, China, under unseen-region testing protocols. It outperforms representative forecasting baselines, achieving a stabilized MAPE of 2.33% and $R^2$ of 0.855 under the fixed unseen-region benchmark. Leave-one-area-out evaluation yields stabilized MAPE values of 2.30%-3.54%. Ablation analysis supports the contributions of the terrain-conditioned graph representation and dual-level static encoding. A historical-event replay of Typhoon Fitow further shows that G-R2R reproduces principal temporal trajectories and node-level spatial patterns of the simulation-derived RNF reference under observed rainfall forcing without target-region retraining. These results support rapid and transferable node-level RNF forecasting across the regional road-network domains represented in this study.