Project Summary
Post-wildfire rainfall-induced debris flows and landslides pose a growing
threat to hillside transportation infrastructure, especially in wildfire-prone
regions like California. This project aims to develop a scalable, data-driven
decision-support framework to identify and retrofit vulnerable road
segments under these cascading hazards. Leveraging recent advances in
machine learning and infrastructure resilience modeling, the proposed
framework integrates post-wildfire soil degradation analysis, synthetic
rainfall scenario generation using Generative Adversarial Networks, and
network performance estimation via a Siamese Graph Convolutional
Network. Hazard scenarios will be evaluated using fragility functions and
simulated impacts on travel time and accessibility, enabling robust
assessment of risk across a range of storm severities and wildfire footprints.
A budget-constrained optimization model will guide the selection of retrofit
strategies—such as retaining walls or drainage improvements—to minimize
expected network disruption. The final product is a web-based
decision-support tool for transportation agencies, with an interactive
interface that visualizes hazard exposure, projected failures, and
cost-effective resilience interventions. Case studies in the Los Angeles
hillside region, including areas impacted by the 2025 Palisades Fires, will be
used to validate and demonstrate the framework. This project supports
California’s climate adaptation and resilience goals by enabling targeted,
proactive investments in transportation safety and continuity.
