AI-Based Decision-Making Framework for Climate Change-Induced Flood Risk Management and Urban Resilience
Year: 2026
Type: article
Abstract: Climate change has intensified the frequency and severity of urban flooding, creating unprecedented challenges for municipalities worldwide. This study develops an AI-based decision-making framework that integrates machine learning, explainable artificial intelligence, and geospatial analytics to enhance urban flood risk management and resilience planning. The framework employs ensemble machine learning algorithms including Random Forest, XG Boost, and Cat Boost for flood susceptibility mapping, coupled with SHAP-based interpretability to decode model predictions and translate them into actionable policy insights at administrative scales. A city-scale digital twin component enables real-time flood impact simulation and cascading infrastructure disruption analysis, while a multi-criteria decision support module prioritizes adaptation investments based on economic risk assessment and equity considerations. The framework was validated using multi-temporal Sentinel-1 SAR data and hydrological records across diverse urban environments, demonstrating superior predictive performance with area under the receiver operating characteristic curve values exceeding 0.95 for ensemble models. Results reveal that elevation, proximity to hydrological features, and impervious surface density consistently emerge as dominant flood drivers, with their relative importance varying significantly across districts. The decision support system successfully transformed pixel-level predictions into ward-level actionable recommendations, enabling evidence-based zoning, drainage infrastructure prioritization, and targeted nature-based solutions deployment. The framework addresses critical gaps in explainability, scalability, and policy relevance that have limited the operational adoption of AI in urban flood governance. The study concludes that integrating AI-driven analytics with participatory decision-making processes can substantially enhance adaptive capacity in flood-prone urban systems, though algorithmic transparency, data equity, and institutional readiness remain essential preconditions for equitable implementation.
Institution: China Three Gorges University
Subfield: Global and Planetary Change
Field: Environmental Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7222580795
FWCI:
Cites:
Cited by:
Related to: