Surrogate modeling for rapid estimation of spatially-resolved flood damage: Application to a coastal region

Flood damage assessment (FDA) is essential for minimizing economic losses and safeguarding communities. Conventional hydraulic model-based FDA approaches are computationally costly, limiting their practicality for real-time emergency response. Therefore, this study introduces BayFlood, a Bayesian-optimized machine learning surrogate model that enables rapid, accurate, and spatially resolved flood damage estimation using only river discharge and tidal level inputs. The framework is trained and validated on a comprehensive dataset of flood events generated from two-dimensional hydraulic simulations of a coastal basin, covering river flow-dominant, storm surge-dominant, and compound flood scenarios. Among multiple learning engines tested, the boosting-ensemble-driven BayFlood achieves the best performance (coefficient of determination = 0.92–0.98; root mean square error = 4–8 %); the model reduces computational time by two orders of magnitude compared with hydraulic modeling, generating damage results within minutes. Monte Carlo uncertainty analysis (1000 runs, 5 % noise level) reveals a mean damage-rate uncertainty of 18 %, confirming model robustness. By effectively combining forecasting efficiency, accuracy, and spatial damage mapping, the BayFlood provides a practical and scalable tool for pre-disaster planning, real-time emergency response, and post-disaster recovery.