A Fuzzy-Bayesian belief network model for coastal flood vulnerability assessment

Flood risk poses a significant threat to coastal areas, with climate change and urbanization intensifying the frequency and severity of flood events. Effective flood risk management requires quantifying vulnerability while accounting for the inherent uncertainty in residential areas exposed to coastal hazards. This study revisits the flood vulnerability index (VIE), which evaluates life loss risk in residential buildings using four microscale criteria, and enhances it with a fuzzy Bayesian belief network (FBBN). The FBBN framework integrates uncertainty and conditional dependencies into the assessment by fuzzifying input criteria and estimating the probability of buildings belonging to various vulnerability classes. The proposed framework was applied to a case study on the island of Noirmoutier (France), evaluating the vulnerability of 21,046 buildings across four flooding scenarios. The FBBN demonstrated good consistency with the original deterministic VIE index, with less than 3% outliers in classification results. Sensitivity analysis identified architectural typology as the most influential criterion, contributing 10%–11% variance reduction across all scenarios. The FBBN provided probabilistic and geo-spatial representations of vulnerability, offering a deeper understanding of risk distribution. These insights enhance decision-making in flood adaptation planning, supporting more targeted and effective mitigation strategies. • The flood vulnerability index (VIE) evaluates life loss risk in coastal buildings based on micro scale criteria. • A fuzzy Bayesian belief network is proposed to integrate uncertainty in the VIE index. • Building vulnerability probabilities are estimated with the proposed approach. • A sensitivity analysis allows to identify the most influencing results. • The probabilistic and geo-spatial results provide deeper and more robust understanding of vulnerability distribution.