A Fuzzy Analytic Hierarchy Process Framework for Estimating Potential Loss of Life in Dam-Failure Scenarios
摘要
Dam failures caused by overtopping, extreme floods, structural deficiencies and operational failures can result in catastrophic downstream consequences, including substantial loss of life (LoL). Accurately estimating potential LoL remains a major challenge in risk assessment of hydraulic structures such as dams. The interaction between flood propagation, population exposure, warning effectiveness, evacuation conditions, and community response introduces substantial epistemic and behavioural uncertainty into consequence estimation. Conventional empirical approaches typically rely on limited variables and fixed fatality-rate assumptions, which may not fully reflect the heterogeneous socio-technical conditions governing downstream vulnerability. In this study, a structured decision-support framework has been developed for LoL estimation in dam-break scenarios using the Fuzzy Analytic Hierarchy Process (FAHP). The framework integrates expert judgement and fuzzy logic to quantify uncertainty across thirteen influencing factors spanning hydraulic, social, infrastructural, and organizational domains. The resulting defuzzified weights represent context-sensitive fatality-rate contributions that can be combined with population-at-risk data for scenario-specific consequence estimation. The framework is demonstrated using nine historical dam-break cases to illustrate the computational procedure and comparative consistency with existing empirical approaches. The results highlight the importance of warning effectiveness, evacuation conditions, and public awareness alongwith hydraulic severity in shaping LoL outcomes. Rather than serving as a statistically calibrated predictive model, the proposed FAHP-based approach provides a transparent and flexible methodology for integrating socio-technical factors into dam-safety risk assessment. The framework can support risk-informed planning, emergency preparedness, and policy evaluation, while allowing context-specific recalibration through expert elicitation.