Toward a New Flood Assessment Paradigm: From Exceedance Probabilities to the Expected Maximum Floods and Damages

Abstract

To assess flood risks, we seek to estimate the probability distribution of the worst possible single-event over a contiguous period of N years rather than the cumulative losses expected over a planning horizon. For this we use the probability distribution FN of extreme flood events over a multi-year period, which is different from using the conventional single-valued exceedance probability of 1/N years. FN can be used to estimate the hazard and then proceed to the estimation of risk, which we define as the “largest expected damage” over the set period. It also allows for a more coherent determination of design values, which descend from fully acknowledging the aleatoric uncertainty of the underlying natural river flow process. The epistemic uncertainty is removed by marginalizing the aleatoric-epistemic uncertainty joint distribution over the parameter space. The advantage of the proposed Bayesian approach, which can be summarized in 12 steps, is demonstrated for the 2021 River Ahr flood in Germany, which caused casualties and huge material damage. Adopting the multi-year maxima extreme value distribution can potentially lead to the reclassification of vulnerability levels for flood-prone areas and reconsideration of current policy-making and flood risk communication.