Probabilistic mapping of life loss due to dam-break flooding

Abstract Assessment of flood damage caused by dam failures is typically performed deterministically on the basis of a single preselected scenario, neglecting uncertainties in dam-break parameters, exposure information, and vulnerability model. This paper proposes a probabilistic flood damage model for the estimation of life loss due to dam-break flooding with the aim of overcoming this limitation and performing a more comprehensive and informative evaluation of flood risk. The significant novelty lies in the fact that the model combines uncertainties associated with all three components of risk: hazard, exposure, and vulnerability. Uncertainty in flood hazard is introduced by considering a set of dam-break scenarios, each characterized by different breach widths and reservoir levels. Each scenario is linked to a probability, which is assumed conditional on the dam-break event. Uncertainty in exposure is accounted for using dasymetric maps of the population at risk for two socio-economic states (representing business and non-business hours of a typical week), along with associated likelihood. Vulnerability to flooding is described through a well-established empirical hazard-loss function relating the fatality rate of the population at risk to the flood hazard, the flood severity understanding, and the warning time; a confidence band provides quantitative information about the associated uncertainty. The probabilistic damage model was applied to the case study of the hypothetical collapse of Mignano concrete gravity dam (northern Italy). The main outcome is represented by probabilistic flood damage maps, which show the spatial distribution of selected percentiles of a loss-of-life risk index coupled with the corresponding uncertainty bounds.