Probabilistic characterisation of coastal storm-induced risks using Bayesian networks

Abstract

A probabilistic estimation of hazards based on the response approach requires assessing large amounts of source characteristics, representing an entire storm climate. In addition, the coast is a dynamic environment, and factors such as existing Background: erosion trends require performing risk analyses under different scenarios. This work applies Bayesian networks (BNs) following the source-pathway-receptor-consequence scheme aiming to perform a probabilistic risk characterisation at the Tordera delta (NE Spain). One of the main differences of the developed BN framework is that it includes the entire storm climate (all recorded storm events, 179 in the study case) to retrieve the integrated and conditioned risk-oriented results at individually identified receptors (about 4000 in the study case). Obtained results highlight the storm characteristics with higher probabilities to induce given risk levels for inundation and erosion, as well as how these are expected to change under given scenarios of shoreline retreat due to Background: erosion. As an example, storms with smaller waves and from secondary incoming direction will increase erosion and inundation risks at the study area. The BNs also output probabilistic distributions of the different risk levels conditioned to given distances to the beach inner limit, allowing for the definition of probabilistic setbacks. Under current conditions, high and moderate inundation risks, as well as direct exposure to erosion can be reduced with a small coastal setback (∼10 m), which needs to be increased up to 20-55 m to be efficient under future scenarios (+20 years).