Uncertainty in flood risk assessment of linear structures: Why correlation matters

Abstract

Estimation of the risk posed by inland fluvial floods to critical linear transportation infrastructures requires quantifying the damage inflicted by flooding on roads and railways. The estimated risk is often quantified in terms of the Expected Annual Damage [EAD] of the given linear structure. The aim of this article is to shed light on the estimation of this number and its uncertainty. The damage function models, which describe the degree of damage given flood intensity, represent a source of uncertainty. In this article we argue that this uncertainty needs to be embedded with a correlation structure in order to be evaluated for aggregated values. To this end, a novel methodology based on the use of Gaussian Processes to model the uncertainty associated with the damage function is introduced. Assuming a spatial correlation structure, parameterized by the so called decorrelation length scale, this framework is applied to estimate the EAD to roads in Portugal. The study shows that the application of an appropriate decorrelation length scale is decisive to the estimated uncertainty of the EAD. The proposed methodology may also be applicable to other types of hazards, if represented by some hazard intensity parameter associated with a damage function.