Modelling Flood Risk of Transport Infrastructure based on Watershed Characteristics

Abstract: Climate change is predicted to increase both general precipitation levels as well as the frequency of intense short-term rainfall events in Sweden. Major transport infrastructure such as roads and railways, which are characterized by long lifetimes and high investment costs, are especially vulnerable for changes in climate. This research aims to identify climate-related vulnerabilities in the transport network in view of Trafikverket’s adaptation to climate change. Specifically, the aim of this research is to identify flood risk of road/rail-stream intersections, based on watershed characteristics. Flooding in Värmland and Västra Götaland in August 2014 serves as the basis on which the models are built. Three different statistical modelling approaches were taken: a partial least square regression, a binomial logistic regression, and artificial neural networks. All three methods perform well, and share of urban land use in the catchment as well as local channel slope at the road-stream intersection were identified as most important parameters in the catchments of the studied area for estimating the probability of flooding. Using the results of the different models together makes it possible to cross-validate their results. A flood thermometer, indicating the level of risk a certain point has for flooding, was introduced to visualize this. This leads to better insights into the results and furthermore allows to use the suggested methods as a complement to Trafikverket’s currently used Blue Spot analysis. It is however essential to improve data collection in order to increase the accuracy and generalizability of the models. A good framework for data collection is therefore essential.