New developments in flood modelling at continental-scale

: Case studies in Europe and the US

Abstract

Floods are the costliest and most deadly class of natural disaster each year, and this situation is not expected to change in the future. Furthermore, the study of floods is relevant for a wide variety of applications ranging from nature conservancy to urban planning and rapid disaster response. Generally, the nature of floods at continental scale has been studied using streamflow data from gauge stations or satellite imagery. Although these approaches can characterise past floods quite well, they do not cover all the locations of interest and do not directly estimate hazards to people and assets. Alternatively, hydrodynamic models specifically designed to capture floodplain hydraulics are an effective option to simulate floods. At continental scale, hydrodynamic models have successfully been able to estimate the risk associated with floods at very high resolutions (i.e. <100 m), however, limited research has been devoted to the understand the evolution of floods over time. This thesis presents new developments in continental sale flood modelling to characterise floods in a multi-decal time window. Firstly, an open-source software package to automatise the input data processing needed for continental scale modelling was developed. Then, a new flood modelling framework that couples streamflow data from a hydrological model with a flood inundation model is presented. The framework was able to predict flood depths over time at continental scale in an efficient way, an approach that to date has not been done before with an inundation model but a routing model. The framework was used to build a European Flood Hindcast where floods hydraulics for 298 basins were mapped over 26 years (1990-2016). The data generated in Europe showed that among the largest basins the Danube, Rhine, Rhone and Elbe were the ones hit the most by floods. Conversely, the least impacted basins in the same period were the Douro, Ebro, Guadiana and Tagus. Building the modelling framework from scratch helped to uncover potential source of errors in the modelling chain, one of which was poor geolocation of rivers. To substantially improve river geolocation, a new data set of river streamlines was generated for the contiguous US based on national high-quality data.

Date of Award	11 May 2021
Original language	English
Awarding Institution	University of Bristol
Supervisor	Paul D Bates (Supervisor) & Jeff Neal (Supervisor)