AbstractFloods pose a major global economic threat, with recorded damages exceeding $1 trillion since 1980. Our ability to mitigate these disasters is dependent on understanding the location, severity and probability of flood hazard: wrought from computer models of flood inundation. These models, though, are traditionally of individual river reaches; built locally by hydraulic engineers with accurate flow, topography and bathymetry data. By leaving the vast majority of the world’s rivers unmodelled, flood risk has historically been largely unquantified globally. To fill this information gap, recent advances in data availability and computational capacity have heralded the advent of global flood models, leading to a number of previously unexplored research questions: How accurate are these emerging models? What are they useful for? How can they be improved?
In this thesis, a large-scale flood hazard model of the US is comprehensively evaluated against thousands of local-scale models as well as observations of real flood events. This procedure uncovered that the skill of large-scale structures is approaching convergence with that of traditional local analyses. Model validation revealed new applications of total-coverage flood inundation models. Flood risk estimates have been updated for the US, finding the low-coverage assemblage of government flood maps underestimates risk considerably. The model also found applicability in indicating potential inundation from incoming storms, a component often neglected by official forecasts. Validation further exposed areas where model development is needed. A solution to the identified issue of poor representation of structural flood defences is presented, ensuring protected areas are modelled as such.
This thesis illustrates the considerable utility of emerging large-scale model structures and offers an improvement to a weakly constrained component within these. A lack of data will inhibit further improvements; thus, an expansion in the scale of publicly-available accurate elevation and bathymetry data is required to replicate these methods globally.
|Date of Award||24 Mar 2020|
|Supervisor||Paul D Bates (Supervisor) & Jeff Neal (Supervisor)|