Abstract/Summary

Flood inundation forecast maps provide an essential tool to disaster management teams for planning and preparation ahead of a flood event in order to mitigate the impacts of flooding. The maps can be used to inform forecast-based financing schemes to release funds ahead of a predicted flood event. Evaluating the accuracy of forecast flood maps is essential for model development and improving future flood predictions. The goal of this thesis is to develop spatial verification methods for deterministic and ensemble flood map forecasts and to improve forecasts using satellite data. Binary verification measures typically provide a domain-averaged score of forecast skill. The skill score is dependent on the magnitude of the flood and the spatial scale of the flood map. In this thesis, a new scale-selective approach is presented to evaluate both deterministic and ensemble forecast flood maps against remotely observed flood extents. The flood-edge location accuracy proves to be more sensitive to variations in forecast skill and spatial scale compared to the accuracy of the entire flood extent. Both the ensemble spatial-skill and spread-skill relationship vary with location and can be linked to the physical characteristics of the flooding event. We find that a scale-selective verification approach can quantify the skill of three systems operating at different spatial scales, so that the benefits and limitations of each system can be evaluated. A new data assimilation framework is presented to update the flood map selection from a static library of flood maps using satellite data, taking account of observation uncertainties. Results show that the flood map selection could be triggered in four out of five sub-catchments tested. The resultant analysis flood map has the potential to be used to trigger a secondary finance scheme during a flood event and avoid missed financing opportunities for humanitarian action. Overall, sensitive spatial verification methods that are location specific and can evaluate ensemble performance will aid future model development for flood inundation prediction.