Suitability of TanDEM-X Data for Flood Inundation Modelling in Small Island Developing States

Abstract

Recent disastrous flood events across Small Island Developing States (SIDS) have reaffirmed the extraordinary risk of flooding in SIDS following extreme rainfall and tropical storms. Estimating flood hazard for disaster risk reduction policy requires simulations of flood extent and water depths from hydrodynamic models. In many SIDS these models have relied upon coarse, global spaceborne Digital Elevation Models (DEMs) such as the ~90m Shuttle Radar Topography Mission data. This has limited the capacity to adequately estimate flood hazard at the localised scale (~10m) suited to many SIDS catchments. Following the release of the global TanDEM-X DEM with a horizontal resolution of ~12m, there is an opportunity to assess whether the finer-resolution TanDEM-X can be utilised to improve flood hazard estimates in SIDS.
The first section of this thesis synthesises the relevant literature on flood risk in SIDS and how flood hazard has been simulated using previous DEMs. The results of this literature review indicate that there is a mismatch between flood risk and capacity to estimate flood hazard in SIDS. A key reason for this is a lack of adequate topographic data for input into a hydrodynamic model used to estimate flood hazard.
The second section of this thesis details and compares methods to process vegetation surface artefacts from the TanDEM-X DSM for input to a hydrodynamic model using the Ba catchment in Fiji as a test case. Seven TanDEM-X DTMs were generated by combining three methods that remove vegetation: Progressive Morphological Filtering and Image Classification of two TanDEM-X auxiliary datasets (Height Error Map and Amplitude). The seven TanDEM-X DTMs were input into the hydrodynamic model LISFLOOD-FP to compare modelled flood extent and water surface elevation with those simulated using the SRTM (v4) and Multi Error-Removed Improved-Terrain (MERIT) DEMs. A model based on an airborne LiDAR DTM was used as a benchmark. The results show that the unprocessed TanDEM-X DSM does not improve flood estimates over the MERIT DTM, but does improve flood estimates over the unprocessed SRTM DSM. The method to remove vegetation that combines Progressive Morphological Filtering with Image Classification of the TanDEM-X Amplitude map has the best fit to the LiDAR model flood extent and water surface elevation estimates in comparison to all other models. The findings indicate the potential for TanDEM-X to improve flood hazard estimates in SIDS when processed using the method developed in this thesis, which should be applied to other SIDS catchments and used to improve flood hazard estimates in flood risk estimations by policy makers.

Date of Award	19 Mar 2019
Original language	English
Awarding Institution	The University of Bristol
Supervisor	Jeff Neal (Supervisor), Paul D Bates (Supervisor) & Joanna Isobel House (Supervisor)