Comparing TanDEM‐X Data With Frequently Used DEMs for Flood Inundation Modeling

Flood risk, particularly in Small Island Developing States, is increasing. Although spaceborne Digital Elevation Models (DEMs) have provided a capacity to model flooding at the global scale, their relatively coarse resolution (~90m)has led to a limited ability to providefine-scalefloodassessments in smaller catchments such as those in Small Island Developing States. Following the release of the TanDEM-X DEM at ~12-m resolution, the aim of this research is to determine whether TanDEM-X can improve flood estimates in comparison to Shuttle Radar Topography Mission (SRTM) andMulti-Error-Removed Improved-Terrain (MERIT) DEMs. Suitablemethods toprocess TanDEM-X to aDigital TerrainModel (DTM) are identified through testingof seven DTMs produced through combinations of different vegetation removal approaches. Methods include Progressive Morphological Filtering and Image Classification of two TanDEM-X auxiliary data sets—a Height Error Map and Amplitudemap. The LISFLOOD-FP hydrodynamicmodel output flood extent andwater surface elevation for the TanDEM-XDTMs, SRTM, andMERIT are compared against the LiDARmodel for a catchment in Fiji. Themain findings show that the unprocessed TanDEM-X has improved predictive capacity over SRTM, but not MERIT. The TanDEM-X processingmethod combining Image Classification of the Amplitude map and Progressive Morphological Filtering produces the DTMwith the highest floodmodel skill in comparison to all testedDEMs. This DTM reports a 12–14 percentage point higher floodmodel skill score thanMERIT and a lower water surface elevation root-mean-square error of 0.11–0.21 m, indicating the suitability of TanDEM-X for floodmodeling. Plain Language Summary Flood risk is increasing almost everywhere, making it vital to identify at-risk areas. Highly accurate elevation data are essential for flood risk estimation, which in high-income countries is usually provided by LiDAR. However, countries such as Small Island Developing States are often reliant on spaceborne elevation data sets due to the high cost of LiDAR, despite experiencing some of the greatest levels of flood risk. These spaceborne data sets have greater errors than LiDAR and often measure vegetation canopy height instead of ground height, reducing the accuracy of flood estimates used by policy makers to assess risk. This paper aims to identify whether newly released spaceborne data set TanDEM-X could improve flood estimates in these areas by comparing flood simulations from a hydrodynamic model using TanDEM-X data with simulations based on other spaceborne data sets and LiDAR for the Ba catchment in Fiji. The results showed that TanDEM-X performs closest to the LiDAR model but only after vegetation removal processing. Further studies should be conducted in other locations, but these results indicate a possible method for improving inundation estimates in data-sparse areas. This should provide useful information for floodmodeling and disaster management communities—essential given predictions of more extreme rainfall and greater exposure on floodplains.