A framework to assess suitability of global digital elevation models for hydrodynamic modelling in data scarce regions

Abstract

With increase in availability of global Digital Elevation Models (DEMs) that are derived from distinctive techniques with varying characteristics, it is essentially required to assess their accuracies prior to use them in the flood inundation modelling, especially in urban catchments. In the present study, firstly, the vertical accuracy of the global DEMs is assessed using datasets of three independent Ground Control Point (GCP) measurements while accounting for the spatial and temporal variabilities with respect to DEM data acquisitions. Next, we investigate their suitability for flood inundation modelling by developing hydrodynamic models. The accuracy of global DEMs, viz. SRTM, MERIT, AW3D30, COPDEM30, TanDEM-X 12, TanDEM-X 90, FABDEM, and CartoDEM are assessed for Lower Tapi Basin (LTB), India using ICESat-1 and ICESat-2 laser altimetry, and GNSS surveyed locations. The potential of TanDEM-X 12 DEM for deriving Digital Terrain Model (DTM) from the Digital Surface Model (DSM) has been explored by adopting Simple Morphological Filter (SMRF) and void-filling approaches. Finally, a comprehensive performance evaluation of these DEMs that includes DSMs and DTMs in flood inundation modelling is carried out by simulating the severe fluvial flood event of magnitude 25,768 m3/sec that occurred in the year 2006 in LTB through 1D-2D coupled Hydro-Dynamic (HD) models using HEC-RAS software. Prior to development of HD models using global DEMs and DTMs as terrain layer, a benchmark model has been simulated for the same event with terrain layer as topographic surveyed contours of the basin. The least flood depth error has been observed for the FABDEM model (RMSE: 1.59 m), followed by the model with terrain topography as the derived “TanDEM-X 12 DTM-AMP-HEM-EVF” (RMSE: 1.88 m). On the other hand, the HD model based on TanDEM-X 12 DTM AMP-HEM-EVF has been able to simulate flood inundation extent better than the FABDEM model for the same flood event in the basin with Critical Success Index (CSI) scores 0.91 and 0.81 respectively. The TanDEM-X 12 DTM-AMP-HEM-EVF terrain-based model indicated consistently superior results in capturing flood inundation extent, leaving behind all the DSMs whose CSI scores are found to be less than 50 %. The global DEMs that are inherently DSMs in nature are unsuitable for flood inundation modelling as they fail to capture the flood pathways in urban topographies with precision. From the systematic framework followed in the present study, it is inferred that the processed TanDEM-X 12 DEM for the removal of artifacts could be a successful replacement to obtain bare-earth topography. In the absence of TanDEM-X 12 DEM, the freely available FABDEM can be used as an alternative source for terrain layer in flood modelling as its results are comparable to the derived terrain topography of TanDEM-X 12 DEM. The study is one of the first to assess the performance of FABDEM in flood modelling of coastal urban floodplain.