Enhancing 2D hydrodynamic flood model predictions in data-scarce regions through integration of multiple terrain datasets

Abstract

Topography highly influences hydraulic model predictions. High-resolution Digital Elevation Models (DEM) are currently used in 2D flood modeling studies to create relatively more accurate flood inundation maps. However, the availability of high-resolution datasets, such as Light Detection And Ranging (LiDAR), remains limited due to cost constraints. Thus, low-resolution global datasets are utilized in data-scarce regions. Merging high and low-resolution terrain datasets will be an alternative approach to improve flood models, and comprehensive analysis of such merged DEMs is lacking. Thus, a new DEM (V-DEM) is developed in this study by incorporating available LiDAR, SRTM, local DEM, and river cross-sectional data. 2D unsteady hydrodynamic model predictions are analyzed using the V-DEM, existing low-resolution global datasets, SRTM and MERIT Hydro, and their modified versions. V-DEM was found to create flood flow predictions with a better Nash–Sutcliffe efficiency, significantly outperforming low-resolution global datasets. In addition, MERIT Hydro showed more than 50% improvement in the Nash–Sutcliffe efficiency over SRTM in flow discharge predictions. There is a 110% improvement in the Nash–Sutcliffe efficiency for hydrologically corrected SRTMs over the original SRTM. When SRTM is merged with LiDAR and hydrologically corrected, the predictions also showed an improvement of 146% over the original SRTM. Moreover, this study highlights that the vertical accuracy of terrain datasets has a more significant effect on the flood model predictions than the horizontal resolution, especially in the high and low-gradient regions of the study area. Overall, this study would benefit flood modelers in developing accurate DEMs, especially in the unavailability of high-resolution data for the entire study area.