Evaluating the capability of uav-derived dems for flood hazard analysis

Abstract

This study evaluates the flood extent, depth, velocity, and flood hazard analyses across 100, 500, and 1000 return periods using 0.1 m, 0.5 m, and 1 m DEM data derived from Unmanned Aerial Vehicles (UAV). The Ulus settlement (T & uuml;rkiye) was selected as the case study area. The primary input data include DEMs produced from the DJI Mavic 2 Pro drone and flood flows derived from the SWAT rainfall-runoff model for the Ulus River and its tributaries. Flood hazard analyses were conducted using the 2D LISFLOOD-FP hydrodynamic model using a constant Manning n value (n = 0.035) to evaluate spatial resolution variability. The findings indicate that as the resolution improves from 1 m to 0.1 m, the model runtime increases significantly (711 times on average), as do average calculation errors. Meanwhile, the spatial distribution of flood extent and hazard classes for people decreases by approximately 0.7% to 9%. However, the differences between resolutions from 0.1 m to 1 m in terms of depth, velocity, and inundation extent are not substantial. Therefore, using a 0.1 m UAV DEM resolution is not cost-effective, especially for large areas, due to the increased model instability caused by surface roughness and the longer processing time required.