Balancing Accuracy and Efficiency for Sustainable Flood Adaptation: Multi-Resolution LiDAR DEM Sensitivity Analysis of Urban Pluvial Flooding in the Gumi Industrial Complex

Urban pluvial flood risk in industrial zones is intensifying under climate change, yet the joint influence of digital elevation model (DEM) resolution, surface roughness heterogeneity, and infiltration capacity on simulation accuracy remains insufficiently characterized. This study presents a comprehensive sensitivity analysis combining five DEM resolutions (0.5, 1, 2, 5, and 10 m), six rainfall scenarios (10- to 200-year return periods plus the observed event of 10 July 2024), and three infiltration rates (5, 10, and 20 mm h−1), yielding 90 simulation cases executed with the open-source GPU solver SynxFlow on an NVIDIA A100 80GB GPU. A spatially distributed Manning’s roughness field (nM = 0.013–0.100 s m−1/3) was derived from the Ministry of Environment land cover product, replacing the conventional uniform-roughness assumption. Model performance was assessed against seven validation gauges (five flooded, two no-flood controls) compiled from contemporaneous news reports, using the 25 m × 25 m patch-maximum simulated depth at each gauge and probability of detection (POD), false alarm ratio (FAR), and critical success index (CSI). The 0.5 m baseline achieved POD = 0.80, FAR = 0.20, and CSI = 0.67 at the 5 cm depth threshold. Coarsening the grid reduced peak depth by up to 37% and flooded area by 5%, with the most rapid degradation occurring between 2 m and 5 m. A 2 m grid retained area error within 2% and volume error within 1% while delivering an approximately 33-fold runtime reduction relative to the 0.5 m baseline; the 10 m grid achieved up to ~1400× speedup, spanning three orders of magnitude across the resolution range. Resolution sensitivity intensified under higher rainfall and lower infiltration, confirming that “adequate” resolution is conditional on event severity. A tiered resolution selection matrix linking application scale, target accuracy, and computational cost is proposed to support evidence-based flood adaptation planning for industrial zones.