Sensitivity and scale dependence of discretization and roughness in the hydrodynamic modeling of surface runoff caused by torrential rainfall

Hydrodynamic surface runoff simulations are an effective method for assessing flash flood risks. In engineering, the lack of observations for model calibration poses a challenge. Therefore, understanding the sensitivity to specific model parameters is crucial for reliable flood protection planning. This study analyzes how surface discretization and roughness affect surface runoff generation and depression storage in a hydrodynamic 2D-model in a southern German alpine region. We compare the runoff generation across five discretization methodologies at 211 selected locations within the model domain. These locations are associated with subcatchments ranging in size from 0.2 to 4 km 2 . The discretization methodologies comprise a one-meter grid refined with survey data, a two-meter grid, a high-resolution and low-resolution irregular mesh and a four-meter grid. These are combined with seven different depth-dependent and constant roughness parameterizations. The sensitivity analysis shows that a higher depth-dependent roughness is needed to achieve comparable results to those of a coarse resolution model. Significant differences were observed with varying roughness parameterizations and meshing approaches. Modest alterations to surface resolution have the potential to yield deviations of up to 20% in maximum runoff. Coarser resolution models tend to create artificial depressions, leading to unrealistic water storage on hillsides. These findings aid in identifying the sources of sensitivity in hydrodynamic surface runoff modeling, especially in ungauged basins and provide guidance on specific model setups. This is particularly relevant given the continued use of coarse-resolution models due to computational constraints and the availability of various roughness parameterizations, while calibration data is scarce. • Sources of sensitivities of hydrodynamic surface runoff modeling are identified. • Depth-dependent roughness with high resolution increases discharge peaks by up to 20%. • Constant roughness has no scale dependency effect. • Change of surface resolution from 1 m to 4 m varies runoff peak by +-50%. • Coarse models tend to create artificial depressions.