Investigating the Effects of SPH Numerical Parameters for Dam-Break Flood Prediction

Researchers regularly perform numerical simulations to study dam-break flooding and predict hydraulic quantities such as Water Surface Elevation (WSE). Mesh-based models, such as HEC-RAS, commonly solve Shallow Water Equations discretized on a computational mesh. As an emerging mesh-free approach for flood prediction, Smoothed Particle Hydrodynamics (SPH) typically involves solving unsteady incompressible Euler flow equations and requires the evaluation of numerical parameters controlling particle resolution, smoothing, dissipation, and time integration. This study examines how the time-stepping scheme, smoothing length, kernel function, interparticle distance, and artificial viscosity coefficient affect WSE predictions in SPH dam-break simulations. The analysis is based on three-dimensional SPH simulations of the Cleveland Dam failure in North Vancouver using DualSPHysics with Light Detection and Ranging (LiDAR)-derived topography. Sensitivity analysis is performed using variance-based Sobol’ indices to quantify the relative influences of numerical parameters on the WSE predictions. The findings reveal that the time-stepping scheme has the largest influence, with percentage differences of 0.87% and 0.63% in the average and maximum WSEs, respectively. The interparticle distance shows a minimal impact on accuracy beyond the optimal resolution of 298,188 particles, while the artificial viscosity coefficient has a negligible impact within the tested range of 0.2 to 0.3. This study suggests appropriate values for SPH numerical parameters for dam-break flooding.