A Flash Flood Numerical Modeling and Forecasting Tool in Mountainous Small Catchments Based on a 2‐D Hydrodynamic Model

Abstract The forecasting and early warning of flash floods in mountainous areas are extremely challenging. Here, we establish an integrated model of Baseflow‐Rainfall‐Interception‐Flood (BRIF) to support efficient numerical modeling and potential forecasting of flash flood in future. The local inertia approximation equations and the heterogeneous parallel computing scheme of CPU + GPU adopted in the BRIF model have realized a high performance and robustness solver, and the generability across frameworks has been completed with low cost and high efficiency. The findings suggest that the local inertia approximation equations remain a viable solution for simulating flash floods using heterogeneous parallel computing methods. For purpose of potential application in forecast and early warning, we develop a model that utilizes land‐use types to determine the underlying surface parameters. The possibility risk of flash flood is proposed to be evaluated by comparing predicted discharge with the peak flow during the return period discharge curve. The efficiency of the BRIF model has been verified by comparing with the analytical solution and multiple real flash flood events. It is shown the speedup (the computational time verse the rainfall procedure) is several to tens of times in a high‐resolution grid. Thus, the current BRIF model proposed has great potential for flash flood forecasting in future.