Coupled GPU-based Modeling of Dynamic-Wave Flow and Solute Transport in Floods with Cellular Automata Framework

Abstract. To achieve efficient modeling of flood flow and subsequent solute transport in real-world flood events, the present study couples the CA-based shallow water flow (SWFCA) solver of Chang et al. (2022) with a CA-based advanced solute transport (ASTCA) solver as a novel coupled approach. This coupled approach is GPU-parallelized by using OpenCL 2.1 under Nvidia CUDA to enhance efficiency. The accuracy of the coupled approach is verified and compared with a popular finite-volume-based coupled approach consisting of a Godunov-type FV-HLLC model for flood flow and a Godunov-type FV-TVD model for solute transport through four test cases. The efficiency of the coupled approach is next assessed through three real-world flood cases with massive computational cells. From the simulated results, the ASTCA solver is found to have better accuracy than the state-of-the-art FV-TVD model. Regarding the efficiency, the SWFCA and ASTCA solvers respectively outperform the FV-HLLC and FV-TVD models by 1.28–1.33 and 2.90–3.33 times. After GPU parallelization, the SWFCA solver, ASTCA solver, and CA-based coupled approach respectively speed up the simulations by 57.64–76.23, 53.55–69.88, and 56.32–74.15 times, which is a remarkable progress as the simulations are performed on a PC with a normal graphic card. Hence, the proposed approach is a useful tool for real-world flood flow and solute transport simulations.