A high-performance approach with cellular automata framework and GPU parallelization for real-time waterflows and pollutant transport simulations of pluvial and fluvial floods

• A new advanced solute transport solver based on CA framework (ASTCA) is proposed. • The ASTCA solver is more accurate than the most accurate Godunov-type FV-TVD model. • A SWFASTCA approach comprising the ASTCA and SWFCA solvers is built. • After GPU parallelization, it outperforms its FV alternative by up to 74.2 times. • It finishes a 24-hour simulation with up to 110 thousand grids within 20 seconds. The present study provides a new tool for real-time high-accuracy water flows and pollutant transport modeling during pluvial and fluvial floods. Firstly, a Cellular Automata (CA)-based advanced solute transport (ASTCA) solver is developed to simulate solute transport due to turbulent diffusion, flow advection, and material decay mechanisms. Constraints are added to prevent unwanted numerical issues in high Pélect number flows from distorting the results. Through three benchmark cases, the ASTCA solver is more accurate than the state-of-the-art Godunov-type FV-TVD model with second-order accuracy. Next, the ASTCA solver is coupled with the well-performing CA-based shallow water flow (SWFCA) solver as a novel CA-based coupled (SWFASTCA) approach. Through real-world pluvial and fluvial flood cases, the ASTCA solver is more reliable than the FV-TVD model. As to the efficiency, the SWFASTCA approach is further GPU-parallelized by OpenCL 2.1 under Nvidia CUDA to see how fast it can be. From the outcomes, the GPU-parallelized SWFASTCA approach is up to 74.2 times faster than its FV-based alternative and can finish a 24-hour simulation with 110 thousand grids within 20 seconds, which is satisfactory as the simulations are performed on a PC without a professional graphics card. Hence, the SWFASTCA approach has a strong potential to become a real-time simulator for water flows and solute transport.