An urban flood inundation model accelerated by the parallel acceleration technology

Abstract

Due to factors such as changes in land use and climate change, floods are increasingly occurring worldwide, resulting in excessive property damage and casualties in urban areas. Numerical simulation techniques can provide valuable support in mitigating urban flood risks. This study developed a coupled flood inundation model for one-dimensional sewer and two-dimensional surface based on parallel acceleration technology. The main findings include: 1. The model was validated in Omihachiman City and Shanghai City, demonstrating satisfactory results in flood inundation simulations and confirming the model's reliability in simulating flood processes. 2. A comparison of simulation times between the surface inundation model's serial version, CPU-accelerated version, and GPU-accelerated version was conducted. The GPU-accelerated version showed significant speed-up compared to the CPU model using the same numerical algorithms, with better performance as computational units increased. 3. The performance of the model is significantly influenced by the underground sewer model.