Author(s): Na Zheng; Zhi Li
Linked Author(s):
Keywords: Coupled 2D-1D simulation; Parallel computing; Urban pluvial flood modeling; Virtual drainage networks
Abstract: Urban pluvial flooding is escalating as a critical hazard, but high-resolution coupled overland flow and pipe drainage simulations remain computationally intensive and heavily reliant on detailed infrastructure data. To address both challenges, we present SERGHEI-SWMM, a parallel urban flood model that integrates a GPU-portable 2D shallow water solver with a 1D drainage module, and evaluate the reliability of virtual drainage networks. The model is evaluated on the Tongji University campus, comparing simulations based on the real drainage system, road-aligned virtual networks, and virtual networks supplemented with off-road secondary pipes. Results demonstrate that off-road secondary pipes significantly enhance the spatial coverage of the drainage system, thereby improving drainage efficiency in off-road depressions. Scalability tests on CPU and GPU clusters reveal substantial speedups for the 2D surface solver, but only marginal improvements for the 1D drainage module — indicating limited benefits of GPU acceleration for drainage simulations at the campus scale. These findings underscore the reliability and practical utility of virtual drainage networks for urban pluvial flood modeling in data-scarce regions.
Year: 2026