A computationally efficient surface-drainage network coupled hydrodynamic model for urban flood simulation

ABSTRACT A diagram summarizes the study, showing a coupled surface-drainage model, validation across four scenarios with NashSutcliffe efficiency over 0.75, and 9.35ՠspeedup via the local time stepping method. Urban flood prediction requires models that are both computationally efficient and physically accurate. This study develops a novel integrated framework that bidirectionally couples a high-performance surface runoff model with an advanced drainage network model utilizing the two-component pressure approach and a 2D-junction strategy. In addition to the widely adopted GPU acceleration, the present model represents the first attempt at applying the local time stepping for urban flood simulation, bringing an additional speedup ratio of up to 9.35ծ The model is validated against four diverse scenarios: the Malpasset dam-break, a mixed-regime pipe flow trial, a full-scale street rainfall–runoff experiment, and an urban flood simulation in Glasgow. The simulation results show excellent agreement with measured data, achieving Nash–Sutcliffe efficiency values exceeding 0.75 in most cases. When applied to the Glasgow urban catchment, the model effectively captures drainage conveyance and overflow dynamics.