Efficient urban flood simulation using a GPU-accelerated SPH model

Urban flooding may lead to significant losses of properties and lives and numerical modelling can facilitate better flood risk management to reduce losses. Flood modelling generally involves seeking numerical solutions to the shallow water equations (SWEs) or one of the simplified forms using the traditional numerical methods including the finite difference method (FDM), finite volume method (FVM) and finite element method (FEM). Recently, a relatively new approach, smoothed particle hydrodynamics (SPH), has also been used to solve the SWEs and encouraging results have been reported. However, the SPH method is computationally too demanding for efficient simulations, which has been one of the major disadvantages dogging its wider applications. This work presents an SPH model that is computationally accelerated by modern graphic processing units (GPUs) for efficient urban flooding modelling. The model's predictive capability and enhanced computational efficiency are demonstrated by application to experimental and field-scale hypothetic urban flood events.