ABSTRACT
Nowadays, numerical simulation in urban hydrology has become an essential tool for the development of risk prevention plans in flooding events characterized by an overloaded sewer system. Urban models need to simulate two different domains (overland and sewer network) paying special attention to the interaction between them in order to accurately predict both flows. In this work, we present the coupling between a 2D overland flow model based on Shallow Water Equations and the SWMM 5 storm sewer model for the routing within pipe networks or drainage systems. The coupling between both models is done in punctual manholes where an exchange is computed depending on the hydraulic conditions of both domains. For the surface flow, these connections are treated as exchange source/sink terms of the mass equation. On the other hand, SWMM considers the manhole discharge as an external inflow/outflow component of each surface-connected node. When the whole model is applied to realistic situations, the computational domain is usually conformed by a huge number of cells. This implies a high computational cost for most of the simulations. In order to maximize the efficiency of the whole computational model, the surface flow code is written in CUDA-C, which allows to perform the simulations using GPU (Graphic Processing Unit) devices, reaching speed-up values larger than 600 in some cases. The coupled model is validated by means of a benchmark test case where a surcharge of the sewer system of the city of Glasgow is assumed, showing a great performance and efficiency. A larger scale case is also shown in order to emphasize the GPU model performance.
