Experimental validation of two-dimensional depth-averaged models for forecasting rainfall–runoff from precipitation data in urban areas

SummaryThis paper presents an experimental validation of two widely used numerical models in urban flood inundation studies, the two-dimensional dynamic and diffusive wave models. Instead of using the common approach in flood inundation modelling, which consists of computing the water depth and velocity fields for a given water discharge, in this study the rainfall intensity is imposed directly in the model, the surface runoff being generated automatically. Both the dynamic and diffusive wave models are implemented in the same unstructured finite volume code, removing in such a way any differences in the numerical discretisation other than the wave approximation used to compute the water velocity. Two different methods for representing buildings are used and compared, the so-called building-block and building-hole approaches. Experimental validation of the models is presented in several simplified laboratory configurations of urban catchments, in which the surface runoff has been measured for different hyetographs. For this purpose, 72 experiments were undertaken in a rainfall simulator, including eight catchment configurations and nine hyetographs. Numerical results show that the dynamic wave model is able to predict the peak discharge and its arrival time, as well as the shape of the outlet hydrograph, while the diffusive wave model gives less accurate results. The experimental validation confirms that, when the geometry of the problem is well defined, depth-averaged dynamic wave models may be used to predict rainfall–runoff from direct precipitation data in urban environments.