Résumé
Simplified 1-D numerical models and more sophisticated 2-D models are usually used for computing the water levels and inundation extent in urban area during flood events. In this paper, a comparison among three different models, with varying degrees of simplifications and assumptions, is presented. The first model relies on 2-D shallow water equations and uses a refined mesh resolution to represent the urban area topography. The second one is a macroscopic (large-scale) 2-D shallow water model, in which buildings are not represented, thus allowing urban flow patterns to be modelled without a need for thorough mesh refinement. The reduction in both water storage and exchange section, due to the presence of buildings and other structures, is modelled by the introduction of porosity in the governing equations. The third model is a 1-D GIS-based kinematic wave model that uses a conceptual scheme to route flow distribution in the crossroads. The selected numerical models are applied to a laboratory experiment of urban flooding and to the October 1988's flood in the city of Nîmes. Results show that the use of 1-D and 2-D macroscopic models leads to a significant reduction of computational effort compared to the classical 2-D model using refined grids, with a similar degree of accuracy in predicting water depth. However, 2-D based refined mesh model provides much reasonably results for flow velocity, and then offers an estimation of flood hazard mapping at a much finer resolution and greater accuracy.