The goal of this paper is to identify the respective influence of different characteristics of urban patterns on urban flooding. A set of 2,290 alternate urban patterns was generated randomly using an urban generator tool providing the geometry of buildings and their relative location to the ground, over a square area of 1 km². Steady 2-D hydraulic computations were performed for these 2,290 different urban patterns with identical hydraulic boundary conditions. The computational time was reduced by using an anisotropic porosity model. This model uses relatively coarse computational cells; but preserves information from the detailed topographic data through the use of porosity parameters. Based on the computed maps of waterdepths for the 2,290 urban patterns, a sensitivity analysis based on a multiple linear regression was performed to outline the most influential urban characteristics. ; Peer reviewed