Current urban washoff models still rely on empirical catchment-scale functions, that have not been substantially updated during the last 40 years. This paper introduce a new approach using the physical model FullSWOF to evaluate urban washoff process. The modelling approach is performed for a Parisian road catchment. Water flow simulation is validated by outlet discharge measurements and local observations of water depth. Water quality modelling of three classes of particles (d50 = 7 μm, 70 μm, and 250 μm) is applied using the Hairsine-Rose model. Analysis of the washoff process at the catchment scale indicates that most (> 90%) of the finest particles are removed at the beginning of a rainfall event, about 10%–20% of medium-sized particles are moved over the latest part of the event, and almost no coarse particles can be transferred into the sewer inlet. Spatial analysis of washoff process reveals that the concentration of suspended solids on road and sidewalk surface is more sensitive to rainfall intensities than that on gutter surface, while coarser particles tend to accumulate in the gutter over the later part of a rainfall event. Investigation of the driving force behind the detachment process indicates that rainfall-driven effects are two orders of magnitude higher than flow-driven effects. Moreover, it is observed that rainfall-driven detachment is considerably decreased with the rising water depth, while flow-driven detachment occurs only in gutter areas. Finally, several controversial arguments on the use of physical models for assessing the washoff process, and perspectives on development of physical urban washoff models are discussed.