As the use of weather radar images is growing up, the link between temporal and spatial variability of rainfall and the hydrological response of catchments becomes an important issue. This growth raises the following question: at which level of detail the spatial distribution of rainfall should be taken into account in rainfall-runoff modeling? This question has been tackled by numerous studies. However no clear conclusions have been established yet. The goal of this PhD is to evaluate in which context a detailed knowledge of rainfall is relevant for the modeling of runoff at the outlet of catchments of about ten to several hundred km² and, this, by simulation. In order to overcome modeling and rainfall data errors, to control the rainfall variability as well as the characteristics and the hydrological behavior of catchments, a simulation chain, including a rainfall simulator, a catchment simulator and a distributed hydrological model, has been used. Two levels of rainfall field knowledge have been considered: a distributed and a mean knowledge. Results of numerous tests show that rainfall field variability can be beneficial in order to estimate outflows. However those benefits highly depend on rainfall and catchments characteristics. A case study, carried out on a real catchment and using rainfall fields measured by weather radar, has allowed us to begin to verify the accuracy of the results obtained by simulation.
