Résumé
Mediterranean catchments are prone to flash floods, caused by heavy rainfall events. The spatial variability of rainfall during these events and the availability of high resolution data have been recognised as important factors for a correct simulation of the hydrological response during flash floods. Several rainfall products are available to describe rainfall input in the Cévennes-Vivarais region: the SAFRAN reanalysis provided by Météo-France (Vidal et al., 2010), radar/raingauges reanalyse built using Kriging with External Drift (Delrieu et al., 2014) and conditional (i.e. respecting the observed values at the rain gauges) simulated rainfall fields using the SAMPO stochastic generator (Renard et al, 2011; Lebois and Creutin, 2013). In this work, we focus on the Ardèche catchment (2300 km2), for which a distributed hydrological model, called Simpleflood, has been set up by Adamovic (2014). The model mesh is composed of 238 sub-catchments, with average size of 10 km2 for which the hydrological response is supposed to be homogeneous. The different rainfall products are processed to derive hourly averages calculated on the model mesh, which are the hydrological model inputs. They are also plotted using the space-time framework of Viglione et al. (2010) using the distance of the mesh to the outlet as space variable and time. We compare several indices describing rainfall spatial variability proposed by Zocatelli et al. (2010) and Emmanuel et al. (2014) using the various rainfall products in order to quantify their similarities/differences in characterizing rainfall space-time variability. We also analyse the impact of using those different rainfall products on the simulated hydrological response.