This work contributes to the evaluation of the dynamics of the human exposure during flash-flood events in the Mediterranean region. Understanding why and how the commuters modify their daily mobility in the Cévennes - Vivarais area (France) is the main objective of the study. To reach this objective, the methodology relies on three steps: i) evaluation of daily travel patterns, ii) reconstitution of road flooding events in the region based on hydrological simulation at regional scale in order to capture the temporality and the intensity of flood and iii) identification of the daily fluctuation of the exposition according to road flooding scenarios and the temporality of mobility patterns. This presentation focuses on step ii), i.e. the implementation of the physically based and non-calibrated hydrological model CVN (Anquetin et al., 2010; Braud et al., 2010; Vannier, PhD 2013) to retrieve the hydrological signatures of past flash-flood events in the Cévennes ¿ Vivarais region. Since the regional scale is investigated, the scales of the studied catchments range from few km2 to few hundreds of km2 where many catchments are ungauged. The model parameters are derived from available information about geology, land use, soil pedology. We found that information on soil was not reliable and that soil water holding capacity was generally underestimated due to the absence of information about altered geological layers. Vannier et al. (2013) regionalized the estimation of catchment-scale soil properties (soil water holding capacity and lateral hydraulic conductivity) by means of stream-flow recession analysis. The use of the corresponding retrieved parameters in the CVN model significantly improves the agreement between observed and simulated hydrological response in gauged catchments. It is therefore taken into account in the regional modeling strategy. Two past events (September 2007; October 2008) are analyzed in terms of temporality of road flooding. The scales of risk (time and location of the road flooding) are compared to observed data of road cuts. A sensitivity study to rainfall estimation investigates the signature of several rainfall products on the simulated discharges. These products are based on radar estimation and/or rain gauge records.