Since 2006, Meteo-France has developed an efficient vigilance system for floods throughout France. Yet, 20000 km out of the 120000 km of the French hydrographic network benefit from this monitoring so far. In order to cope with the needs of the remaining network for a flood warning system, Irstea and Meteo-France have developed AIGA (Adaptation d’Information Géographique pour l’Alerte en Crue). AIGA is a flood warning system based on a simple distributed hydrological model run at a 1km² resolution using radar rainfall information (Javelle, Demargne, Defrance, Pansu, & Arnaud, 2014). The warnings, produced every 15 minutes, result of the comparison of these real time data with statistical data to obtain warning levels. These statistical data come from the SHYREG method which is a regional flood frequency method that is based on an hourly stochastic rainfall generator coupled with a simplified distributed rainfall–runoff model (Arnaud et al., 2014). The flood warnings thus obtained provide information about the severity of the flow in Mediterranean regions but so far don’t take into account the local stakes. To verify the relevance of the warnings regarding these stakes, we first have undertaken to compare them with flood damage (C. Fouchier, 2012). Unfortunately, as many authors before us (Bourguignon, 2014; Vinet F., 2011), we have observed the lack of a comprehensive and consistent database on this matter. In France, some databases with partial coverage exist as a result of various research projects such as the HYDRATE European project which collected information about flash floods in Europe and focused, in France, on the Cevennes-Vivarais area (Gaume et al., 2009). Furthermore, the PREDIFLOOD project focused on floods inducing road breaks in the Gard administrative area (Naulin, 2012). However, databases with a larger coverage report with errors. For instance, in 2014, the international EM-DAT database has recorded 5 deaths due to the floods (D. Guha-Sapir, 2015) while we have documented 20 for the same period. In order to collect and complete the existing ones, we have thus started to build an exhaustive database compelling all damage due to flash floods in the French Mediterranean regions since 2011 (Saint-Martin, 2014). Once these data obtained and compared with the warnings, we initiated to correlate them with another type of data, namely the vulnerability of territories to flash floods. However, since no existing method has been able to give satisfactory information about the vulnerability levels of territories on such a large scale, we have undertaken to develop one. In this poster, we offer a new method for a systematic and automatized analysis of the vulnerability of territories that we described as “the external vulnerability”. This concept corresponds to the susceptibility of a stake to be damaged, depending on the only characteristics of its nature. The first comparisons between the warning levels and the localisation of the impacts sometimes show significant discrepancies. However, these discrepancies appear to be explained by the vulnerability levels of the studied area. Indeed, we observe that the impacts are concentrated among the territories with a high-level of “external vulnerability” regardless of the warnings. We may thus be able to weight the levels of the AIGA warnings based on the vulnerability level of the area surrounding the streams for which the warnings are issued. This way, the warnings would characterize not only the flood hazard but more inclusively the risk of flooding.