Warning ! What makes a meteorological information relevant during a flash-floods warning process ?

Warning remains a main way to reduce losses during flash-floods. Because of uncertainty and multiple space-time scales, it is very difficult to issue a warning everyone finds satisfactory. In spite of many progresses realized in the past 30 years, the warning for flash floods is currently still criticized because of its non-efficiency. The September 2005 event in the Gard region in France gave a new example of its limits. We argue that the main reason why we are not able to improve further flash-flood warnings is rooted in the usual warning approach. Now, we should look at the warning with an other point of view and conduct comprehensive studies about warning processes. It means that, with a integrated approach (Montz and Gruntfest, 2002), warning should be considered from meteorological data acquisition (Morss and Ralph, 2007), until people realise something is about to happen. We will define warning as a process, that transforms perceived environmental phenomena into significant signals for people. In this communication, we will focus on the interactions between the French national weather service (Meteo France) and the other stakeholders of the warning process. The key question is : what makes a Meteo-France signal relevant according to the meteorology forecasters themselves / hydrology forecasters / mayors / inhabitants ? We will use data from our research field: the Vidourle basin, in the Gard region in France, where severe flash-floods occurred in the past decade. My talk will be based upon cross-analysis of: Interviews with warning process stakeholders: weather and flood forecasters, mayors, department chief, emergency services, inhabitants... Operational data produced during the September 2005 event : “vigilance” maps from Meteo France, the mayor teams hand-recorded informations, etc.. Further on, we will discuss the cause and consequences linked to the advent of commercial companies, who offer decision-making help to mayor teams.