We know that global climate change has many local impacts on environment and populations, particularly through increased flooding. These affect more and more great cities. The consequences are devastating with important stakes on several plans: society, human, economic, financial and environmental. Reduce damage and protect populations at risk of flooding is essential. This relationship "nature-society" can only be interdisciplinary as factors to control are numerous and complex. The objective is to provide an early warning system for floods that combining the waterside communities and their needs, the political decisionmakers and different scientific fields (geography, remote sensing, hydraulic, mathematics, modeling...).The experience takes place on two ponds hillsides tributaries of the Garonne in the south of France. Variable areas (500 to 1000 km ²) with lengths of 50 to 70 km, these two rivers converge to on the large city of Toulouse. But what prognostic issue without understanding the past and present of the contemporary evolution of land use, rural landscapes and the growth of urban areas ; all these key factors of flood hazards and the spatial extension of water during floods. The retrospective spatial analysis by means of satellite images with fine resolutions allows to establish a reliable, perceptible and quantifiable diagnosis to meet the needs of populations. All these data are then integrated into a chain of forward-looking models.