In Mediterranean Europe, one of the richest biodiversity hotspots, GCM models forecast a faster climate change than in most other regions. Water availability is there a key factor driving vegetation distribution and growth. We designed a new bioclimatic model computing a water-availability index for Mediterranean forests. The model was calibrated in France by means of a vegetation index, based on the flora composition of 325 representative old forests. Eighty percent of the variance of the vegetation index was explained by the model. We used this model to quantify the potential impact of climate change scenarios on flora (see below), and verified that these predictions fitted the flora turnover observed in the last 15 years. Climate change vs 1961-96 mean parameters => (Predicted Flora turnover)+1 / +2°C mean annual t° => (7 / 12 %) - 10 / - 20% in spring or summer rainfall => (3 / 6 %) +2°C nd -20% spring rain (IPCC B2 for 2050 in the study area) => (18%) +1.4°C -15% spring -32% summer rainfall (1998-07 for the study area) => (25%).A new flora census performed in 2008 on the plots initially surveyed between 1995 and 1997 to calibrate our model confirmed that the flora turnover was ongoing, with a sharp increase of water-stress-resistant species and the reduction in number and cover of water-demanding species.The maps of our bioclimatic index at local as well as at regional scale showed that the potential distribution area of species is very fragmented in the landscape: drought tolerant species are already scattered in rocky and South-facing sites within the hinterland, ready to spread from these bases. Numerous favorable sites allow mesophilous species to survive in dry or hot areas, although such sites may become rarer and smaller in time. Our model, including variables describing precisely landscape structure, allows the assessment of climate change impact on flora diversity at a biologically realistic scale, our basic pixel corresponding to the average dispersal distance of a majority of species.