Résumé
This article presents a probabilistic approach to quantify the relationships between three millions of botanical observations and monthly average values of day and night temperatures, cumulated values of precipitations and number of rainy days, recorded for a period of 50 years by 574 climatic stations in France. The probabilistic calibration provides a catalogue of 2.861 taxa indicating the average climate of this period. The calibration validation, which is based upon the calculation of the difference between the climate estimated by plants and the climate measured by climatic stations (Accuracy = 88%), according to various vegetation communities (Stability = 96%), underline that plants are precise and stable indicators of climate variables. The classification of these behaviours determines phytoclimatic groups of which spatial representation shows climatic gradients of flora. Thus, these results could be used to simulate flora and vegetation distribution in France, according to the global climatic change scenarios.