Traiter l'incertitude des projections climatiques (essai)

International audienceThe uncertainties associated with climate projections cannot be ignored when those projections are used. They arise from uncertainties surrounding developments for greenhouse gases, uncertainties arising from the climate models and the impact models (hydrology, biodiversity etc.), as well as the natural variability of the climate. To describe these different sources of uncertainty, and take them into account is not easy for engineers and managers, who are more accustomed to reasoning in a determinist framework. This article aims to demonstrate that statistics offers a number of approaches which make it possible to use a given multiscenario, multimodel climate projection. The simplest propose a descriptive summary of the information, while the more complex approaches can test the significance of the expected changes, to separate the different causes of uncertainty and measure each one’s contribution to the overall variability. The approach is illustrated by a set of projections for temperature, precipitation and mean snow depth for a mountainous region of the French Alps. It can be easily applied to any variable whose projections have been generated through an impact model which simulates the consequences of a set of climate projections for a given sector defined in social and ecosystem terms