Approche multi-échelle pour l'évaluation de la pluie dans les modèles climatiques régionaux.Etude dans le sud-est de la France

摘要

Climate projections for the end of the century indicate drier summers and more intense precipitation in the Mediterranean.In this respect, the characterization of rainfall in the region is necessary to understand its future changes.Regional climate models (RCM) are essential tools to understand the regional climate and to project its future evolution.This thesis aims at characterizing and evaluating rainfall simulated by RCM in Southeast France, typical of the mediterranean coastal regions.The description of observed rainfall at the same scales as RCM is a prerequisite.Observed and simulated rainfall is described in terms of values and spatial and temporal structure of occurrence and intensity.Weather types are used to explore the relation between rainfall features and the underlying physical processes.In the RCM, the annual total volume of water precipitated over the study region is closed to the observed values.However, this total volume is not distributed the same in space and time in the RCM simulations and in the observations.Among the multiple facets of the rainfall climatology, this study highlights the influence of the relief and of the solar cycle both in the triggering and in the intensity of rain.Rain appears to occur preferentially in the late afternoon, in connection with the daytime heating of the surface.However, the maximum of rain occurrence simulated by the RCM is earlier than in the observations, suggesting a too quick response of the convection scheme to the diurnal cycle of surface fluxes, in relation to the absence of transition between shallow and deep convection in most schemes.Besides, within the study region, the orographic forcing appears to be quite different for the two ranges of the domain and is much more pronounced over the Cévennes.The RCM reproduce the influence of the topography on rainfall features.Yet, the contrast between plains and mountains is more pronounced in the models than in the observations, especially when the large-scale forcing is weak and the convection is prevailing in the study area.The contrast accentuation between plain and relief in the models seems to be due to a too high sensitivity of the convection schemes to the air mass convergence and instability favored by the relief.