Scoping report on downscaling of climate model outputs for CRE estimation | Rapport d'orientation sur la régionalisation de sorties de modèles climatiques pour l'estimation des énergies liées au climat

[Departement_IRSTEA]Eaux [TR1_IRSTEA]ARCEAU [ADD1_IRSTEA]Hydrosystèmes et risques naturels

Inglés. This report gives an overview of methods currently used for generating space-time scenarios of multiple surface meteorological variables from outputs of climate models. Conversily to number of recent review publications on the topic of downscaling, the present report focuses on meteorological variables which drive Climate Related Energies (CRE). Climate Related Energies considered here are presented in a first introduction section. They are mainly wind, solar, hydroelectricity power. Requirements for generating relevant scenarios at regional scale for those variables are highlighted. We next review the different downscaling methods, dynamical, statistical and hybrid ones, as well as post-processing methods of climate model outputs. Postprocessing methods include bias correction and perturbation ones. A special attention is paid on statistical downscaling methods including perfect prog methods, weather generators. We also make a focus on the critical issue of multivariate and multiple site generation. Strenghts and limitations of the different methods are reviewed in the third section. A fourth section is dedicated to methods evaluation. We present some classical evaluation approaches / scores mostly refering to a climatological evaluation. We also present chronological and impact oriented evaluations. The critical issue of evaluating model temporal transferability for application in a modified climated context is also treated. Specific evaluations for probabilistic predictions are also reviewed. In section 5, we give an overview of key issues relative to uncertainties associated to future predictions obtained from climate models and downscaling methods. Recommandations for the potential improvement of such approaches and of the relevance of associated generated scenarios are finally drawn in the conclusion.