Evaluation of regional climate model temperature and precipitation outputs over Scandinavia

Abstract

Present-day precipitation and temperature simulations from regional climate models (RCMs) are compared with 2 re-analysis datasets, a hindcast dataset and observations over Scandinavia. The selected RCM data consist of 25 runs from the ENSEMBLES project, the re-analysis datasets are ERA-40 and ERA-Interim from ECMWF, the hindcast is NORA10 from MET Norway, and observations consist of E-OBS gridded data as well as data from meteorological stations. We studied the interannual variability and mean annual temperature and precipitation cycle for 1981 to 2000 and for 2001 to 2012 for 5 locations in Norway and Sweden: Oslo, Bergen, Trondheim, Tromsø and Östersund. The results show rather large differences between models and observations, demonstrating the need for bias adjustment of results from climate models. A model ranking is provided to indicate which model gives the best representation of present climate over Scandinavia. The performance measure used is the root-mean-square deviation of mean and standard deviation of monthly values at the 5 selected locations. The regional models RACMO2 and RCA show the smallest deviations from observed climate. Among the top-ranking model runs, most were driven by the global model ECHAM5 and some by a version of HadCM3. These 2 GCMs are also present among the worst-performing GCM-RCM combinations, indicating that selection of RCMs is crucial. For robust projections, an ensemble mean is still needed. © Inter-Research 2014.