Temperature and Precipitation Bias Patterns in a Dynamical Downscaling Procedure over Europe during the Period 1951–2010

Abstract

The Weather Research and Forecasting (WRF) mesoscale meteorological model is used to dynamically downscale data from the Goddard Institute for Space Studies (GISS) atmospheric general circulation model (GCM) CMIP5 version (Model E2-R) over Europe at a 0.25° grid size resolution, for the period of 1951 to 2010. The model configuration is single nested with grid resolutions of 0.75° to 0.25°. Two 30-year datasets are produced for the periods of 1951–1980 and 1981–2010, representing the historic and current periods, respectively. Simulated changes in climate normals are estimated and compared against the change derived from the E-OBS gridded dataset at 0.25° spatial analysis. Results indicate that the model consistently underpredicts the temperature fluctuations observed across all subregions, indicative of a colder model climatology. Winter has the strongest bias of all seasons, with the northeastern part of the domain having the highest. This is largely due to the land–atmosphere interactions. Conversely, spring and summer have the lowest regional biases, owing to a combination of low snow cover (relative to winter) and milder radiation effects (as opposed to summer). Precipitation has a negative bias in most cases, regardless of the subregion analyzed, due to the physical mechanism employed and the topographic features of each region. Both the change in the number of days when the temperature exceeds 25 °C and the change in the number of days when precipitation exceeds 5 mm/day are captured by the model reasonably well, exhibiting similar characteristics with their counterpart means.