• English 
  • Spanish 

Ability of statistical and dynamical downscaling approaches to reproduce mean and heavy precipitation at different spatial scales over Europe

The present study compares statistical downscaling methods (SDMs) and dynamical downscaling with nested regional climate models (RCMs) at different target spatial scales over two European regions. All downscaling products were developed in the framework of Euro-CORDEX and use the ERA-Interim reanalysis as boundary conditions (in RCMs) and as large-scale predictor data (in SDMs). Downscaled precipitation is evaluated at two horizontal resolutions (0.11º and 0.44º) using simulations from the Euro-CORDEX RCM ensemble and SDS methods trained against gridded observational data. The objectives of the work are: (1) to compare both downscaling methodologies, to identify their respective strengths and weaknesses at (2) different spatial scales and considering (3) mean and heavy daily precipitation, and (4) to assess the impact of bias correction methods.
We focus on two target areas in Europe: Continental Spain and the European Alps, where high-resolution and high-quality gridded observational data sets are available. Moreover, these areas represent a variety of climate conditions. Over Spain, the Spain011/044 data sets are considered, which exactly match the Euro-CORDEX grid over Spain. The Alpine data set has been derived from the EURO4M gridded precipitation product provided by MeteoSwiss at 5km resolution. Observations were spatially averaged to match the Euro-CORDEX grid cells. In order to make the SDM results comparable, they were calibrated using the gridded observations as predictand and model performance was evaluated in a cross-validation framework. Bias correction methods and an additional SD step are applied to the RCM data to also incorporate observational information. The ability of the high-resolution RCM simulations (0.11°) to represent observed precipitation is furthermore assessed at their skillful scale, by aggregating the 0.11° grid to the 0.44° resolution and evaluating the added value with respect to the low resolution (0.44°) runs.