Comparison of machine learning statistical downscaling and regional climate models for temperature, precipitation, wind speed, humidity and radiation over Europe under present conditions

There are two main approaches to downscale global climate projections: dynamical and statistical downscaling. Both families have been widely evaluated, but intercomparison studies between the scarce, usually limited temperature precipitation. In this work, we present a comparison downscaling model (SDM) based on machine learning six regional models (RCMs) from EURO-CORDEX, for five variables of interest: temperature, precipitation, wind, humidity solar radiation under climatic conditions. The study is conducted at continental scale over Europe, with spatial resolution 0.11° daily data. SDM RCMs driven by ERA-Interim reanalysis, observations taken gridded dataset E-OBS. Several aspects evaluated: series, mean values extremes, patterns also temporal aspects. Additionally, multivariable index (fire weather index) derived fundamental has included. better scores than all evaluated metrics only few exceptions, mainly related an underestimation variance. After bias correction, both similar results, no significant differences among them. Results presented here, combined low computational expense SDMs availability some CORDEX domains, should motivate consideration same level as official providers information, its inclusion in reference sites. Nonetheless, further analysis crucial such impact long-term trends or sensitivity different methods being models, needed.