Analogue methods and <scp>ERA5</scp> : Benefits and pitfalls

Perfect prognosis statistical downscaling relies on the relationships established using observational data for predictands and predictors. Predictors are often retrieved from reanalyses, which considered pseudo-observations. The impact of choice a reanalysis dataset performance method is usually overlooked, as global reanalyses frequently assumed to be equivalent last few decades data-rich regions such Europe. However, it was recently shown that can have bigger skill than predictor variables. Generally, processed by more recent atmospheric models assimilate perform best. This work aimed at assessing extent potential gains use ERA5, following its release, compared other reanalyses. assessment carried out six variants analogue methods, techniques, predict daily precipitation 301 stations across Switzerland. ERA5 proved one best performing different methods. Due availability, we recommend 20CR applications starting between 1851 1900, CERA-20C those 1900 1950, after 1950. high spatial resolution (0.25°) turned trap simple calibration techniques. domains over fields need optimized, high-resolution grids come along with numerous sub-optimal local solutions. An enhanced procedure, thus, must used. Besides risk poorly-calibrated domains, also requires much higher computational time no gain in skill, provided predictors relevant synoptic scale. Although should choice, lower provide performance.