What global reanalysis best represents near-surface winds?

Since global reanalysis datasets first appeared in the 1990s, they have become an essential tool to understand the climate of the past. The wind power industry uses those products extensively for wind resource assessment, while several climate services for energy rely on them as well. Nowadays various datasets coexist, which complicates the selection of the most suitable source for each purpose. In an effort to identify the products that best represent the wind speed features at turbine hub heights, five state-of-the-art global reanalyses have been analysed: ERA5, ERA-Interim, the Japanese 55-year Reanalysis (JRA55), the Modern Era Retrospective Analysis for Research and Applications-2 (MERRA2), and the National Centers for Environmental Prediction (NCEP)/National Center for Atmospheric Research (NCAR) Reanalysis 1 (R1). A multi-reanalysis ensemble approach is used to explore the main differences amongst these datasets in terms of surface wind characteristics. Then, the quality of the surface and near-surface winds is evaluated with a set of 77 instrumented tall towers. Results reveal that important discrepancies exist in terms of boreal winter seasonal means, interannual variability (IAV), and decadal linear trends. The differences in the computation of these parameters, which are mainly concentrated inland, reach up to the order of magnitude of the parameters themselves. Comparison with in situ observations shows that the ERA5 surface winds offer the best agreement, correlating and reproducing the observed variability better than a multi-reanalysis mean in 35.1% of the tall tower sites on a daily time-scale. However, none of the reanalyses stands out from the others when comparing seasonal mean winds. Regarding the IAV, near-surface winds from ERA5 offer the values closest to the observed IAV.