Atmospheric reanalyses are the only source of spatial and temporal gridded wind information at wind turbine height providing data over several decades in the past. The application potential of reanalyses in the renewable energy sector depends strongly on the quality of the meteorological quantities. While global reanalyses have a resolution of typically 50 km, new regional reanalyses COSMO-REA6 and COSMO-REA2 have about 6 km and 2 km horizontal grid spacing, respectively. Here, we investigate the added value of the new regional reanalyses for the renewable energy sector, especially their application potential for site assessment. Four well established wind towers in Europe are used as reference for this purpose. We find regional reanalyses performing significantly better or at least similar to global reanalyses. Especially marginal distributions show significant improvements e.g. the most extreme temporal wind changes (ramp rates) at typical hub-heights are underrepresented by global reanalyses between -80 and -43% while COSMO-REA2 represents them with relative errors between -14 and +9%. Considering biases, mean absolute errors, and correlations most significant improvements occur close to ground and in areas with complex terrain. Moreover, vertically extrapolated wind measurements which are commonly used for site assessment show a stronger site dependency in their performance than reanalyses. (C) 2019 The Authors. Published by Elsevier Ltd.