An evaluation of wind speed profiles in model-based reanalyses using ground-based measurements of high quality in the context of wind energy generation

Wind speed from atmospheric reanalyses is often used as input for modelling wind energy production in energy systems analysis. While some studies compare energy generation of wind turbines to those modelled from reanalysis data sets for specific sites, such analyses are usually aggregated to regional or national levels. However, nationwide evaluations using high quality wind speed measurements at heights relevant for modern wind turbines are still scarce. This paper presents a detailed comparison of high quality wind speed measurements of tall profiles with different reanalysis datasets at more than 75 locations in Germany measured by lidars and masts. Among the evaluated model-based products are the regional reanalysis COSMO-REA6, the global reanalysis ERA5 and the new European reanalysis CERRA. They are evaluated at different measurement heights using statistical analysis. All sites include measurement heights above 100 m and are suited for wind energy applications. This evaluation dataset provides good coverage of the relevant terrain ranging from offshore to the low mountain regions. Measurement locations are distributed all over Germany. Data was collected over multiple years (2012 2023) and measurement durations at individual locations range from months to multiple years. Many of the measurements were carried out adhering to the current standards used in wind resource assessment or have comparable quality. Thus, the dataset allows for a unique and comprehensive evaluation of the reanalysis datasets with respect to the representation of geographic and topographic features as well as seasonal patterns in the context of wind energy generation. To address current advancements in wind power generation, our analysis focuses on heights above 100 m to reflect the height of modern wind turbines. First analysis results using ERA5 and COSMO-REA6 indicate a distinct effect of the terrain on the model skill. Both reanalyses have a small median bias across all measurements with larger variations seen for ERA5. There is a height dependency in the bias of the wind speed, with positive (negative) biases for lower (higher) orographic measurement heights i.e. the terrain height at which the lidar or mast is installed. The bias varies depending on the elevation of the measurement position in hilly/mountainous terrain. A clear correlation can be observed for the bias and the difference of the terrain height at the measurement location and the orographic height of the assigned model grid box. While for elevated lidar/mast positions (higher than the model grid cell) a clear tendency towards higher measured wind speeds can be observed the effect vanishes for measurement sites close to the orographic model height.