T. Spangehl et al.: Intercomparing the quality of recent reanalyses for offshore wind farm planning 113 (2023), COSMO-CLM is evaluated and used to study cluster- scale wake effects of wind farms in the North Sea. A horizontal grid spacing of 0.0275? (? 3 km) is used together with 50 vertical levels and 25 s time stepping. The simulation starts in December 1969 with 13 months spin-up time and is extended regularly close to real- time. Currently, the most commonly used variables of this dataset are available for the time period 1971–2019 on the DWD node of the ESGF (Earth System Grid Federation) as version V2022.01 (https://esgf.dwd.de/projects/dwd-cps/ hoklisim-v2022-01, last access: 6 November 2023, Brienen et al., 2022). The model domain covers 461x481 grid points and is centered around Germany. The forcing at the lat- eral boundaries has been updated every hour for the ERA5 period in a direct nesting approach. For ERA40, 6-hourly boundary data from an intermediate nesting simulation at 0.11? (? 12 km) have been used, which had been run using the same configuration as in the EURO-CORDEX-CMIP5 simulations with COSMO-CLM (see e.g. Kotlarski et al., 2014; Vautard et al., 2021). For this study, the wind com- ponents which have been interpolated inside the model to the 100 m a.g.l. are investigated. 2.2.2 NEWA (New European Wind Atlas) The New European Wind Atlas (NEWA) covers the 30-year period from 1989 to 2018. The Weather Research and Fore- casting (WRF) model was used together with ERA5 as driv- ing reanalysis for a series of one-way nested simulations. The simulation’s design is optimised to represent wind speed dis- tributions in complex terrain. Three nested domains with a resolution of 27, 9 and 3 km were used. Spectral nudging was applied in the outer domain to incorporate the observed large-scale atmospheric patterns. The simulations consist of 7 d periods using a spin-up of 24 h to achieve equilibrium of the mesoscale flow with the terrain. In the vertical the model incorporates 61 levels with the model top at 50 hPa. Data is available for 30 min intervals for wind energy relevant pa- rameters. Details are given by Hahmann et al. (2020) and Dörenkämper et al. (2020). Data for 2005–2018 is publicly available via the website https://map.neweuropeanwindatlas. eu/ (last access: 6 November 2023). The WRF model output was further downscaled to create the microscale atlas. In the present study the mescoscale data at 3 km resolution is used. 2.3 FINO observations 2.3.1 FINO (Ger. Forschungsplattformen in Nord- und Ostsee) The FINO research platforms facilitate the exploration of off- shore conditions and help to determine the effects of off- shore wind energy development on marine flora and fauna. Masts have been erected on the working platforms of FINO1, FINO2 and FINO3, on which the most important meteoro- logical parameters, in particular wind speed and direction at different heights, are measured. In addition, a complete set of hydrographic data is collected. Moreover, the forces induced by wind and waves are measured in the foundation area (Lei- ding et al., 2016). In the present study measurements of wind speed and di- rection at FINO1 are used. Wind direction data is taken from the highest measurement level available for this parameter at 91 m a.m.s.l. (above mean sea level). The wind speed time se- ries stems from the top anemometer at 102 m a.m.s.l. At this height the effect of the mast on the wind speed measurement is assumed to be small. The mast effect is corrected by apply- ing a mast correction to the wind speed data (Leiding et al., 2016). The mast correction depends on the wind direction. The simultaneous wind direction measurement of the wind vane at 91 m is included in each corrected wind speed value in order to be able to carry out the mast correction (UL Inter- national GmbH, personal communication, 2022). Moreover, measurements are influenced by a lightning protection cage which leads to slightly lower wind speeds in 4 narrowly pro- nounced wind direction sectors (Leiding et al., 2016). Here we estimate the overall error for all wind directions to be less than 1 %. The time series of wind speed and direction con- sist of 10 min averages. Hourly data at full hours is used to enable comparability with available model output data. Ad- ditional information on measurement uncertainties and data availability is indicated by Leiding et al. (2016). 2.4 Satellite observations 2.5 Copernicus Marine Environment Monitoring Service or Copernicus Marine Service (CMEMS) Monthly averaged near-surface wind speed is obtained from CMEMS. The CMEMS wind product WIND_GLO_PHY_L4_MY_012_006, https://doi.org/10.48670/moi-00185, is used. The prod- uct incorporates scatterometer observations to correct for persistent biases in hourly ERA5 model fields. Bias correc- tions are based on scatterometer observations from Metop-A, Metop-B, Metop-C ASCAT (0.125?) and QuikSCAT Sea- Winds (0.25?). The bias corrections are calculated over 20 d centered around the time of interest. Therefore, averaging hourly wind speeds from this product over a month includes some observations (10 d) from both the previous and next month (PUM, CopernicusMarineService, personal com- munication, 2023). The product provides stress-equivalent Level-4 wind components at 10 m at 0.125 and 0.25? horizontal spatial resolution and covers the period from August 1999 to February 2023. The stress-equivalent wind does not rely on the assumption of neutral stability (de Kloe et al., 2017). In the present study data at 0.125? horizontal spatial resolution is used. Hourly near-surface wind speed is calculated from components. Monthly values are obtained from the hourly wind speed by arithmetic averaging. In a previous version of the manuscript the https://doi.org/10.5194/asr-20-109-2023 Adv. Sci. Res., 20, 109–128, 2023