T. Spangehl et al.: Intercomparing the quality of recent reanalyses for offshore wind farm planning 111 satellite imaginary, research aircraft measurements and nu- merical model simulations that the wakes from large wind farms can be several tens of kilometres in length under sta- ble atmospheric stratification (Platis et al., 2018). Wakes of more than 100 km length are seen for very large wind farms in large eddy simulations (Maas and Raasch, 2022) under certain meteorological conditions. Moreover, high-resolution climate model simulations em- ploying a wind farm parameterization that considers the wind turbines as a sink of kinetic energy (KE) and source of turbu- lent kinetic energy (TKE) at rotor height indicate a distinct effect of offshore wind farms on power generation down- stream of wind farms as well as on the near-surface wind field and climate (Akhtar et al., 2021, 2022). In the present study the quality of recent regional reanaly- ses for offshore wind farm planning in Germany’s EEZ of the North Sea is assessed and compared to the quality of simula- tions without data assimilation by using in-situ observations from FINO and satellite-based data products as observational reference. The structure of the manuscript is as follows. Section 2 introduces the reanalysis, simulation and observational data. In addition, the statistical methods used for evaluation are briefly discussed. Results are presented in Sect. 3 focusing on interannual to multi-annual variability of wind speed, fre- quency distribution of wind speed and wind direction, and as- sessment of extreme wind speeds during wind storm CHRIS- TIAN at near-turbine hub height using FINO1 measurements as reference. Wind storm CHRISTIAN, 27–29 October 2013, led to severe damage in Western and Central Europe. Fur- thermore, the spatial variability of near-surface wind speed using different satellite-based data products as reference is assessed. Section 4 provides discussion of results and con- clusions. 2 Data and methodology In this section the used regional and global reanalyses, the simulations without data assimilation, the in-situ observa- tions, and satellite-based data products are described. More- over, the methods used to process (time series manipulation, regridding) and statistically analyse (histogram, shape pa- rameters, box-plot) the data are briefly explained. 2.1 Reanalyses 2.1.1 COSMO-REA6 The regional reanalysis COSMO-REA6 was developed within the Hans-Ertel-Centre for Weather Research (Kas- par et al., 2020). COSMO-REA6 is based on DWD’s for- mer operational NWP model COSMO. The model is well documented and has intensively been used by the meteo- rological community. The regional reanalysis is based on COSMO model version 4.25 which was operational at DWD from 26 September to 12 December 2012. The model do- main is adjusted to match the EURO-CORDEX region. The configuration used incorporates a horizontal resolution of 6 km with a non-hydrostatic model formulation. In the ver- tical the terrain following hybrid coordinate system con- sists of 40 main levels with 10 levels in the lowest 1000 m. The top level is located at 22 700 m (? 40 hPa). The nu- merical core of the model is integrated with a time step of 50 s. For data assimilation COSMO-REA6 employs a New- tonian relaxation scheme (nudging) to combine prognostic model variables with observations. Observations of wind, temperature, humidity, geopotential and station pressure are assimilated stemming from radiosondes, SYNOP stations, ships, buoys or aircrafts. No satellite data is assimilated by COSMO-REA6. In addition, an external analysis scheme is used for (i) snow depth, (ii) sea surface temperature and sea ice and (iii) soil moisture. The soil moisture scheme uses 2 m temperature observations for the derivation of optimized soil moisture fields. It should be noted that the assimilation of non-conventional observations such as satellite data (ra- diances) is not always possible by the nudging technique, as the observations have to be available in the model space rather than the observation space (Bollmeyer et al., 2015). Data from the global reanalysis ERA-Interim (6-hourly data) serves as boundary for COSMO-REA6. More details on the configuration and results of the regional reanalysis system are specified by Bollmeyer et al. (2015). Data of COSMO- REA6 is publicly available as part of DWD’s open data (https://opendata.dwd.de/climate_environment/REA, last ac- cess: 6 November 2023). An overview of evaluation studies and application examples with a focus on renewable energy is illustrated by Kaspar et al. (2020). 2.1.2 COSMO-REA2 COSMO-REA2 is a convective-scale reanalysis nested into COSMO-REA6. The model domain constitutes a slightly en- larged version of COSMO-DE thereby covering Germany and adjacent areas. COSMO model version 5.00.2 is used. The horizontal resolution is about 2 km. There are 50 verti- cal levels with the model top at 22 km. In addition to the con- tinuous nudging of conventional observations a latent heat nudging scheme is used. A detailed description of COSMO- REA2 is given by Wahl et al. (2017). 2.1.3 COSMO-REA6 Generation 2 (COSMO-R6G2) In order to extend the COSMO-based reanalysis and to pro- vide regional reanalysis data with short delay from real time DWD is currently producing a successor of COSMO- REA6 (COSMO-REA6 Generation 2, hereafter referred to as COSMO-R6G2) using ERA5 as boundary conditions and a newer model version (Kaspar et al., 2020). Hourly data from ERA5 is used as lateral boundaries. Moreover, the benefit of incorporating ERA5T to generate near real-time https://doi.org/10.5194/asr-20-109-2023 Adv. Sci. Res., 20, 109–128, 2023