112 T. Spangehl et al.: Intercomparing the quality of recent reanalyses for offshore wind farm planning data has to be evaluated. COSMO model Version 5.04d was operationally used by DWD from 12 December 2016 to 23 March 2017 and is the latest model version that was oper- ationally used with nudging. Therefore, COSMO-R6G2 uses COSMO model version 5.04d4. Here we present results from a first simulation stream for 2018. 2.1.4 ERA5 ERA5 is the fifth generation ECMWF atmospheric reanaly- sis and covers the period from 1940 to present. It is based on the Integrated Forecasting System (IFS) Cy41r2 which was operational in 2016 (Hersbach et al., 2020). The spatial horizontal resolution is 0.28? or 31 km (spectral truncation T639). In the vertical there are 137 levels from the surface to the model top located at 0.01 hPa or 80 km. 4DVar is used for the assimilation of a variety of conventional and satellite based observational data. ERA5 includes information about uncertainties for all variables at reduced spatial and temporal resolutions. Quality-assured monthly updates of ERA5 are published within 3 months of real time. Preliminary daily updates of the dataset are available to users within 5 d of real time. A detailed description and evaluation results are given by Hersbach et al. (2020). Access to the data is provided by Meteorological Archival and Retrieval System (MARS) and Copernicus Climate Data Store (CDS). 2.1.5 Copernicus European Regional Re-Analysis (CERRA) The CERRA system builds on the HARMONIE script sys- tem cycle 40h1.2. Several changes and optimisations in the script system have been made, compared to the reference version of HARMONIE, to make the model run more ef- ficiently in a re-analysis production environment. The AL- ADIN synoptic scale physics scheme is used including sev- eral updates from later cycles that are backported to fit with cy40h1.2. The model runs with a 5.5 km horizontal grid spac- ing and with 106 vertical levels. The model domain is some- what larger than the Euro-CORDEX domain. It runs with a 3 h cycle producing 6 h forecasts at all analysis times ex- cept at 00:00 and 12:00 UTC where 30 h forecasts are pro- duced. Information about sea surface temperature and sea ice are obtained from the Operational Sea Surface Temperature and Sea Ice Analysis (OSTIA) database (Donlon et al., 2012; Stark et al., 2008). The boundary information is taken from the ERA5 output. Upper air observations are introduced into the model through a three-dimensional variational (3D-Var) data assimilation scheme (e.g. Gustafsson et al., 2001; Lind- skog et al., 2001; Brousseau et al., 2008). Included obser- vations are the conventional observations, i.e. observations from SYNOP stations, ships, buoys aircrafts and radioson- des, together with satellite radiances from the early Mi- crowave Sounding Unit (MSU) to the latest Infrared Atmo- spheric Sounding Interferometer (IASI). In addition, ground- based Global Navigation Satellite Systems (GNSS), radio occultation GNSS, scatterometer winds, atmospheric motion vectors and surface observations from local sources are used in the analyses. A new way of creating and updating the background error covariance matrix is used in order to fol- low the current weather regime and to allow for possible evolutions in the weather over the reanalysis period (El-Said et al., 2022). The surface assimilation uses an optimal in- terpolation assimilation scheme for the surface observations (e.g. Taillefer, 2002; Seity et al., 2011). The observations in- cluded are relative humidity and temperature at 2 m height obtained from SYNOP stations as well as the snow water equivalent. Also, for the surface assimilation additional ob- servations from local sources are included. In the present study the 10 and 100 m wind speed and direction of CERRA are used (Schimanke et al., 2021b, a). The following data is used from the deterministic system (CERRA-DET). Monthly averages of the 10 m wind speed are calculated from the 3- hourly analyses (hereafter referred to as CERRA-an). The hourly 100 m wind speed and direction are obtained from forecasts (CERRA-fc) using lead time hours T + 1, T + 2 and T + 3 (T = analysis time step). Direct comparison of CERRA-an and CERRA-fc reveals only small differences. In addition to CERRA-DET, hourly data of 100 m wind speed from the ensemble of data assimilations (CERRA-EDA, lead time hours T +1, . . . , T +6 used) is analysed. CERRA-EDA is a 10-member ensemble of 3D-Var data assimilations with a 6 h cycle and 11 km horizontal grid spacing. 2.2 Simulations without data assimilation: climate simulation (COSMO-CLM) and wind atlas (WRF) 2.2.1 High resolution simulation with COSMO-CLM (HoKliSim-De) HoKliSim-De (“High resolution COSMO-CLM climate sim- ulation with ERA reanalysis forcing for Germany”) is a nearly 50-year dataset which downscales the European re- analysis datasets ERA40 (Uppala et al., 2005, for the years 1971–1978) and ERA5 (Hersbach et al., 2020, for 1979– 2019) on a higher-resolution grid for Germany. The regional climate model COSMO-CLM (COSMO model in CLimate Mode; Rockel et al., 2008; Steger and Bucchignani, 2020) is used for this downscaling in a convection-permitting setup. The COSMO-CLM is the climate version of the limited-area weather forecast model COSMO (Baldauf et al., 2011; Doms et al., 2013) and it is the community model of the German regional climate research community jointly further devel- oped by the CLM-Community (http://www.clm-community. eu, last access: 6 November 2023). It has been proven to be suitable for regional climate model simulations at grid scales between 1 and 50 km in Central Europe in numerous studies (e.g. Berg et al., 2013; Kotlarski et al., 2014; Brienen et al., 2016; Ban et al., 2021). In a recent study by Borgers et al. Adv. Sci. Res., 20, 109–128, 2023 https://doi.org/10.5194/asr-20-109-2023