124 T. Spangehl et al.: Intercomparing the quality of recent reanalyses for offshore wind farm planning when compared to observations. A similar behaviour is ev- ident for MAM. Here the observations show maximum oc- currences at sector E and sector SW. The maximum at sec- tor SW seen in observations is shifted to sector WSW in all model-based products. The maximum at sector E is cap- tured by the model-based products though slightly underes- timated. This underestimation goes along with slightly too frequent occurrences of wind directions in sector ESE. In JJA the observations show a clear maximum at sector WSW. Here only COSMO-REA6 shows maximum occurrences at the same sector. In CERRA and HoKliSim-De this maximum is shifted to sector W. In DJF observations show maximum occurrences at sector SSW. A similar maximum is apparent in COSMO-REA6 and HoKliSim-De. In ERA5 this maxi- mum is shifted to sector SW. The seasonal variations shown (Fig. A1) are similarly represented by marine observations (ships, buoys, platforms) for the North Sea (Bundesamt für Seeschifffahrt und Hydrographie, 2023a). Figure A1. Wind rose of hourly wind speed and direction at FINO1 for heights at or near 100 m. Results are shown for seasons (column 1 to 4 panels) DJF, MAM, JJA and SON. First row panels show results obtained from hourly FINO1 measurements (10 min averages) of wind speed at 102 m and wind direction at 91 m. Consequtive rows show results at FINO1 grid points of model-based products (row 2 panels) COSMO-REA6, (row 3 panels) ERA5, (row 4 panels CERRA-fc and (row 5 panels) HoKliSim-De. The time period is 2004–2009. Adv. Sci. Res., 20, 109–128, 2023 https://doi.org/10.5194/asr-20-109-2023