Mesoscale weather systems and associated potential wind power variations in a midlatitude sea strait (Kattegat)

Abstract

Abstract. Mesoscale weather systems cause spatiotemporal variability in offshore wind power, and insight into their fluctuations can support grid operations. In this study, a 10-year model integration with the kilometre-scale atmospheric model COnsortium for Small-scale MOdelling – CLimate Mode (COSMO-CLM) provided a wind and potential power fluctuation analysis in the Kattegat, a midlatitude sea strait with a width of 130 km and an irregular coastline. The model agrees well with scatterometer data away from coasts and small islands, with a spatiotemporal root-mean square difference of 1.35 m s−1. A comparison of 10 min wind speed at about 100 m with lidar data for a 2-year period reveals very good performance, with a slight model overestimation of 0.08 m s−1 and a high value for the Perkins skill score (0.97). From periodograms made using the Welch's method, it was found that the wind speed variability on a sub-hourly timescale is higher in winter compared to summer. In contrast, the wind power varies more in summer when winds often drop below the rated power threshold. During winter, variability is largest in the northeastern part of the Kattegat due to a spatial spin-up of convective systems over the sea during the predominant southwesterly winds. Summer convective systems are found to develop over land, driving spatial variability in offshore winds during this season. On average over the 10 summers, the mesoscale wind speeds are up to 20 % larger than the synoptic background at 17:00 UTC with a clear diurnal cycle. The winter-averaged mesoscale wind component is up to 10 % larger, with negligible daily variation. Products with a lower resolution like ERA5 substantially underestimate this ratio between the mesoscale and synoptic wind speed. Moreover, taking into account mesoscale spatial variability is important for correctly representing temporal variability in power production. The root-mean square difference between two power output time series, one ignoring and one accounting for mesoscale spatial variability, is 14 % of the total power generation.