%0 journal article %@ 0280-6495 %A Bierstedt, S., Huenicke, B., Zorita, E. %D 2015 %J Tellus A %P 29073 %R doi:10.3402/tellusa.v67.29073 %T Variability of wind direction statistics of mean and extreme wind events over the Baltic Sea region %U https://doi.org/10.3402/tellusa.v67.29073 %X It is not clear to what extent the variations of seasonal mean winds and seasonal extreme winds are related. We investigate this relationship for the Baltic Sea area by analysing two regional climate gridded data sets, coastDat2 and HiResAFF, for the periods 1948–2009 and 1850–2009, respectively. Both data sets are based on regional climate simulations incorporating information from observations with the aim of reproducing the observed trajectory of climate variables. We compare the wind direction distribution of mean and extreme wind events by analysing seasonal wind roses. Mean wind directions display a more isotropic distribution, with a seasonally varying maximum. Extreme winds are much more constrained to south-westerly and westerly directions. The co-variability in time between the wind speed along the dominant directions of seasonal mean and the seasonal extreme winds was investigated using a complex correlation coefficient. This coefficient enables the simultaneous investigation of the co-variability of two-dimensional variables, for example wind. This coefficient is small for all seasons, indicating a very weak co-variance in time between seasonal mean and seasonal extremes. Hence, deviations in the direction of the mean wind are not a good indicator for deviations in the direction of extreme winds. We also assess the spatial structure and temporal variability of mean and extreme wind statistics using a principal component analysis. The principal components exhibit no significant long-term trends over the simulation periods, although multidecadal trends are detected for some periods and seasons. In recent decades, wintertime mean and extremes shifted to a more south-westerly direction. In the other seasons, no trends in wind directions are detected. We also investigate the possibility that seasonal patterns of extreme winds might persist over several adjacent seasons. No such persistent patterns can be identified, and hence extreme winds in one season are not useful to predict extreme winds in the following seasons.