This paper investigates the relationship between wind turbine noise annoyance, exposure indicators, operational characteristics and environmental variables. A six-month field experiment at an industrial site near a residential area includes regular on-line annoyance reports, continuous 1/3-octave band noise level registrations, periodic sound recordings, data on electricity production per minute and meteorological observations. Here the risk of high annoyance does not only depend on the angular blade velocity, but also on the wind turbines' nacelle position relative to the location of the dwellings, i.e. the wind direction. This directivity effect can be captured when noise parameters such as the background noise level caused by other sources and a so-called fluctuation-indicator are introduced, the latter calculated from the 1/3-octave band spectra to quantify the periodic part of wind turbine noise. In addition, the calculated turbine's specific emission levels are closely related to the angular blade velocity, and an important parameter to predict the risk of high annoyance. Finally, these results suggest that operational restrictions based on wind direction together with the angular blade velocity might help to reduce noise annoyance while preserving cost-effectiveness.