Noise from wind turbines is of concern in the planning process of new wind farms, and accurate estimations of immission noise levels at residents nearby are required. Sound propagation from wind turbine to receiver could be modelled by a simplified standard model assuming constant meteorological conditions, by an engineering method taking atmospheric and ground propagation conditions into account, or by a more exact model. Epidemiological studies have found a higher frequency of annoyance due to wind turbine noise than to other community noise sources at equal noise levels, indicating that the often used simplified model is not sufficient. This paper evaluates the variation of immission sound levels under the influence of meteorological variation and explores if the prediction of levels could be improved by taking the effect of wind speed on sound propagation into account. Long-term sound recordings and measurements at a distance of 530 m from a wind turbine show that the simplified standard model predicts the average sound pressure levels satisfactorily under downwind conditions, and that a more complex propagation model might not be needed for wind turbine noise at a relatively short distance. Large variations of sound immission levels at the same wind speed were however present. Statistical analysis revealed that these variations were influenced by meteorological parameters, such as temperature, static pressure and deviation from ideal downwind direction. The overall results indicate that meteorological factors influence the noise generated by the wind turbine rather than the sound propagation.