Wind turbine noise is often mentioned as one of the limitations restraining the deployment of wind energy in rural areas. So far, the mechanisms behind the noise annoyance induced by wind turbines have been studied mostly using field recordings. However, optimizing the design, location and operation of modern wind turbines to both increase energy production and limit noise pollution clearly benefits from accurate prediction models. Work in this field has received much interest and physics-based models of wind turbine noise have been developed. Recently, such a model has been proposed by the authors where equivalent sources representing the leading-and trailing-edge noise emission of the blades are characterized using Amiet's theory and RANS simulations. These sources are then propagated in the far-field using standardized engineering models suited to take into account topography, ground properties, atmospheric and weather effects. Furthermore, an auralization technique allows the generation of audio signals for 3D noise rendering and perceptual evaluation. In this paper, the performance of the proposed auralization system is evaluated through listening tests comparing recorded and auralized wind turbine noise. Recorded samples were obtained on an existing site, with dedicated instrumentation to measure the environmental noise under different wind conditions at several locations.