Input parameters of outdoor sound prediction models are related to environmental phenomena, such as atmospheric conditions and ground properties, which are variable in both time and space. In order to obtain reliable predictions, it is essential to get information on uncertainties by quantifying the sensitivity of numerical or analytical models to their input parameters, and thus determine the inputs that will be the main source of uncertainties. This paper focuses on ground parameters impact on sound propagation considering wind turbine noise. First, the implementation of ground roughness in a parabolic equation model validated against scale model measurements and analytical solution is proposed. Then, the sensitivity of the model to its ground parameters is performed with the Morris' screening method in order to access their relative influences. Three parameters are considered: the ground absorption through the airflow resistivity, the ground roughness through the roughness height, and correlation length. Results clearly show that the variations of ground roughness induce non-negligible differences in sound pressure levels regarding the ground absorption, even for high height sound source, i.e., nongrazing incidence.