Abstract: It is difficult to select an optimal atmospheric dry refractivity profile model from the existing models for practical applications. As the atmospheric wet refractivity profile model has a lower precision, the demand to construct an improved model of atmospheric refractivity profile using measured sounding data based on wavelet-based clustering technology and EOF(Empirical Orthogonal Function) analysis method arises. Results showed that the improved model is more accurate than the existing models. The part of dry refractivity profile model can fully consider the local climatic features, and wet part has a higher precision (about 6 unit of atmospheric refractivity) than the existing models by subdividing weather types into four categories. Meanwhile, the weather classification method can be more easily understood by non-meteorological professionals to promote the application of the improved model.