Abstract: As an important parameter in atmospheric water vapor inversion by GNSS, weighted average temperature directly affects the inversion accuracy of atmospheric Precipitable Water Vapor (PWV), and establishment of a localization-weighted average temperature model is helpful to improve water vapor inversion accuracy. Based on the analysis of the relationship between the weighted average temperature and the surface meteorological elements by using Hong Kong sounding station’s data from 2012 to 2015, single-factor model and muti-factor model of the regional weighted average temperature are established with the least square principle which is used to explore the optimal regression equation coefficients. The results show that the accuracy of the multi-factor model is higher than that of the single-factor model, but it is not significant. The Bevis empirical formula does not meet the accuracy requirements when applied to the Hong Kong region; the accuracy and applicability of the model are analyzed and compared, indicating that the accuracy of the model established in this paper is high, and it can better meet the high-precision requirements of water vapor remote sensing.