Abstract: Aiming at the accuracy of calculation results of atmospheric precipitation affected by the precision of atmospheric weighted mean temperature in the groundbased GPS water vapor retrieval process, the regression analysis method is used to study the sounding data of Hong Kong from 2006 to 2016, and the singlefactor and multi-factor atmospheric weighted mean temperature calculation models for Hong Kong are constructed. The two models are used to predict the 2017 weighted mean temperature. Compared with the results of various empirical formulas and true values, the deviations between the two models and true values account for 80.72％ and 85.26％ respectively in the range of -5 K~5 K, which is obviously better than other empirical formulas. Moreover, the accuracy of the weighted mean temperature calculated by seasonal modeling has not been significantly improved, but the accuracy of the calculated results can be significantly improved by diurnal modeling. Therefore, in order to improve the accuracy of the water vapor retrieval calculation, the local meteorologic data in the air should be used to construct a local weighted mean temperature calculation model, which is of great significance for improving the accuracy of GPS total water vapor retrieval.