Using observation data from 945 ground-based Global Navigation Satellite System（GNSS）stations in the summer of 2023 in China，the double-difference（DD）network algorithm and Precise Point Positioning （PPP）algorithm were used to invert the Precipitable Water Vapor（PWV）of the atmosphere. The accuracy and stability characteristics of the two methods for inverting PWV in different climate regions of China were studied and analyzed using PWV from radiosonde stations at the same location and ERA5 reanalysis data as reference values. The results show that compared with the PPP solution，the DD solution has a stronger correlation with the PWV of radiosonde and ERA5 data，a more concentrated bias frequency distribution，a higher probability of peak area，and a smaller range of bias. Compared with the Radio-PWV，the average biases for the DD-PWV and PPP-PWV are -0. 1 mm and 1. 1 mm，respectively，and the corresponding root mean square error（RMSE）are 2. 4 mm and 3. 1 mm，respectively. While compared with the ERA5-PWV，the average Bias of the DD-PWV and PPP-PWV are -0. 2 mm and 0. 1 mm，respectively，and the corresponding RMSE are 2. 7 mm and 3. 2 mm， respectively. The average RMSE of the DD solution is less than 3 mm，indicating that the PWV inverted by the DD network method has higher accuracy and stability. The accuracy of GNSS detection of water vapor is generally better in the western non monsoon region than in the eastern monsoon region. The RMSE of the DD solution is more concentrated around the median in various climate regions，while the PPP solution shows different levels of accuracy at different GNSS stations. The PPP solution has a greater degree of dispersion in accuracy and strong instability in temperate and subtropical monsoon climate regions with sufficient water vapor and low detection accuracy.