Abstract: Precipitable water vapor (PWV) derived from ground-based GNSS inversion, serves as a critical parameter for atmospheric water vapor content and plays a vital role in rainfall monitoring and forecasting. The rapid advancement of four systems GPS, BeiDou Navigation Satellite System (BDS), Galileo, and GLONASS has provided abundant data sources for water vapor inversion. However, the accuracy of water vapor inversion varies across these four systems, the PWV accuracy of different systems varies across time periods, and the precision of any single system is not consistently optimal, and the direction of errors is not uniform. Therefore, this paper proposes a multi-system GNSS PWV fusion method based on time-weighted least squares (TWLS). Fusion experiments were conducted using PWV inversion data from 17 GNSS stations in Hong Kong during 2023. Results demonstrate that the proposed TWLS method significantly enhances water vapor inversion accuracy. Compared to sounding station PWV at UTC 00:00, the TWLS method reduces root mean square error (RMSE) from 2.3 mm to 0.69 mm, an improvement of approximately 70% and decreases mean absolute error (MAE) from 1.68 mm to 0.4 mm, an improvement of approximately 76.2%. Compared to three traditional methods ordinary least squares (OLS), weighted least squares (WLS), and Bayesian model averaging (BMA), the TWLS method’s PWV at UTC 00:00 improved RMSE by 68.8%, 70.1%, and 71.6%, respectively; MAE improved by 75.5%, 76.5%, and 77.7%, respectively. Using the fifth generation ECMWF reanalysis (ERA5) PWV as the reference value, compared to the three traditional methods, the TWLS-fused PWV achieved the lowest RMSE of 2.1 mm and the lowest MAE of 1.4 mm at all observation stations.