Abstract. Water vapor is the engine of the weather. Owing to its large latent energy, the phase changes of water vapor significantly affect the vertical stability, structure and energy balance of the atmosphere. Many techniques are used for measuring the water vapor in the atmosphere such as radiosondes, Global Navigation Satellite System (GNSS) and water vapor radiometer (WVR). In addition, the method that uses European Centre for Medium-range Weather Forecasts (ECMWF) data is an important method for studying the variations in precipitable water vapor (PWV). This paper used both GNSS PWV and ECMWF PWV to establish a city-level local PWV fusion model using a Gaussian Processes method. The results indicate that by integrating the precipitable water vapor obtained from GNSS and ECMWF data, the accuracy of fusion PWV is improved by 1.89 mm in active tropospheric conditions and 2.61 mm in quiescent tropospheric conditions compared with ECMWF-PWV, reaching 3.87 mm and 3.97 mm, respectively. Furthermore, the proposed fusion model is used to study the spatial and temporal distribution of PWV in Hong Kong. It is found that the accumulation of PWV corresponds to monsoon and rainfall events.