Abstract:   Aiming at the problem that due to the easy variability of water vapor in the atmosphere and the lack of the high temporalspatial resolution water vapor data, the forecast level of the storm is not high. The accuracy of GPS water vapor retrieval is analyzed and discussed. Using the data of CORS in Hong Kong, the time series of 1 hour atmospheric precipitable water vapor can be obtained by GAMIT software, and then compare it with the PWV obtained from sounding data and actual precipitation. The results show that: GPS／PWV and Radio／PWV has a very good consistency in the overall trend of change, and the correlation coefficient is greater than 09. The accuracy of GPS／PWV is matched with PWV obtained from sounding data, and the mean deviation is less than 1mm, and the root mean square error is less than 3mm. The atmospheric precipitable water vapor from GPS has a good correspondence with the actual precipitation,can accurately monitor the process of water vapor change and can be used for monitoring and forecasting water vapor.