The precipitable water vapor (PW) data at 30min intervals are analyzed and ass imilated into the WRF model for the “7·10” Beijing heavy rainfall event that occurred in the period of 911 July, 2004. Observational results show that the temporal variation of the GPS/PW is an instructive factor for the analysis and p rediction of this heavy rainfall event. Using the WRF model and its 3DVar system with twoway and twonested grids of 12/4km, three 36h numerical experiments are conducted to investigate the impacts of PW with the va riation analyses at 1200 UTC of July 9 as the initial fields. Numerical experime ntal results show that the assimilation of GPS/PW observations significantly hel ps to improve the moisture analysis at the initial time, then helps to improve forecasts of the location, intensity and evolution of this heavy rainfall event