Abstract: A rapid updated cycling assimilation and forecast system which is capab le of assimilating various kinds of the observations including the data from the local automatic weather stations (AWS) and the Global Position System （GPS）precipitable water vapor （PWV） data in the Beijing area with 3hour updated time interval was built in the Institute of Urban Meteorology (IUM), CMA. To evaluate the impact of assimilating the local GPSPWV data, two parallel experiments with and without assimilating the GPSPWV observations into the WRFbased rapid updated cycling assimilation and forecast system for the urban area were performed for 28 precipitation incidents occurring in August 2006 and in summer of 2007. From the verification scores of precipitation forecast, it's revealed that the ETS scores for almost all precipitation thresholds were improved with the local GPSPWV data assimilated, especially during the first 0-6 hours, i.e. the impact caused by assimilating the GPSPWV data was more evident during the initial stage of integration. It is found that the forecasted precipitation area was enlarged with the assimilation of GPSPWV. The underprediction of precipitation for the larger thresholds such as 10 mm/6 hours and 25 mm/12 hours was ameliorated as the consequence of assimilation of GPSPWV data. However, the assimilation of the local GPS/PWV observations in the Beijing area does not produce significant improvement in the forecast skills for both the surface and upperair prognostic variables. This is to be expected, since the distribution of the groundbased GPS sites is very narrow, and moisture is the only affected variable among these variables above. But it can be found that the forecast performance for the conventional elements in the Beijing local area is significantly improved during the first 6 hours forecast owing to the accumulation of the assimilation effects by the rapid updated assimilating. In other words, the rapid local change information of moisture unable to be captured by conventional observations can be assimilated into the model's initial conditions via the rapid updated cycling assimilation of the very local GPS/PWV data in the interval of 3 hours, which is also consistent to the improved precipitation forecast skill during the first 6 hours forecast. Therefore, the assimilation of local GPSPWV data would not only improve the forecast quality in the current cycle, but also help to prepare better background for the next cycle assimilation. It is due to this kind of accumulation of local data assimilation effects leading to the better forecast performance. In additi on, according to the results of a convective case study, it is shown that how muc h or even whether the assimilation of GPSPWV observations may bring to better precipitation forecast, is still up to the general effects of the assimilation with the other different kinds of observations incorporated. Moreover, the accumulated assimilation effects of local GPSPWV observations would be able to create a circumstance more favorable to the formation and sustaining of local convection in the model, thus causing improved forecast skills for convective precipitation.