Accurate information of the spatial distribution and temporal variations of water vapor is essential for the study of short-term severe weather phenomena, such as localized heavy rainfalls that often cause serious damages during the summer season on the Korean Peninsula. In particular, a good description of the initial conditions for the three-dimensional water vapor field is crucial for the simulation of convective systems and situations leading to heavy rainfall events in numerical model. Since the emergence of a new technique to retrieve an amount of atmospheric water vapor from ground-based GPS observation, numerous meteorological studies using precipitable water (PW), slant wet delay(SWD) and zenith total delay(ZTD) derived from GPS measurements have been conducted and showing encouraging results in various fields of meteorology, especially for improving moisture fields of numerical weather prediction from development (Kuo et al. 1993) to operational use (Gutman et al. 2003).

In this study, an impact of ground-based GPS-PW and ZTD assimilation on a rainfall forecast is examined. A series of assimilation experiments were performed using the WRF (Weather Research and Forecasting) model and its three-dimensional data assimilation (WRF 3DVAR) system. PW and ZTD data derived from 39 GPS sites over the Korean Peninsula is assimilated with cold-start and cycling techniques.

The result shows that the assimilation of GPS-PW is effective in reducing the error of PW in the initial condition. Both negative bias and the rms error of the PW contained in original analysis (background) are significantly reduced after the assimilation of GPS-PW. As a results, the improved moisture field in initial condition lead to positive effect on the forecast of rainfall initiation at the right time and intensity. The additional results of assimilation of GPS- PW and ZTD will be presented in more detail.