This is a preliminary study on improving the precipitation forecast using precipitable water vapor from GPS data and SSM/I precipitation estimates in a physical initialization procedure. The ground-based GPS data provided by UCAR will be used to correct the NCEP humidity analyses. Then, a rainfall nudging scheme will adjust the vertical humidity profiles using the new humidity analyses and the SSM/I estimates. In this preliminary study, only a few sites over the United States will be used for performing these humidity analyses. However, one can still evaluate the sensitivity of the precipitation short-term prediction to this initial field. Unlike the well known precipitable water assimilation procedures, the moisture field will not be relaxed toward individual station observations or/and grid-point analyses by means of the Newtonian nudging technique. In the physical initialization approach, only vorticity, divergence and surface pressure spectral coefficients are relaxed toward to the initial conditions. The CPTEC atmospheric general circulation model, with the new physical initialization procedure implemented, will be used to forecast the precipitation fields in this study.