Abstract: Some experiments were made to determine the relative importance of the meteorological variables in the initial condition for the numerical rain storm prediction. A time integration has been made with a 5 L-FLM, using a complete set of realistic initial conditions. This is called as a "control run". Next, the integration is repeated, but one of the initial variable is modified. Both resuits are compared. If the solutions are found to be essentially same, it may be concluded that the variable is redundant. Three variables: moisture, boundary wind and surface pressure are tested. It is found that the initial condition of the moisture and boundary layer wind are redundant. They will be adjusted to the "control run" in 6-12 hrs because of dynamical coupling. While the initial surface pressure is important in the numerical rain storm prediction. Using modified surface pressure as an initial variable, the predicted charts are quite different from that of the control run.