Abstract: A simple quasi-geostrophic barotropic vorticity equation model is used as the dynamic frame of the model in this paper. Considering that there are many random errors in model's initial values of meteorological data, and that it is not perfectly complete about model's physical processes (for example, take no account of the interaction between atmosphere and underlying surface, radiation, etc.),the random forced term is added to the model, and the Monte-Carlo method with random initial values is used. A statistical-dynamic integrated model is thus built up, and a numerical forecasting experiment of 500 hPa monthly mean height field of January 1983 has been carried out. The experiment result proves that the forecasting result of the model, considering random forcing and random initial values at the same time,is better than that by the pure dynamic model, the random initial value model and the random forced model.