Abstract: In this paper, a multiple-grid sia-sea coupled stochastic dynamic model whose main variables are height field and sea temperature field is designed and deduced according to the principles of dynamics and thermeodynamics of the. atmosphere, moreover, and used for forecasting experiments on 500hPa height field. The integral coefficients in this model are valued out by inversion from historical data, and the samples of the forecasting experiments are independent.Presuming the stochastic terms in the equations to be white noises, a series of tests on different orders of magnitude of white noises is performed. The results show that the model is more efficient when it adds white noises than it doesn't,which means that the stochastic effects of secondary factors in the process of air-sea interaction should not be ignored. This model is of certain forecasting. abibity and suitable for long-term forecast to atmospheric elements in local districts.