Two dynamical-statistical spectral models for forecasting the 500 hPa monthly mean height (MMH) in the Northern Hemisphere are described. One is based on the annual variation of MMH; the other on the monthly variation. First, the spectral equations with unknown coefficients are given by means of variation method.Then,making use of past data, the unknown coefficients are estimated by the least square method. Finally, solutions of the two models are obtained by simplex method. It is shown that the skill scores of the models are better than that of persistence forecasting.