Abstract: Satellite mobile channel can be described as a fading channel model based on Markov process. Firstly, the predictability of satellite mobile channel is analyzed in this paper. Then a long term prediction method is proposed based on weighting prediction. The method resamples the channel sample values with a sample rate 2 times of the Markov state transferring speed, and uses the correlation of the channel states and Markov state transfer probability to predict the future channel state and outputs the channel prediction value of the long term future according to autoregression prediction model. Simulation results show that by using the proposed method the mean square error can reach about 10-2 when signal noise ratio is relatively high, moreover, the system average bit error rate can be reduced and the system throughput can be increased. This method can be used to instruct the adaptive transmission and adaptive resource assignment of satellite mobile communications.