Abstract: Based on the observed 500 hPa height, the monthly scale predictability of short-term climate variation by the method of Analysis of Variance (ANOVA) is studied, and the influences of the predictability of different scale movements on the whole predictability are also examined. The predictability of 500 hPa height simulated by the ECMWF T63 spectrum model and forecasted by the National Climate Center T63 spectrum model with the same method is researched. The results show that the atmospheric predictability of two models is much smaller than the actual atmospheric predictability, especially for the movements of wave number 0-3, which have the most important influence on the whole atmospheric movement. Therefore there is plenty of potential skill to improve the climate model if the forecasting of wave number 0-3 movements is improved.