Abstract: Using the theories and methods of seasonal time series model, and the monthly average atmosphere temperature from 1935 to 1945 of Yangling Demonstration Zone Zhangjiagang to build up the seasonal time series model, then make use of the model to carry on the simple forecast to the average air temperature of each month in 1946. The results show: the relatively errors of the monthly average atmosphere temperature’s prediction is bigger in Jan, Feb and Dec in 1946, and it is smaller from March to Nov in the same year, one of the reason is that annually the temperature of the northern China is lower during Jan, Feb and Dec, and the small fluctuation may cause bigger relatively error, but the temperature is higher from March to Nov, and the relatively error is smaller. Furthermore, the model is more accurate for short forecast, but for longer forecast, the error compares greatly. Therefore, this model is suitable for short forecast.