In order to reduce the effect of year temperature variations on effective accumulated temperature forecasting
error，we calculated the effective accumulated temperatures of 12 years separately and the average level in two effective
accumulated temperature ranges ( 10 － 30℃，10 － 35℃ ) with a single sine model in Farm 121，Xinjiang． We then
predicted eclosion peaks of Helicoverpa armigera ( Hübner) in different years based on the average level． Finally，we
corrected the number of forecasted error days through the difference between the average temperature of the forecasted year
and that of the 12 year period ( date from effective accumulated temperature above 0℃ to eclosion peak) ． The results
indicate that the prediction error was larger when there were larger gaps between single years and 12 year average
temperatures． A linear regression model of all generations was significant level ( P ＜ 0． 05 ) ． In the two effective
accumulated temperature ranges，the number of corrected forecasted error days was fewer than uncorrected in all
generations． The method was more effective for overwintering generations; historical coincident rates of all generations
were 83. 33% ，100% ，100% and 100% ，100% ，93. 33% ，respectively． This method can improve forecasting
accuracy，especially for generations and regions in which average temperatures vary greatly between years． It also can
provide a theoretical foundation for reducing the forecasting error for many other pests．