Abstract: In order to make relevant agricultural precaution in time and reduce the impacts of the cold disaster on cotton production,the cotton heat indexes were calculated and the cool injury indexes were defined in the main cotton production regions in Xinjiang Uygur Autonomous Region,by using the data of mean daily air temperature,cotton growing period and local cotton yields from 1961 to 2005.The occurrence status and extend of the cool injury were well decided in main cotton regions in Xinjiang during the predicted year.Considering 74 types of atmospheric circulation characteristics(ACC),the stepwise regression models of cotton heat indexes were set up,which was monthly calculated from the cotton seeding to stopping growth in the main cotton regions of Xinjiang.The results showed that each model could well forecast heat status in each region during the cotton growing period,and the all of the accuracy rate was above 90%.The characteristics of polar vortex in high latitude before the predicted month had important effect on the heat conditions of cotton growth in Xinjiang.The validation showed that the predicting accuracy rate for cotton heat indexes was effectively improved and the cold years were predicted more accurately by the method of rolling forecasting monthly.This study provided a favorable basis for the arrangement and measures of the cotton production.

Key words: Cotton, Cotton, Heat index, Stepwise regression, Rolling forecast