Based on the analysis of temperature data from 60 meteorological stations in Jiangsu Province during the period 1961-2010 by using the Mann Kendall method, it was found out that both daily mean temperature and effective accumulative temperature during the growing period of winter wheat prior to winter have significant increasing trends in parallel with global warming. The statistical analysis on the agro meteorological records in different regions has shown that the sowing date of winter wheat is significantly correlated with the total and effective accumulative temperatures prior to the winter. The correlation between sowing date and effective accumulative temperature has passed the test at the 0.001 significance level. According to coupling between the atmosphere and oceans and the fact that the atmospheric circulation characteristics (ACC) of 500 hPa can indicate the weather patterns and control the weather conditions, a batch of effective sea surface temperature (SST) and ACC predictors of accumulative temperature prior to the winter were selected using the optimum correlation technique. These SST and ACC predictors are independent each other and have stable and significant correlations with the accumulative temperature. Then the models for predicting the accumulative temperature prior to the winter were developed. Calibrations and validations show that these developed models are able to predict the accumulative temperature prior to the winter with satisfied accuracy. They can be routinely operated to predict the accumulative temperature prior to the winter and to determine the optimum sowing date of winter wheat one to two months in advance.