Abstract: Based on annual winter wheat powdery mildew（WPM）data from 1990 to 2010，and 74 monthly atmospheric circulation characteristics data from National Climate Center of China，the key circulation characteristic indices（Atmospheric Circulation Indices，ACI），which significantly correlated with WPM area in Hebei province，were determined through Pearson correlation analysis and stepwise regression analysis.Based on these key factors，the WPM area prediction model for Hebei province was established.At first，Model I and Model II were established based on the relationship between WPM area and ACI，then Model III and Model IV were established by Bayesian classification. All these models were validated by comparing predicted WPM value and actual WPM records.The results showed that both prediction models of WPM area and WPM level were available in the end of previous December and in the end of April.The predicted WPM area was consistent with the real records，both for the historical data during 1990-2010 and the real time records during 2011-2013.The accuracy of the WPM level prediction model was 81.0% and 90.5% respectively at the end of previous December and the end of April.The maximum error of the WPM level prediction model in real time prediction for 2011-2013 was one level.The results indicated that the atmospheric circulation characteristics were significant indicators to WPM and could provide basic to predict to winter wheat powdery mildew in long term.

Key words: Winter wheat powdery mildew, Atmospheric circulation characteristic, Predict, Bayesian classification model