Abstract:Pu'er city is located in the Southwest of Yunnan Province.The onset date of summer rainfall is very important for the crop and tea production in Pu'er.Based on the variation of interannual increment of onset date of Pu'er rainy season and the atmospheric circulation and physical process affecting the onset date of the rainy season,five key predictors with physical significance are selected by using the year-to-year increment approach.The five predictors are the circumpolar circulation in the Southern Hemisphere in January,the South Pacific high in February,the sea level pressure from the Bay of Bengal to the South China Sea in April,the sea ice in northern Canada in the previous winter,and the snow depth in Iranian plateau in the previous winter.A prediction model of the onset date of the rainy season in Pu'er is established by using the multiple linear regression analysis method.The cross-validation tests for the period 1967-2017 and the independent hindcast for the period 1998-2017 are performed to validate the prediction model.In the cross-validation test,the correlation coefficient in interannual increment between predicted and observed onset date of the rainy season is 0.84,and the relative root mean square error is 24%.In the independent hindcast,the relative root mean square error of interannual increment of onset date of the rainy season is 15%,and the prediction error of onset date of the rainy season is less than 7 d,showing that the prediction model can well reproduce the change trend of onset date of the rainy season from 1967 to 2017.