中国南方冬季最低温延伸期预报方法尝试

A possible approach for the extended-range forecast of winter minimum temperature over Southern China

未来10~30天的延伸期预报一直是当前的难点问题.本文针对对极端低温事件异常敏感的中国南方地区,构建了预报时效为25天的冬季最低温的有效延伸期预报模型.利用1979/1980—2019/2020年冬季日平均再分析资料,本文研究发现,在季节内尺度,相比于原始变量,利用当候值减去其5候前值的增量方法可以放大和延长预测信号.中国南方地区的土壤温度和平流层50 hPa位势高度作为影响南方冬季最低温的关键因子,它们通过影响西伯利亚地区环流异常、北大西洋涛动以及欧亚遥相关型,对中国南方冬季最低温进行调制.由于两个增量形式的关键因子具有50天的准周期性,存在超前5候的显著自相关,超前5候增量形式的土壤温度和平流层极涡异常可以有效预报我国南方最低温增量.将预报的最低温增量,加上5候前的观测,得到的独立试报时段(2009/2010—2019/2020年冬季)的最低温与观测的时间相关系数为0.45,极端低温日命中率为57%.本研究使用的增量方法可以为当前极端天气事件的延伸期预报提供新思路.

Extended-range forecast of the next 10~30 days has been a persistent challenge.In this study,a leading 25 days extension prediction model is constructed for the winter minimum temperature(Tmin)over southern China where is extremely sensitive to extreme low temperature events.The winter daily datasets during 1979/1980—2019/2020 are from the fifth generation ECMWF reanalysis.The pentad increment method is suggested to build the forecast model,in which 5 pentad increment is calculated by subtracting the leading 5 pentad value from the current day value.Compared to the original format,the pentad increment format can amplify and extend the prediction signal at the intraseasonal scale.The soil temperature and the 50 hPa geopotential height in the stratosphere are chose as predictors.They can modulate the Tmin by influencing the circulation anomalies in Siberia,the North Atlantic Oscillation,and the Eurasian teleconnection pattern.Based upon the quasi-50-days periodicity and 5 pentad significant autocorrelation,above two predictors in the pentad increment format can skillfully predict the pentad increment of Tmin at leading 5 pentads(PI_Tmin).The predicted PI_Tmin is add to the observation at 5 pentad ago to obtained final predicted Tmin.During the independent hindcast period of 2009/2010—2019/2020 winter,the correlation skill of Tmin is 0.45 and the hit rate for extreme cold days is 57%.The incremental method used in this study can provide new ideas for the extension period forecasting of current extreme weather events.