华北—东北地区南部汛期降水影响因子分析及季节预报模型的建立

Screening of predictors and development of a statistical prediction model for flood-season precipitation in North—southern Northeast China

华北—东北地区南部夏季受东亚夏季风的影响,频繁发生干旱和洪涝等灾害,但华北—东北地区南部汛期降水成因复杂,其季节预测因子的选择和预报模型还需进一步研究。由经验正交函数分解(EOF)得到1981—2020年华北—东北地区南部汛期(7—8月)降水前两个模态(分别为整体一致型、南北相反型的空间结构),基于信息流特有的因果关系挑选前两个模态的预报因子,并通过多元逐步回归进一步筛选预报因子、建立季节统计预报模型。第一模态降水对应的主成分时间序列(PC1)的预报因子经过筛选有3个,分别为超前5个月的热带西印度洋海表温度(SST)、超前6个月的西西伯利亚平原850 hPa经向风以及超前2个月的热带中太平洋850 hPa经向风。第二模态降水对应的主成分时间序列(PC2)的预报因子经筛选有4个,分别为超前2个月的南印度洋中部SST、超前3个月的鄂霍次克海地区向外长波辐射(OLR)、超前2个月的热带西太平洋850 hPa经向风以及超前9个月的北大西洋涛动指数(NAO)。基于前两模态的回报/预报时间序列的重构场与观测场的时间相关系数(TCC)全场平均达到0.46,实际两个模态的时间序列重构场与观测场空间相关系数(PCC)的年际变化较大,而回报/预报时间序列重构场和观测场年际变化与其类似,其40 a平均PCC为0.35,说明前两个模态重构拟合较好的年份预报技巧也较高。

Drought and flood disasters occur frequently in North China,where summer precipitation is under strong influence of the East Asian summer monsoon.Further studies are necessary to analyze predictors of summer precipitation and its seasonal prediction in North China.The first two leading modes of summer(July—August)precipitation in North China during 1981—2020 are the whole anomaly mode and dipole mode obtained by empirical orthogonal function(EOF)analysis.Based on the causality of the information flow method,the predictors of the first two leading modes of summer precipitation in North China are selected first and further screened by multiple stepwise regression,and a statistical prediction model for summer precipitation in North China is then established.Three predictors are selected for the first leading mode,i.e.,5-month-lead sea surface temperature(SST)in the western tropical Indian Ocean,6-month-lead 850 hPa meridional wind in the West Siberian plain,and 2-month-lead 850 hPa meridional wind in the tropical mid Pacific.Four predictors are selected for the second leading mode,i.e.,2-month-lead SST in the central southern Indian Ocean,3-month-lead outgoing longwave radiation(OLR)in the Sea of Okhotsk,2-month-lead 850 hPa meridional wind in the tropical West Pacific,and 9-month-lead North Atlantic Oscillation(NAO).The area averaged time correlation coefficient(TCC)between the reconstructed field based on the hindcast/forecast PCs of the first two modes and observations is 0.46.The pattern correlation coefficient(PCC)between the reconstructed field based on the hindcast/forecast PCs of the first two modes and observations show a large interannual variation,and the 40-year average PCC is 0.35,which is similar to the inter-annual variation of PCC between the reconstructed field based on actual PCs of the first two modes and observations.It is found that the prediction model performs well in the years when the precipitation can be reconstructed by the first two modes.Finally,possible mechanisms for the impacts of these predictors on summer precipitation in North China are preliminarily discussed.