中文摘要：

用1979—2000年逐日长江下游降水的50—80d低频分量和全球850hPa低频纬向风主成分,构建了长江下游降水低频分量的次季节预测的扩展复数自回归模型（ECAR）。这种基于数据驱动建模的气候预测方法,不仅能在复数空间上反映全球环流主要低频主分量和长江下游降水低频分量之间的时滞变化信息,而且能更好地描述气候系统的主要分量在低维空间中的变化规律。对2001—2014年长江下游降水低频分量进行次季节逐日变化回报试验的结果表明,50—80d时间尺度的长江下游低频降水分量的预测时效可达52d左右,预报能力明显优于自回归模型（AR）,而且6—8月的预报技巧最高。基于全球环流主要50—80d振荡型的发展和变化以及与长江下游低频降水相关的时间演变,对于提前50—60d预报长江下游地区持续多（少）雨过程很有帮助（尤其是夏季）,其中,东亚经向三极子型（EAT）是影响长江下游地区季节内降水变化的最主要的环流因子之一。

英文摘要：

Low-frequency component of daily rainfall in the lower reaches of the Yangtze River Valley （LYRV） and the princi- pal components of the global 850 hPa low-frequency zonal wind for the period of 1979 - 2000 were used to develop the extended complex autoregressive model （ECAR） for subseasonal forecast of the 50- 80 d low-frequency rainfall component in the LYRV. This type of climate forecast method, which is based on a data-driven model, can not only reflect the lagged variation information between the principal low-frequency component of global circulation and the low-frequency component of rainfall over the LYRV in a complex space, but also can well describe the variation of the principal component of the climate system in a low dimensional space. A 14 a hindcast was conducted in the recent study for subseasonal forecast of the low-frequency compo- nent of rainfall in the LYRV during the period from 2001 to 2014. These experimental results show that this ECAR model has a good skill for the forecast of the 50 - 80 d low-frequency component of the rainfall in the LYRV at a lead time of approximately 52 d. The ECAR model performs much better than the traditional autoregressive （AR） model, and it performs best in June - August. Hence, the development and variation of the major 50 - 80 d oscillation of global circulation and the temporal evolution of the relationship between the oscillation and the low-frequency component of rainfall in the LYRV are very helpful for in- traseasonal forecast of variations in rainfall in the LYRV at a lead time of 50 - 60 d, particularly in the summer. The 50 - 80 d oscillation of circulation associated with the East Asian meridional tripole pattern （EAT） is considered to be the main factor that affects the subseasonal variation of rainfall in the LYRV.