中文摘要：

2～4周的延伸预报是近年来国际上天气和气候业务预报发展的一个重要方向。本文以江淮梅雨区降水为例,在利用集合经验模态分解（EEMD）及多变量EOF方法获取梅雨区降水及其影响系统低频信号的基础上,采用最优子集回归方法、经验波传播（EWP）方法及全球海气耦合模式产品,对梅雨季节内演变的延伸期预报方法进行了预报和试验,以期为建立延伸期预报业务提供科学依据。试验结果表明：（1）大气季节内振荡对梅雨区降水的延伸预报具有重要的应用价值,可能是联系天气过程和异常的重要系统。（2）通过EEMD方法提取前期降水演变及影响因子的季节内振荡信号,采用最优子集回归统计学方法对梅雨区逐候降水量演变进行超前30天预报是有可能的。（3）EWP经验动力方法对热带ITCZ活跃异常的未来40天东传可能具有较好的预报效果,还可能较好地预报出延伸期的梅雨区风场距平演变,具有一定应用价值。（4）全球海气耦合动力模式输出产品在延伸期环流形势趋势预报及20天左右的MJO指数预报方面有一定的参考价值。

英文摘要：

2 - 4 weeks extended range forecast （ERF） has become an important research area for operational devel- opments of both weather forecast and climate prediction in the last decades. In order to provide scientific clues to es tablish ERF, by taking Meiyu over the Yangtze- Huaihe basin as an example, based on intraseasonal oscillation （ISO） signals of rainfall, key influential systems and strong signals of tropical MJO extracted by Ensemble Empiri- cal Mode Decomposition （EEMD） and multi-variable EOF, optimal subset regression （OSR） statistical method, em- pirical wave propagation （EWP） dynamical method, and global ocean-atmosphere coupled model output are respec- tively employed to explore ERF of rainfall variation over the Meiyu region. Experimental results suggest that （1） the ISO may be an important system contacting weather processes and weather anomaly, which can be potential sig- nals of the extended range forecast for pentad rainfall over the Meiyu region. （2） LFO （Low Frequency Oscillation） predictors extracted from rainfall in the Meiyu region during earlier period by EEMD can be used to forecast pentad rainfall amount in the Meiyu region about 30 days in advance by the OSR statistical method. （3） For the EWP meth- od in dynamical sense, it can well forecast the eastward propagation of the tropical ITCZ 40 days in advance. At the same time, the EWP method has potential ability of forecasting wind anomaly over the Meiyu region. （4） The global ocean- atmosphere coupled model output is a worthwhile reference for the ERF of atmospheric circulation and MJO index 20 days in advance.