提高汛期降水过程的延伸期预报能力是目前天气预报和气候预测发展的重要方向。本文以上海梅汛期降水为例，利用非传统滤波方法提取多变量季节内分量，分析了梅汛期季节内候降水异常及其相联系的延伸期关键低频信号，进一步综合多变量低频信号建立了梅汛期候降水异常延伸期预报方法，并开展了多年的回报和试报检验。结果表明：①梅汛期候降水异常季节内分量具有显著的40～60 d低频振荡周期，与降水异常实况具有显著的正相关和较高的符号一致率;②梅汛期季节内候降水异常与超前10～35 d的热带及中高纬低频信号有关，主要包括：热带MJO（Madden Julian Oscillation）自阿拉伯海的向东传播、西太平洋副热带高压季节内活动的西北向传播、PNA（PacificNorth American）遥相关型的季节内位相转换以及东北亚冷空气的持续性异常影响;③综合上述多变量低频信号建立了延伸期候降水异常预报模型，对提前10～35 d的延伸期候降水异常的季节内分量具有预报技巧，也能较好地预报实际的候降水异常趋势。
Improving the capability of extendedrange forecast of precipitation during the Meiyu season has become an important research area for operational developments of both weather forecast and climate prediction. By taking the Meiyu season in Shanghai as an example, this paper analyzes the characteristics of intraseasonal pentad anomalies of Meiyu precipitation and its associating key lowfrequency signals on the extendedrange scale through investigating multivariable intraseasonal components extracted by nontraditional filtering. An extendedrange forecast model of pentad precipitation anomalies during the Meiyu season is further established by integrating the multivariable lowfrequency signals, and the performances of the forecast model are evaluated by the hindcasts and forecast experiments. Results show that: (1) The intraseasonal component of pentad precipitation anomaly during the Meiyu season has significant features of 40 to 60 day lowfrequency oscillation, which also has an significant positive correlation and high sign consistency rate with the observed precipitation anomaly. (2) The intraseasonal pentad anomaly of Meiyu precipitation is related to the low frequency signals from both tropics and the middlehigh latitudes, such as the eastward propagation of tropical MJO from the Arabian sea, the northwestward propagation of the western Pacific subtropical high, the intraseasonal phase conversion of PNA (PacificNorth American) teleconnection and the persistent anomaly influences of cold air activities in Northeast Asia. (3) The extendedrange forecast model of the pentad precipitation anomalies by integrating the above multivariable lowfrequency signals is statistically skillful to forecast the intraseasonal component of pentad precipitation anomalies with a leading time of 10 to 35 days. It also shows outstanding capability in predicting the trend of the observed pentad precipitation anomaly.