自回归统计模型在延伸期预报中的应用

An Autoregressive Statistical Model for Extended-Range Prediction

主要参考美国Jones等的统计预报模型,结合本地区研究特点,建立5-10月的夏季预报模型。利用1998—2007年的NCEP逐日700 hPa风场资料,使用向量场的自然正交分解所得的主成分,结合自回归及多元回归等统计方法,建立低频（30-50天）统计模型,用最近5候的前五个主成分来预报未来5候的时间主成分,最后将空间部分与所得的未来时间部分还原为低频风场。结果表明,低频经向风预报与低频纬向风相比较好,该模型对未来3-5候长江中下游的预报结果有参考价值。

Using the Jones＇ autoregressive statistical model and 700 hPa wind of NCEP/NCAR reanalysis data in 1998--2007, and considering the characteristic of the regional climate, the summer forecast model for the extended-rang （10-30 days） are constructed for operation from May to October, in which firstly the main components of the wind fields are chosen by means of Empirical Orthogonal Function analysis （EOF）, then the lastest five pentad main components （PCs） are utilized to predict the coming five pentad PCs and finally the corresponding wind fields are derived from combination of the PC, forecast models use. The results indicate that the forecast skill scores for the low-frequency zonal winds seem to be larger, implying the coming the third to fifth pentad predictions thus given along mid- and lower-reaches of the Yangtze River.