用于ENSO预测的一种广义典型混合回归模式及其预报试验

A GENERAL I ZED MODEL USING CANONICAL MIXED REGRESS I ONAND FORECASTING TEST TO ENSO

提出一种基于主分量典型相关分析（ＰＣ－ＣＣＡ）的广义典型混合回归模式，用于建立ＮＩＮＯ海区ＳＳＴ预报方案．该模式引入ＥＥＯＦ、ＰＲＥＳＳ准则和集成预报等技术思想，在优选物理因子，确定最佳模式参数的基础上，对ＮＩＮＯ海区海温指数所作的超前１—４季度预报试验取得优良效果．试验表明，该模式方案性能稳定，其总体预报技术水平已达到美国ＮＯＡＡ／ＮＷＳ／ＮＣＥＰ／气候诊断公报（ＣＰＣ）所用同类模式水平．而本模式方案预报同类产品所需因子数远少于ＣＰＣ方法。这就有可能为建立我国的ＥＮＳＯ业务监测系统提供有益的基础。

In terms of the method of a variant of principal component--canonicalcorrelation analysis proposed by Barnett and Preisendorfer (PC-CCA), a schemeto forecast SSTA over the area of Nino is explored, using the on criterion of PRESS.the idea of EEOF and consensus forecast. The prediction test has got excel lentresults in the leading forecast of 1-4 seasons of SST index over Nino areas,which are based on fine physical factors and optimal parameters of the model.Experiments indicate that the forecast skill is fine and stable. and beingsimilar to that of identical model in Climate Diagnostics Bulletin (publishedby NOAA/NWS/NCEP), a fact which appears quite promising in the ENSO operationalforecast of National Climate Center.