MSSA-SVD典型回归模型及其用于ENSO预报的试验

文中提出了一种基于多通道奇异谱分析 (MSSA)的广义典型混合回归模式。其基本思想是 ,利用MSSA SVD提取预报因子场和预报变量场的显著耦合振荡信号 ,对它们的前几个显著典型分布型建立多元线性统计气候预报模式。经对Nino海区各季海温距平所进行的短期气候预测试验表明 ,其预报效果优于其它统计预报方案 ,从而为探索ENSO预测方法提供了一种新的思路

A PREDICTION EXPERIMENT BY USING THE GENERALIZED CANONICAL MIXED REGRESSION MODEL BASED ON MSSA-SVD FOR ENSO

A generalized Canonical Mixed Regression Model based on MSSA SVD is presened to prediction of ENSO. The MSSA is a Multichannel Singular Spectrum Analysis and the SVD is Singular Value Decomposation. The basisc idea of the method is that (1) the prominent coupled oscillation signals are segregated between the forecasted fields and the forecastor fields by using MSSA SVD method;(2) the generalized Canonical Mixed Regression Model is constructed according to the first several prominent coupled oscillation patterns for ENSO prediction. The results of statistical forecast test based on the MSSA SVD method show that the predictional model possesses more advantages and better effects than other statistical prediction methods.