全球平均海温演变的奇异谱分析

The Singular Spectrum Analysis for Global SSTA

用奇异谱分析（Ｓｉｎｇｕｌａｒｓｐｅｃｔｒｕｍａｎａｌｙｓｉｓ，以下简称ＳＳＡ）研究了１３６年（１８５６－１９９１年）全球平均海温季距平演变特征，并对各主分量作了最大熵谱分析。结果表明，海温演变除了具有８１年周期外，还有一些非常明显的周期振动，这些周期主要是准１０年利２．６－６年。从各主分量的方差贡献来看，前１５个主分量的方差贡献之和是总方差的７８％。从特征向量的分布函数来看，第１、２特征向量表示了海温序列的线性特征，它们的方差之和可以占到总方差的４８％，其合成曲线基本上可以描述海温百年尺度内数年或数十年中的总体平均变化趋势；第３及其以后的特征向量表示了海温的波动规律，前４个主分量的合成曲线能很好地拟合海温年变化特征。

The global SSTA for past 136years(1856-1991)was studied with singular spectrum analysis,then a maximum entropy spectrum analysis for serveral principal components was conducted.The result shows that SSTA has serveral very evident periods of quasidecade and 2 6-6 years besides 81 years.The first fifteen principal components have a variance contribution of 78%.Eigenvector 1,2 stand for linear characteristic of SSAT,their variance contribution is 48%,and their resultant curve basically described average thenes during serveral years and serveral decades.Eigenvector 3 and afterwards eigenvectors stand for wave characteristics of SSTA.The first four principal components can better fit SSTA yearly change characters.