EOF方法与多元回归方法对热带SST拟合的比较  

The Comparison Between EOF and Multi-Regression Method on the Fitting of Tropical SST

作　　者：钟贻森[1] 翟方国[1] 

机构地区：[1]中国海洋大学,山东青岛266100

出　　处：《中国海洋大学学报（自然科学版）》2009年第S1期319-325,共7页Periodical of Ocean University of China

基　　金：国家重点基础研究发展计划项目(2005CB422302)资助

摘　　要：基于热带降水数据,本文比较了EOF模型与多元回归模型在海表面温度(SST)拟合估计上的优劣,拟合结果显示2种模型均能很好的体现赤道东太平洋ENSO现象引起的SST显著异常,而对其他海域变化的拟合则误差较大。对比2种模型的结果,尽管降水与SST两者的第一模态主成分之间存在95%的线性相关,但是忽略高模态特别是第二模态的EOF方法的拟合逊于包含多模态的多元线性回归方法,从而表明高模态的影响大于线性假设拟合的影响。进一步通过SVD分解方法分析降水与SST各模态相关性,结果表明占约13%变化的第二模态降水与SST存在相似的空间分布,因此纳入更多模态是多元回归模型更优的主要原因。Based on the tropical precipitation data,two methods EOF method and multi-regression method are compared in the fitting of SST.The result implies both of these two methods can make a good estimation on the East Pacific SST anomaly due to ENSO,but produce relatively large error on other areas.The comparison between these two model results shows that the first modes of SST and precipitation have 95% linear correlation,however,the skill of EOF method without high modes is not as good as multi-regression method which includes more modes.This presents the effect of high modes overwhelms the assumption of linear model.SVD composition is further used to analyze the correlation between modes of SST and precipitation,which also says that the second modes of SST and precipitation have similar spatial distribution.Therefore,introducing more modes is the main reason why multi-regression method does better than EOF method.

关 键 词：EOF 多元回归 ENSO SST拟合 SVD 

分 类 号：P731.11[天文地球—海洋科学]