Multivariate data analysis

Abstract

This report describes some of the multivariate data analysis techniques used in meteorology and oceanography and provides the necessary background and software to undertake appropriate analyses of a given data set. The multivariate data analysis techniques discussed are for the identification of simple patterns within a data set and between data sets, and for studying the characteristics of moving patterns in space-time data sets. Among the data analysis techniques discussed are Empirical Orthogonal Function (EOF) Analysis, Complex EOF Analysis, and Canonical Correlation Analysis, which are useful in the description and prediction of climate. Results of EOF and Complex EOF analysis of TOPEX/Poseidon derived sea level height anomaly data in the Indian and Pacific Oceans at inter-annual time scales are presented. Results of Canonical Correlation Analysis of a sample data set is also presented.