Abstract
A method for analyzing the linear relationship field is presented. The empirical orthogonal function technique is utilized directly on the field of time cross-correlation coefficients between two meteorological fields, rather than on the time variation of a spatial meterological field itself, as in the usual approach. A set of eigenvectors and their corresponding amplitude coefficients is obtained for each eigenvalue. The eigenvector is dependent on the spatial characteristics of the first meteorological field and the amplitude coefficient on the spatial characteristics of the other meteorological field. Because the method maximizes areas of high linear relationship on each of the corresponding grids of the original fields' dominant modes, the areas of significant relationship can be isolated in terms of the explained mean-square correlation.
The technique is illustrated on the zero-lag cross-correlation field between monthly mean Northern Hemisphere sea-level pressure and United States temperature for the midseason months of January, April, July and October during the 59-year period from 1912 to 1970.
