Abstract

Canonical correlation analysis (CCA) will evaluate the degree of correlation between two multivariate data sets and will uncover patterns of correlation between the two data sets. Here CCA is applied to the multivariate solar wind data set and the multivariate geomagnetic index data set. CCA creates a new set of solar wind variables and a new set of Earth variables. The first of the new solar wind variables can be used as a solar wind driver function for the magnetosphere; the conjugate Earth variable can be used as an Earth vector to describe global geomagnetic activity in the magnetosphere-ionosphere system. The CCA-generated driver functions are found to be superior in accuracy to other driver functions in the literature. CCA of the combined data sets provides information about (1) differences in the solar wind driving of high-latitude geomagnetic indices (AE, AU, AL, and polar cap index) versus magnetospheric-convection geomagnetic indices (Kp and midnight boundary index), (2) the properties of electric field-based driver functions versus reconnection-based driver functions, and (3) improvements to solar wind/magnetosphere correlations produced by time averaging the solar wind clock angle. The CCA process tends to focus on the magnetospheric-convective indices over other indices: this may indicate that there is more predictable variance in the global-convective indices than in the others.