On the application of Exploratory Data Analysis for characterization of space weather data sets

This paper presents an overview of the mathematical foundations for techniques in Exploratory Data Analysis (EDA) for the purpose of investigating the relationships among the numerous variables in large sets of multivariate space weather data. Specifically, we cover techniques in Principal Components Analysis (PCA) and Common Factor Analysis (CFA). These techniques are illustrated using space weather activity indices collected during the year 2002 and the corresponding noon-time hmF2 data from the International Reference Ionosphere (IRI). A CFA is used to categorize the activity indices, and a PCA is used to derive two macro-indices of activity to ascertain the strength of solar and geomagnetic activity. These macro-indices are then used to compare and contrast IRI’s noon-time hmF2 values at six different geographic stations. It was found that the correlation between hmF2 and the macro-indices more accurately represented the variation of this correlation with latitude found in previous studies than if we used an isolated conventional index, such as SSN and AE. We also found that the daily maximum value of the Polar Cap Index was dependent on both solar and geomagnetic activity, but the closely-related cross-Polar Cap Potential was solely associated with elevated levels of geomagnetic activity, which is a unique result compared to previous studies. We argue that the discrepancy can be explained by the difference in experiment designs between the two studies. This paper demonstrates the usefulness of EDA in space weather studies of large multivariate data sets.