Comparison of the solar variations of the geomagnetic field at the Coimbra Magnetic Observatory (COI) obtained by different methods: effect of the solar and geomagnetic activity

TL;DR

This study compares traditional and PCA-based methods for analyzing daily geomagnetic variations at Coimbra Observatory, revealing PCA's effectiveness in extracting certain variations and highlighting the need for supplementary analysis for others.

Contribution

It demonstrates the capability of PCA to automatically extract specific geomagnetic variations from observational data, with insights into seasonal and decadal activity influences.

Findings

PCA effectively extracts Sq variations from Y and Z components.

Extraction of SD variations requires supplementary analysis.

Seasonal and decadal variations influence PCA classification of geomagnetic activity.

Abstract

Here we present the results of the comparative analysis of the regular daily variations of the geomagnetic field, Sq and SD, obtained from the observations with two methods: (1) using the so-called geomagnetically quiet day (QD) to calculate mean daily variations and (2) using the principal component analysis (PCA) to decompose original series into main variability modes. The original geomagnetic field series were obtained at the Coimbra Magnetic Observatory (COI, Portugal) in 01.01.2007-31.12.2017. The data were analyzed separately for each of 12 months. The geomagnetic field components X, Y and Z were analyzed independently. The Sq and SD variations obtained with QD were compared with the principal components from PCA (PCs, daily variation of the geomagnetic field of different types) using the correlation analysis. In this study only the PCA ability to extract daily variations similar to ones acquired using the standard approach with QD was analyzed. The results for X, and Y and Z are essentially different. The Sq variation is always filtered to the 1st PCA mode for Y and Z, whereas the SD variation is filtered either to the 2nd or 3rd mode. For X the Sq variation tends to be filtered to the 2nd mode, while the SD variation is more often filtered to the 1st mode. The number of the PCs classified either as Sq or SD follows seasonal and decadal variations of the geomagnetic activity. The analysis showed that PCA can automatically extract the Sq variation from the observations of Y and Z; however complementary analysis (for example, a comparison to a reference curve) is needed to identify a PC corresponding to the SD variation. For X component, the automatic extraction of the Sq or SD variations is not possible, and the complementary analysis is always needed. The dataset used in this paper is described in detail in a companion paper by Morozova et al. (2021a).