Assessment of the TEC climatology based on spatial and temporal characteristics of TEC model coefficents

Abstract

The construction of a realistic reference for the quiet-time ionosphere is part of many ionospheric studies, like the analysis of long-term changes, ionospheric storms or coupling from below. Typically, regression models, Fourier and other harmonic fits are used to generate such quiet time models. But, what can the temporal and spatial analysis of the model coefficients teach us about the ionospheric climatology? And what do we know about the natural spread of the data? Global maps of Total Electron Content (TEC) provided by the International GNSS Service (IGS) are used to construct a local time dependent regression model for each grid point of the global maps with a method robust against outliers. In addition, the tolerance interval is computed to capture the spread of the data. The variability of the regression coefficients of the local models is assessed with respect to its physical meaning. A seasonal symmetry around the solstices with the well-known semiannual variation is clearly visible in the linear model coefficients. However, the two peaks of the semi-annual variation in March and October do not have exactly the same magnitude, what can be associated with a different magnitude of geomagnetic activity. The reconstruction of TEC from the quiet-time model coefficients allows the comparison against common empirical models like the International Reference Ionosphere (IRI) and the Neustrelitz Electron Denstiy Model (NEDM), as well as a recently published machine learning model. We use the tolerance interval to assess, how well these models predict within the natural spread of the data.