Empirical orthogonal function analysis and modeling of the ionospheric peak height during the years 2002-2011

The ionospheric peak electron density height (h_m) is one of the most important ionospheric parameters characterizing high-frequency radio wave propagation conditions. In this paper, a global h_m model based on the empirical orthogonal function (EOF) analysis method is constructed by using Global Navigation Satellite Systems ionospheric radio occultation measurements from COSMIC and CHAMP as well as global ionosonde data during the years 2002-2011. The variability of h_m can be well represented by the several EOF base functions E_k and the corresponding coefficients P_k. The rapid convergence of EOF decomposition makes it possible to use only the first four EOFs components, which express 99.133% of total variance in this study, to construct the empirical model. The variations of h _m with respect to the magnetic latitude, local time, season, and solar cycle have been studied, and the EOF-based h_m model has been validated through comparisons with the International Reference Ionosphere (IRI) model and other observation. The evaluations indicate that the EOF and IRI model give better h_mF₂ at middle and high latitudes than those at low latitudes. Since the limited data were used in the EOF model during high solar activity years, its accuracy degrades to some extent. During nighttime of spring, summer, and winter in the auroral zone, the h_m derived from the EOF model may range from 90 to 150 km because of the reduction of h _m, which is due to particle precipitation, whereas the IRI model does not include this reduction. During the periods of low solar activity, the F₂ peak heights (h_mF₂) from the EOF model are in good agreement with the observed data, while the IRI model tends to overestimate the h_mF₂ in the middle and high latitudes. Key Points A global hm model has been constructed by EOF analysis using observed data The hm model can capture the main features with only a few EOF series The hm model agrees better with the observed data than IRI model