The Improved Two-Dimensional Artificial Neural Network-Based Ionospheric Model (ANNIM)

Abstract

An artificial neural network-based two-dimensional ionospheric model (ANNIM) that can predict the ionospheric F2-layer peak density (NmF2) and altitude (hmF2) had recently been developed using long-term data of Formosat-3/COSMIC GPS radio occultation (RO) observations (Sai Gowtam & Tulasi Ram, 2017a, https://doi.org/10.1002/2017JA024795). In this current paper, we present an improved version of ANNIM that was developed by assimilating additional ionospheric data from CHAMP, GRACE RO, worldwide ground-based Digisonde observations, and by using a modified spatial gridding approach based on the magnetic dip latitudes. The improved ANNIM better reproduces the spatial and temporal variations of NmF2 and hmF2, including the postsunset enhancement in equatorial hmF2 associated with the prereversal enhancement in the zonal electric field. The ANNIM-predicted NmF2 and hmF2 exhibit excellent correlations with ground-based Digisonde observations over different solar activity periods. The ANNIM simulations under enhanced geomagnetic activity predict the depletion of NmF2 at auroral-high latitudes, and enhancement over low latitude to midlatitude with respect to quiet conditions, which is consistent with the storm time meridional wind circulation and the associated neutral composition changes. The improved ANNIM also predicts a significant enhancement in hmF2 around auroral latitudes due to increased plasma scale height associated with particle and Joule heating during storm periods. Further, the ANNIM successfully reproduces the coherent oscillations in NmF2 and hmF2 with recurrent cororating interaction region-driven geomagnetic activity during the extreme solar minimum year 2008 and can distinguish the roles of recurrent geomagnetic activity and solar irradiance through controlled simulations.