Single station modelling of ionospheric irregularities using artificial neural networks

An empirical model of ionospheric irregularities over Mbarara (MBAR, 30.7∘E geographic longitude, 0.6∘S geographic latitude, 10.22∘S geomagnetic latitude) based on Artificial Neural Networks (ANNs) is developed using Global Navigation Satellite System (GNSS) derived Total Electron Content (TEC) data from 2001–2022. This long term data helped to study the climatology of the trends, the diurnal, seasonal, and solar activity dependence of ionospheric irregularities. We used the rate of change of TEC index (ROTI) to quantify the strength of irregularities. The input space consisted of time of the day (Hr), day of the year (doy), z-component of the Interplanetary magnetic field (IMF Bz), symmetric horizontal component of the ring current (SYM-H), solar activity factor (F10.7P), and vertical E×B drift, all of which are thought to influence irregularity occurrence, though with different percentage contributions. Of these inputs, Hr, doy, and F10.7P constituted the primary input parameters (PIP). We investigated the contribution of each input to the ROTI changes by developing seven models adding an input to the PIP at each time. The greatest contributor to the modelling results was SYM-H with a percentage contribution of ≈2