Neural network-based prediction techniques for global modeling of M(3000)F2 ionospheric parameter

Abstract

In recent times neural networks (NNs) have been employed to solve many problems in ionospheric predictions. This paper illustrates a new application of NNs in developing a global model of the ionospheric propagation factor M(3000)F2. NNs were trained with daily hourly values of M(3000)F2 from various ionospheric stations spanning the period 1964–1986 with the following temporal and spatial input parameters: Universal Time, geographic latitude, magnetic inclination, magnetic declination, solar zenith angle, day of the year, A16 index (a 2-day running mean of the 3-h planetary magnetic ap index), R2 index (a 2-month running mean of sunspot number), and the angle of meridian relative to the subsolar point. The performance of the NNs was verified by comparing the predicted values of M(3000)F2 with observed values from a few selected ionospheric stations and the IRI (International Reference Ionosphere) model (CCIR M(3000)F2 model) predicted values. The results obtained compared favourably with the IRI model. Based on the error differences, the result obtained justifies the potential of the NN technique for the predictions of M(3000)F2 values on a global scale.