Assessing the performance of artificial neural networks to predict ionospheric TEC over Nigeria during different space weather events

Abstract

The ionosphere model is essential to satellite-based systems to accurately correct the ionospheric error encountered by satellite signals en route. The Levenberg–Marquardt backpropagation (LMBP) algorithm in the artificial neural network (ANN) was used in this work to predict the total electron content (TEC) within the trough of equatorial ionization anomaly (EIA) over Nigeria. Two sets of data were used over the period of three consecutive years (2011–2013) of high solar activity. The first set was used as an input to the ANN model and the second set of data was used as a target. Seventy percent of the data sets were used to train the network, 15% of the data were used for validation, and 15% used for testing. The performance of the model was assessed during specific quiet and disturbed geomagnetic conditions. The regression analysis of the model output was optimized by minimizing a cost function of the mean square error (MSE). The results of the errors, regression, and comparative analyses have revealed that the ANN model is able to predict accurate and reliable TEC that compares well with the actual experimental data at any geophysical conditions. Hence, this model would be useful to forecast TEC over Nigeria to a reliable threshold.