Forecast of Ionospheric TEC during Solar Storms in Low and Mid-Latitudes Using Kriging and Recurrent Neural Network

Abstract

Geomagnetic storms are disorders in Earth’s magnetic field triggered by solar activity. This research attempts to foretell the total electron content (TEC) using the Kriging and AI model in both low and mid-latitude stations during strong geomagnetic storms that happened on March 17, 2015 and February 3, 2022. This research paper focuses on predicting and analysing TEC anomalies in the ionosphere during the solar storm by using three models: ordinary kriging (OK), cokriging (CoK) and recurrent neural network (RNN). The predicted TEC values by the models are justified with the TIEGCM and KMPCA models. Parameters like RMSE, CC, MAE, and MAPE were applied to assess the execution of predictive models and to quantify the accuracy of predictions. The average RMSE for TEC predicted in the low-latitude region ranges from 4.90 to 5.41, 5.85 to 6.26 and 8.50 to 9.90 for the OK, CoK, and RNN models, respectively. Likewise, the average RMSE for TEC predicted in the mid-latitude region ranges from 1.81 to 4.04, 1.91 to 4.24 and 2.77 to 5.38 for the OK, CoK, and RNN models, respectively. The performance evaluation parameters show that the OK performs better than the CoK and RNN models.