Prediction of Ionospheric TEC Using RNN During the Indonesia Earthquakes Based on GPS Data and Comparison with the IRI Model

Total electron content (TEC) is a significant descriptive measure for the ionosphere of the earth. Due to either the sun’s activity like solar flare or the positive hall effect caused during earthquake (EQ), the oxygen atoms of the ionosphere split into oxygen ions and electrons increasing the electron content in the ionosphere which causes a rise in the TEC value, thus causing the delay in the signals coming from the satellite to the earth. TEC is associated with the Sun’s parameter and geomagnetic indices. In this research, parameters such as planetary K and A-index (Kp and Ap), Radio flux at 10.7 cm (F10.7), Sunspot number (SSN), and IONOLAB true TEC values were collected for the BAKO IGS network station situated in Indonesia (− 6.45° N, 106.85° E) for predicting TEC variations during EQ days occurred in the years 2004 and 2012. A total of three months of TEC data from the BAKO station during the years 2004 and 2012 were used for the developed Recurrent Neural Network (RNN) model in order to predict the TEC before and after the EQ days. For the year 2004, the model has an average Root Mean Square Error (RMSE) and Correlation Coefficient (CC) of 6.79 and 0.90. Also, for the year 2012, during April it has the average RMSE and CC of 8.90 and 0.94. For the same year in August month, the model has the average RMSE and CC of 8.70 and 0.94. The performance of the model is also evaluated using linear regression scatter plot. The Pearson’s R value calculated from the scatter plot is 0.92, shows that the model has good correlation with the true TEC.