Bi-LSTM based vertical total electron content prediction at low-latitude equatorial ionization anomaly region of South India

Abstract

In this study, we consider Bi-directional Long Short Term Memory (Bi-LSTM) model based Vertical Total Electron Content (VTEC) prediction over Thanjavur (Geographic 10.72˚ N, 79.01˚ E, Geomagnetic 2.34˚ N, 152.19˚ E) Global Positioning System (GPS) station. This station is located at low latitude Equatorial Ionization Anomaly (EIA) region of 2˚ geomagnetic dip latitude and has unique ionospheric dynamics. In this region, the VTEC prediction is crucial and challenging for space weather and the Sixth Generation (6G) Internet of Space (IoS) application to support early warning systems and future spatial data transmissions. A Deep Learning (DL) model based on Bi-LSTM was developed and trained for F10.7 and Dst index for predicting the VTEC. This study highlights the prediction of VTEC for any day that includes solstice and equinox time frames. The Bi-LSTM has an improvement of 28 % in mean absolute error (MAE), 48% in mean square error (MSE) and 24% in root mean square error (RMSE) as compared to the conventional Long Short Term Memory (LSTM) network. Hence, this Bi-LSTM model can be helpful to predict the VTEC in the EIA region and may be helpful to extrapolate over the unmeasured grid region of ocean and land.