Modeling the Total Electron Content variation in the Ionosphere for Improved Positional Accuracy

• Published in: 2020 12th International Conference on Electronics, Computers and Artificial Intelligence (ECAI) pp. 1 - 7
• Main Authors: Iyer, Sumitra, Mahajan, Alka
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01.06.2020
• Subjects: Computational modeling; Feature extraction; Global navigation satellite system; GNSS; Ionosphere; Predictive models; Range error; Receivers; Support vector machine; Support vector machines; Total Electron Content; Training
• DOI: 10.1109/ECAI50035.2020.9223247

The total electron content (TEC) in the ionosphere produces a group delay in the propagating signal of a navigational satellite and produces a range error. This error in Global Navigation Satellite Systems (GNSS) is corrected by the fixed Klobuchar ionospheric model in the single-frequency receivers. The performance of the Klobuchar error correction is lowered due to the irregular and complex distribution of TEC in the equatorial and low latitude regions. This is further affected in the equinox month of March and during geomagnetic disturbances. In this paper, model using the Support Vector Machine (SVM) is proposed to predict the ionospheric TEC range half an hour in advance. The TEC thus predicted can be further used for estimating a suitable correction to minimize the range error. The model can apriori estimate the range of mean value of TEC at t+δt using the input features at time t. The model is trained using the TEC values of all days, Disturbed Storm Index (Dst), Auroral Electrojet index (AE), time of the day and extracted features of TEC. The performance of the model is also compared by changing the input features and its accuracy is measured using precision, recall, F score and the receiver operating characteristic (ROC) curve. It is observed that the model performance is improved by adding extracted features of TEC and is able to predict the mean TEC range half an hour in advance with better accuracy. The model is also tested with unseen storm day TEC data and is found to perform well with 80% accuracy. The equatorial anomaly region near the geomagnetic equator is chosen for the study. The TEC data is obtained from the International GNSS Service (IGS) dual-frequency receiver located at IISC, Bengaluru, India for the equinox month of the year 2015 for training the model.