Global Ionospheric Tomography Using Virtual Station Multi-Source Observation Fusion Based on Machine Learning Method

Effective space weather monitoring, exploration endeavors, and radio signal applications rely on precise global ionospheric electron density modeling. In this study, we developed a novel three-dimensional computerized ionospheric tomography model. This model employs machine learning techniques to create virtual stations, enabling the integration of multi-source observations for tomography modeling. To enhance the utilization of real observational data and increase the accuracy of virtual observations, we implemented the virtual observations in each candidate area using a machine learning ensemble algorithm, specifically Extreme Gradient Boosting (XGBoost). We validated the tomography model with added virtual station data using observations from 100 GPS stations globally. The results indicate that incorporating virtual observations improved the accuracy of the independent reference station's differential slant total electron content (dSTEC) by approximately 0.45 TECU compared to previous tomography results. This method offers the potential to overcome challenges in global tomography modeling.