| Name: Dr. Sumanjit Chakraborty |
| Affiliation: Indian Institute of Geomagnetism, Mumbai, India |
| Conference ID: ASI2026_30 |
| Title: A user-friendly Neural Network framework for single-to-multi-day prediction of global ionospheric TEC: First Results |
| Abstract Type: Poster |
| Abstract Category: Sun, Solar System, Exoplanets, and Astrobiology |
| Author(s) and Co-Author(s) with Affiliation: Sumanjit Chakraborty(Indian Institute of Geomagnetism, Navi Mumbai - 410218, India), Gopi K. Seemala(Indian Institute of Geomagnetism, Navi Mumbai - 410218, India) |
| Abstract: Accurate prediction of global ionospheric Total Electron Content (TEC) is crucial for mitigating the impacts of space weather on satellite navigation, communication, and radio propagation systems. In this work, we present a data-driven framework for single-to-multi-day TEC prediction using a Neural Network (NN) model trained on the drivers related to the Solar wind–Magnetosphere–Ionosphere coupling physics. This model feeds on multi-day solar wind and Interplanetary parameters (magnetic field Bz, bulk velocity Vsw, proton density ⍴ and temperature Tp), geomagnetic activity index (Ap), the solar activity proxy F10.7 flux, autoregressive TEC terms, and diurnal harmonic components to capture local time dependence. A user-friendly prediction pipeline has been developed that supports seamless ingestion of multiple daily files and robust temporal alignment across heterogeneous data sources. Predictions are generated over any user-defined date ranges, enabling efficient investigation of quiet and disturbed ionospheric conditions. Model performance is evaluated using the coefficient of determination (R²), and the results (R² ~ 0.98) demonstrate that this NN model can successfully capture diurnal variability as well as geomanetic storm-time TEC enhancements/depletions driven by solar wind and geomagnetic forcings. The framework automatically saves all diagnostic figures and numerical outputs using standardized date-based naming conventions, ensuring reproducibility and traceability of results. The flexibility to analyze multi-day to multi-week intervals within a single execution makes the system suitable for event-based studies. Therefore, this work highlights the effectiveness of machine-learning approaches for ionospheric TEC prediction when combined with rigorous data handling and user-oriented design, and it provides a scalable and extensible platform for space weather research with potential applications in real-time ionospheric monitoring and forecasting over the globe. |
