The accuracy and reliability of Global Navigation Satellite System (GNSS) applications are affected by the Earth‘s ionosphere. Accurate and timely corrections of ionospheric effects and early warning information in the presence of space weather are therefore crucial for GNSS applications. To model the impact of space weather, a complex chain of physical dynamical processes between the Sun and the ionosphere need to be taken into account. These nonlinear processes and relationships can be approximated using Machine Learning (ML). However, the fact that ML models are often treated as black-box models, puts the reliability of the predictions into question. Assessing the quality and confidence of ML results in the form of an uncertainty quantification is an important step towards an improved interpretation of ML results for the development of a “trustworthy” model. This study presents novel ML approaches to forecast vertical TEC (VTEC) by utilizing state-of-the-art supervised learning techniques. The learning algorithms used in this study range from Decision Tree-based, such as Random Forest, Adaptive Boosting and Gradient Boosting to Artificial Neural Networks. Both aleatoric and epistemic uncertainties of the achieved results are quantified with different approaches, including ensemble modeling, modifying the objective cost function accordingly or applying Bayesian learning. These approaches show potential for forecasting VTEC in different ionosphere regions during quiet and storm periods, while providing the uncertainties of the results.     «

The accuracy and reliability of Global Navigation Satellite System (GNSS) applications are affected by the Earth‘s ionosphere. Accurate and timely corrections of ionospheric effects and early warning information in the presence of space weather are therefore crucial for GNSS applications. To model the impact of space weather, a complex chain of physical dynamical processes between the Sun and the ionosphere need to be taken into account. These nonlinear processes and relationships can be approxi...     »