Machine Learning Algorithm for the Influence of Space Weather Events of Geomagnetic Activity on GNSS Services

This study proposes a machine learning-based model that utilises the Disturbance storm-time (Dst) index to categorise space weather events in sub-equatorial regions and assess their impact on Global Navigation Satellite System (GNSS) Positioning, Navigation, and Timing (PNT) performance. Performance was benchmarked using various indicators derived from a confusion matrix for each Dst class defined in the study. Several algorithms from various families, including probabilistic methods, tree-based methods, ensemble learning methods, and a deep neural network (NN), using Long Short-Term Memory (LSTM) and bidirectional or convolutional layers, were evaluated on historical Dst-index-labelled GNSS data. The LSTM and convolutional-layer methods achieved the highest accuracy, but other methods had much lower inference times. The study presents a novel framework for classifying the impact of geomagnetic activity on GNSS services, thereby contributing to the development of an early-warning system for PNT degradation and to the establishment of a tailored GNSS ionospheric correction model for space weather conditions.