Creating a Machine Learning database to characterize high-rate ionospheric scintillation signatures in high latitudes

Ionospheric scintillation in Global Navigation Satellite System (GNSS) signals propagating through the high-latitude ionosphere have been described in recent years through modeling and data assimilation. However, linking ionospheric irregularities in different scale sizes directly to the generated scintillation signatures and their features remains a challenge. This study focuses on the two most characteristic regions of the nighttime high-latitude ionosphere, the auroral oval and the polar cap. It is expected that the predominant irregularity mechanisms leading to scintillation in these two regions are significantly different. Thus, we present a Machine Learning (ML) approach to distinguish high-rate scintillation signatures in signal phase and amplitude from the polar cap vs. the auroral oval to gain a deeper understanding of the two regions. We describe the process of preparing the data from various stations and storm cases in the Northern hemisphere for the analysis and how to detect significant scintillation events. Then the data set was prel-abeled into events that were located either clearly inside the auroral oval or the polar cap based on mean energy electron flux observations. Finally, after comparing the signatures in phase and amplitude through hierarchical clustering, a decision tree model was trained to classify the signatures into their corresponding characteristic region. Based on theoretical studies and additional observations from other instruments, the differences between signatures from the auroral oval vs. the polar cap were explored for scintillation event time series and power spectra to create a database with further insight into characterizing high-latitude ionospheric irregularities.