Using Machine Learning to link ionospheric irregularity types and scintillation signatures in high latitudes

In the high-latitude ionosphere, disruptions in radio signals are generated due to different types of ionospheric irregularity structures. We expect the radio wave scintillation in Global Navigation Satellite System (GNSS) signals to look different depending on which irregularity structuring mechanism was present in the ionosphere, like shear instabilities, electron precipitation, or steep density gradients. Since each characteristic region of the high-latitude ionosphere has particular, more dominating irregularity sources present, the GNSS signatures are also assumed to differ in each region. On this underlying assumption that scintillation event signatures can be distinguished based on the ionospheric region they were recorded in, a Machine Learning model was trained to separate GNSS time series in the auroral oval from the polar cap with a supervised clustering and decision tree approach. Based on the database of GNSS scintillation events from the auroral oval vs. the polar cap, the events were expanded to include more irregularity-specific information like spectral parameters of the phase and amplitude time series, plasma drift, electron energy flux, and electron density. The most likely irregularity mechanism during the time of the event was determined by analyzing the dynamics and ionospheric conditions around the scintillation event in detail. To use this database to understand favorable ionospheric conditions for certain irregularity types, this study explores different input data sets from different instruments and resolutions.