Categorizing Characteristic Regions of Nightside High-Latitude Ionospheric Irregularities Using a Machine Learning Approach.

We are making use of the vast amount of publicly available GNSS data to develop a data driven supervised machine learning model to categorize to a predefined set of characteristic high-latitude ionospheric irregularity nightside regions. The goal of this model is to predict whether the scintillating instance of high-frequency data was recorded in the auroral oval vs. the polar cap with a high level of confidence. The source regions were determined using the SUSSI instruments onboard the DMSP satellites. The model is trained and tested with events extracted by thresholding the low-rate data from receivers from both characteristic nightside regions. As input parameters of the model serve the Power Spectral Densities (PSD) of the events since they provide context about the size and velocities of the irregularities. The model is expected to identify potential distinct characteristics of each predefined source region and in a second step to predict the region based on a given phase or amplitude time series. This will allow us to further understand of the characteristics of ionospheric irregularities and their sources.