Speaker
Description
At high-latitude regions on the Earth, auroral displays exhibit strong spatiotemporal variability yet fall into broad morphological classifications. Studies have shown that different types of aurora can exhibit distinct particle precipitation energies and fluxes, convection electric field configurations, and preferential occurrence rates across different levels of geomagnetic activity. The variable electrodynamic forcing imposed by these structures can generate plasma density irregularities, temperature variations, and ground magnetic perturbations that can pose risks to communication systems and infrastructure sensitive to geomagnetically induced currents. Traditionally, meso-scale auroral classes are identified through visual inspection of all-sky imager (ASI) data. However, this approach requires manually sorting through millions of images introducing inherent uncertainties due to subjective and limited categorizations, especially when multiple classes may be present. Furthermore, the determination of electrodynamic properties and impacts necessitate acquiring data from multiple co-located instruments. To address these challenges, we employ a self-supervised contrastive learning algorithm to systematically extract characteristic features from the all sky imager data located at Poker Flat, Alaska during intervals of increased geomagnetic activity. We identify occurrence rates of different labels with respect to ground magnetic field signatures to investigate their association with substorm phases. We use measurements from the Poker Flat Incoherent Scatter Radar to characterize the electric fields, average energies, and energy fluxes associated with each morphological class and quantify the temperature variations in the Ionosphere. We compare our clustering methodology with classifications reported in the literature and present a statistical analysis of the resulting labels, establishing a framework for characterizing and assessing the geoeffectiveness of distinct auroral forms.