Automated Segmentation of the Dayside Aurora

The aurora is the most visible manifestation of space weather, driven by charged particle precipitation (mainly electrons) into the upper atmosphere. Abrupt precipitation events can disrupt ionospheric electron density and currents, leading to ground-induced currents that may cause power grid failures, high-frequency communication blackouts, GPS disruptions, and spacecraft anomalies. Different models used to predict or understand these events require access to reliable and expansive data under different conditions. While the nightside aurora is relatively easy to detect, accurately identifying auroral emissions, and their associated energies, on the dayside is challenging due to extreme light pollution. In 2018, the Global Observation of Limb and Disk (GOLD) mission launched and features an ultraviolet spectrograph onboard a geostationary satellite at 47.5° W longitude. GOLD provides continuous scans of the northern and southern hemispheres that contain UV auroral signatures within them. By utilizing classical computer vision, unsupervised machine learning algorithms, and signal processing techniques we are able to accurately segment pixels of the dayside aurora. This approach offers a promising path forward to the generation of a reduced dataset. This dataset will span 6+ years of data under various solar conditions which can be used for analysis or potential inclusion and training in traditional or machine learning models.