Development of a new aurora model for the small- and meso-scale structures through deep-learning methods and their influence on the upper atmosphere

The energy inputs related to the meso- and small-scale auroral structures can be significant to the simulations of global ionosphere/thermosphere system simulations and space weather forecast. However, our knowledge about those structures is limited and the specification of them in general circulation models (GCM) is poor. Using deep learning methods, a new auroral model including small- and meso-scale structures has been developed, which has two components, mean energy (Eo) and energy flux (Q) of precipitating electrons. 4 years (2002-2005) Global Ultraviolet Imager (GUVI) data and Zhang-paxton model are used to build the deep learning algorithm of the mean energy (Eo) and energy flux (Q) of precipitating electrons. The IMF Bz and Kp indices have been used to label the data and to drive the model. Due to the high spatial resolution ( 0.15⁰ in Lat and 1.5⁰ in Lon) of auroral images produced by the new model, the polar distribution of small- and meso-scale structures in aurora has been constructed for the first time. Through coupling the new model with Global Ionosphere-Thermosphere Model (GITM), the impact of those structures on the energy input estimation and ionosphere/thermosphere variation has been examined as well.