Patterns of Suprathermal Electron Flux in the Near-Earth Space Environment: Statistical Learning using THEMIS Data

Suprathermal (~0.5-100keV) electrons in the near-Earth space environment play an important role in inner magnetosphere dynamics and are a contribution to hazardous spacecraft radiation environments. The source of electrons to this region is the plasma sheet. Variations of electron flux in the near-Earth plasma sheet are highly driven by upstream interplanetary conditions, and they are correlated with geomagnetic activity. However, the systematic dependencies of suprathermal electron flux to interplanetary variations, as well as predictive capabilities of local flux enhancement, remain elusive. Utilizing data from 2008-2020 of the Time History of Events and Macroscale Interactions during Substorms (THEMIS) mission, we generated statistical maps showing the median electron flux from 2-12 RE binned by various solar wind, IMF, and geomagnetic index values. These maps show the systematic and climatic averages of suprathermal electron flux in the near-Earth space environment in the near-equatorial plane. They reveal the radial and magnetic local time dependence of electron flux for various interplanetary conditions - such as solar wind speed, pressure, and IMF BZ - and during geomagnetic quiet and disturbed periods as indicated by the AE, SYM-H, and Kp indices. We statistically investigated the increase in suprathermal electron flux in the plasma sheet during substorm dipolarizations during years 2008-2009. We fail to find a dependence of the magnitude of local electron flux enhancement based on substorm strength. Additionally using THEMIS data, we introduce a machine-learned model that accurately makes predictions of electron flux in the plasma sheet using solar and interplanetary features. Our model is a feed-forward neural network using mainly OMNI inputs with included time history of four hours. We use the SHAP algorithm to determine the relative contributions of the features to model outputs. The most important contributions to model predictions are the solar wind speed, IMF BZ, and solar EUV flux. We discuss the physical impacts of these features and outline steps for model improvement, further scientific research, and forecasting use.