Recent Advances in the SHELLS Model: Specifying High-altitude Electrons using Low-altitude LEO Systems

We describe a deep learning artificial neural network model of the near-Earth space environment. The model uses inputs of geomagnetic indices and LEO electron flux measurements from the NOAA POES spacecraft. The outputs are MagEIS electron fluxes from NASA's Van Allen Probes. We focus on recent improvements to the model architecture including integration of L-shell and B-mirror dependencies into a single neural network. These improvements allow for better spatial and temporal coverage. We present initial results, demonstrating the ability to reproduce out-of-sample MagEIS observations throughout the outer belt. Finally, we discuss the utility of the model in specifying the radiation environment for both historical periods as well as the post-Van Allen Probes era.