Item

Abstract

Highly energetic electrons within the Earth's outer radiation belt pose a significant threat to satellite operations. Specifically, relativistic electrons can penetrate through satellite shielding and induce deep dielectric charging, resulting in onboard electronics damage. To address this issue, we developed a neural network-based model for relativistic electron fluxes (with energies >1.8 MeV) using data from the REPT instrument onboard the Van Allen Probes mission between 2012-2019. The model takes the geomagnetic index hp-30 and solar wind parameters as input. By training the models solely on ground-based indices, and comparing their accuracy to models incorporating solar wind parameters, we show that the relativistic electron flux can be reconstructed with high fidelity. Specifically, the model has close-to-zero bias and >95% correlation on both training, validation and unseen test intervals. We discuss the potential applications of the model in combination with physics-based radiation belt codes and compare it to current state-of-the-art simulations.