Reconstructing Equatorial Electron Flux Measurements From Data Collected at Low‐Earth‐Orbit Resolved in Magnetic Local Time

Abstract

We present an ensemble of five artificial neural network (ANN) models, extending and upgrading a previous model that reconstructs >30 keV equatorial electron flux measurements from low‐Earth‐orbit (LEO) satellite observations. We use the global coherent nature of the radiation belts to map data collected from the Polar Orbiting Environmental Satellites (POES) to that of the Van Allen Probes (formerly known as the Radiation Belt Storm Probes; RBSP) in near‐equatorial geostationary transfer orbit (GTO). The LEO‐GTO mapping procedure and machine learning (ML) modeling is based on our magnetic conjunction framework. We identify over 76,500 conjunctions per POES and RBSP spacecraft and use roughly 27,000 conjunctions to train our ML model. Our conjunction training data set for the five POES spacecraft, each in their own magnetic local time (MLT) swath, is comprised of the POES electron flux measurements at two viewing angles, the LEO ephemeris data, the geomagnetic activity AE index, and fitting coefficients from the pitch angle distribution of each conjunction. We present our models' out‐of‐sample results, with each POES spacecraft performing with R ²  > 0.63. The updated ANNs maintain the ability to nowcast and reconstruct inferred equatorial GTO observations from LEO data, now MLT resolved. This update to our models allows for LEO‐GTO mapping beyond the end of the RBSP mission in 2019 and showcases the potential for future, real‐time radiation belt monitoring solely from LEO constellations.