Computational study demonstrates reliable imputation of missing medium-energy electron flux data in Low Earth Orbit, highlighting potential for near-real-time magnetospheric monitoring.
Key Points
Develop a machine learning imputation framework to reconstruct missing pitch-angle-resolved medium-energy electron flux observations from the Fengyun-3E satellite in Low Earth Orbit.
Trained eight Multi-Layer Perceptron (MLP) neural network models using satellite orbital data and electron flux measurements taken near a 90° local pitch angle.
Targeted missing observations near 0° and 180° local pitch angles for electron energy channels spanning 280–600 keV.
Achieved a correlation coefficient r of at least 0.915 between reconstructed values and actual observations across unseen test datasets.
Demonstrated a maximum Root Mean Square Error (RMSE) of 0.110 on a logarithmic scale across evaluated energy channels.