Using Machine Learning to Understand Electron Pitch Angle Evolution of the Van Allen Radiation Belts

Abstract

Van Allen radiation belt dynamics are difficult to diagnose due to the multitude of physical processes that can simultaneously drive energetic electron acceleration, transport and loss. Different processes can drive a specific energy-dependent electron pitch angle distribution (PAD). Signatures of these physical processes are difficult to identify in large amounts of data; therefore, a machine learning approach is developed to classify energetic electron distributions which have been driven by different mechanisms. We employ a unique amalgamation of unsupervised machine learning techniques on 7 years of Van Allen Probe Relativistic Electron-Proton Telescope and Magnetic Electron Ion Spectrometer data to identify the typical types of plasma conditions across an extensive energy spectrum, each with a distinctly shaped energy-dependent PAD. Six PADs successfully describe 93% of outer belt relativistic electrons, two each of: pancake, butterfly, and flattop; while three PADs describe 94% of sub-relativistic electrons: flattop and two pancake morphologies. We investigate the occurrence and storm-time evolution of each PAD through 45 geomagnetic storms. We find new populations of PADs, including unusual “shadowing-like” and wave-particle interaction (WPI) signatures at low-L, and radial diffusion and substorm activity signatures at higher-L, where radial diffusion processes swamp WPI-dominated PADs through geomagnetic storms. The spatiotemporal variability of similarly shaped distributions is not energy-dependent, suggesting that processes that drive these PADs are also non-energy dependent, ruling out WPI as their primary driver. This technique could be applied to other space plasma regions and datasets from inner heliospheric missions such as Parker Solar Probe and Solar Orbiter, to planetary magnetospheric missions such as Juno. PAD characterisation enables researchers to determine the physical mechanisms that drive radiation belt dynamics, which is ultimately required for improved radiation belt models and forecasting techniques.

Date of Award	22 May 2025
Original language	English
Awarding Institution	Northumbria University
Supervisor	Jonathan Rae (Supervisor), Clare Watt (Supervisor), Sarah Bentley (Supervisor) & Jasmine Mann (Supervisor)