Using Machine Learning to Diagnose Relativistic Electron Distributions in the Van Allen Radiation Belts

Abstract

The behaviour of relativistic electrons in the radiation belt is difficult to diagnose as their dynamics are controlled by simultaneous physics processes, some of which may be still unknown. 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. A series of unsupervised machine learning tools has been applied to 7-years of Van Allen Probe Relativistic Electron Proton Telescope data to identify 5 different typical types of plasma conditions, each with a distinctly shaped energy dependent pitch angle distribution (PAD). The PADs at lower energies have shapes as expected from previous studies -either butterfly, pancake or flattop, providing evidence that machine learning has been able to reliably classify the relativistic electrons in the radiation belts. Further applications of 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 magnetospheres and the JUICE mission. Understanding pitch angle distributions across the heliosphere enables researchers to determine the physical mechanisms that drive pitch angle evolution and investigate their spatial and temporal dependence and physical properties.

Automatic Encoding of Unlabeled Two Dimensional Data Enabling Similarity Searches: Electron Diffusion Regions and Auroral Arcs

“…Unsupervised techniques, those that do not require manual labeling, have become more popular for use with higher dimensional datasets in the last few years. However, such methods are mostly used to group (“cluster”) similar observations, for example, to identify observations in distinct magnetospheric regions based on plasma moments (e.g., Bloch et al., 2020; Innocenti et al., 2021) or distributions (e.g., Bakrania et al., 2020; Killey et al., 2023). For two dimensional data, techniques such as auto‐encoders may be used to reduce the dimensionality of the data.…”

“…For two dimensional data, techniques such as auto‐encoders may be used to reduce the dimensionality of the data. This allows similar observations to be grouped together in the smaller dimensional space—a space that is assumed to be representative of the original data (e.g., Bakrania et al., 2020; Killey et al., 2023). However, typically such methods directly use the pixel values in the data and thus care must be taken if a key feature of the image is known to move within an image or vary in brightness.…”