Using Dimensionality Reduction and Clustering Techniques to Classify Space Plasma Regimes
Bakrania, Mayur R., Rae, Jonathan, Walsh, Andrew P., Verscharen, Daniel, Smith, Andy W.
Abstract
Collisionless space plasma environments are typically characterized by distinct particle \npopulations. Although moments of their velocity distribution functions help in distinguishing \ndifferent plasma regimes, the distribution functions themselves provide more comprehensive \ninformation about the plasma state, especially at times when the distribution function \nincludes non-thermal effects. Unlike moments, however, distribution functions are not \neasily characterized by a small number of parameters, making their classification more \ndifficult to achieve. In order to perform this classification, we propose to distinguish between \nthe different plasma regions by applying dimensionality reduction and clustering methods to \nelectron distributions in pitch angle and energy space. We utilize four separate algorithms to \nachieve our plasma classifications: autoencoders, principal component analysis, mean shift, \nand agglomerative clustering. We test our classification algorithms by applying our scheme \nto data from the Cluster-Plasma Electron and Current Experiment instrument measured in \nthe Earth’s magnetotail. Traditionally, it is thought that the Earth’s magnetotail is split into \nthree different regions (the plasma sheet, the plasma sheet boundary layer, and the lobes), \nthat are primarily defined by their plasma characteristics. Starting with the ECLAT database \nwith associated classifications based on the plasma parameters, we identify eight distinct \ngroups of distributions, that are dependent upon significantly more complex plasma and field \ndynamics. By comparing the average distributions as well as the plasma and magnetic field \nparameters for each region, we relate several of the groups to different plasma sheet \npopulations, and the rest we attribute to the plasma sheet boundary layer and the lobes. We \nfind clear distinctions between each of our classified regions and the ECLAT results. The \nautomated classification of different regions in space plasma environments provides a useful \ntool to identify the physical processes governing particle populations in near-Earth space. \nThese tools are model independent, providing reproducible results without requiring the \nplacement of arbitrary thresholds, limits or expert judgment. Similar methods could be used \nonboard spacecraft to reduce the dimensionality of distributions in order to optimize data \ncollection and downlink resources in future missions.