Automatic Region Identification over the MMS Orbit by Partitioning n-T space

TL;DR

This paper presents a novel machine learning method that automates the identification of magnetospheric regions in space plasma data with high accuracy, aiding space mission data analysis and interpretation.

Contribution

The paper introduces a new approach using SVM to partition plasma parameter space for region identification, combining automation with interpretability.

Findings

Achieved 99.9% accuracy in region classification.

Automated method reduces manual data sorting effort.

Provides scientific insights through interpretable boundary fitting.

Abstract

Space plasma data analysis and mission operations are aided by the categorization of plasma data between different regions of the magnetosphere and identification of the boundary regions between them. Without computerized automation this means sorting large amounts of data to hand-pick regions. Using hand-labeled data created to support calibration of the Fast Plasma Instrument, this task was automated for the MMS mission with 99.9% accuracy. The method partitions the number density and ion temperature plane into sub-planes for each region, fitting boundaries between the sub-planes using a machine learning technique known as the support vector machine. This method presented in this paper is novel because it offers both statistical automation power and interpretability that yields scientific insight into how the task is performed.