Classifying 8 Years of MMS Dayside Plasma Regions via Unsupervised Machine Learning
Vicki Toy‐Edens, Wenli Mo, Savvas Raptis, D. L. Turner
Abstract
Abstract The Magnetospheric Multiscale (MMS) mission has probed Earth's magnetosphere, magnetosheath, and near‐Earth solar wind for over 8 years. We utilize an unsupervised learning algorithm, Gaussian mixture model clustering, along with feature generation and simple post‐cleaning methods to automatically classify 8 years of MMS dayside observations into four plasma regions (magnetosphere, magnetosheath, solar wind, and ion foreshock) at 1‐min resolution. With these plasma regions distinguished, we have also identified boundary surfaces (e.g., magnetopause, bow shock). We validate our results on manually generated and rule based region labels described in the literature. We report overlap rates in our cluster determined magnetopauses and bow shocks against Scientist‐in‐the Loop (SITL) identified transitions and published databases. Our features are general and our model is extensible, potentially making it applicable to observational data from multiple other missions.