Characterizing Plasma Sheet Flows With Machine Learning Identification

Abstract Bursty bulk flows, injections, and related flows are a key component of plasma energy and flux transport during substorms. Mixed observational features require researchers to carefully define their quantitative criteria in studies of such events, a process that requires presuming some quantity of the output events. In this work we present a novel supervised random forest classification model for identifying substorm related flows in the plasma sheet. Our model allows a more complete view of event variety, and avoids the presumptive nature of data‐derived quantitative classification. We observe no correlation between event velocities and geomagnetic activity, however an uneven formation of large‐scale dipolarization in the tail region during strong substorms is statistically present in our data. A bi‐modality also appears in the event occurrence location during strong substorms. Some model limitations and future work is discussed.