Large-Scale MMS Survey using Rapid Unsupervised Detection of Events (RUDE) Methodology

This work discusses the application of a generalizable technique, based on computationally inexpensive statistical and machine learning methods, for the rapid identification of geophysical events in large observational datasets. Specifically, Dynamic Principal Components Analysis (D-PCA) and One-Class Support Vector Machines (OC-SVMs) are used to generate an alternative representation of time series inputs, which is subsequently clustered to identify outliers. Preliminary studies utilizing this technique demonstrate its ability to identify geophysical events using only a single data product, or with combinations of different data products. Further, this method has been shown to be generalizable to a variety of missions and input data products, and its computational efficiency makes it suitable for rapid data analysis tasks on the ground or for implementation on spaceflight hardware.

Here, we discuss the results of this event detection methodology when applied to four years of data from the Magnetospheric Multiscale mission (MMS). These results show a high statistical incidence of events detected near boundary crossings such as the magnetopause, as well as other interesting features occurring throughout near-Earth space, illustrating the potential for this tool to provide powerful data reduction capabilities for large-scale surveys, as a method of in-situ data prioritization in missions’ back orbits, or as a supplement to region-of-interest definitions for telemetry-limited missions.