Identifying MESSENGER Magnetospheric Boundary Crossings Using a Random Forest Region Classifier

Abstract We present a new list of bow shock and magnetopause crossings based on automated region classification for the MESSENGER (Mercury Surface, Space Environment, Geochemistry and Ranging) mission. We fit a random forest model to magnetometer and ephemeris data to classify the solar wind, magnetosheath, and magnetosphere regions surrounding Mercury. The random forest was highly accurate when predicting the testing data set, with an accuracy of . We apply this model to the orbital phase (March 2011 to April 2015) of the MESSENGER mission, and determine crossings automatically where changes in region classification occur. We apply several post‐processing steps to this list of crossings and determine 19,989 bow shock crossings and 11,414 magnetopause crossings for the MESSENGER mission. We find good visual comparison with the spatial distribution of these new crossings and existing data sets of boundary crossing intervals.