Machine Learning Classification of Interplanetary Coronal Mass Ejections using Satellite Accelerometers

Space weather phenomena is a complex area of research as there are many different variables and signatures that are used to identify the occurrence of solar storms and Interplanetary Coronal Mass Ejections (ICMEs), with inconsistencies between databases and solar storm catalogues. The identification of space weather events is important from a satellite operation point of view, as strong geomagnetic storms can cause orbit perturbations to satellites in low-earth orbit. Satellite-borne accelerometers have contributed significantly to advancing the knowledge of the thermospheric response to geomagnetic storms for the past few decades. The GRACE, CHAMP, CASTOR-D5B, ACE, and SWARM satellites are all examples of spacecraft housing accelerometer-based experiments utilized for atmospheric density derivation, and extensive work has been done to understand the affects of space weather events on the Earth’s neutral density utilizing these instruments. Obtaining themorspheric density from raw accelerometer data requires modelling and complicated derivations, which is both time consuming and requires significant computation power and time. The question is then asked, is it possible to examine satellite response to space weather events using only the raw accelerometer data? This work presents a new way of utilizing the raw, unprocessed satellite accelerometer data from the GRACE-A satellite to identify Interplanetary Coronal Mass Ejections. Using decision-tree based binary classification models, statistical features are extracted from the accelerometer data of the GRACE-A satellite. Random Forest and Extremely Randomized Trees machine learning classifiers are trained using the data features and achieve an accuracy of greater than 80%. The results yield two main conclusions: accelerometer data can be utilized as an additional characteristic signature for ICMEs, and a binary classifier can be trained on spacecraft accelerometer data for space weather prediction applications.