Plasma is the most abundant form of matter in our Universe, and magnetic reconnection is the most energetic process that happens in plasma. During reconnection, oppositely directed magnetic field lines become stretched and break yielding characteristic signatures in spacecraft magnetometer and plasma sensor data.
This project will focus on the use of machine learning for feature identification of reconnection signatures. Collaboration with Al and machine learning experts across the Turing Institute will enable us to address the specific challenges which spacecraft data bring, including, but not limited to (i) data gaps, (ii) variable viewing (due to changing spacecraft orbits) and (iii) class imbalance (due to the sporadic nature of reconnection). We also aim to work to change the culture in the Space Physics community to embrace the power of machine learning to unlock the potential of vast, rich, complex data sets.
The proposed work will explore strategies to tackle these academic and societal challenges, with a view to building more ambitious large-scale research programmes in the medium term which fully utilise machine learning techniques to bring Space Physics data analysis in line with the state-of-the-art in automated feature identification.
Principal Investigator: Dr Caitriona Jackman (Southampton)