Plasmapause Observations from a Data-Driven Model of the Magnetospheric Electric Field

The inner magnetosphere hosts distinct features such as the plasmasphere and its plasmapause boundary that interact between the denser, inner layer of the magnetosphere and the outer region. Due to the complexity of magnetospheric dynamics, scientists often rely on models of the inner magnetospheric electric field (IMEF) and electric potential for better understanding. However, existing models struggle to accurately reproduce inner magnetosphere electrodynamics, especially in times of high geomagnetic activity. Here, we present a physio-temporal analysis of the first Machine Learning Inner Magnetospheric Electric Field (ML-IMEF) model with the aim to advance the state of physics-based modeling of the magnetosphere through improved accuracy and predictive capabilities. ML-IMEF is a multi-layer deep neural network trained on electric field data from multiple instruments onboard NASA’s Magnetospheric Multiscale (MMS) mission where we train our model with the time history of location data and geomagnetic indices. The result of the IMEF is a global, dynamic and time-dependent model of the IMEF where we resolve the electric potential contours through the solving an inverse problem. We evaluate the modeled electric field and potential during varying geomagnetic storms, including the May 2024 Gannon Storm, and compare the plasmapause boundary with other models, such as the Moldwin et al. (2002) empirical plasmapause model. Furthermore, we explore magnetospheric characteristics of our model in relation to meso-scale electric field features, such as electric potential patterns and last closed equipotential (LCE) lines.