Magnetopause location modeling using machine learning: inaccuracy due to solar wind parameter propagation

• DOI: 10.3389/fspas.2024.1390427
• Publication Date: 2024-05-15T05:11:11Z
• Authors: M. Aghabozorgi Na...; F. Němec; G. Pi; Z. Němeček; J. Šafránková; K. Grygorov; J. Šimůnek; T.-C. Tsai
• Keywords: magnetopause crossings; solar wind parameter propagation; QC801-809; Astronomy; Geophysics. Cosmic physics; 0103 physical sciences; magnetopause empirical model; model inaccuracies; QB1-991; 01 natural sciences; magnetopause machine learning model

An intrinsic limitation of empirical models the magnetopause location is a predefined shape and assumed functional dependences on relevant parameters. We overcome this using machine learning approach (artificial neural networks), allowing us to incorporate general, purely data-driven dependences. For training testing developed network model, data set about 15,000 crossings identified in THEMIS A-E, Magion 4, Geotail, Interball-1 satellite subsolar region used. A cylindrical symmetry around direction impinging solar wind assumed, dynamic pressure, interplanetary magnetic field magnitude, cone angle, clock tilt corrected Dst index are considered as The effect these parameters revealed. performance model compared with other models. Finally, we demonstrate discuss inaccuracy due inaccurate information based measurements near L1 point. This imposes theoretical limit precision predictions, that our closely approaches.