Initial Condition Estimation in Flux Tube Simulations using Machine Learning

Abstract

Space weather has become an essential field of study as solar flares, coronal mass ejections, and other phenomena can severely impact Earth's life as we know it. The solar wind is threaded by magnetic flux tubes that extend from the solar atmosphere to distances beyond the solar system boundary. As those flux tubes cross the Earth's orbit, it is essential to understand and predict solar phenomena' effects at 1 AU, but the physical parameters linked to the solar wind formation and acceleration processes are not directly observable. Some existing models, such as MULTI-VP, try to fill this gap by predicting the background solar wind's dynamical and thermal properties from chosen magnetograms and using a coronal field reconstruction method. However, these models take a long time, and their performance increases with good initial guesses regarding the simulation's initial conditions. To address this problem, we propose using varied machine learning techniques to obtain good initial guesses that can accelerate MULTI-VP's computational time.