D-SURGE · Data-driven Simulations for Understanding Reconnection and GEomagnetism

Understanding and predicting energization of particles in space plasmas, e.g. during sub-storms in magnetosphere, is a profound scientific challenge which is hampered by the complex multi-scale interactions in plasma. At present there is no global model which reliably captures the microscopic effects of energetic electrons which participate in matter and energy transfer in the Earth's space environment. These electrons may precipitate and present hazard for our space born assets such as satellites and space missions. The goal is to build a global physics-based magneto-fluid model of plasma with AI component that represents such microscopic physics via machine learning algorithms trained on particle in cell (PIC) simulations. By introducing physics-informed data-driven models, we aim to bridge the gap between local and global scales, offering unprecedented accuracy in simulating space weather phenomena. Such data-driven models will allow us to better understand and predict the evolution of magnetic storms crucial for safeguarding modern technological infrastructure. What makes this goal now possible in this interdisciplinary project are the advances in scientific machine learning in meteorology which the PI plans to bringing from environmental data science to plasma physics as well as the PI's current involvement in energy-conserving Particle in Cell simulations and previous experience in kinetic and magneto-fluid modelling. These innovations will lead to physics-informed data driven model that will be validated with space mission observations. If successful, we will have multi-scale multi-fluid models of plasma that will uncover magnetospheric loading by solar wind and the mechanisms behind acceleration of particles during magnetic sub-storms.