Machine learning-based emulator for physics-based simulation of auroral current system
Published in Space Weather, 2024
Physics-based auroral simulations, such as Japanese REProduce Plasma Universe (REPPU) code, are not practically fast enough for the purpose of real-time space weather forecast, even using the designated super computers. Here we developed a million-times-faster emulator to surrogate the outputs of the physics-based simulation, using the machine-learning technique called Echo State Network. The newly developed emulator, the surrogate model for REPPU auroral Ionosphere version 2 (SMRAI2) enables us to realize the real-time forecast of the auroral current system.
Recommended citation: Kataoka, R., A. Nakamizo, S. Nakano, and S. Fujita (2024). "Machine learning-based emulator for physics-based simulation of auroral current system" Space Weather. 21, e2023SW003720 https://doi.org/10.1029/2023SW003720