The machine learning technique has been developed to predict various plasma quantities including total plasma density and plasma wave amplitude in earth's magnetosphere using controlling indices. We…

The Trustees of Boston University received a $837K research grant from a civilian federal agency (Shared Services Center) with a completion date of December 31, 2024, to develop and validate advanced machine learning and simulation techniques for predicting plasma wave distributions in Earth's magnetosphere. The project leverages existing machine learning models to forecast global distributions of whistler mode waves (chorus, hiss, and magnetosonic waves), EMIC waves, and ULF waves using solar wind parameters derived from OMNI data. The research integrates observational data from THEMIS and Van Allen Probes satellites to refine predictions and analyze radiation belt electron flux variations in response to solar wind drivers.

The technical approach combines machine learning predictions with established simulation methodologies, specifically utilizing the UCLA 3D diffusion code to model electron evolution through resonant wave-particle interactions. Boston University will validate simulation results against satellite observations using established validation metrics and will benchmark machine learning-derived plasma wave distributions against empirical models based on recent satellite statistics. This research supports the development of improved space weather forecasting capabilities by connecting solar wind drivers to radiation belt dynamics through accurate characterization of the global plasma wave environment.

The machine learning technique has been developed to predict various plasma quantities including total plasma density and plasma wave amplitude in earth's magnetosphere using controlling indices. We will extend this technique to predict the global distribution of whistler mode waves (chorus, hiss, magnetosonic wave), emic and ULF waves using solar wind parameters. The solar wind parameters are available from the omni data, and the dataset of various plasma waves and electron density are available from the themis and van allen probes measurements. We will use the van allen probes observation of energetic electrons to analyze the radiation belt electron flux variations. After identifying typical events showing the electron response to the solar wind drivers, we will simulate the radiation belt flux evolution using the global plasma wave distribution predicted by the machine learning technique. The ucla 3D diffusion code has been developed and fully validated for simulating the electron evolution due to resonant wave-particle interactions. For comparison, we will also perform the radiation belt simulation using the empirical plasma wave model, which is available from the recent satellite statistics. The simulation results will be validated against the satellite observations using validation metrics.