Predicting characteristics of bursty bulk flows in Earth’s plasma sheet using machine learning techniques

Bursty bulk flows (BBFs) play a crucial role in transporting energy, mass, and magnetic flux in the Earth's magnetotail, particularly in the earthward direction. However, their impulsive nature and small spatial scale present significant challenges for in-situ observation, as only a limited number of spacecraft operate within the vast expanse of the magnetotail. Consequently, studying their statistical characteristics is a highly demanding task, and accurately predicting their behavior remains a distant goal. In this study, we analyze key characteristics of BBFs and apply regression-based models to predict their parameter behaviorUsing observational data from the THEMIS mission collected between 2007 and 2023, we conducted a feature analysis on parameters associated with BBFs evolution, including velocity, magnetic field, electric field, temperature, density, pressure, and specific entropy indices. Through statistical techniques, we identified parameters exhibiting predictable patterns during BBF events, distinguishing them from background conditions. Furthermore, we used XGBoost regression model, optimized for different parameter combinations, to forecast BBF duration, physical parameters’ average minimum, and peak intensity. This study also tested combinations of parameter predictions across instruments. When using observed background value in parameter combination, our models achieved Mean Absolute Percentage Errors of under 35% for critical variables, including Bz, Btotal, plasma pressure, and ion temperatures, and ion specific entropy and so on. Additionally, we observed BBF duration’s spatial distribution trends: it peaked at approximately X=-13Re, while decreasing with increasing Z distance from the plasma sheet, showing dawn-dusk asymmetry consistent with prior observations. This work highlights the potential of regression methods in forecasting BBFs characteristics and offers insights into their spatial behavior, supporting enhanced prediction capabilities in magnetospheric studies. Future research will aim to improve accuracy with enriched datasets.