Abstract:
The Earth’s radiation belts consist of energetic charged particles trapped by the geomagnetic field. This radiation environment, exhibiting rich dynamical variations, is known to be particularly hazardous to spacecraft and is difficult to predict because of its complex physical process. This seminar will discuss the radiation belts starting from a machine-learning (ML) model, which accurately reproduces the electron dynamics driven only by geomagnetic indices and solar wind parameters. The model’s performance surpasses all physical models. However, the machine learning model is a “black box”, we will then conduct a feature attribution method named SHAP(SHapley Additive exPlanations) to further understand the capabilities and the information of this ML model. The framework has the ability to identify what parameter controls, and how they control the different dynamics. In addition, we will use such a framework to answer a long-standing question: why electron flux can experience depletion or acceleration after a geomagnetic storm? At last, we will demonstrate how the insight we learn in the ML model can be used to help us build a general physics-based model and the physical understanding from the results.