[P-EM13] Dynamics of the Inner Magnetospheric System

[PEM13-P06] Forecast of the Dst index using the XAI

*西野 幹志¹、三好 由純¹ (1.名古屋大学宇宙地球環境研究所)

Geomagnetic storms are caused by the large enhancements of ring current, and severe space weather events often happen during the storm time. The Dst index is a measure of the intensity of geomagnetic storms, and making its prediction crucial. There are two different types of geomagnetic storms: CME (Coronal Mass Ejection) driven storms and CIR (Corotating Interaction Regions) driven storms. CMEs are phenomena that cause plasma to be ejected from the corona into interplanetary space and CIRs are regions of compression caused by the stream interface region. The large magnetic storms are caused by CMEs. The CME-driven storms are moderate and have long-lasted recovery phase. To improve Dst index forecast, we developed a forecast model using a recurrent neural network (RNN) with a long short-term memory (LSTM) architecture. Furthermore, we applied Shapley Additive exPlanation (SHAP), a type of eXplainable Artificial Intelligence (XAI), which is a powerful method for identifying key parameters in time-series analysis. Our neural network model used solar wind velocity, IMF-Bz, and solar wind density as input paremeters. We also compared CME-driven storms and CIR-driven storms using XAI techniques. Our analysis reveals that IMF-Bz is the most important driver of geomagnetic storms, with stronger southward interplanetary magnetic fields leading to more intense magnetic storms. Furthermore, solar wind speed plays a crucial role in the recovery phase of CIR-associated storms.