Predicting the SYM-H index using the ring current energy balance mechanism

The geomagnetic disturbance index SYM-H is primarily governed by the total kinetic energy of ring current particles. Consequently, the energy balance mechanism of the ring current provides a foundation for constructing an SYM-H evolution equation for predictive purposes. This study builds upon a modeling approach introduced by Ji et al. (2023) to develop an algebraic equation for predicting the SYM-H index based on the equilibrium between energy injection and ring current loss. A fully connected neural network determines the loss term in the model, while the energy injection function is derived from established solar wind–magnetosphere energy coupling functions, with its scale factor treated as a free-fitting parameter to optimize observational predictions. The model, trained on solar wind and SYM-H data from 20 magnetic storms, effectively predicts the SYM-H index one and two hours in advance, achieving root mean square errors (RMSE) of 6.7 nT and 8.9 nT, respectively. These results reflect a 7% improvement for the 1-hour model and a 6% improvement for the 2-hour model compared to the previous version. Moreover, the scale factors for the solar wind parameters in the energy coupling function align with prior observations of the magnetotail current sheet, reinforcing the conclusion that the ring current energy predominantly originates from the current sheet. The neural network-determined lifetimes of ring current particles vary with the SYM-H index, approximating six hours during the fast recovery phase and exceeding 10 hours in the slow recovery phase. This variation is consistent with a transition in dominant ring current particles from oxygen ions to protons during intense storms.