Abstract The electron flux in the outer radiation belt is primarily governed by interactions between the solar wind and magnetosphere. While geomagnetic indices are widely used for modeling and forecasting, their real‐time availability is limited. In contrast, solar wind parameters can be continuously monitored by a single satellite at the point, providing better timeliness and accessibility. This work investigates whether solar wind parameters alone can effectively predict radiation belt electron flux variations and how prediction accuracy depends on geomagnetic activity, energy, and forecast lead time. We propose a dual‐module machine learning framework to forecast electron flux under two configurations: (a) using both geomagnetic indices and solar wind parameters, and (b) using only solar wind parameters. Based on nearly seven years of Van Allen Probes observations, we train separate models for each energy channel (235–909 keV) and forecast lead time (nowcast to 24 hr). Including geomagnetic indices improves prediction efficiency to 0.813–0.896, while models using only solar wind parameters still achieve 0.793–0.848. Forecast performance, however, declines during extreme geomagnetic storms, likely due to the limited availability of storm‐time data. These results demonstrate the feasibility and limitations of developing radiation belt forecasting systems based solely on single‐satellite solar wind observations, and suggest that improved performance during extreme geomagnetic conditions may require expanded training data sets or hybrid physics‐machine learning approaches.
Li et al. (Wed,) studied this question.