Real‐Time Forecasting of Pitch Angle‐Resolved Differential Electron Flux at MEO Using Solar Wind and Geomagnetic Indices

We present a real‐time operational machine learning model based on convolutional neural network for forecasting pitch angle‐resolved differential electron flux at Medium Earth Orbit (MEO) altitudes in the outer radiation belt. Unlike previous studies, our model is designed for immediate operational use, relying exclusively on real‐time accessible solar wind parameters and geomagnetic indices, without requiring non‐real‐time available inputs such as AE or AL indices or direct electron flux measurements. Our model captures multi‐timescale electron flux variations, making it well‐suited for dynamic space weather conditions. During a 10‐month test period in 2019, the model achieves a prediction efficiency (PE) of 0.79–0.88 for the 2‐hr forecast and a PE above 0.7 for the 1‐day forecasts at energies below 1.8 MeV. However, model performance declines for >2 MeV electrons at lower L ‐shells, a direct result of imbalanced data distribution caused by the distinct dynamics of ultra‐relativistic electrons within the plasmasphere, as discussed in the main text. By providing a real‐time forecasting framework, this model enhances space weather monitoring capabilities and can serve as a valuable tool for assessing radiation environment at MEO altitudes, especially when direct particle observations are unavailable.