Cross‐Calibrated FY‐4A and GOES‐15 Data for >2 MeV Electron Flux Prediction at Geosynchronous Orbit: A Machine Learning Approach

This is the authors' abstract. We don't add key points for this paper.

High‐energy electrons in Earth's outer radiation belt pose a significant threat to orbiting spacecraft, accurately forecasting these fluxes is crucial for mitigating potential damage. Therefore, it is necessary to fuse data from different satellites to eliminate systematic discrepancies between satellite data sets, thereby enabling the development of more accurate and reliable predictive models. This study presents a cross‐satellite calibration method to address systematic discrepancies in >2 MeV electron flux measurements from the FY‐4A and GOES‐15 satellites. By aligning magnetic local times (MLT) through a time‐shift operation and excluding solar proton contaminations, we ensure high‐quality data for calibration. The data are divided into 24 MLT intervals to determine optimal calibration factors, significantly improving consistency between the two data sets. Using the calibrated data, we generate a merged data set spanning 13 years, encompassing a complete solar activity cycle. This extended data set is utilized to develop a neural network‐based nowcasting model for predicting >2 MeV electron integral flux at geosynchronous orbit (GEO). Our results demonstrate that cross‐calibration with expanding the data set's temporal coverage enhances predictive performance for both quiet and active geomagnetic periods. The improved accuracy arises from increased training data volume, comprehensive representation of solar activity cycles. These findings underscore the importance of cross‐satellite calibration and long‐term data sets for advancing space weather forecasting capabilities. This study highlights the value of multi‐source data integration and advanced modeling techniques in understanding and mitigating radiation belt hazards.