Global Ionospheric TEC Forecasting Driven by a Unified Physical Index System via the Spectral Whitening Method

Abstract

Accurate forecasting of global ionospheric total electron content (TEC) is important for Global Navigation Satellite Systems (GNSS) performance and satellite communications. This study proposes a driver framework based on the spectral whitening method (SWM). We combine a new high‐latitude geomagnetic disturbance index with previously developed SWM‐based solar extreme ultraviolet and geomagnetic‐response indices and use them to drive a deep learning model for global TEC forecasting. Unlike existing driving indices (e.g., AE, SYM‐H, F10.7) derived with different procedures, the SWM‐based drivers are generated with a consistent signal‐extraction method designed to isolate non‐stationary disturbances. Throughout the year, the Pearson correlation coefficients between the SWM index group and the existing driving indices remain greater than 0.7. Seasonally, the SWM‐driven model yields Root Mean Square Error (RMSE) of 2.13–3.49 TECU, compared to 3.68–4.46 TECU for the IRI‐2020 model. For spring and winter specifically, the SWM‐driven RMSE are 2.63 and 2.76 TECU, compared to 2.64 and 2.79 TECU for the traditional‐index‐driven model. During recursive forecasts extending to 144 hr for the 2021 geomagnetic storm, the SWM‐driven model records RMSE of 4.57 TECU at 96 hr and 4.74 TECU at 144 hr, whereas the traditional‐index‐driven model yields 4.65 TECU and 4.94 TECU, respectively. These results suggest that SWM‐derived driver indices can improve forecast stability under disturbed conditions.