Research on Ionospheric Scintillation Effects and Prediction Model in East Asia Based on COSMIC-1 Occultation Dataset

In this study, the temporal and spatial distribution characteristics of ionospheric scintillation in the East Asian sector are statistically analyzed based on S4 data provided by the COSMIC-1 occultation dataset and solar–terrestrial spatial environment parameters from 2007 to 2018. The results show that scintillation activity has an obvious distribution pattern with local time: the frequency gradually increases from 17:00 in the evening, with the peak concentrated at 22:00–01:00 at night; in terms of seasonal variation, scintillation activity is highest in spring and fall, followed by summer, and lowest in winter; and, regarding annual variation, it is highly correlated with the solar activity. Further analyses show that scintillation activity is strongly correlated with geomagnetic activity. On this basis, this study constructs a two-layer LSTM deep learning model based on weighted regression to realize S4 numerical forecasting for the next 1 h in the middle- and low-latitude regions of China, using F10.7, Kp, Dst, sunspot number, solar wind vertical velocity, and historical S4 values as inputs. The model demonstrates robust predictive performance on the validation dataset containing 8760 samples, with a mean squared error of 0.00546 and an absolute error that is distributed within the interval [−0.2, 0.2] 98% of the time, indicating strong accuracy and robustness. These results suggest that the proposed model provides a high-precision tool for ionospheric scintillation warning.