Global Ionospheric Total Electron Content Prediction Based on Spatiotemporal Network Model

Ionospheric delay is an important source of error in the Global Navigation Satellite System, and the ionosphere total electron content (TEC), as an important parameter describing the ionospheric properties, directly affects the magnitude of the ionospheric delay. Accurate modeling and forecasting of ionospheric TEC can serve to correct ionospheric delay and improve the accuracy of satellite navigation positioning. However, most of the current ionospheric TEC forecasting studies are only related to the temporal or spatial dimensions, without combining the temporal and spatial information of the global ionospheric TEC, ignoring the spatial and temporal autocorrelation, which seriously limits the forecast accuracy of the global ionospheric TEC. Therefore, we construct a spatiotemporal network forecast model with convolutional long short-term memory with spatiotemporal memory to achieve spatiotemporal forecasting of global ionospheric TEC by learning the temporal variation characteristics and spatial features of global ionospheric TEC and further suppresses the gross error and noise in ionospheric TEC data by using Huber loss function to improve the forecast accuracy of global ionospheric TEC. Compared with the one-day forecasts of global ionospheric TEC from the convLSTM model and the CODE’s spherical harmonic model, the average RMSE of the forecast results of our method is improved by 13.65% and 13.96%, respectively, which can be effectively applied to ionospheric delay error correction and improve the accuracy of satellite navigation and positioning.