The application of deep learning technology to ionospheric prediction has become a new research hotspot. However, there are still some gaps, such as the prediction effect with different input solar and geomagnetic activity parameters, and the forecast accuracy with different prediction methods as well as the validation of long period data results, to be filled. We developed an ionospheric long short‐term memory network (Ion‐LSTM) with multiple input parameters to predict the global ionospheric total electron content (TEC). Two solutions with different ionospheric data based on Ion‐LSTM were assessed, namely spherical harmonic coefficients (SHC) and vertical TEC (VTEC) prediction solution. The results show two solutions, both perform well in accuracy and stability. The input of the geomagnetic activity index improves the prediction effect of the model in the storm period. For the 1‐ and 2‐day‐predicted global ionospheric maps (GIMs) from 2015 to 2020, the root mean square error (RMSE) of SHC prediction solution is 1.69 TECU and 1.84 TECU while that of the VTEC prediction solution is 1.70 TECU and 1.84 TECU, respectively. Over 70% of the absolute residuals are within 3 TECU in high solar activity and over 96% in low solar activity. Further, by comparing the predicted results between Ion‐LSTM and conventional methods (e.g., Center for Orbit Determination in Europe (CODE) predicted GIMs), the evaluation results show that the RMSE of Ion‐LSTM is 0.7 TECU lower than that of CODE predicted GIMs under different solar and geomagnetic activities. Additionally, the accuracy of the Ion‐LSTM prediction results decreases slightly as the input time span increases.

In recent years, deep learning has been extensively used for ionospheric total electron content (TEC) prediction, and many models can yield promising prediction results, particularly under quiet conditions. Owing to the ionosphere's intricate and dramatic changes during geomagnetic storms, the high‐reliable prediction of the storm‐time ionospheric TEC remains a challenging problem. In this study, we developed a new deep learning‐based multi‐model ensemble method (DLMEM) to forecast geomagnetic storm‐time ionospheric TEC that combines the Random Forest (RF) model, the Extreme Gradient Boosting (XGBoost) algorithm, and the Gated Recurrent Unit (GRU) network with the attention mechanism. Seven features in 170 geomagnetic storm events, including the three components Bx, By and Bz of interplanetary magnetic field (IMF), the Kp and Dst indices of geomagnetic activity data, the F_10.7 index of solar activity data and global TEC data, were used for modeling. The test set results showed that the DLMEM model can reduce the root mean square errors (RMSE) by an average of 43.6% in comparison to our previously presented model Ion‐LSTM, especially during the recovery period of geomagnetic storms. Furthermore, compared to Ion‐LSTM, the RMSE values of the low‐, middle‐ and high‐latitude single‐station forecast TEC can be greatly decreased by 33%, 53% and 59%, respectively. It was shown that the new model allows for more precise short‐term global ionospheric forecasting during geomagnetic storms, enabling real‐time monitoring of ionospheric changes.

To provide backup and supplementation for the Global Navigation Satellite System (GNSS), Doppler shift from Low Earth Orbit (LEO) satellites can be used as signals of opportunity to provide positioning, navigation, and timing service. In this contribution, we first investigate the model and performance of instantaneous velocity determination and positioning with LEO satellites. Given a LEO constellation with 288 satellites, we simulate Doppler shift observations at nine multi-GNSS experiment stations. Owing to the lower orbit, the performance of LEO velocity determination is much more sensitive to the initial receiver position error than that of GNSS. Statistical results show that with the initial receiver position error increased from 0.1 to 10 m, the Root Mean Square Errors (RMSEs) increase from 0.73 to 2.65 cm/s, 0.68 to 2.96 cm/s, and 1.67 to 4.15 cm/s in the east, north, and up directions, respectively. The performances with GPS are compared with GPS + LEO, and it is found that LEO Doppler shift observations contribute to GPS velocity determination. As for LEO Doppler positioning, even if more than 30 visible LEO satellites are available, the position dilution of precision values can reach several hundreds. Assuming that the error of LEO Doppler measurements is 0.01 m/s, the instantaneous Doppler positioning accuracy can achieve about a few meters, which is comparable to that of GNSS pseudorange positioning. A constant velocity model is adopted for state transition. Static LEO Doppler positioning results show that an accuracy at centimeter to decimeter level can be achieved after solution convergence. For a static simulated kinematic positioning test, the RMSEs range from a few decimeters to several meters in different regions by giving different constraints. For a dynamic positioning test, the RMSEs are about 2–3 m in high latitude region.

Tropospheric delay error is independent of the signal's frequency and has strong spatiotemporal variation. It is one of the most severe error sources of satellite navigation and spatial measurement. In view of the limitation of global zenith tropospheric delay (GZTD) model considering the altitude coefficient β as a constant and ignoring its spatiotemporal variation, this paper analyzes the spatiotemporal distribution of zenith tropospheric delay (ZTD) with altitude variation coefficient β based on the meteorological reanalysis data provided by the European Centre for Medium‐Range Weather Forecasting from 2011 to 2015. The global altitude coefficient β model is established by using the trigonometric function and seventh‐order spherical harmonics function model, and then the global tropospheric delay model R_GZTD (reconstruction GZTD) is reconstructed by using Gbeta (Global beta β) model. The R_GZTD model gives high‐accuracy global distribution of tropospheric delays without meteorological parameters. The results show that the internal accuracy of the R_GZTD model is 3.22 cm, which has a good fitting effect. This paper uses the tropospheric delay products in 2016–2017 provided by the International GNSS Service, the tropospheric delay calculated by the European Centre for Medium‐Range Weather Forecasting reanalysis data, and the radiosonde ZTD data in 2016–2017 as external compliance check data. The results show that the accuracy of the R_GZTD model is better than that of the GZTD model, UNB3m model, and the global pressure and temperature 2 wet model in the global and regional scope. Especially in areas with a higher altitude, the correction effect of the R_GZTD model is more significant. The root‐mean‐square error is 8.5% smaller than that of the GZTD model in the range of 0–500 m, 14.6% smaller in the range of 500–1,000 m, 12.5% smaller in the range of 1,000–2,000 m, and 16.8% smaller in the range above 2,000 m. The accuracy with the increasement of height is due to the fact that the R_GZTD model takes account of the spatiotemporal variation of β.