Explore the vast range of titles available.

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 F10.7 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.

The total electron content (TEC) is an important parameter for characterizing the morphology of the ionosphere. Modeling the ionospheric TEC accurately during the storm time could contribute to the operation of global navigation satellite systems (GNSS), satellite communications, and other applications. This study uses an image-based convolutional long short-term memory (ConvLSTM) network with multichannel features to forecast ionospheric TEC during the quiet periods and storm periods. The sunspot number (SSN), solar wind velocity ( V sw ), Dst, and Kp geomagnetic indices are firstly fed into the model as the channel features to improve generalization performance. Based on the variation of the Dst index, we have collected gridded TEC maps from 2011 to 2018 with a 1-h interval from the global ionospheric maps (GIM) as the data set including quiet periods and storm periods of ionospheric TEC. The performance of the ConvLSTM model in forecasting TEC is also compared with other deep learning models such as LSTM, gated recurrent unit (GRU), and LSTM-CNN. Furthermore, the accuracy consistency of the ConvLSTM model during the different phases of the storm period is also evaluated for the different output steps of predicted TEC maps. The optimal combination of input features for the model is also investigated during the storm period. Testing results show that the ConvLSTM network with multichannel features has good prediction performance for quiet periods and storm periods by incorporating both solar and geomagnetic activity indices. The statistical indicators show that the ConvLSTM model performs well with lower mean absolute error (MAE), root mean square error (RMSE), and larger correlation coefficient (R) compared with other methods. We have demonstrated that the model with a larger prediction step has worse prediction performance at the low-latitude area, especially during the storm period. In our future work, the larger TEC data set and more solar and geomagnetic indices will be investigated. Highlights An image-based convolutional long short-term memory (ConvLSTM) network with multichannel features for forecasting ionospheric TEC. Solar and geomagnetic indices as the input features for improving the performance of the model. The ConvLSTM model has good prediction performance for quiet and storm periods by incorporating both solar and geomagnetic activity indices.

Specification and forecast of ionospheric parameters, such as ionospheric electron density (Ne), have been an important topic in space weather and ionospheric research. Neural networks (NNs) emerge as a powerful modeling tool for Ne prediction. However, heavy manual adjustments are time consuming to determine the optimal NN structures. In this work, we propose to use neural architecture search (NAS), an automatic machine learning method, to mitigate this problem. NAS aims to find the optimal network structure through the alternate optimization of the hyperparameters and the corresponding network parameters within a pre‐defined hyperparameter search space. A total of 16‐year data from Millstone Hill incoherent scatter radar (ISR) are used for the NN models. One single‐layer NN (SLNN) model and one deep NN (DNN) model are both trained with NAS, namely SLNN‐NAS and DNN‐NAS, for Ne prediction and compared with their manually tuned counterparts (SLNN and DNN) based on previous studies. Our results show that SLNN‐NAS and DNN‐NAS outperformed SLNN and DNN, respectively. These NN predictions of Ne daily variation patterns reveal a 27‐day mid‐latitude topside Ne variation, which cannot be reasonably represented by traditional empirical models developed using monthly averages. DNN‐NAS yields the best prediction accuracy measured by quantitative metrics and rankings of daily pattern prediction, especially with an improvement in mean absolute error more than 10% compared to the SLNN model. The limited improvement of NAS is likely due to the network complexity and the limitation of fully connected NN without the time histories of input parameters.

The ionosphere poses a significant source of error in satellite‐based navigation systems for aviation and radio communication applications. Accurate estimation of the total electron content (TEC) can effectively mitigate the impact of such errors. However, constrained by observational techniques, the acquisition of global ionospheric TEC in practical applications relies heavily on high‐precision forecasting products. In this study, we construct a global ionospheric forecasting model based on the global ionospheric TEC products disseminated by the International GNSS Service (IGS) and deep learning. We incorporate an attention module to extract global spatiotemporal features from historical ionospheric data and employ these features to predict the TEC values over the next 24 hr. Additionally, we select a long short‐term memory (LSTM) model and a ConvLSTM model as baseline models for comparison, conducting experiments under varying solar activity conditions. The experimental results demonstrate that RA‐ConvLSTM model outperforms the other two models in quantifying the performance of the models. During high solar activity years, the bias and Root Mean Square Error (RMSE) of RA‐ConvLSTM model are −0.0298 TECU and 3.8980 TECU, respectively, while during low solar activity years, these values are 0.0905 TECU and 1.5059 TECU, marking a notable improvement over the comparative models. Furthermore, by contrasting the precision of the three forecasting models during geomagnetic storms, the RA‐ConvLSTM model exhibits the least fluctuations in accuracy, indicative of a higher degree of stability in its forecasting outcomes.

To achieve accurate forecasting of the peak height of the ionospheric F2 layer (hmF2), we propose a hybrid deep learning model of improved seagull optimization algorithm (ISOA) optimized long short‐term memory (LSTM) model based on a complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN) theory. The hybrid model decomposes the hmF2 time data into multiple subsequences through CEEMDAN and reconstructs the subsequences by sample entropy and correlation coefficient into high and low‐frequency sequences, which effectively shortens the calculation time of the model. Then, we determine the optimal hyperparameters of the LSTM models through ISOA, achieving high‐precision forecasting of the hmF2. In single‐step forecasting, the forecasting values of the hybrid model in diurnal and seasonal changes are highly consistent with the observation, which can better capture the severe changes in the hmF2. The model's RMSE, MAE, MAPE, and CC evaluation metrics are 15.86, 11.03 km, 4.76%, and 0.93 in the test set. Compared to IRI, GRU, and LSTM models, taking RMSE as an example, the forecasting accuracy of the models increased by 65.24%, 29.89%, and 29.60%, respectively. In multi‐step forecasting, the proposed model is better at forecasting the changing trend of hmF2, and the forecasting accuracies are significantly better than the IRI model. The data from multiple stations also verified the applicability of the proposed model for hmF2 forecasting. The above results indicate that the hybrid model has high accuracy in hmF2 short‐term forecasting and good applicability in multiple multi‐step forecasting, which can further improve the accurate forecasting of space weather.

In this paper, we forecasted the ionospheric total electron content (TEC) over the region of China using the bidirectional convolutional gated recurrent unit (BiConvGRU) model. We first generated the China Regional Ionospheric Maps (CRIMs) using GNSS observations provide by the Crustal Movement Observation Network of China (CMONOC). We then used gridded TEC maps from 2015 to 2018 with a 1 h interval from the CRIMs as the dataset, including quiet periods and storm periods of ionospheric TEC. The BiConvGRU model was then utilized to forecast the ionospheric TEC across China for the year 2018. The forecasted TEC was compared with the TEC from the International Reference Ionosphere (IRI-2016), Convolutional Long Short-Term Memory (ConvLSTM), Convolutional Gated Recurrent Unit (ConvGRU), Bidirectional Convolutional Long Short-Term Memory (BiConvLSTM), and the 1-day Predicted Global Ionospheric Map (C1PG) provided by the Center for Orbit Determination in Europe (CODE). In addition, indices including Kp, ap, Dst and F10.7 were added to the training dataset to improve the forecasting accuracy of the model (-A indicates no indices, while -B indicates with indices). The results verified that the prediction accuracies of the models integrated with these indices were significantly improved, especially during geomagnetic storms. The BiConvGRU-B model presented a decrease of 41.5%, 22.3%, and 13.2% in the root mean square error (RMSE) compared to the IRI-2016, ConvGRU, and BiConvLSTM-B models during geomagnetic storm days. Furthermore, at a specific grid point, the BiConvGRU-B model showed a decrease of 42.6%, 49.1%, and 31.9% in RMSE during geomagnetic quiet days and 30.6%, 34.1%, and 15.1% during geomagnetic storm days compared to the IRI-2016, C1PG, and BiConvLSTM-B models, respectively. In the cumulative percentage analysis, the BiConvGRU-B model had a significantly higher percentage of mean absolute error (MAE) within the range of 0–1 TECU in all seasons compared to the BiConvLSTM-B model. Meanwhile, the BiConvGRU-B model outperformed the BiConvLSTM-B model with lower RMSE for each month of 2018.

Accurate modeling of ionospheric variability is critical for space weather forecasting and GNSS applications. While machine learning approaches have shown promise, progress is hindered by the absence of standardized benchmarking practices and narrow test periods. In this paper, we take the first step toward fostering rigorous and reproducible evaluation of AI models for ionospheric forecasting by introducing IonoBench: a benchmarking framework that employs a stratified data split, balancing solar intensity across subsets while preserving 16 high-impact geomagnetic storms (Dst ≤ −100 nT) for targeted stress testing. Using this framework, we benchmark a field-specific model (DCNN) against state-of-the-art spatiotemporal architectures (SwinLSTM and SimVPv2) using the climatological IRI 2020 model as a baseline reference. DCNN, though effective under quiet conditions, exhibits significant degradation during elevated solar and storm activity. SimVPv2 consistently provides the best performance, with superior evaluation metrics and stable error distributions. Compared to the C1PG baseline (the CODE 1-day forecast product), SimVPv2 achieves a notable RMSE reduction up to 32.1% across various subsets under diverse solar conditions. The reported results highlight the value of cross-domain architectural transfer and comprehensive evaluation frameworks in ionospheric modeling. With IonoBench, we aim to provide an open-source foundation for reproducible comparisons, supporting more meticulous model evaluation and helping to bridge the gap between ionospheric research and modern spatiotemporal deep learning.

The ionospheric state holds significant implications for satellite navigation, radio communication, and space weather; however, precise forecasting of the ionosphere remains a formidable challenge. To improve the accuracy of traditional forecasting models, we developed an enhancement model based on the CODE and IRI forecasting methods, termed the Global Ionospheric Maps Forecast Enhancement Model (GIMs-FEM). The results indicated that by extracting the GIM features from existing forecasts and incorporating additional proxies for geomagnetic and solar activity, the GIMs-FEM provided stable and reliable forecasting outcomes. Compared to the original forecasting models, the overall model error was reduced by approximately 15–17% on the test dataset. Furthermore, we analyzed the model’s performance under different solar activity conditions and seasons. Additionally, the RMSE for the C1pg model ranged from 0.98 TECu in the solar minimum year (2019) to 6.91 TECu in the solar maximum year (2014), while the enhanced GIMs (C1pg) model ranged from 0.91 to 5.75 TECu, respectively. Under varying solar activity conditions, the RMSE of GIMs-FEM for C1pg (C2pg) ranged from 0.98 to 6.91 TECu (0.96 to 7.26 TECu). Seasonally, the GIMs-FEM model performed best in the summer, with the lowest RMSE of 1.9 TECu, and showed the highest error in the autumn, with an RMSE of 2.52 TECu.

The critical frequency of ionospheric F2 layer (foF2) is an important ionospheric characteristic parameter. In this paper, a deep learning model based on Bidirectional long short‐term memory (BiLSTM) and attention mechanism is implemented for predicting the foF2 parameter. The inputs of models are the foF2 of globally available ionospheric ionosonde stations, geographic longitude and latitude, world time (UT), geomagnetic activity index, and solar activity index from 2015 to 2017. The superiority of the model is analyzed from different latitudes, seasons, and geomagnetic conditions. The results show that the prediction performance of the Bidirectional long short‐term memory model based on attention mechanism (BiLSTM‐Attention) is better than other models. The performance of the prediction model is optimal at high latitudes. The root mean square error (RMSE) and correlation coefficient (R) of the BiLSTM‐Attention model are 0.539 MHZ and 0.908 MHz at high latitudes, respectively. In terms of RMSE, it is 25.243%, 18.209%, and 11.203% lower than those of the international reference ionosphere (IRI), LSTM, and BiLSTM models, respectively. The prediction results of the four seasons show that the models are more applicable in winter. Compared with the IRI model, the RMSE of the BiLSTM‐Attention model in spring, summer, autumn, and winter is reduced by 24.344%, 21.181%, 25.058%, and 30.948%, respectively. The prediction effect of the BiLSTM‐Attention model is improved in the magnetic quiet period, the magnetic moderate period and the magnetic storm period. Also, the improvement effect is more obvious in the magnetostatic day, and the RMSE is reduced by 27.462% compared with the IRI model.

This paper presents forecast and investigation of the variation in ionospheric Total Electron Content (TEC) during the solar eclipses (SEs) of December 2019 and June 2020 using three different methods: Deep Autoregressive model (DeepAR), Feed-Forward Neural Network (FFNN), and Ordinary Kriging-based Surrogate Model (OKSM), and the TEC data predicted by DeepAR, FFNN, and OKSM were compared with the actual TEC during the observation days. The study was conducted based on GPS data taken from the IISC receiver located in Bangalore, India, during the SEs which happened on 26.12.2019 and 21.06.2020. The TEC data were examined to assess the effect of solar eclipses on TEC values. Eighty-day prior TEC data for the IISC station are gathered from IONOLAB servers along with the other parameter data like Dst, Ap, F10.7, and Kp taken from OMNIWEB servers which were used to predict TEC. The reliability of the forecasted results is evaluated using numerical factors like Normalized Root Mean Square Error (NRMSE), Correlation Coefficient (CC), Root Mean Square Error (RMSE), Mean Absolute Error (MAE), and R-squared. The study demonstrates the usefulness of combining multiple methods for analyzing TEC variations during SEs and highlights the potential of OKSM, FFNN, and DeepAR models for studying TEC variation in the same context. The findings may be useful for satellite broadcasting and navigational services and for further research into the influence of solar eclipses on the TEC changes.