One‐Day Forecasting of Global TEC Using a Novel Deep Learning Model

The Swiss Federal Office of Topography is currently building up the Automated GPS Network Switzerland, called AGNES. This GPS network will serve various purposes, like national surveying, engineering surveying, navigation, meteorology, and research.For wide-area differential GPS (WADGPS), in particular precise real-time positioning using the GPS phase data of one or more reference stations of the AGNES network, one may expect that atmospheric refraction plays the crucial role, since the reference stations may be rather widely separated in Switzerland, typically by 50-100 km. This assumes that orbits and “absolute” tropospheric refraction for at least one of the sites are known with sufficient precision.In the time period of the next solar maximum, the ionospheric refraction will probably be the most crucial error source, in particular for resolving the initial phase ambiguities, which is mandatory when aiming at a coordinate accuracy on the centimeter level. Nowadays, global and regional total electron content (TEC) maps, like those produced by the Center for Orbit Determination in Europe (CODE), are available and may be used to model the mean ionospheric refraction. However, on time scales of few minutes, ambiguity resolution usually suffers from short-term ionospheric fluctuations which are not accounted for by such TEC maps.Significant height differences with respect to the nearest reference stations may also occur, especially in mountainous areas like Switzerland. This implies that not perfectly modeled tropospheric refraction may cause biases too.We study the impact of tropospheric as well as ionospheric biases on WADGPS results and investigate whether it is possible to extract atmospheric information from networks of the AGNES-type for correcting data of mobile GPS receivers in (near) real time.KeywordsPermanent GPS networksWAD-GPSatmosphere modeling

In this paper, we propose a global ionospheric total electron content (TEC) maps (GIM) prediction model based on deep learning methods that is both straightforward and practical, meeting the requirements of various applications. The proposed model utilizes an encoder-decoder structure with a Convolution Long Short-Term Memory (ConvLSTM) network and has a spatial resolution of 5° longitude and 2.5° latitude, with a time resolution of 1 h. We utilized the Center for Orbit Determination in Europe (CODE) GIM dataset for 18 years from 2002 to 2019, without requiring any other external input parameters, to train the ConvLSTM models for forecasting GIM 1, 2, and 3 days in advance. Using the CODE GIM data from 1 January 2020 to 31 December 2023 as the test dataset, the performance evaluation results show that the average root mean square errors (RMSE) for 1, 2 and 3 days of forecasts are 2.81 TECU, 3.16 TECU, and 3.41 TECU, respectively. These results show improved performance compared to the IRI-Plas model and CODE’s 1-day forecast product c1pg, and comparable to CODE’s 2-day forecast c2pg. The model’s predictions get worse as the intensity of the storm increases, and the prediction error of the model increases with the lead time.

Timely, reliable and comprehensive global observation information is essential for space weather research. However, limited observation technology hinders the consecutive global coverage of observation data. For the integrity and continuity of the global observation data, deep learning can obtain a global Ionospheric total electron content (TEC) map by fusing multi‐source TEC maps. Different from the previous methods, in the study, a deep learning hybrid model (RFGAN) based on Dual‐Discriminator Conditional Generative Adversarial Network (DDcGAN) and Free‐Form Image Inpainting with Gated Convolution (Deepfill v2) is proposed to fuse the Massachusetts Institute of Technology (MIT)—TEC, International Global Navigation Satellite System TEC (IGS‐TEC) and altimetry satellite TEC. Throughout the RFGAN structure, we use an autoencoder model with gated convolution to inpaint the missing parts of MIT‐TEC and altimetry satellite TEC. Meanwhile, DDcGAN fuses the inpainted MIT‐TEC (MIT'‐TEC) and IGS‐TEC to get a global TEC map with high accuracy. To a certain extent, we inpainted the ocean area of MIT‐TEC through RFGAN. At the same time, RFGAN keeps the consistency of RFGAN‐TEC and MIT‐TEC in the continent area. Our proposed deep learning hybrid model can be easily extended and widely applied to other fields of space science, especially in addressing observational data loss and multi‐source data fusion.

Research on total electron content (TEC) empirical models is one of the important topics in the field of space weather services. Global TEC empirical models based on Global Ionospheric Maps (GIMs) TEC data released by the International GNSS Service (IGS) have developed rapidly in recent years. However, the accuracy of such global empirical models has a crucial restriction arising from the non-uniform accuracy of IGS TEC data in the global scope. Specifically, IGS TEC data accuracy is higher on land and lower over the ocean due to the lack of stations in the latter. Using uneven precision GIMs TEC data as a whole for model fitting is unreasonable. Aiming at the limitation of global ionospheric TEC modelling, this paper proposes a new global ionospheric TEC empirical model named the TECM-GRID model. The model consists of 5183 sections, corresponding to 5183 grid points (longitude 5°, latitude 2.5°) of GIM. Two kinds of single point empirical TEC models, SSM-T1 and SSM-T2, are used for TECM-GRID. According to the locations of grid points, the SSM-T2 model is selected as the sub-model in the Mid-Latitude Summer Night Anomaly (MSNA) region, and SSM-T1 is selected as the sub-model in other regions. The fitting ability of the TECM-GRID model for modelling data was tested in accordance with root mean square (RMS) and relative RMS values. Then, the TECM-GRID model was validated and compared with the NTCM-GL model and Center for Orbit Determination in Europe (CODE) GIMs at time points other than modelling time. Results show that TECM-GRID can effectively describe the Equatorial Ionization Anomaly (EIA) and the MSNA phenomena of the ionosphere, which puts it in good agreement with CODE GIMs and means that it has better prediction ability than the NTCM-GL model.

Mapping high spatial resolution ionospheric total electron content by integrating Time Series InSAR with International Reference Ionosphere model

Validation of CAS’s final global ionospheric maps during different geomagnetic activities from 2015 to 2017

Mapping of the Ionospheric Total Electron Content over the East African Low–Latitude Region

Two major earthquakes separated by ∼5 minutes occurred in the same fault in Peru at depths of 606.2 and 620.6 km on November 24, 2015. By using Global Ionospheric Maps (GIMs) from the Center for Orbit Determination in Europe (CODE) and a broadly used statistical method, differential Vertical Total Electron Content (VTEC) maps were derived. Two positive ionospheric anomalies were clearly identified in the differential VTEC maps 2 and 1 day prior to the day of the earthquakes. These anomalies were located inside the earthquakes’ preparation regions defined by the Dobrovolsky equation. On the other hand, due to the low-latitude nature of the seismic events, the Equatorial Ionization Anomaly (EIA) shape was also analyzed. A third positive disturbance was revealed between November 20 and 21, 2015. For the aforementioned anomaly and the one on November 22 (2 days before the earthquakes), an enhancement of the VTEC was observe through the considerable modification of the EIA shape into a well-defined double-crest with a trough. By looking into the Dst and Kp indices, the geomagnetic conditions starting on November 20 until the 24 were very quiet; thus, it is considered that the three detected anomalies are precursors to the earthquake doublet. Moreover, it is suggested that the mechanism at work that produced the positive disturbances is the air ionization through the release of radon from the Earth’s crust.

Obtaining ionospheric Total Electron Content (TEC) with high spatial and temporal resolution is crucial for satellite navigation and space weather monitoring. Global Ionospheric Maps (GIMs) are key sources of TEC data. However, the temporal resolution of GIMs from ionosphere associate analysis centers does not satisfy the requirements for space weather monitoring. To address this issue, the Massachusetts Institute of Technology (MIT) Haystack Observatory has provided high‐time‐resolution TEC maps since 1998, using data from ∼5,000 Global Navigation Satellite System (GNSS) tracking stations. However, our research reveals a significant systematic bias in MIT's TEC maps. Unlike the biases between GIMs, it exhibits an unexpected pattern of variation and is also present in comparisons with radio occultation TEC. Further analysis indicates that the systematic bias in MIT TEC decreases with increasing geomagnetic latitude and is correlated with solar activity. In low‐latitude regions, the bias can reach up to 30 TECU, while in high‐latitude regions, it can drop to nearly 0 TECU. During diurnal and seasonal variations, the systematic bias in MIT TEC follows TEC patterns but appears unaffected by small‐scale ionospheric structures. Finally, we analyze the factors potentially contributing to the systematic bias and find that Differential Code Biases (DCBs) provided by MIT exhibits errors of a magnitude comparable to the TEC bias, with a strong correlation between the two. Through further investigation, we confirm that the systematic bias in MIT TEC primarily stems from its DCB estimation strategy and choice of mapping function, and this bias persists both before and after MIT updated its DCB estimation approach.

Abstract. Turkey is a country located in the middle latitude zone, where tectonic activity is intensive. Recently, an earthquake of magnitude 6.5 Mw occurred offshore in the Aegean Sea on 24 May 2014 at 09:25 UTC, which lasted about 40 s. The earthquake was also felt in Greece, Romania, and Bulgaria in addition to Turkey. In recent years, ionospheric anomaly detection studies have been carried out because of seismicity with total electron content (TEC) computed from the global navigation satellite system's (GNSS) signal delays and several interesting findings have been published. In this study, both TEC and positional variations have been examined separately following a moderate size earthquake in the Aegean Sea. The correlation of the aforementioned ionospheric variation with the positional variation has also been investigated. For this purpose, a total of 15 stations was used, including four continuously operating reference stations in Turkey (CORS-TR) and stations in the seismic zone (AYVL, CANA, IPSA, and YENC), as well as international GNSS service (IGS) and European reference frame permanent network (EPN) stations. The ionospheric and positional variations of the AYVL, CANA, IPSA, and YENC stations were examined using Bernese v5.0 software. When the precise point positioning TEC (PPP-TEC) values were examined, it was observed that the TEC values were approximately 4 TECU (total electron content unit) above the upper-limit TEC value at four stations located in Turkey, 3 days before the earthquake at 08:00 and 10:00 UTC. At the same stations, on the day before the earthquake at 06:00, 08:00, and 10:00 UTC, the TEC values were approximately 5 TECU below the lower-limit TEC value. The global ionosphere model TEC (GIM-TEC) values published by the Centre for Orbit Determination in Europe (CODE) were also examined. Three days before the earthquake, at all stations, it was observed that the TEC values in the time period between 08:00 and 10:00 UTC were approximately 2 TECU above the upper-limit TEC value; 1 day before the earthquake at 06:00, 08:00, and 10:00 UTC, the TEC values were approximately 4 TECU below the lower-limit TEC value. Again, by using the same 15 stations, positional variation investigation for before and after the earthquake was undertaken for the AYVL, CANA, IPSA, and YENC stations. As a result of the conducted analysis, positional displacements were seen before and after the earthquake at the CANA station, which is the nearest station to the earthquake centre. Before and after the earthquake, positional displacements were observed as 10 and 3 cm respectively.

Total electron content (TEC) serves as a key parameter characterizing ionospheric conditions. Accurate prediction of TEC plays a crucial role in improving the precision of Global Navigation Satellite Systems (GNSS). However, existing research have predominantly emphasized spatial variations in the ionosphere, neglecting the periodic changes of the ionosphere with the diurnal cycle. In this paper, we propose a TEC prediction model, which simultaneously considers both spatial and temporal characteristics to extract spatiotemporal features of ionospheric distribution. Additionally, we integrate several space weather element datasets into the prediction model framework, allowing the generation of multiple space weather feature values that represent the influence of space weather on the ionosphere at different latitudes and longitudes. Moreover, we apply Gaussian process regression (GPR) interpolation to geomagnetic data to characterize impact on the ionosphere, thereby enhancing the prediction accuracy. We compared our model with traditional image-based models such as convolutional neural networks (CNNs), convolutional long short-term memory networks (ConvLSTMs), a self-attention mechanism-integrated ConvLSTM (SAM-ConvLSTM) model, and one-day predicted ionospheric products (C1PG) provided by the Center for Orbit Determination in Europe (CODE). We also examined the effect of using different numbers of space weather feature values in these models. Our model outperforms the comparison models in terms of prediction error metrics, including mean absolute error (MAE), root mean square error (RMSE), correlation coefficient (CC), and the structural similarity index (SSIM). Furthermore, we analyzed the influence of different batch sizes on model training accuracy to find the best performance of each model. In addition, we investigated the model performance during geomagnetic quiet periods, where our model provided the most accurate predictions and demonstrates higher prediction accuracy in the equatorial anomaly region. We also analyzed the prediction performance of all models during space weather events. The results indicate that the proposed model is the least affected during geomagnetic storms and demonstrates superior prediction performance compared to other models. This study presents a more stable and high-performance spatiotemporal prediction model for TEC.

This paper first applies a prediction model based on self‐attention memory ConvLSTM (SAM‐ConvLSTM) to predict the global ionospheric total electron content (TEC) maps with up to 1 day of lead time. We choose the global ionospheric TEC maps released by the Center for Orbit Determination in Europe (CODE) as the training data set covering the period from 1999 to 2022. Besides that, we put several space environment data as additional multivariate‐features into the framework of the prediction model to enhance its forecasting ability. In order to confirm the efficiency of the proposed model, the other two prediction models based on convolutional long short‐term memory (LSTM) are used for comparison. The three models are trained and evaluated on the same data set. Results show that the proposed SAM‐ConvLSTM prediction model performs more accurately than the other two models, and more stably under space weather events. In order to assess the generalization capabilities of the proposed model amidst severe space weather occurrences, we selected the period of 22–25 April 2023, characterized by a potent geomagnetic storm, for experimental validation. Subsequently, we employed the 1‐day predicted global TEC products from the Center for Operational Products and Services (COPG) and the SAM‐ConvLSTM model to evaluate their respective forecasting prowess. The results show that the SAM‐ConvLSTM prediction model achieves lower prediction error. In one word, the ionospheric TEC prediction model proposed in this paper can establish the ionosphere TEC of spatio‐temporal data association for a long time, and realize high precision of prediction performance.

Second-Order Ionospheres’ Effect in Precise Point Positioning During the Solar Minimum and the Solar Maximum Periods in Turkey

Electron density retrieval from occulting GNSS signals using a gradient-aided inversion technique

The ionosphere is a critical factor for the performance of a wide range of communication and navigation systems. Sudden significant changes in its electron density can cause degradations in the performance of these technical systems. The variability of the ionosphere is mainly driven by solar EUV radiation, but solar wind can also modify the ionosphere significantly for periods of geomagnetic storms. The modelling of the solar wind driven variability of the ionospheric electron density is still an open challenge because of the very complex nature of the ionosphere response to the solar wind input. This study presents an attempt to reproduce the variability of the Total Electron Content (TEC) during one of the most recent extreme geomagnetic storms, which occurred on 10 May 2024.18 years of TEC maps (2005-2023) provided by the International GNSS Service (IGS) are analysed for potential correlations with the popular geomagnetic index Kp. The analysis differentiates local and UT dependencies. The correlation results show a clear latitudinal dependence and hemispheric asymmetry. A linear regression model is generated for those conditions, where a significant correlation is detected. This statistical model is used to reproduce the storm in May 2024. The results are compared with the actual IGS TEC maps observed during the storm.