https://agupubs.onlinelibrary.wiley.com/doi/pdf/10.1029/2020SW002600

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

EXtreme Gradient Boosting over Decision Trees (XGBoost or XGBDT) is a powerful tool to model a wide range of processes. We propose a new approach to create a global total electron content model, using machine-learning-based techniques, in particular, gradient boosting. The model is based on the Global Ionospheric Maps computed by Universitat Politecnica de Catalunya with a tomographic-kriging combined technique (UQRG). To reduce the problem complexity, we used empirical orthogonal functions (EOFs). The created model involves the first 16 spatial EOFs. For training and validation we used the 1998–2016 data sets, and the 2017 data as a test data set. To drive the model, we used the following features: (1) geomagnetic activity indexes (Kp, Ap, AE, AU, AL) and solar activity indexes (R, F10.7); (2) derivative values from these indexes such as the mean value and standard deviations within the last 12 h, last 11 days, and last 40 days; (3) day of the year (DOY); (4) averaged EOFs for given Kp and UT, and those for a given DOY and UT. The validation data set revealed the following hyperparameters for XGBoost learning: number of trees is 100, tree depth is 6, and learning rate is 0.1. Comparisons with the NeQuick2, Klobuchar, and GEMTEC models show that machine learning achieves higher accuracy for the 2017 test data set. The global averaged root-mean-square errors and mean absolute percentage errors were about 2.5 TECU and 19% for the nonlinear GIMLi-XGBDT model, about 4 TECU and 30–40% for NeQuick2, GEMTEC, and the linear model GIMLi-LM, and about 5.2 TECU and 73% for the Klobuchar model. A 4-fully-connected-layer artificial neural network provided a higher error (3.28 TECU and 27.7%) as compared to GIMLi-XGBDT. For all models mentioned, the error peaked in the equatorial anomaly region. The solar activity increase does not affect the error of the nonlinear GIMLi-XGBDT model. However, an increase in geomagnetic activity strongly affects that model.

GIMLi: Global Ionospheric total electron content model based on machine learning

Space weather describes varying conditions between the Sun and Earth that can degrade Global Navigation Satellite Systems (GNSS) operations. Thus, these effects should be precisely and timely corrected for accurate and reliable GNSS applications. That can be modeled with the Vertical Total Electron Content (VTEC) in the Earth’s ionosphere. This study investigates different learning algorithms to approximate nonlinear space weather processes and forecast VTEC for 1 h and 24 h in the future for low-, mid- and high-latitude ionospheric grid points along the same longitude. VTEC models are developed using learning algorithms of Decision Tree and ensemble learning of Random Forest, Adaptive Boosting (AdaBoost), and eXtreme Gradient Boosting (XGBoost). Furthermore, ensemble models are combined into a single meta-model Voting Regressor. Models were trained, optimized, and validated with the time series cross-validation technique. Moreover, the relative importance of input variables to the VTEC forecast is estimated. The results show that the developed models perform well in both quiet and storm conditions, where multi-tree ensemble learning outperforms the single Decision Tree. In particular, the meta-estimator Voting Regressor provides mostly the lowest RMSE and the highest correlation coefficients as it averages predictions from different well-performing models. Furthermore, expanding the input dataset with time derivatives, moving averages, and daily differences, as well as modifying data, such as differencing, enhances the learning of space weather features, especially over a longer forecast horizon.

/pdf/ensemble-machine-learning-of-random-forest-adaboost-and-1n7k76i5.pdf

Ensemble Machine Learning of Random Forest, AdaBoost and XGBoost for Vertical Total Electron Content Forecasting

This paper presents a review on the PECASUS service, which provides advisories on enhanced space weather activity for civil aviation. The advisories are tailored according to the Standards and Recommended Practices of the International Civil Aviation Organization (ICAO). Advisories are disseminated in three impact areas: radiation levels at flight altitudes, GNSS-based navigation and positioning, and HF communication. The review, which is based on the experiences of the authors from two years of running pilot ICAO services, describes empirical models behind PECASUS products and lists ground- and space-based sensors, providing inputs for the models and 24/7 manual monitoring activities. As a concrete example of PECASUS performance, its products for a post-storm ionospheric F2-layer depression event are analyzed in more detail. As PECASUS models are particularly tailored to describe F2-layer thinning, they reproduce observations more accurately than the International Reference Ionosphere model (IRI(STORM)), but, on the other hand, it is recognized that the service performance is much affected by the coverage of its input data. Therefore, more efforts will be directed toward systematic measuring of the availability, timeliness and quality of the data provision in the next steps of the service development.

Space Weather Services for Civil Aviation—Challenges and Solutions

The ionospheric total electron content (TEC) is an important ionospheric parameter, and it is widely utilized in research such as space weather prediction and precise positioning. However, it is still challenging to develop an ionospheric TEC prediction model with high accuracy. In this study, a new ionospheric TEC model over China was developed using the bidirectional long short-term memory (bi-LSTM) method and observations from 257 ground-based global navigation satellite system (GNSS) stations in the Crustal Movement Observation Network of China from January 2018 to December 2021. The root mean square errors of the bi-LSTM-based model’s 1 and 2 hr ahead predictions on the test data set (from June 2021 to December 2021) are 1.12 and 1.68 TECU, respectively, which are 75/50/32% and 72/48/22% smaller than those of the IRI-2016, artificial neural network and LSTM-based models, correspondingly. The bi-LSTM-based model shows the best performance, which is most likely due to the fact that the sequence information in both forward and backward directions is taken into consideration in the new model. In addition, the diurnal variation, seasonal variation of the ionospheric TEC, and variations under geomagnetic storm conditions are successfully captured by the bi-LSTM-based model. Moreover, the TEC maps resulting from the bi-LSTM model agree well with those obtained from the final ionospheric product from the Chinese Academy of Sciences. Hence, the new model can be a good choice for the investigation of the spatiotemporal variation trend in the ionosphere and GNSS navigation. 

https://onlinelibrary.wiley.com/doi/pdfdirect/10.1029/2022SW003103

An Investigation of Ionospheric TEC Prediction Maps Over China Using Bidirectional Long Short‐Term Memory Method

We introduce a novel empirical model to forecast, 24 h in advance, the Total Electron Content (TEC) at global scale. The technique leverages on the Global Ionospheric Map (GIM), provided by the International GNSS Service (IGS), and applies a nonlinear autoregressive neural network with external input (NARX) to selected GIM grid points for the 24 h single-point TEC forecasting, taking into account the actual and forecasted geomagnetic conditions. To extend the forecasting at a global scale, the technique makes use of the NeQuick2 Model fed by an effective sunspot number R12 (R12eff), estimated by minimizing the root mean square error (RMSE) between NARX output and NeQuick2 applied at the same GIM grid points. The novel approach is able to reproduce the features of the ionosphere especially during disturbed periods. The performance of the forecasting approach is extensively tested under different geospatial conditions, against both TEC maps products by UPC (Universitat Politecnica de Catalunya ) and independent TEC data from Jason-3 spacecraft. The testing results are very satisfactory in terms of RMSE, as it has been found to range between 3 and 5 TECu. RMSE depend on the latitude sectors, time of the day, geomagnetic conditions, and provide a statistical estimation of the accuracy of the 24-h forecasting technique even over the oceans. The validation of the forecasting during five geomagnetic storms reveals that the model performance is not deteriorated during disturbed periods. This 24-h empirical approach is currently implemented on the Ionosphere Prediction Service (IPS), a prototype platform to support different classes of GNSS users.

/pdf/neural-network-based-model-for-global-total-electron-content-4arno81p76.pdf

Michele Fiocca

Papers

Neural network based model for global Total Electron Content forecasting