An increasing number of terrestrial‐ and space‐based radio‐communication systems are influenced by the ionospheric space weather, making the ionospheric state increasingly important to forecast. In this study, a novel extended encoder‐decoder long short‐term memory extended (ED‐LSTME) neural network, which can predict ionospheric total electron content (TEC) is proposed. Useful inherent features were automatically extracted from the historical TEC by LSTM layers, and the performance of the proposed model was enhanced by considering solar flux and geomagnetic activity data. The proposed ED‐LSTME model was validated using 15‐min TEC values from GPS measurements over one solar cycle (from January 2006 to July 2018) collected at 15 GPS stations in China. Different assessment experiments were conducted in different geographical locations and seasons as well as under varying geomagnetic activities, to comprehensively evaluate the model's performance. These comparative experiments were conducted using an ED‐LSTM, a traditional LSTM, a deep neural network, autoregressive integrated moving average, and the 2016 International Reference Ionosphere models. The results indicated that the ED‐LSTME model is superior to the other statistical models, with R² and root mean square error values of 0.89 and 12.09 TECU, respectively. In addition, TEC was adequately predicted under different ionospheric conditions, and satisfactory results were obtained even under geomagnetically disturbed conditions. These results suggest that the prediction performance could be significantly improved by utilizing auxiliary data. These observations confirm that the proposed model outperforms several state‐of‐the‐art models in making predictions at different times and under diverse conditions.

The paper presents an analysis by using the methods of Eddy field calculation mean and wavelet maxima to detect seismic anomalies within the outgoing longwave radiation (OLR) data based on time and space. The distinguishing feature of the method of Eddy field calculation mean is that we can calculate "the total sum of the difference value" of "the measured value" between adjacent points, which could highlight the singularity within data. The identified singularities are further validated by wavelet maxima, which using wavelet transformations as data mining tools by computing the maxima that can be used to identify obvious anomalies within OLR data. The two methods has been applied to carry out a comparative analysis of OLR data associated with the earthquake recently occurred in Haiti on 12 January 2010. Combining with the tectonic explanation of spatial and temporal continuity of the abnormal phenomena, the analyzed results have indicated a number of singularities associated with the possible seismic anomalies of the earthquake and from the comparative experiments and analyses by using the two methods, which follow the same time and space, we conclude that the singularities observed from 19 to 24 December 2009 could be the earthquake precursor of Haiti earthquake.

Anomalies of multi-parameters (outgoing longwave radiation, surface latent heat flux, air temperature, relative humidity, and air pressure) before the 12 May 2008 Wenchuan M_s =8.0 earthquake were discussed in order to obtain the seismic precursors. Multi-parameter data were computed based on multi-year background data. The results indicated that these parameters had significant variations prior to this event. The anomaly of outgoing longwave radiation was observed firstly, which gives an early warning. Next were air temperature, relative humidity, and air pressure, which had quasi-simultaneous variations in the basin and the mountain region close to the epicenter. The last was surface latent heat flux, which happened the day before this event. The characteristics of the parameter variations for this event are similar to other earthquakes, which is demonstrated in some other publications. The variations can be attributed to solid earth degassing and chemical reactions in the atmosphere. Additionally, the emission of gases from solid earth into the atmosphere could be ascribed to the tectonic stress of the Wenchuan earthquake.

In this article, a new method for seismic ionospheric Global Navigation Satellite System (GNSS) total electron content (TEC) based anomaly detection using a deep learning framework is proposed. The performance of the proposed encoder–decoder long short-term memory extended model is compared with those of five other benchmarking predictors. The proposed model achieves the best performance ( R ² = 0.9105 and root-mean-square error (RMSE) = 2.6759) in predicting TEC time series data, demonstrating a 20% improvement in R ² and 39.1% improvement in the RMSE over the autoregressive integrated moving average model. To detect the pre-earthquake ionospheric disturbances more accurately, a reasonable strategy for determining anomaly limits is also proposed. Finally, the method is applied to analyze the pre-earthquake ionospheric TEC disturbance of the 2016 Xinjiang M _s 6.2 earthquake. By excluding the effects of solar activity and geomagnetic activity, obvious ionospheric anomalies could be observed, occurring during 4–8 days prior to, and on 1 day before, the earthquake. Negative anomalies tended to occur in the earlier period, whereas positive anomalies occurred closer to the earthquake time, with increasing anomaly intensity with temporal proximity. Furthermore, confusion analysis is used in this article to verify the reliability of the proposed model. The proposed model achieves significant improvements in predicting GNSS TEC time series and is shown to advance the performance of earthquake anomaly detection technology.

The low-altitude satellite DEMETER recorded many cases of ionospheric perturbations observed on occasion of large seismic events. In this paper, we explore 16 spot-checking classification algorithms, among which, the top classifier with low-frequency power spectra of electric and magnetic fields was used for ionospheric perturbation analysis. This study included the analysis of satellite data spanning over six years, during which about 8760 earthquakes with magnitude greater than or equal to 5.0 occurred in the world. We discover that among these methods, a gradient boosting-based method called LightGBM outperforms others and achieves superior performance in a five-fold cross-validation test on the benchmarking datasets, which shows a strong capability in discriminating electromagnetic pre-earthquake perturbations. The results show that the electromagnetic pre-earthquake data within a circular region with its center at the epicenter and its radius given by the Dobrovolsky’s formula and the time window of about a few hours before shocks are much better at discriminating electromagnetic pre-earthquake perturbations. Moreover, by investigating different earthquake databases, we confirm that some low-frequency electric and magnetic fields’ frequency bands are the dominant features for electromagnetic pre-earthquake perturbations identification. We have also found that the choice of the geographical region used to simulate the training set of non-seismic data influences, to a certain extent, the performance of the LightGBM model, by reducing its capability in discriminating electromagnetic pre-earthquake perturbations.