A short‐term (30 days before an earthquake) prediction of an earthquake is a big challenge in seismology. As a first step, we apply deep learning to the ionospheric total electron content (TEC) data between 2003 and 2014 to detect the seismo‐ionospheric precursors of M ≥ 6.0 earthquakes in Taiwan. The bidirectional Long Short‐Term Memory (Bi‐LSTM) network is employed to use observed input data (features) to obtain the sequential TEC variations. The five input features are sequential vectors of TEC, the geomagnetic index Dst, the solar activity index F10.7, sunspot number (SSN), and solar emission index Lyman‐α. The daily values of F10.7, SSN, and Lyman‐α are converted into hourly values, depending on the solar elevation angle. The calculated hourly TEC variations can be more precisely predicted with this data conversion. We calculate the normalized difference of errors between two 15‐day adjacent stages as the “relative error”. Three trained models with the best discrimination between the relative errors of earthquake and no‐earthquake cases are chosen as classifiers. These three classifiers are then used to have a majority vote to declare whether the 30‐day period is related to the preparation of an earthquake or not. The results show that all 22 positive cases (earthquakes) are successfully predicted, giving a true positive rate of 100%. Among the 19 negative cases (normal cases), 10 of them are true negative.” Overall, a high accuracy of 78.05% is obtained.

Several lithosphere–atmosphere–ionosphere coupling (LAIC) models have been proposed to explain observed precursors before strong earthquakes, such as atmospheric conductivity anomalies and ionospheric total electron content (TEC) variations. These coupling models include (a) radon ionization leading to charged aerosols and change of load resistance in the global atmospheric electric circuit, (b) an electric coupling model of stressed rock–atmosphere–ionosphere based on experimental evidence of stressed rock currents, and (c) ionosphere dynamics with an imposed zonal electric field. Here, we summarize our modeling of stressed-rock–atmosphere–ionosphere coupling for the LAIC process. A physical mechanism of electrical coupling between the atmosphere and surface charge in the fault region is proposed. We also discuss other possible seismic anomalies before and after the earthquake and suggest that in future electric field measurements are used to validate the possible existence of electric activities associated with an earthquake preparation region.

The monthly, seasonal, yearly, and latitudinal variations of ionosphere tidal waves are studied based on the total electron content (TEC) data from global ionosphere maps. We use two‐dimensional (longitudinal, −180°E to 180°E; and time, 10 days) Fourier transform of TEC variations along different latitude to obtain all wave modes for the period from 18 November 2002 to 15 October 2014. The wave modes can be classified as westward propagating (eastward propagating and standing) waves with 85.8–97.3% (2.3–10.0% and 0.4–4.2%) of total power. The wave modes are found to consist of migrating waves with power 75–94% and nonmigrating waves with 5.8–25%. The migrating waves are dominated by diurnal migrating tide (DW1, 69.9–89.4%), semidiurnal migrating tide (SW2, 0.5–4.3%), and terdiurnal migrating tide (TW3, 0.2–1.9%). Also, the total wave power and powers of DW1 and SW2 peak along magnetic latitude 10°N and decrease with increasing magnetic latitude. The power of TW3 peaks along the magnetic equator and decreases with increasing magnetic latitude. We also found that the wave powers of DW1 and SW2 obtained from TEC data are very similar to the wave powers obtained from the ionospheric density with a profile of sech² solitary wave. The observed migrating amplitudes are stronger in equinoxes and weaker in solstices. For the nonmigrating tides, we simultaneously identify the wave modes and wave powers, which contribute to wave‐1, wave‐2, wave‐3, wave‐4, and wave‐5 peak structures in the local time frame. The total wave power associated with TEC variations increases with solar activity (sun spot number).

This study conducts a comprehensive analysis of observations related to the M_L6.2 Hualien earthquake that struck eastern Taiwan on 6 February 2018, focusing particularly on gamma-ray emissions and total electron content (TEC) as earthquake precursors. Prior research has shown that significant gamma-ray enhancements are frequently detected at the YMSG (Yangmingshan gamma-ray) station prior to major earthquakes in eastern and northeastern Taiwan, suggesting that gamma-ray anomalies may serve as reliable indicators for identifying seismic precursors in this area. Our findings reveal a significant rise in gamma-ray emissions at the YMSG station from 19 January to 4 February 2018, which corresponds to a precursor period of approximately 18 days before the Hualien earthquake. Positive and negative TEC anomalies were observed in Taiwan on 20–21 January and 5 February, respectively, and may be considered as ionospheric precursors to the earthquake. Additionally, deep-learning techniques applied to TEC data facilitate the detection of ionospheric precursors associated with the Hualien earthquake, enabling forecasts of an approaching seismic event. Collectively, these observations indicate that all identified anomalies are regarded as short-term precursors, explicable through the theoretical framework of lithosphere–ionosphere coupling (LIC).