A Method for Detecting Ionospheric TEC Anomalies before Earthquake: The Case Study of Ms 7.8 Earthquake, February 06, 2023, Türkiye

Abstract

The ionospheric anomalies before an earthquake may be related to earthquake preparation. The study of the ionospheric anomalies before an earthquake provides potential value for earthquake prediction. This paper proposes a method for detecting ionospheric total electron content (TEC) anomalies before an earthquake, taking the M_S 7.8 earthquake in Türkiye on 6 February 2023 as an example. First, the data of four ground-based GNSS stations close to the epicenter were processed by using the sliding interquartile range method and the long short-term memory (LSTM) network. The anomaly dates detected by the two methods were identified as potential pre-earthquake TEC anomaly dates after eliminating solar and geomagnetic interference. Then, by using the sliding interquartile range method to process and analyze the CODE GIM (Center for Orbit Determination in Europe, Global Ionospheric Map) data from a global perspective, we further verified the existence of TEC anomalies before the earthquake on the above TEC anomaly days. Finally, the influence of the equatorial ionospheric anomaly (EIA) on the TEC anomaly disturbance was excluded. The results show that the ionospheric TEC anomalies on January 20, January 27, February 4, and February 5 before the Türkiye earthquake may be correlated with the earthquake.

1. Introduction

Earthquakes are usually huge natural disasters, and their sudden, strong, wide impact seriously endangers the safety of human property. Due to the heterogeneity of the Earth’s internal structure and the complex physical mechanisms, earthquake prediction is still a difficult problem in contemporary earth science [1]. There is a lithosphere–ionosphere coupling, and studies on its mechanism can be found in reference [2]. As a branch of lithosphere–ionosphere coupling, the study of ionospheric disturbance and seismic ionospheric-coupling mechanisms during earthquake gestation has become a hot spot in the study of earthquake precursors [3]. Common seismic monitoring methods depend on ground-based observation data, and the seismogenic process also influences the Earth’s atmosphere; especially, some ionospheric disturbances are associated with earthquakes. Leonard and Barnes [4] found anomalies in the ionosphere before the 1964 Alaska earthquake and believed that ionospheric anomalies may be correlated with seismogenic processes. Large- and medium-sized earthquakes, especially strong earthquakes, are often preceded by abnormal phenomena of ionospheric disturbances. Such anomalies are usually local, and their magnitude increases with the increase in the earthquake magnitude [5]. Although changes in the ionosphere are not observed before all earthquakes, ionospheric disturbances can be one of the potential precursors of possible earthquakes [6].

At present, much research on ionospheric anomalies before earthquakes has been conducted by local and international scholars. According to a study on the French DEMETER satellite, many large earthquakes are accompanied by ionospheric disturbance anomalies of different degrees before they occur. Antsilevich [7] found an increase in the ionospheric TEC over the seismic area of the 1966 Tashkent Ms 7.5 earthquake. Calais and Minster [8] used GPS data to study and find that the ionospheric TEC abnormal disturbance occurred within 30 min after the 17 January 1994 Northridge earthquake. Pulinets [9] found that the projection position of the area with the most evident ionospheric anomalies on the ground usually does not coincide with the epicenter. Before the 2008 Yutian earthquake, several parameter anomalies in the ionospheric ion temperature, ion density, and magnetic field near the epicenter were also observed [10]. Prior to the 2005 Sumatra earthquake, ionospheric disturbances occurred in several parameters in the equatorial region [11]. Before the 2010 Chile earthquake, the electric field, magnetic field, and some plasma parameters over the epicenter showed anomalous characteristics of synchronous disturbance [12]. Given multiple parameters with abnormal characteristics at the same time, such as before the 2008 Wenchuan earthquake, the electron density, electron temperature, and oxygen ion density near the epicenter all change dramatically, and the electromagnetic radiation was enhanced [13]. Li et al. calculated the ionospheric disturbance anomalies before moderate and strong earthquakes worldwide, and the results showed that the seismic-related ionospheric disturbance mainly appeared one week before the earthquake and had an evident trend of migration to the epicenter [14]. Tong et al. used the sliding interquartile range method to find positive anomalies in the ionospheric TEC before the M7.2 earthquake in Mexico [15]. Before the 2008 Gaize earthquake, the ion temperature rise and electric field synchronous disturbance were abnormal [16]. Apart from such coseismic and postseismic ionospheric anomalies, Heki [17] found ionospheric TEC enhancement starting ~40 min before the 2011 Mw9.0 Tohoku-oki earthquake by using the Japanese dense GNSS network GEONET (GNSS Earth-Observation Network). He also confirmed similar TEC enhancements before the 2004 Sumatra–Andaman (Mw9.2), the 2010 Maule (Mw8.8), and the 1994 Hokkaido Toho-oki (Mw8.3) earthquakes, and later the 2007 Bengkulu earthquake (Mw8.5), Southern Sumatra [18]. Heki and Enomoto [19] further added the main shock (Mw8.6) and the largest aftershock (Mw8.2) of the 2012 North Sumatra (Indian Ocean) earthquake and the 2014 Iquique earthquake (Mw8.2). At this time, the number of earthquakes showing similar precursory ionospheric anomalies became eight, and their Mw ranged from 8.2 to 9.2. They include all the earthquakes with Mw8.5 or more in this century, with just one exception: the 2005 Nias earthquake (Mw8.6), where plasma bubble signatures hampered detections of near-field ionospheric disturbances.

In China, strong ionospheric disturbances can be observed before many strong earthquakes of magnitude 7 or above. Examples include the Mani earthquake in 1997, the Kunlun Mountain Pass earthquake in 2001 [20], the Wenchuan earthquake in 2008 [21], and the Yushu earthquake in Qinghai Province in 2010 [22]. Many scholars have also analyzed the statistical characteristics of ionospheric anomalies in China. Xu et al. [23] examined the data of 14 strong earthquakes of magnitude 7.0 or above in China and found that among them, 85.7% of the earthquakes had preceding ionospheric anomalies, most of which occurred within 7 days before the earthquake, and only the M8.1 earthquake in the west Kunlun Pass in 2001 had anomalies 14 days before the earthquake. All the anomalies were concentrated between 11:00 and 17:00 local time, and the amplitude of the ionospheric disturbance exhibits varying degrees of nonlinear correlation with the earthquake magnitude, focal depth, and epicenter distance. Liu et al. [24] studied 56 M ≥ 6.0 earthquakes in China from 1998 to 2012 by using the total electron density content released by the global ionospheric map (GIM), and the results showed that the TEC decreased substantially in the afternoon 2 to 9 days before the earthquake. Liu et al. [25] analyzed the GPS TEC data of 39 M ≥ 6.0 earthquakes in China from 1998 to 2010 and found that the change trends in the four directions above and around the epicenter were similar from 15 days before the earthquake to the day of the earthquake. The negative anomaly was slightly higher than the positive anomaly 3 to 5 days before the earthquake, and the most noticeable anomaly was not located directly above the epicenter but drifted toward the magnetic equator, with a spatial influence scale of about 15°. The positive anomaly appeared on the 14th and 10th day before the earthquake, located to the southwest of the epicenter. The negative anomaly was more apparent in the southeast direction on the 5th day before the earthquake. Zou et al. used principal component analysis and the sliding interquartile range method to analyze the anomaly characteristics of the ionospheric TEC before earthquakes with magnitudes greater than Ms 7.0 (the Yutian earthquake, Japan earthquake, Wenchuan earthquake, and Yushu earthquake) [26]. Ke et al. [27] conducted a statistical analysis on the GIM TEC data of 24 M5–8 earthquakes in China from 2003 to 2013 and found evident ascending and descending anomalies, and the magnitude of the anomalies before the earthquake was higher than that after the earthquake. Not all the anomalies were ascending or descending within 5 days before the earthquake, and the occurrence of anomalies was not continuous. The magnitude of the TEC anomalies was not linearly correlated with the magnitude of the earthquake, and the magnitude of the TEC anomalies on the day of the earthquake was not considerably higher than that on other days before and after the earthquake. Liu et al. [28] used GIM TEC data to further analyze the temporal and spatial distribution characteristics of the electrical high-rise anomalies of 62 M ≥ 6.0 earthquakes in China from 1998 to 2015. The TEC decline occurred mainly during UT18–22 (midnight to early morning local time) 4 to 5 days before 37 earthquakes of magnitude 6.0 to 6.5, UT01–04 (morning) 3 to 6 days before 18 earthquakes of magnitude 6.5 to 7.0, and UT04–10 (noon to afternoon) 3 to 5 days before the earthquake. However, an uptick was noted in UT08–12, 18 to 20 days before the seven earthquakes of magnitude 7.0 or greater. The results of the statistical analysis show that the preseismic anomalies of China’s strong earthquakes are complex, with prevalent positive and negative ionospheric anomalies, and the anomaly distribution is more concentrated in the week before the earthquake, but some earthquake anomalies appear in earlier periods such as 10 to 14 days or even in longer periods such as 18 to 20 days. Liming He conducted an analysis by comparing the observed vertical total electron contents (VTEC) with reference curves, using data from nearby Global Navigation Satellite Systems stations [29]. The study revealed that eight earthquakes within the magnitude range of Mw7.0–8.0 exhibited possible preseismic ionospheric anomalies. This research provides new insights into the relationship between ionospheric disturbances and seismic activity, offering valuable insights into the potential of pre-earthquake ionospheric monitoring as a predictive tool. A study by Feng et al. on ionospheric electron content found that ionospheric anomalies also existed before volcanic eruptions, and this anomaly was not related to equatorial ionospheric anomalies [30].

In this paper, a method for detecting ionospheric TEC anomalies before earthquakes is proposed, taking the M_S 7.8 earthquake in Türkiye on 6 February 2023 as an example. On 6 February 2023, two consecutive 7.8 magnitude earthquakes struck southern Türkiye at 4:17 a.m. and 1:24 p.m. local time, with epicenters located in the Gaziantep and Kahramanmaras provinces, approximately 100 km apart. The earthquakes resulted in significant damage and casualties, affecting 10 provinces in Türkiye and some regions in neighboring Syria. As of April 6, it was reported that the earthquakes had caused 50,399 fatalities and 107,207 injuries in Türkiye, with 16 million people affected, including 1.8 million Syrian refugees.

2. Data Sources

This paper adopts ground-based GNSS station data from IGS, and the access address is http://cddis.nasa.gov/archive/gnss/data/daily/ (accessed on 19 August 2023). Figure 1 shows that the locations of the two earthquakes are (37.15°N, 36.95°E) and (38.00°N, 37.15°E). TEC data from four GNSS stations (bshm, zeck, tubi, and mers) close to the epicenter from 1 December 2022 to 11 February 2023 are selected for an anomaly analysis. The locations of the stations are shown in

Table 1. GNSS station location information.

Station	Latitude	Longitude
bshm	32.779°N	35.020°E
zeck	43.788°N	41.565°E
tubi	40.787°N	29.451°E
mers	36.566°N	34.256°E
monp	32.890°N	116.420°W

. Note that only GPS satellites that appear above these stations are used to calculate the vertical TEC, abbreviated as GPS-TEC. From a global perspective, the global ionosphere maps (GIM) released by IGS are adopted, and the access address is http://cddis.nasa.gov/archive/gnss/product/ (accessed on 19 August 2023). Additionally, the ring current index Dst of the magnetic storm, the three-hour magnetic situation index Kp, the solar 10.7 cm wavelength radio flux F10.7, and the solar wind speed SW (flow speed) from January 18 to February 11 are selected to reflect the solar activity level and geomagnetic field conditions, as shown in Figure 2. F10.7 is an index for solar radiation flux used to measure the intensity of solar radiation. KP is a geomagnetic index used to indicate the strength of geomagnetic activity. SW and AP are two indices used to measure geomagnetic activity. SW is the storm index, derived from observations of magnetic field variations at magnetic observatories, while AP is the auroral electrojet index, estimated based on measurements of the geomagnetic-induced currents in polar regions. DST is the Disturbance Storm Time index, used to measure the level of disturbance in the Earth’s magnetic field. BZ represents the north–south component of the solar magnetic field, indicating the interaction between the solar wind and the Earth’s magnetic field.

3. Method

The method of this paper is shown in Figure 3. First, the data of four ground-based GNSS stations close to the epicenter were processed by using the sliding interquartile range method and the long short-term memory network. The anomaly dates detected by the two methods were identified as potential pre-earthquake TEC anomaly dates after eliminating solar and geomagnetic interference. Then, by using the sliding interquartile range method, we further verified, from the global point of view, the existence of an abnormal TEC disturbance over the earthquake area during the detected date. Finally, the influence of the EIA on the TEC anomaly disturbance was excluded.

4. Result Analysis and Discussion

4.1. Sliding Interquartile Range Method Was Used to Analyze TEC Anomalies in a Single Station

Data from four ground-based GNSS stations near the epicenter provided by IGS were selected to investigate the ionospheric TEC anomalies over GNSS stations before the earthquake by using the sliding interquartile range method, as shown in Figure 4, Figure 5, Figure 6 and Figure 7. In this paper, a time window of 27 days is chosen to process the data, excluding the influence of the ionospheric period. TEC data from 22 December 2022 to 17 January 2023 were obtained, and their values were divided into four equal parts after ranking from smallest to largest. Put them in order as the time series $x_1, x_2, x_3\dots \dots x_{27}$ (x₁ represents the first day, x₂ represents the second day, and so on) and then calculate the upper quartile $Q_1$, middle quartile $Q_2$, lower quartile $Q_3$, and interquartile:

$$IQR:\{\begin{array}{c}{Q}_1=x_7\\ Q_2=x_{14}\\ Q_3=x_{21}\end{array}$$

(1)

$$IQR=Q_3-Q_1$$

(2)

The upper and lower boundary is determined to be Equation (3):

$$\{\begin{array}{c}TE{C}_{up}=Q_2+1.5IQR\\ TE{C}_{low}=Q_2-1.5IQR\end{array}$$

(3)

$TE{C}_{up}$ and $TE{C}_{low}$ are used to determine whether the ionospheric TEC is abnormal. When the value of the ionospheric TEC exceeds the upper or lower boundaries, it is considered that the ionospheric TEC is anomalous.

Figure 4 shows the TEC time series data and TEC anomalies over the bshm station. A positive anomaly peak of more than 15 TECU appeared 18 days (January 18) before the earthquake. The earthquake on February 6 was preceded by a positive anomaly approaching 15 TECU; positive anomalies of less than 2.5 TECU appeared on January 20, January 22 to January 24, January 26, January 27, January 31, and 4 days before the earthquake, and the duration was short. On January 19 and January 21, a positive anomaly of more than 5 TECU appeared, with a large variation. Negative anomalies occurred on January 23, January 25–January 28, and February 1 for short periods of time, with a peak negative anomaly approaching 1 TECU.

Figure 5 shows that the TEC anomaly changes at the tubi station and bshm station are consistent, and the outliers and peaks are small on the whole. Eighteen days before the earthquake (January 18), a positive anomaly of more than 14 TECU appeared, reaching the peak value of the positive anomaly and lasting about 12 h. A positive anomaly of less than 2.5 TECU was observed 3 days before the earthquake (February 3). A positive anomaly approaching 1 TECU occurred on January 28, and a negative anomaly occurred on January 19–21, January 23, and February 1–3. A negative anomaly peak near 1 TECU occurred on January 26.

Figure 6 shows a negative anomaly peak of more than 1 TECU at the zeck station 2 days before the earthquake (February 4). Eighteen days before the earthquake, the peak value of the negative anomaly approaching 0.5 TECU and positive anomaly approaching 12 TECU appeared successively, and the duration of the positive anomaly exceeded 12 h. The ionospheric TEC increased abnormally when the second earthquake occurred at 10:24 (UT) on February 6. Positive anomalies occurred from January 18 to January 22, January 26 to January 31, February 2, and February 4 to February 6. Negative anomalies occurred on January 18, January 22, January 23, January 25, January 27, January 30, and February 1 to February 4.

Figure 7 shows that a positive anomaly approaching 16 TECU appeared at the mers station 18 days before the earthquake. Before the earthquake on February 6, the ionospheric TEC increased abnormally, with a positive anomaly approaching 13 TECU. Positive anomalies occurred from January 18 to January 22, January 24, January 26, January 31, and February 2 to February 6. Negative anomalies occurred on January 23, January 25 to January 31, February 3, and February 4.

4.2. LSTM Was Used to Analyze Single-Station TEC Anomalies

Ionospheric TEC data over the same four stations (bshm, zeck, tubi, and mers) were also analyzed by using LSTM. LSTM is a method based on a neural network, which has a better ability to analyze outliers and can process multivariable time series data to predict the value of future time points. Therefore, it is often used to analyze ionospheric anomalies before earthquakes. LSTM cited three gate structures to filter information from input neurons, namely the forgetting gate, input gate, and output gate. These three gate structures mainly control and protect information through sigmoid (σ) and tanh functions. More details of this method for time series forecasting are described in [31].

The TEC observation data $x_1, x_2, x_3, \cdots , x_{27}$ of four GNSS stations from 22 December 2022 to 17 January 2023 were obtained and utilized in this paper, and the ionospheric TEC anomaly values were calculated. We set $m_i$ as the mean value of the outliers at each station, and the standard deviation is ${\sigma }_i$($i=1, 2, 3$). $m_i\pm 1.5{\sigma }_i$ is taken as the upper and lower boundary of the ionospheric TEC outliers, and the ionospheric TEC anomalies from January 18 to 6 February 2023 are analyzed. When the ionospheric TEC anomalies exceed the upper and lower boundary, it is considered that the ionospheric TEC anomalies occur at that time. The upper and lower bounds are determined by the following formulas:

$$\{\begin{array}{c}TE{C}_{up}=m_i+1.5{\sigma }_i\\ TE{C}_{low}=m_i-1.5{\sigma }_i\end{array}$$

Figure 8 compares the actual observed TEC values of the GNSS station and the predicted TEC values by using LSTM from January 18 to 11 February 2023. From January 18 to January 21, January 25, January 27, January 30, January 31, and February 2 to February 5, the actual observed TEC value of the GNSS station had a large deviation from the predicted LSTM value. To analyze the variance further, as shown in Figure 9, the mean value of the outliers at each station was calculated as $m_i$, and the standard deviation was ${\sigma }_i$ ($i=1, 2, 3$). $m_i\pm 1.5{\sigma }_i$ was taken as the upper and lower boundaries of the ionospheric TEC outliers. When the ionospheric TEC outliers exceed the upper and lower boundaries, the ionospheric TEC is abnormal. An analysis of the processed data from January 18 to 6 February 2023 reveals evident anomalies in the ionospheric TEC from January 18 to January 21, January 25, January 27, January 30, January 31, and February 2 to February 5, which are the same as the sliding interquartile range method analysis results.

By using the sliding interquartile range method and LSTM to process the data of the ground-based GNSS station, when the two methods indicate a remarkable TEC anomaly on a certain day, the date is identified as a potential anomaly. This comprehensive dual validation method is designed to enhance the robustness and accuracy of the analytical results. To further ensure the accuracy of the analysis, interference factors such as the solar activity and geomagnetic activity were also considered and eliminated in this study. Figure 2 shows that the solar wind speed changed substantially on January 23, January 26, January 30, and February 1 before the earthquake. According to the Space Environmental Prediction Center, solar X-ray flares occurred on January 18, January 19, January 22, January 25, and January 26, so the above periods of solar activity are considered high. According to the geomagnetic activity index, on January 20, January 24, January 29, February 2 from 03:00 to 00:00 on February 3, and before February 4 from 03:00 to 00:00 on February 6, the Kp index was between 0 and 3, the Ap index was between 0 and 10, and the Dst was stable near 0. The solar wind speed was stable with F10.7, so the geomagnetic activity and solar activity level in the above period are low. In the final determination of anomaly dates, these dates that are considerably affected by solar and geomagnetic activity were excluded individually to ensure that the identified anomaly dates were, with a high probability, primarily associated with seismic activity.

Table 2. Sliding interquartile range method, LSTM analysis of anomalies, and solar and geomagnetic activity levels are summarized.

Date	Sliding Interquartile Range (TECU)	LSTM	BZ (nT)	Kp	DSt (nT)	Ap (nT)	SW (km/s)	F10.7 (sfu)	Solar Activity	Geomagnetic Activity Level
1.18	+15	Anomaly	3.9	40	−10	27	445	213	Solar flare	Active
1.19	+5	Anomaly	3.8	27	−6	12	423	219	Solar flare	Active
1.20	+2.5	Anomaly	6	27	4	12	444	211	None	Stable
1.21	+5	Anomaly	5.5	40	4	27	503	202	None	Active
1.22	+2.5	None	4	33	1	18	445	192	Solar flare	Active
1.23	−1	None	9	30	7	15	516	183	Solar wind anomaly	Active
1.24	+2.5	None	4.1	13	3	5	446	175	None	Stable
1.25	−1	Anomaly	6.7	30	4	15	431	167	Solar flare	Active
1.26	+2.5	None	7.1	27	7	12	551	146	Solar wind anomaly	Active
1.27	+2.5	Anomaly	6	33	4	18	574	141	None	Stable
1.28	−1	None	2.8	30	7	15	552	133	None	Active
1.29	−1	None	2.5	20	8	7	499	133	None	Stable
1.30	−1	Anomaly	5.4	27	25	12	475	132	Solar wind anomaly	Active
1.31	+2.5	Anomaly	3.1	33	7	18	484	133	None	Active
2.1	−1	None	0.7	27	14	12	430	130	Solar wind anomaly	Active
2.2	+2.5	Anomaly	3.7	33	0	18	413	131	None	Active
2.3	+2.5	Anomaly	5.5	30	18	15	357	131	None	Active
2.4	+2.5	Anomaly	2.9	27	2	12	382	135	None	Stable
2.5	+2.5	Anomaly	6.4	20	26	7	358	140	None	Stable

shows that the TEC anomalies derived from the two methods at the same time, with low levels of solar and geomagnetic activity, were identified as potentially earthquake-related anomalies on four dates: January 20, January 27, February 4, and February 5 (black bold text).

In order to verify whether the TEC anomaly only exists near the epicenter, this study selected the station (monp in North America) far from the epicenter for research. Figure 10 shows that there are no TEC anomalies at the monp station on the above anomalous dates (January 27, February 4, and February 5). LSTM is also used to analyze the data of stations far away from the epicenter. Figure 11 shows the comparison between the predicted data and the actual observation data. The analysis method and boundary of Figure 12 are the same as those in the previous section. It can be seen from Figure 12 that the deviation of the monp station on January 27, February 4, and February 5 is within the allowable range, which means that there are also no TEC anomalies at that monp station on those days. In fact, other methods can also be used to analyze single-station TEC anomalies, such as the ACEVS [32,33], etc.

4.3. Global Ionospheric TEC Anomaly Analysis

These four anomalous dates (January 20, January 27, February 4, and 5 February 2023) were further explored and validated from a global perspective. CODE GIM data can provide a global observation perspective, and the sliding interquartile range method is selected for anomaly analysis. As a dynamic, robust statistical method, the sliding interquartile range method is suitable for single-station ionospheric anomaly analysis, as well as for global ionospheric anomaly analysis. Figure 13, Figure 14, Figure 15 and Figure 16 show the distribution of CODE TEC anomalies on January 20, January 27, February 4, and 5 February 2023 after using the sliding interquartile range method, where the green five-pointed star represents the epicenter location and the black curve represents the magnetic equator.

Figure 13 shows the global ionospheric TEC anomaly distribution on 20 January 2023 when a broad range of positive anomalies appeared on the global ionospheric TEC. At 04:00 (UT), a positive anomaly appeared in the east direction of the epicenter and gradually approached the epicenter with the passage of time. At 06:00 (UT), a positive anomaly occurred at the TEC, with an anomaly value approaching 5 TECU, and the anomaly disappeared gradually until 08:00 (UT). A positive anomaly appeared in the southwest direction of the epicenter at 12:00 (UT) and gradually weakened over time. The positive anomaly deepened southwest of the epicenter at 18:00 (UT) and continued to do so until 22:00 (UT). Figure 14 shows the global ionospheric TEC anomaly distribution map on 27 January 2023, and a positive anomaly appeared in the due west direction near the epicenter at 02:00 (UT). At 04:00 (UT), a positive anomaly appeared in the southwest direction of the epicenter, and the anomaly area expanded. By 06:00 (UT), the range and the value were reduced. By 08:00 (UT), the anomaly disappeared gradually. From 10:00 (UT) to 18:00 (UT), no evident TEC anomalies were near the epicenter. A positive anomaly appeared southwest of the epicenter at 20:00 (UT) and dissipated over time at 22:00 (UT). Figure 15 shows the 2D global TEC anomaly distribution map before the earthquake occurred on 4 February 2023, with no apparent TEC anomaly near the epicenter from 00:00 (UT) to 14:00 (UT). Since 14:00 (UT), a large area of positive anomalies was in the northeast direction of the epicenter. With the passage of time, the positive anomaly area at 16:00 (UT) approached the epicenter, the anomaly value increased, and the anomaly gradually disappeared until 20:00 (UT). Figure 16 shows that a small range of positive anomalies appeared in the east direction of the epicenter at 00:00 (UT), which gradually deepened and approached the epicenter until 02:00 (UT). A small positive anomaly occurred due west of the epicenter at 06:00 (UT) and dissipated at 08:00 (UT). A small positive anomaly appeared in the southwest direction of the epicenter at 12:00 (UT). At 14:00 (UT), a broad range of positive anomalies appeared in the north and northeast directions of the epicenter and gradually spread westward over time. At 18:00 (UT), a positive anomaly approached the epicenter. At 20:00 (UT), positive TEC anomalies were on the west and east sides of the epicenter, which lasted until 22:00 (UT) and gradually disappeared.

The above analysis results reveal TEC anomalies over the earthquake area on January 20, January 27, February 4, and 5 February 2023, which validates the conclusions previously obtained on single stations. Because the earthquake occurred in the area near the ionospheric equatorial anomaly (EIA) of the north hump, the EIA may also be one of the causes of ionospheric disturbances. To exclude the influence of the EIA, GIMs containing EIA changes in the corresponding time were plotted in this paper, as shown in Figure 17, Figure 18, Figure 19 and Figure 20. Figure 17 shows that the ionospheric TEC anomaly at the epicenter from 06:00 (UT) to 14:00 (UT) on 20 January 2023 was affected by the shift in the equatorial anomaly peak but had a low correlation with the equatorial ionospheric anomaly peak. Figure 18 shows that the peak value of the equatorial anomaly on January 27 was weak, which had a minute effect on the ionospheric TEC anomaly. Figure 19 shows that the TEC anomaly near the epicenter on February 4 was weakly correlated with the equatorial anomaly. Figure 20 shows that the ionospheric TEC anomaly near the epicenter on February 5 was weakly correlated with the equatorial ionospheric anomaly. In a word, the global ionospheric TEC anomalies on January 20, January 27, February 4, and February 5 were very slightly affected by factors such as the shift in the equatorial anomalous peak.

5. Conclusions

A method for detecting ionospheric TEC anomalies before an earthquake is proposed, which was applied to the M_S 7.8 earthquake in Türkiye on 6 February 2023. The key point of this paper is to use multisource data and a variety of methods to comprehensively discuss the ionospheric anomalies before an earthquake. Firstly, GNSS station data with different distances from the epicenter are used in this paper, and the sliding interquartile method and LSTM data are used for processing. Considering the interference of solar activity and geomagnetic activity, the ionospheric anomaly date before the earthquake is preliminarily analyzed. Considering the influence of the equatorial anomaly, a more accurate ionospheric anomaly before the earthquake can be obtained by a comprehensive analysis. It has a certain degree of significance for seismic detection. In summary, the presence of ionospheric anomalies over the epicenter on January 20, January 27, February 4, and 5 February 2023, which exclude the influence of solar activity, geomagnetic activity, and the EIA, may be correlated with the breeding earthquake.

There is a paper that also studied the ionospheric anomalies before the earthquake in Türkiye [34]. In that paper, five different methods, namely the median, Kalman Filter, artificial neural network (ANN)–multilayer perceptron (MLP), LSTM, and ant colony optimization (ACO), are used to detect the seismoanomalies in the time series of the TEC before the earthquake in Türkiye. All these methods show outstanding anomalies in the period of 10 days (January 27) before the earthquake, which is also found in this paper. Kalman Filter and ANN-MLP methods also resulted in the finding that those ionospheric anomalies appeared 1 day (February 5) and 2 days (February 4) before the earthquake, which is consistent with the conclusion of this paper. However, that paper [34] also concluded that there were ionospheric anomalies 3 (February 3) and 7 days (January 30) before the earthquake, which is different from this paper. In this paper, those days (February 3 and January 30) are also detected by both the sliding interquartile method and LSTM. After considering solar and geomagnetic activity, these two days are ruled out.