Construction of a Fine Extraction Process for Seismic Methane Anomalies Based on Remote Sensing: The Case of the 6 February 2023, Türkiye–Syria Earthquake

Abstract

Identifying seismic CH₄ anomalies via remote sensing has been verified as a legitimate method. However, there are still some problems, such as unknown reliability due to the complex characteristics of seismic anomalies. In this study, a multi-dimensional and multi-scale methane seismic anomaly extraction process for remote sensing was constructed with the Robust Satellite Technique (RST) based on the Atmospheric Infrared Sounder (AIRS) CH₄ data and then applied to the 2023 Türkiye–Syria earthquake. This study obtained the two-dimensional temporal–spatial distribution of methane anomalies and temporal variation in the anomaly index. Based on this, the three-dimensional profile structure of the 8-day methane anomaly was extracted to determine the reliability of the anomaly. Finally, based on the daily methane anomaly, combined with atmospheric circulation and backward trajectory analysis as auxiliary tools, the influence of air mass migration was excluded to enhance the accuracy of CH₄ anomaly determination. The results show that the three-dimensional anomalous structure is consistent with the geological characteristics of tectonic activities, and it appears as a “pyramid” or “inverted pyramid” type in a three-dimensional space. The anomalies caused by air mass migration can be eliminated by combining them with synoptic-scale circulation motion. The time series calculated at the epicenter or a certain point in a region may not accurately reflect the influence of regional or specific tectonic activity in the atmosphere. Thus, the optimal determination of the range and magnitude of atmospheric anomalies caused by tectonic activities is a difficult task for future research.

1. Introduction

According to the United States Geological Survey (USGS), a 7.8 magnitude earthquake occurred on February 6 at 1:28 UTC (4:28 local time), with an epicenter of 37.23°N, 37.01°E, and a focal depth of 10 km, followed by a 7.5 magnitude earthquake occurring 96 km (38.01°N, 37.2°E) far from the epicenter of the previous strong earthquake 9 h later at 10:24 UTC (13:24 local time), forming a typical and rare double earthquake sequence. The Türkiye–Syria earthquake doublet was the world’s largest earthquake in 2023.

In the process of earthquake preparation and occurrence, complex physical and chemical changes may occur in the huge hypocenter and are often accompanied by phenomena such as subsurface fluid change [1], surface deformation [2], atmospheric gasses [3] and thermal emissions [4,5,6], and ionospheric space environment changes [7,8,9,10]. Scientists have proposed a Lithosphere–Atmosphere–Ionosphere Coupling (LAIC) model for earthquake research, which aims to unify the observed precursor signals of various disciplines qualitatively or semi-quantitatively. The hypothesis of LAIC explains the synergy between different ground surfaces, atmosphere and ionosphere processes, and anomalous variations, when different gasses leaking from the crust can serve as carrier gasses for radon. The increase in radon release makes air ionization. Then, condensation of water vapor is accompanied by latent heat exhalation. This coupling (LAIC) mechanism is detected through various channels [11,12,13,14,15]. After the Türkiye–Syria earthquake, geologists and geophysicists carried out a great deal of field investigation and analysis. Akhoondzadeh et al. [16] found that significant pre-earthquake anomalies were observed on the surface, atmosphere, and ionosphere about 10 days before the earthquake. At present, in terms of the lithosphere, An Qi et al. [17] fused optical and SAR offsets and obtained the three-dimensional deformation field of the 2023 Turkey earthquake sequence. The results showed that the Mw 7.8 earthquake ruptured four segments of the EAF, with a cross-fault deformation of over 5 m. And the two ≥ Mw 7.7 earthquakes were both left-lateral strike-slip earthquakes, causing predominant E-W surface deformation and slight vertical deformation. In terms of atmosphere, Tang et al. [18] found that the thermal infrared was significantly enhanced on the east and west sides and the north of the epicenter one month before the earthquake by studying the anomalies of thermal infrared Outgoing Longwave Radiation (OLR). Jing et al. [19] proposed that the atmospheric pressure variation caused by the ground motion of seismic waves is a possible reason for the quasi-coseismic variations in ocean, atmosphere, and snow parameters. In terms of the ionosphere, Artem Vesnin et al. [20] analyzed the ionospheric disturbance effect based on Global Navigation Satellite System (GNSS) data and ionosondes. Moreover, Tang et al. found that the Total Electron Content (TEC) in the region indicated a large area of low-intensity anomalies that appeared in the epicenter area, which lasted from half a year before the earthquake to the time of the earthquake’s occurrence. Akhoondzadeh found obvious TEC anomalies based on the Global Positioning System (GPS) about 10 days before the earthquake [21], and Maletcki also found an obvious TEC response after the earthquake [22]. These scientists have studied the Türkiye–Syria earthquake doublet from different aspects, covering a wide area and achieving certain results. However, existing research on the Türkiye–-Syria earthquake doublet lakes exploration into gas, including CH₄.

The variation in the gas parameters in an earthquake, as an intermediate link in the LAIC model, has crucial significance. Previous studies have shown that CH₄ in hydrocarbon gasses has a close relationship with seismic activities, and the emission concentration of CH₄ within a tectonic area can reflect the intensity of tectonic stress and is an important object for seismic monitoring [23]. The Mw9.1 and Mw8.6 earthquakes in Sumatra on 26 December 2004 [24] and 28 March 2005, the Wenchuan earthquake on 12 May 2008, the Lushan earthquake on 20 April 2013 [25,26], the Alxa Left Banner earthquake in Inner Mongolia province on 15 April 2015 [27], the Jiuzhaigou Earthquake in Sichuan province on 8 August 2017, the Medog Earthquake in Xizang province on 24 April 2019, and the Yutian earthquake in Xinjiang province on 26 June 2020 [28] all exhibited methane anomalies before their occurrences. Fischer’s paper in Nature Geoscience also illustrated that gas hydrates are commonly found in seismically active areas [29]. Wang et al. [30] conducted a time series analysis of the methane anomaly index for 21 earthquakes with magnitudes of 6 or above in the Sichuan and Yunan regions since 2008 and found that CH₄ anomalies appeared in 18 earthquakes. The above studies indicate that methane anomalies can reflect seismic activity to a certain extent.

Currently, commonly used satellite data for methane seismic monitoring mainly include AIRS data from NASA. They are suitable for continuous monitoring in fixed areas because of their long accumulation, high time resolution, good spatial–temporal continuity, and high quality. The most commonly used seismic anomaly extraction algorithms are Robust Satellite Techniques (RSTs) [24,25,31,32,33,34,35]. In addition, column concentration products are often used in the early studies of gas anomalies. However, these products have some limitations in reflecting the geological characteristics of earthquakes and excluding anomalies caused by air mass migration. Considering the above reasons, Cui introduced the gradient anomaly index, which is only suitable for analyzing the adjacent layer of the atmosphere and cannot directly reflect the vertical change characteristics of the methane multilayer atmosphere triggered by the earthquake [26]. The three-dimensional structure of gas is an important interpretation symbol of seismic anomalies [36,37], which is helpful to accurately extract and identify seismic anomalies and deeply understand the geochemical mechanism of seismic gasses. How to fully combine the characteristics of two-dimensional (2D) and three-dimensional (3D) data, establish a complete seismic gas anomaly extraction and identification process, and better serve the seismic monitoring business are difficulties.

For this purpose, in this study, taking the Türkiye–Syria earthquake doublet as an example, we first obtained the two-dimensional spatial distribution and three-dimensional profile distribution of CH₄ anomalies by processing CH₄ data, and then combined them with the atmospheric circulation field and trajectory analysis obtained by meteorological data processing to explored the extraction and analysis methods of atmospheric seismic anomalies, further analyzing the possible mechanism of earthquake-induced gas anomalies. The structure of this paper is as follows. The tectonic background is described in Section 2, and data and methods are discussed in Section 3. Moreover, the results, discussion, and conclusion are explained in Section 4, Section 5, and Section 6, respectively.

2. Tectonic Background

Türkiye is located at the junction of the Eurasian plate, the African plate, and the Arabian plate and is controlled by these three large plates and the smaller Anatolian plate [38], bordered by the Black Sea to the north and the Mediterranean to the south (Figure 1). The Arabian Plate and the African plate are moving northward, pushing against the Anatolian plate. In this region, the combined action of these three plates forms the North Anatolian Fault (NAF) and the East Anatolian Fault (EAF). The NAF and the EAF run through Türkiye and converge in eastern Türkiye. From a seismological perspective, the intricate interaction among these plates increases the complexity of seismic activity in the area. The Türkiye–Syria earthquake doublet occurred near the East Anatolian Fault. The tectonic evolution of this region is dominated by the effects of the northward subduction of the African plate beneath western Türkiye and the continental collision of the Caucasus and eastern Türkiye [38,39,40]. This region is one of the most seismically active and rapidly deformed regions within the continents [41]. Research has indicated high levels of seismic risk as a result of moderate-to-high seismic hazards in the Gaziantep region (the city near the epicenter) coupled with the significant vulnerability of existing structures at some locations in the Gaziantep city center [42,43]. Previous studies have shown that the surface rupture of this earthquake doublet extended more than 300 km along the northeast–southwest direction, and the two epicenters were about 96 km apart, belonging to different seismogenic faults [39,41]. The first earthquake event is located on the NE-SW trending EAF, with left-lateral strike-slip motion, and the second event is located on the EAF and its accessory structure, the Surgu fault. The focal mechanism solution shows that both earthquakes are left-lateral strike-slip, accompanied by small vertical movements [39,40,41,44]. The aftershocks are located on the EAF and the NAF and its branches.

3. Data and Methods

3.1. Data

The CH₄ data used in the study of seismic anomaly information extraction are mainly from the Atmospheric Infrared Sounder (AIRS), a hyperspectral sensor mounted on the American Earth-observation satellite, infrared AQUA. AIRS is an infrared spectrometer/radiometer that covers the 3.7–15.4 μm spectral range with 2378 spectral channels. The AIRS dataset comes with a spatial resolution of roughly 45 × 45 km at nadir [46]. This sensor scans the entire globe, with passages twice a day (night and day). The geophysical parameters have been averaged and binned into 1 × 1 deg grid cells, from −180.0 to +180.0 deg longitude and from −90.0 to +90.0 deg latitude, with high temporal and spatial resolution and high spatial coverage. It is available at the NASA Goddard Earth Sciences Data and Information Services Center (DISC) (https://disc.gsfc.nasa.gov/datasets?keywords=AIRS&sort=source&page=1) (accessed on 24 March 2023). In this study, we select the descending–orbiting CH₄ volume mixing ratio (VMR) data from version 6, level 3 standard gridded product of 8 days, and the version 7 and level 3 standard gridded daily product. Considering that the thermal infrared spectroscopy used in the AIRS is more sensitive to the upper troposphere, a data level of 200–700 hPa was used in this study.

The European Centre for Medium-range Weather Forecasts (ECMWF) reanalysis dataset v5 (ERA5) has been used for examining the global climate and weather for the past 8 decades. The data of ERA5, mean sea level pressure, vertical velocity, and u zonal component and v meridional component of wind, were selected to analyze the atmospheric circulation and then eliminate non-seismic anomalies caused by air mass movement The spatial resolution of the data is 0.25° × 0.25°, and the temporal resolution is 1 h. The average of the hourly data in a day is selected as the daily data in this study. The data can be obtained from CDS (https://cds.climate.copernicus.eu/#!/home) (accessed on 20 November 2023). In this study, sea level pressure data were used to obtain sea level pressure field, u zonal component and v meridional component of wind were used to obtain horizontal wind field, and vertical velocity was used to obtain vertical wind direction. The above data are used for background atmospheric circulation analysis of gas anomalies.

3.2. Anomaly Extraction Algorithm

A robust estimator of Thermal Infrared Remote sensing (TIR) anomalies (RETIRA) index was used for gas anomalous extraction. The RETIRA index is based on the general approach, the Robust Satellite Technique (RST) [34,35]. RETIRA was proposed and first used for thermal anomalous extraction; this index is based on a multi-temporal analysis of several years of a historical dataset of satellite observations acquired in similar observational conditions [47,48,49]. This approach filters out noisy contributions to the satellite-measured signal due to variable, observational, and meteorological conditions. It is thought to isolate (if any) possible pre-seismic anomalous patterns from the most important noisy contributions to the measured signal. The RETIRA index can be computed as follows [47,50]:

$$\Delta G\left(x,y,t\right)=G\left(x,y,t\right)-G\left(t\right)$$

(1)

$${\mu }_{\Delta G}\left(x,y\right)={\displaystyle \sum }_{i=1}^N\Delta G_i\left(x,y,t\right)/N$$

(2)

$${\sigma }_{\Delta G}\left(x,y,t\right)=sqrt\left\{{\displaystyle \sum }_{i=1}^N{\left[\Delta G\left(x,y,t\right)-{\mu }_{\Delta G}\left(x,y\right)\right]}^2/\left(N-1\right)\right\}$$

(3)

$$RETIRA\left(x,y,t\right)={\displaystyle \frac{\Delta G\left(x,y,t\right)-{\mu }_{\Delta G}\left(x,y\right)}{{\sigma }_{\Delta G}\left(x,y\right)}}$$

(4)

where $G\left(x,y,t\right)$ indicates the gas value measured at time t, corresponding to a location at (centered on) the coordinates $\left(x,y\right)$, and $G\left(t\right)$ denotes the spatial average of the investigated area at time $t$. $\mathsf{\Delta }G\left(x,y,t\right)$ is the value of the difference between the punctual value of gas value at the location $\left(x,y\right)$ and the acquisition time $t$ and its spatial average $G\left(t\right)$. ${\mu }_{\Delta G}\left(x,y\right)$ indicates the reference fields for the location $\left(x,y\right)$, defined as a time average gas value; ${\sigma }_{\Delta G}\left(x,y,t\right)$ is the standard deviation of historical records collected under the temporal constraint. For this study, $N$ was defined as 20 years, from 2003 to 2022; $t$ is the time of the measurement acquisition with $t\in \tau$, where $\tau$ defines the homogeneous domain of the satellite imagery collected in the same time slot of the day and period (month) of the year. $RETIRA\left(x,y,t\right)$ is the anomaly value at time $t$, corresponding to a location at (centered on) the coordinates $\left(x,y\right)$. The bigger the absolute value of RETIRA, the more evident the anomaly. The differential variable $\Delta G\left(x,y,t\right)$ instead of $G\left(x,y,t\right)$, is expected to reduce possible contributions due to short-term climate variation and/or season-to-season changes over the years [51].

3.3. Identification of the CH₄ Anomaly

After the calculation of RETIRA, the next step is to identify the CH₄ anomalies and correlate them with earthquake occurrences. In this study, a CH₄ anomaly is defined by the RETIRA value of each pixel, which at time t exceeds the selected threshold; further conditions will be applied to confirm the possible correlation with the earthquake. Only if the following conditions are satisfied can it be concluded that the anomaly is possibly related to the earthquake:

• The $RETIRA\left(x,y,t\right)$ > 2 [26,30,34,35]. When the value exceeds 2 times the standard deviation, it falls within the abnormal category. The threshold is set to 2 in this study.
• Spatial persistence: The CH₄ anomalies cluster together and are not isolated, being part of a group covering 2° × 2°. The location of the anomaly cluster is near the epicenter or the fault zone. The maximum distance of an abnormal cluster cannot exceed $R={10}^{0.43M}$ [52], where M is the earthquake’s magnitude and R is the unit of kilometers.
• Temporal persistence: 3 months before the earthquake, the seismic anomalies occur at least three times in succession or several times in intervals.
• Vertical spatial distribution: Considering that the earthquake is caused by tectonic activity in the Earth’s interior, the earthquake anomaly presents a positive pyramid shape from bottom to top. In a bottom–up inverted triangle, the lower area is small, increasing layer by layer.

Based on the above anomaly identification, in order to rule out that the gas anomaly is caused by atmospheric migration, we consider the atmospheric circulation motion. The horizontal and vertical wind fields can directly reflect the trend in airflow, and the mean sea level pressure can reflect the situation of the high or low pressure in the region and describe whether there is a weather system (cyclone, etc.) and its strength. We can use The National Oceanic and Atmospheric Administration (NOAA) Air Resources Laboratory’s (ARL) Hybrid Single-Particle Lagrangian Integrated Trajectory model (HYSPLIT) [53] to analyze the trajectory of air mass. This model is one of the most widely used atmospheric transport and dispersion models for computing simple air parcel trajectories and complex transport and dispersion. One of its most common model applications is to determine the origin of an air mass through backward trajectory analysis [54,55].

The technical process is shown in Figure 2. We summarize the process into three parts: Data Collection, Anomaly Extraction, and Anomaly Judgement. The specific details are as follows:

• Data Collection: Long-term anomaly monitoring found that most anomalies appeared three months before the earthquake. In daily monitoring, we often select 8-days’ product data about half a year before the earthquake.
• Anomaly Extraction: The 2D CH₄ anomalies extraction mainly include spatial and temporal distribution and anomaly index time series.
• Seismic Anomaly Judgement: According to the CH₄ anomaly, the seismic CH₄ anomaly is determined comprehensively from the three-dimensional point of view, combined with daily product data, wind field, and backward trajectory analysis.

4. Results

This section presents analysis of the results based on the technical process (Figure 2), which mainly includes two-dimensional time–spatial distribution analysis and three-dimensional vertical distribution analysis.

4.1. Two-Dimensional Time–Spatial Distribution Analysis

According to the empirical equation of the earthquake preparation zone put forward by Dobrovolsky et al. [52], the influential zone of an earthquake is roughly computed via equation $R={10}^{0.43M}$, where $R$ is the radius of the influential earthquake zone in the km scale, and $M$ is the magnitude of the EQ. The impact radius of this earthquake is about 20°.

The 8-day average background field at 300 hPa in the study area is shown in Figure 3. The distribution of CH₄ has obvious seasonal characteristics. The CH₄ VMR showed a high value in August and then gradually weakened. Most of the high value areas appear near the fault zone and in the northern part of the study area. The spatial and temporal distribution of RETIRA of the research region five months before and one month after the earthquake are shown in Figure 4. The result indicated that the concentration of CH₄ before the earthquake doublet break the distribution features of the background field (Figure 3). In late October and early November, a large area anomaly occurred in the north of the NAF and the south of the EAF, which lasted for two cycles before dissipating. It was not until early December that a large CH₄ anomaly reappeared in the epicenter and surrounding areas, mainly in the west of the EAF and the northeast of the NAF. The anomalies then converged at the intersection of the NAF and the EAF. Finally, they passed through the NAF, and high-value anomalies appeared in the southern section of the EAF, lasting until the end of December. From 29 December to 5 January, the area of the CH₄ anomaly decreased, and a low-value anomaly appeared near the epicenter. From 6 January onwards, high-value anomalies appeared again in the NAF and EAF; the anomaly area continued to increase, and the high-value anomaly converged toward the epicenter, reaching the peak value in 22–29 January 2023 about 10 days before the earthquake, and was mainly located in the north of the NAF and the EAF and the latter’s south. Abnormal weakening occurred from 30 January to 6 February near the epicenter of the earthquake cycle. After the earthquake, on 15 February, large-area anomalies appeared in the EAF and west of it, followed by large-area anomalies on the EAF and both sides, corresponding to the location of aftershocks, which also indicates the correlation between methane anomalies and seismic activities. The anomalies disappeared again on 3 March. The whole anomaly shows the process of “enhancement—reduction—enhancement—reduction (earthquake time)—enhancement—reduction after the earthquake”. Figure 5 is a magnitude frequency chart of earthquakes during the investigated time from 23 August 2022 to 3 March 2023. Compared with before the earthquake, after the main earthquake, there were large-area low-value anomalies in the north, east, and west sides of the epicenter. Still, the duration was short, which may be related to the release of post-earthquake gas along the fault caused by continuous aftershocks on the two fault zones. Small-scale CH₄ concentration anomalies may also be due to some leaks from oil pipelines [16].

The mean value of RETIRA in the area of possible influence of earthquake preparation was calculated, and the results are shown in Figure 5. The results show two significant peaks of the RETIRA index in the time series, both of which are preceded by a continuous increase in the index for three or more time-cycles. After the two peaks, the index decreases, and two earthquakes of magnitudes more than six occur in the corresponding cycle of the low-value points. The first peak corresponds to an earthquake that occurred 15 km from Duzce (40.8°N, 31°E), Türkiye, with a magnitude of 6.1 on 23 November. In analogy, the CH₄ anomalies that grew continuously before the earthquake (from 6 January onwards) may be related to the earthquake doublet. There are also continuous high RETIRAs in the post-earthquake time, which may be related to aftershocks. Compared to the same period in the past 12 years, the time series of the average CH₄ concentration increased significantly in 2023, especially three cycles before the origin time of the earthquake (6 January 2023), and the high values persisted up to the earthquake (Figure 6). The time series of the average RETIRA index in the study area shows the same trend (Figure 6). The results of Figure 5 and Figure 6 correspond well with Figure 4.

4.2. Three-Dimensional Vertical Distribution Analysis

The gas anomalies caused by earthquakes should be bottom–up in theory; that is, the gas source is in the lithosphere. Therefore, in this paper, we introduce 8-day three-dimensional data to analyze the spatial structure characteristics of anomalies from a three-dimensional perspective and discuss whether the structure conforms to the seismogenic hypothesis. The main method is to calculate the RETIRA index at different levels according to Formulas (1)–(4) in Section 3.2 and present it in three-dimensional form. Figure 7 shows the three-dimensional spatial–temporal distribution of the methane RETIRA index on an 8-day scale. According to anomaly determination rules in Section 3.3, there are two main types of gas anomalies caused by earthquakes in a vertical space: the “pyramid” type and the “inverted pyramid” type.

The result in Figure 7 shows that persistent and large-area CH₄ anomalies occurred at 200–700 hPa in the epicenter and its surrounding areas at the beginning of December. The time the anomalies occurred and the location changes in anomalies at each level correspond to the two-dimensional spatiotemporal distribution of RETIRA in Figure 4 and reflect the changes among anomaly levels and the spatial distribution characteristics. The time variation in RETIRA in three-dimensional space is as follows: emergence–enhancement–reduction–enhancement to peak–reduction to the onset of the earthquake. In the vertical direction of the study area, from 5 December to 28 December and from January 6 to the day of the earthquake, the RETIRA gradually weakened with increasing height, and the three-dimensional structure was a “pyramid” type. From 29 December to 5 January, RETIRA gradually increased with height, and the three-dimensional structure was an “inverted pyramid” type.

5. Discussion

5.1. Exploration of Mechanism of Methane Anomaly Formation

During the seismogenic process, the accumulation of ground stress causes the deformation of the surface and the fracture of rocks, thus forming cracks. Under pressure, underground gasses, including CH₄, migrate to the crust through new cracks, forming gas anomalies above the incoming epicenter and surrounding areas. Many studies show a clear connection of gas release with earthquake preparation process. When geochemical gasses are released to the surface, it can undergo photochemical reactions in the atmosphere, causing other gas concentrations to change. In addition, the increase in methane may also trigger a local greenhouse effect, resulting in changes in surface temperatures. Geochemical gasses can ionize air and affect factors such as OLR or relative humidity or trigger atmospheric gravity waves that affect the ionosphere [56,57]. The study of these geochemical gasses is of great help in understanding the LAIC model. Figure 4 shows that high-value anomalies mainly occurred at the fault or fracture junction, and the anomaly positions were also shifted possibly due to the change in rock permeability before and after the earthquake. For example, the pre-earthquake anomalies were mainly located on the west side of EAF and the north side of NAF, but many CH₄ anomalies appeared on both the east and west sides of EAF after the earthquake. CH₄ has the characteristics of low density and easy diffusion. After escaping from the ground to the surface, it diffuses into the atmosphere, causing gas anomalies in three-dimensional space. Moreover, due to the structural activities themselves, the gas release has the characteristics of intermittent and uneven speed, coupled with the instantaneous spatial characteristics of satellite transit scanning, making the gas appear as a “pyramid” type and an “inverted pyramid” type in three-dimensional space. In the earthquake preparation stage, when the tectonic activity is enhanced, a large area of the surface is activated, and methane can escape from the surface in a large area simultaneously. The near-surface anomaly has a wide range and a large concentration, and the concentration decreases with height, showing the anomaly “pyramid” type in three-dimensional space. When tectonic activity is relatively weakened, regional methane release decreases. Methane is released along the relatively weak area, and its emission form is similar to that of multiple “point sources” (relative to the region); the gas emission rate decreases, and the scope of near-surface anomalies becomes smaller. Meanwhile, methane accumulates at a higher level, the anomalies increase with height, and the anomalies appear as an “inverted pyramid” in three-dimensional space.

However, the three-dimensional structure of methane anomalies is susceptible to the influence of atmospheric circulation, resulting in non-seismically induced pyramidal features. For example, a methane-carrying air mass drifting in from a distance is affected by a high-pressure anticyclone. The airflow is dispersed and sinks, resulting in the methane anomaly “pyramid” structure. In order to study the correspondence between methane and the seismogenic process in detail and to exclude the anomalies caused by other factors such as meteorology, the daily methane volume mixing ratios of the 17 days before the earthquakes and 6 days after the earthquakes were further selected, the RETIRA index was calculated, and its spatial structure characteristics were analyzed. Figure 8 shows abnormal methane emissions from 20 January to 12 February, with obvious differences in the emission intensity. The anomalies on 23 January and 27 January were “pyramid”-type anomalies, showing the lowest surface concentration of methane anomalies, and the three-dimensional structure increased with height. The methane anomalies on 20 January, 21 January, 22 January, 25 January, 26 January, and 29 January were “inverted pyramid”-type anomalies, showing the lowest surface concentration of methane anomalies and an increase in three-dimensional structure with height. This further demonstrates the intermittency and inhomogeneity of the Earth’s gas emissions.

The anomaly extraction algorithm can reduce the influence of regular changes in topography, geomorphology, and seasons. Still, short-term atmospheric motions can also cause changes in air masses, which interfere with identifying seismic anomalies. According to the air mass track of the abnormal high value point and the methane volume mixing ratio data, we can determine whether the methane anomaly is transported from other high value areas or from this area, and then determine the relationship between the methane anomaly and the earthquake. In order to exclude the influence of air mass migration, this study took the methane anomaly, which migrated to the northwest with increasing height, as an example to explore the source of the northwest-direction methane anomaly in the upper troposphere on 27 January. The coordinates [34.5°E, 41°N] were selected as the anomalous high-value candidate point, and the 850 hPa wind field, the mean sea level pressure, and the latitudinal and longitudinal vertical velocity profiles of 27 January were selected for analysis, as shown in Figure 9. The 850 hPa wind field in Figure 9a shows a cyclone to the west of the epicenter, and the anomalous high point was in the east–north direction of the cyclone, with prevailing south winds. Figure 9b,c shows strong upward airflow in the vertical direction of abnormal candidate points. In addition, the point of abnormal high value was taken as the starting position of the HYSPLIT model to conduct backward trajectory analysis, as shown in Figure 10. After 24 h of tracing, the air mass was located at [6°W, 32.4°N] and had an altitude of 9100 m (about 300 hPa). The blue trajectory in Figure 10 represents the migration path of the air mass within 24 h. Although there is a lack of measured value in the migration path of the air mass, the location of the air mass is not in the area with a high CH₄ volume mixing ratio at 00:00 and after 12:00 on 26 January, which can exclude the regional CH₄ anomaly caused by the air mass migration.

The RST algorithm uses years of data to construct a background field, which can partially eliminate the influence of natural sources such as seasonal changes and surface vegetation, and has strong stability and regularity. However, RST is a statistical algorithm, and small-sample data may have a certain impact on the result of the algorithm, which will interfere with the selection of threshold values. The data we used in this paper include daily product data and 8-day product data. The time continuity of daily product data is better than 8-day product data, but there are a lot of missing measurements due to the satellite orbit. The 8-day product data are not continuous, but the coverage is wide, and the stability is good. In general, the 8-day methane volume mixing ratio data result from data accumulation, which can better highlight the characteristics of the time series after a short period of gas accumulation and show a more complete pyramid structure. By using the methods in this paper, methane anomalies, especially in the 8-day scale, can reflect the seismogenic characteristics to a certain extent, and the three-dimensional spatial distribution of methane anomalies can better reflect the geological characteristics of earthquakes. Combined with atmospheric circulation movements, non-seismic anomalies can be excluded. However, due to the limitations of the satellite observation model, the daily methane volume mixing ratio data used in this paper have some shortcomings, which cannot cover the global area, so these methods cannot be applied to every earthquake case. In addition, the factors excluded by these methods are limited, and the universality of this process and the degree of response to different earthquake magnitudes still need in-depth statistical analysis. In the future, with the enrichment of products and the improvement and innovation of methods, daily scale data can be used in earthquake monitoring in a refined way, which will help to understand the geochemical mechanism of earthquakes in depth.

5.2. Earthquake Case Supplement and Verification

In order to further test the availability of the anomaly extraction press in this paper, we selected the 2021 Alaska earthquake as a case for analysis. According to the USGS, the 8.2-magnitude earthquake struck south of the Alaska Peninsula at 06:15 UTC on 29 July 2021, with an epicenter of 55.36°N, 157.89°W. We use this seismic event as a case to validate the methods in Section 3. Figure 11 shows the three-dimensional spatial–temporal distribution of the methane RETIRA index on an 8-day scale. The result shows that, in the vertical direction of the study area, from May 16 to June 8, the RETIRA gradually weakened with increasing height, and the three-dimensional structure was a “pyramid” type. From June 17 to August 3, the RETIRA gradually increased with height, and the three-dimensional structure was an “inverted pyramid” type. Analyzing the spatial structure characteristics of the RETIRA index of daily methane data, as shown in Figure 12, indicated that the anomalies were strongest at the surface on July 13 and July 23 and gradually weakened with the height, indicating the “pyramid” type of anomalies. The anomalies showed the lowest concentration at the surface from July 14 to July 21, and the three-dimensional structure exhibited the “inverted pyramid” type. It is worth noting that the anomaly three-dimensional structure corresponding to June 9 in Figure 11 was difficult to define because the northwest part of the study area showed a “pyramid” type, but the south part of the study area showed a “inverted pyramid” type. In order to explore the formation of this phenomenon, the coordinates [45°N, 161°W] were selected as the anomalous high-value candidate point, and the 850 hPa wind field, the mean sea level pressure, and the latitudinal and longitudinal vertical velocity profiles of July 9 were selected for analysis, as shown in Figure 13. The 850 hPa wind field in Figure 13a shows a cyclone near the anomalously high point. The central flow of the cyclone converges and rises, and this may be one of the reasons for the anomaly three-dimensional structure on June 9 in Figure 11. Figure 13b,c shows strong upward airflow in the vertical direction of abnormal candidate points. In addition, the point of abnormal high value was taken as the starting position of the Hysplit model to conduct backward trajectory analysis, as shown in Figure 14. After 24 h of tracing, the air mass was located at [36.8°N, 175°W] and had an altitude of 2400 m (between 700 and 850 hPa). The blue trajectory in Figure 14 represents the migration path of the air mass within 24 h. The migration path of the air mass is mostly in the area of low-methane volume mixing ratio, which can exclude the regional methane anomaly caused by air mass migration. The analysis of the Alaska earthquake case further proves the applicability of the anomaly extraction process in this study.

5.3. Study on the Area of Influence of a Methane Anomaly

At present, it is still difficult to carry out a whole-area earthquake study based on satellite remote sensing, and most of the current research involves earthquake case analysis, with varying ranges of the selected study areas. The main reference is the empirical relationship between the circular size area of mechanical/thermal/electromagnetic precursors and the magnitude of the final earthquake, $R={10}^{0.43M}$ (M is the magnitude), proposed by Dobrowolsky et al. [52]. For anomalously high values, or time series maps of regional anomaly time series analysis, the current applications are time series maps of epicenter anomalies [58] and time series maps of averages [26,30]. It is not clear how the satellite data of different spatial scales reflect tectonic activity, and whether they correspond to Dobrovolsky’s empirical relationship remains to be investigated. In Figure 15, we plotted the time series of the RETIRA index in different regions with the epicenter as the center and the 4° radius as the interval.

The results show the following: (1) In terms of anomaly amplitude, it is difficult to single out a fixed threshold to determine the anomaly, and comprehensive analysis should be combined with the time series characteristics of the selected region. For example, the anomaly value greater than 2 is the statistical feature of an anomaly for a single pixel, and the regional mean is the average value of all pixels in the selected region. Figure 15 shows that 2.5 (red dotted line) can better highlight the seismogenic anomaly value than 2 (green dotted line), but it does not show that the higher the value, the better. How to determine the threshold is still a challenge. To a certain extent, using the time series mean value of the regional mean value in the selected period can reflect the regional tectonic activity (dotted black line), but it is not necessarily the best choice. If risk monitoring is carried out based on the regional statistical abnormal characteristics, it is still necessary to research the threshold determination method. (2) The epicenter of an earthquake is not necessarily the best representation of tectonic activity. The empirical formula $R={10}^{0.43M}$ illustrated by Dobrowolsky et al. [52] is suitable for selecting a gas anomaly monitoring area for seismic remote sensing. The main reason for this may be that the preparation and occurrence of earthquakes is a regional, comprehensive, and complex geophysical and geochemical evolution process, and a single point does not usually reflect the seismic characteristics accurately. As shown in Figure 15, with the increase in the radius, the abrupt change in the anomaly index decreases. However, the trend in the auxiliary impending earthquake’s high value becomes increasingly prominent, and the anomaly features are not obvious beyond a certain regional radius (20° in this study).

6. Conclusions

In this study, we used remote sensing satellite data, combined with atmospheric circulation and backward trajectory analysis to analyze the Türkiye–Syria earthquake doublet, with the purpose of establishing a fine extraction process and determination rules for methane anomalies. In addition, we used the Alaska earthquake to test. The results show that this process and rules can be used for seismic remote sensing monitoring and methane anomaly extraction, and they can provide ideas and means for eliminating non-seismic anomalies. The main conclusions are summarized as follows:

• The focus of this study is on the fine extraction of earthquake anomaly information, trying to construct a process to provide reference for the extraction of earthquake anomalies. This study takes a typical strong earthquake as an example to analyze, and the multi-dimensional and multi-scale seismic methane anomaly extraction process constructed in this study can be used for remote sensing seismic monitoring and is more in line with the seismogenic mechanism.
• The anomaly extraction process and determination rules were applied to the case analysis of the 2023 Türkiye–Syria earthquake doublet, and the methane anomalies possibly related to the earthquake were extracted. Methane anomalies appeared on the EAF and NAF or seismogenic structures in the epicenter and the surrounding area two months before the earthquake, especially in the northern and southern locations, and the anomalies lasted until three days before the earthquake. The anomalous structure conformed to the geological characteristics of tectonic activity and was manifested as a “pyramid” or “inverted pyramid” type in three-dimensional space. The anomalies caused by air mass migration could be eliminated by combining them with atmospheric circulation motion.
• Dobrowolsky et al.’s empirical formula, $R={10}^{0.43M}$, is suitable for regional seismic remote sensing gas-anomaly monitoring, but the time series of the epicenter or a certain point in the region does not fully reflect the characteristics of regional tectonic activity. The optimal determination of the range and magnitude gas anomalies caused by tectonic activities is a difficult task for future research.