Analysis of Spatiotemporal variations in mid-upper tropospheric methane during the Wenchuan Ms8.0 earthquake by three indices

This research studied the spatiotemporal variation in methane in the mid-upper troposphere during the Wenchuan earthquake (12 May, 2008) using AIRS retrieval data and discussed the methane anomaly mechanism. Three indices were proposed and used for analysis. Our results show that the methane concentration increased significantly in 2008, with an average increase of 5.12 ∗ 10−8, compared to the average increase of 1.18 ∗ 10−8 in the previous five years. The Alice and Diff indices can be used to identify methane concentration anomalies. The two indices showed that the methane concentration distribution 5 before and after the earthquake broke the distribution features of the background field. As the earthquake approached, areas of high methane concentration gradually converged towards the west side of the epicenter from both ends of the Longmenshan fault zone. Moreover, a large anomalous area was centered at the epicenter eight days before the earthquake occurred, and a trend of strengthening, weakening and strengthening appeared over time. The Gradient index showed that the vertical direction obviously increased before the main earthquake, and the value was positive. The gradient value is negative during coseismic 10 or postseismic events. The gradient index reflects the gas emission characteristics to some extent. We also determined that the methane release was connected with the deep crust-mantle stress state, as well as microfracture generation and expansion. However, due to the lack of any technical means to accurately identify the source and content of methane in the atmosphere before the earthquake, an in-depth discussion has not been conducted, and further studies on this issue may be needed.

surface ruptures that were more than 60 km long, as indicated. The Longmenshan fault zone has a high dipping angle (more than 50 • − 60 • ) near the surface and a low angle at depth (15 − 20 km). This listric shape favors significant strain or energy accumulation, forming large earthquakes. This earthquake was characterized by slow strain accumulation, a long recurrence interval and significant damage power. It is a new type of earthquake that deserves further study (Zhang et al., 2008).
The variation in soil gas concentration serves as a useful tool for monitoring earthquakes. Numerous field investigations 5 have indicated that large amounts of gases are emitted from active fault zones before, during and after great earthquakes.
The increased gas emanation from the Earth's crust in the vicinity of active tectonic faults is triggered by a chain of physical processes and chemical reactions from the ground surface. This complex chain leads to geochemical, atmospheric, ionospheric and magnetospheric anomalies (Dobrovolsky et al., 1979;Pulinets and Ouzounov, 2011;Ouzounov et al., 2007). The number of pre-earthquake thermal, surface latent heat flux and outgoing longwave radiation anomalies apparently results from earthquakerelated gas emission from the lithosphere (Tronin et al., 2002;Dey et al., 2004;Tronin, 2006;Ouzounov et al., 2007;Tramutoli et al., 2013). Thermal (Zhang et al., 2010), ionospheric (Lin, 2012(Lin, , 2013Zhu et al., 2010), electromagnetic (Zhang et al., 2011a), and aerosol (Qin et al., 2014) anomalies were found before the Wenchuan earthquake of May 12, 2008. 5 As the second most important greenhouse gas after carbon dioxide (CO 2 ), methane (CH 4 ) is approximately 20 times better at warming the atmosphere than CO 2 by weight and plays an important role in atmospheric chemistry. Some people have speculated that thermal abnormalities before earthquakes are related to the release of CH 4 . The CH 4 release mechanism has been investigated and results indicate strong methane emissions when friction is applied to marl-type rock (Martinelli and Plescia, 2005;Italiano et al., 2008). Yue indicated that the Wenchuan earthquake of May 12, 2008, was caused by the 10 rapid migration and expansion of a large amount of highly pressurized and dense CH 4 gas in crustal rock masses (Yue, 2013).
Sample results indicated that CH 4 was discharged from a shallow reservoir through faults or fractures caused by the earthquake (Zheng et al., 2013). This study choose to study the spatiotemporal variations of CH 4 during the Ms8.0 Wenchuan earthquake from satellite observations.
CH 4 satellite observations have started to be applied to seismological studies over recent years. The total column of CH 4 15 associated with the 12 May 2008 Wenchuan earthquake was investigated using satellite data from the AQUA Atmospheric Infrared Sounder (AIRS), and this work indicates that a large amount of CH 4 was emitted from underground into the atmosphere along the Longmenshan Fault Zone from approximately 1.5 months before to 3 months after the earthquake, and the closer to the epicenter, the larger the amount of emitted gas. The peak values were found at intersection areas (Yue, 2013;Cui et al., 2017 (Zhang et al., 2011b;Xiong et al., 2015). 25 Systematic observation of the vertical variation of CH 4 is scarce. The focus of this study is to examine the spatiotemporal variation of CH 4 in the mid-upper troposphere during the Wenchuan earthquake (12 May, 2008) using AIRS retrieval data and to discuss the mechanism of the methane anomaly. Three indices were proposed and used for analysis: the Absolute Local Index of Change of the Environment (ALICE) was used for anomaly detection; the Vertical Concentration Gradient (Gradient) was proposed to study the vertical variation; the Successive Differential Value (Diff) can show the time variation. These three 30 indices analyzed the spatial and temporal distribution of CH 4 before and after the earthquake from horizontal, vertical and time scales helping us understand lithospheric and atmospheric interactions during seismic activity.
The Version 6 and Level 3 standard gridded product of 8-day CH 4 volume-mixing ratios in a descending model (local nighttime), with 1 * 1 degree of spatial resolution, were obtained from the NASA Goddard Earth Sciences Data and Information Services Center (DISC) (http://disc.gsfc.nasa.gov/AIRS/index.shtml/). The peak sensitivity of the AIRS to methane retrieval occurs at 300 hPa, and the channels near 7.6 µm are most sensitive to the middle to upper troposphere (Xiong et al., 2015).

5
Therefore, the volume mixing ratio of the middle troposphere (400 hPa, approximately 5 km), the upper middle troposphere (300 hPa, approximately 7 km) and the upper troposphere (200 hPa, approximately 11 km) in the descending orbit data was used.
To extract the methane spatiotemporal anomalies before and after the earthquake, three parameters were applied: the Absolute Local Index of Change of the Environment (ALICE) (Tramutoli, 1998;Cui et al., 2017) the Vertical Concentration 10 Gradient (Gradient) and the Successive Differential (Diff), all carried out on the basis of eliminating the multiyear background.
The background field can partially remove the influence of natural sources such as seasonal changes and surface vegetation, effectively capturing emergency information such as earthquakes, reducing "nonseismic anomalies" to a certain extent, providing criteria for the extraction of seismic anomalies, and reducing the misjudgment and leakage of seismic anomalies. It was calculated by Eq 1: where G i (x, y, t, l) means the gas value in the l layer of the atmospheric pressure, measured at time t, corresponding to a location at (centered on) the coordinates (x, y); G ref (x, y, t, l) means the reference fields for the study area, defined as a time average gas value; σ(x, y, t, l) is the standard deviation of historical records collected under the temporal constraint. For this 20 study, N was defined as 5 years, from 2003 to 2007; t is the time of the measurement acquisition with t ∈ τ , where τ defines the homogeneous domain of the satellite imagery collected in the same time-slot of the day and period(month) of the year; l was defined as 3 layers (400 hPa, 300 hPa and 200 hPa).
The Absolute Local Index of Change of the Environment (ALICE) is calculated by Eq. 3 (Tramutoli, 1998;Cui et al., 2017): 25 The Vertical Concentration Gradient (Gradient) is used to characterize the vertical variation of gas and is calculated by Eq 4 to Eq 7 : G(x, y, t, ∆l) = G(x, y, t, l) − G(x, y, t, l + 1) where Gradient(x, y, t, l) stands for vertical concentration gradient value at level l, measured at time t, corresponding to a location at (centered on) the coordinates (x, y); G(x, y, t, ∆l) means the vertical difference of adjacent layers; G ref (x, y, t, ∆l) 5 means the reference fields of the vertical difference for the study area, defined as a time average gas value; σ(x, y, t, ∆l) is the standard deviation of historical records collected under the temporal constraint; l was defined as 3 layers (400 hPa, 300 hPa and 200 hPa).
Successive Differential Value (Diff) refers to the difference of the adjacent time gas value and was calculated by Eq. 8 to Eq 11 : 3 Results

Reference field
The 8-day average background field at different CH 4 heights in the study area was obtained and is shown in Fig.2, 3, and 4.
These figures show that the CH 4 concentration in the study area decreased significantly from the middle troposphere to the upper, and the average methane concentrations (from Mar 9th to Aug 15th, 2008) in each stratum were 1.794 PPM, 1.775 PPM and 1.762 PPM for middle, upper middle, and upper troposphere, respectively. This indicates that the middle troposphere is significantly affected by terrestrial emission sources.
The distribution of CH 4 has obvious seasonal characteristics and the seasonal cycles at different heights are similar. The CH 4 mixing ratio has a weak high value during Mar, and then decreases during April-May. It starts to increase in June and 5 stays high during July and August, especially over the Sichuan basin. The relatively high values at different times are mostly located at a structural confluence or at tectonic plate boundaries except for late July and August. Earth gas emissions may be the main cause of the high long-term values at that location (http://www.climatechange2013.org/images/report/P36Doc3_ WGI-12_Approved-SPM.pdf). Why does the CH 4 begin to increase in early summer? Studies have shown CH 4 emissions significantly correlate with temperature. When the temperature is < 15 • the Earthseldom produces CH 4 , and when the tem-10 perature reaches 35 • , CH 4 production can be many times higher than at 25 • (KazuyukiYagi and KatsuyukiMinami, 1990;Thomas et al., 1996;Holzapfel-Pschorn et al., 1986;Saarnio et al., 2010;Saarnio and Silvola, 1999). Therefore, the CH 4 concentration in the troposphere is higher during July and August. The CH 4 rice paddy emission, local emissions (such as from gas leakage or energy use), and transport (such as the meridional and zonal advection, convection from the lower troposphere, or stratosphere-troposphere exchange) also increase in early summer (Xiong et al., 2010).

15
The results of the background field analysis show that the CH 4 concentration is greatly influenced by the season, structure and underlying surface, and they present a typical spatial and temporal distribution of methane in southwest China (Zhang et al., 2011b). Gas emanates from the earth's crust continuously, even without earthquakes.

Anomalous Gas
The three indices mentioned above have been applied to identify the spatial and temporal variation of mid-upper tropospheric 20 methane anomalies associated with the Wenchuan Ms 8.0 earthquake using AIRS data. showed NE distribution. Then, the high-value areas began to weaken on August 8. This circle was probably caused by the big aftershocks on July 24th and August 1st.

Vertical Concentration Gradient index result
The Vertical Concentration Gradient stands for the vertical rate of change. Figures 8 and 9 show the value changing obviously at the Bayan har block and plate boundary. The highest values appeared at the Bayan har block, the XSH, ANH and DLS faults and lasted until May 12th before the earthquake. The values were positive before May 12th. They weaken and even turn negative after May 12th, especially during July 23rd to August 7th. The overall shape is still closely related to the fault and 5 basin edge. We will discusses this later.

Successive Differential Value
The trend of the reflected gas concentration over time was reflected by the difference change, and the result of 20080222 represented the difference between February 22nd and the 14th. The places with rapid changes mainly occur at plate boundaries and tectonic intersections. Figure 10 shows that the LMS fault zone began to show an abnormally high value on March 25th,

Discussion
Background field analysis can help us understand the gas distribution in the study area and has a certain reference role for the extraction and analysis of late abnormal changes. We have found that the CH 4 concentration is greatly influenced by season, structure and underlying surface, and presents a typical spatial and temporal distribution in southwest China. Comparing our work with the GPS velocity field, the high value area of the background field corresponds to the area of high GPS velocity 20 (Wu et al., 2015). Long-term stress accumulation causes deep earth gases to be expelled along fissures or structural weaknesses (Wang et al., 2017).
Three indices have been used to see if anomalous CH 4 could be identified. The CH 4 spatial variations were most likely caused by geological processes and/or the action of crustal stress in the lithosphere resulting in the Wenchuan earthquake.
A large amount of CH 4 was emitted from underground into the atmosphere along plate boundaries and tectonic belts from 25 approximately 1.5 months before to 3 months after the earthquake, and the closer to the epicenter, the larger the amount of emitted gas. A large anomalous area occurs centered at the epicenter eight days before the earthquake. A crucial question in this field of research refers to how can we link an individual precursor with a distinctive stage of the earthquake preparation.
The generation of such a seismic anomaly requires physical and chemical transformations which occur in a spatially extended preparation (activation) zone of an impending earthquake. Earthquakes exhibit in general complex correlations in time, space 30 and magnitude. It is widely accepted that the observed earthquake scaling laws indicate the existence of phenomena closely Figure 10. Distribution of the successive differential value at 300 hPa associated with the proximity of the system to a critical point (Varotsos et al., 2019). Therefore, such a requirement is satisfied during the appearance of the "critical window", i.e., the epoch during which the short-range correlations have evolved to longrange ones in an extended area, where the "critical radius R" is given by the empirical relation logR ≈ 0.5M , where M is the earthquake magnitude (Bowman et al., 1998). Notice, based on the recently introduced concept of the "natural time" by Varotsos and his colleagues (Varotsos et al., 2011). it has been shown that the foreshock seismic activity that occurs in the 5 region around the epicentre of the upcoming significant shock a few days up to one week before the main shock occurrence, behaves as critical phenomenon.
Therefore, the hypothesis that the large anomaly in methane eight days before the earthquake occurred corresponds to the critical point-window of the earthquake preparation process cannot be excluded. Accumulated experimental evidence supports the aforementioned hypothesis as follows:

10
The earthquake preparatory process has various facets which reflect correspondingly different precursors. Importantly, precursors emerge during the same period. A few days up to one week before the main shock occurrence, while they behave as critical phenomena, as well.Characteristically, such as precursors are: (i) ULF magnetic field variations recorded by groundbased magnetic observatories before significant earthquakes,e.g., Potirakis et al., 2018). (ii) MHz fracture induced MHz EM anomalies (Eftaxias et al., 2018). The generation of such a seismic anomaly also requires physical 15 and chemical transformations which occur in a spatially extended preparation (activation) zone of an impending earthquake.
Characteristic precursors are athe short-lived seismo-ionospheric EM precursors and EM anomalies rooted in preseismic LAI-coupling (Pulinets et al., 2003;Pulinets and Boyarchuk, 2004). Pulinets et al.(Pulinets et al., 2003) have provided strong evidence for the occurrence of ionospheric precursors well before the main shock: ionospheric precursors within a 5 days before the seismic shock were registered in 73% of the cases for earthquakes with a magnitude 5, and in 100% of the cases for 20 earthquakes with a magnitude 6.
To further verify that the increase in CH 4 concentration is related to earthquakes, we compared and analyzed the variation of overall methane concentration and anomaly index in the study area for nonseismic years (Figures 11 and 12 increased significantly due to seasonal influence from June 21st, showing a rapid increase in the slope of the curve (Table 1).
Although there was a certain increase in 2008, the curve slope was relatively small (Table 1), which may be due to the large amount of CH 4 released by the aftershock earthquake. The increasing trend of CH 4 concentration caused by seasonal change weakened between Mar 25th and June 21st, 2008.

30
We also compare the ALICE result with 2008 and 2007 (Figures 12 and 13 Figure 11. Comparison of the average CH4 VMR in the study area from 2003 to 2008 (300 hPa) activity, these effects may be local and can be eliminated by calculation with Equations 1 to 11. Therefore, the identified anomalies of CH 4 VMR can be considered earthquake-related (Cui et al., 2017). CH 4 is an important greenhouse gas next to carbon dioxide, and previous work shows that the obvious anomalies of thermal infrared brightness temperature appeared on April 25th and were mainly distributed in the LMS fault zone and its southern region (Zhang et al., 2010). These results correspond well with our work.
Compared with the nonseismic years from 2003 to 2007 (no earthquake >Ms 5.0), the study area before and after the Wenchuan earthquake showed a significantly increased CH 4 anomaly. It is well known that CH 4 is in a supercritical or critical 5 state at high temperature and pressure 10-20 km below the surface, and its physical and chemical properties are relatively stable during periods of weak tectonic activity. However, with active geological activity, this state is also very easy to break.
For example, the sudden expansion of the volume of cracks or cavities caused by the change of tectonic stress field can easily change the internal fluid state and cause the supercritical fluid phase transition (Liu and Du, 2000). As soon as the phase transition process begins, fluid volume and fracture expansion are triggered, resulting in a large amount of fluid spilling over 10 the surface and spreading into the atmosphere, which is then observed by ground-based instruments or satellites (Martinelli and Plescia, 2005;Italiano et al., 2008). The migration or release of underground fluids such as CH 4 is inevitably completed by other seismic processes in the preearthquake period, which is also an important way for energy conversion and accumulation from the deep part of the crust and mantle to the shallow part of the crust. Therefore, it can be inferred that the phase transition process of underground fluids 30 such as CH 4 was mainly completed before the earthquake, so there was an abnormal increase before the earthquake. During the earthquake and coseismic activity, the abnormal increase of fluid concentration such as CH 4 may also occur, which is mainly caused by local fluid release in the brittle layer of the upper crust (Wang et al., 2017). The range is mainly limited to the vicinity of the earthquake rupture zone, which is not on the same order of magnitude as the regional CH 4 release from the deep part. For example, the gradient shows that the values are positive before May 12th and become weak and even turn negative after May 12th ( Figures 8 and 9 ). The CH 4 in the upper layer may be deep gas released before the earthquake, and locally released after the earthquake decreases, so the concentration of the lower layer decreases and the gradient becomes negative.
The iteration of positive and negative values can also indicate geogenic emissions.
The aforementioned results seem to support the hypothesis that the observed anomaly in terms of spatiotemporal variation in methane is rooted in the stage of critical point-epoch of the earthquake preparation process.

Conclusions
This study has examined the spatiotemporal variation in CH 4 in the mid-upper troposphere during the Wenchuan earthquake (May 12th, 2008) using AIRS retrieval data and discusses the mechanism of the methane anomaly. The CH 4 concentration increased significantly in 2008 with an average increase of 5.12 * 10 −8 , compared with 1.18 * 10 −8 in the past five years. Three indices were proposed and used. The Absolute Local Index of Change of the Environment (ALICE) was used for anomaly detection; the Vertical Concentration Gradient (Gradient) was proposed to study vertical variation; the Successive Differential Value (Diff) shows time variation. The three indices analyzed the spatial and temporal distribution of CH 4 before and after the earthquake from horizontal, vertical and time scales. Through this work, the following conclusions are obtained: The background field of CH 4 is greatly influenced by season, structure and underlying surface and presents a typical spatial 5 and temporal distribution of methane in southwest China. Gas emanates from the earth's crust continuously, even without earthquakes.
The CH 4 concentrations in the upper and middle atmosphere before and after the Wenchuan earthquake have a certain temporal and spatial variation. The ALICE and Diff indices could be used to identify the CH 4 concentration anomaly. The research results indicate that the CH 4 concentration distribution before and after the earthquake breaks the distribution features of the 10 background field. The CH 4 concentration distribution starts from both ends and gradually gathers around the epicenter. Moreover, a large anomalous area occurs centered at the epicenter eight days before the earthquake, and a trend of strengthening, weakening and strengthening appears over time.
The Gradient method can reflect the change in gas concentration in the vertical direction. The results show that the vertical direction obviously increases before the main earthquake, and the value is positive. The gradient value is negative during 15 coseismic or postseismic events. It may be that the gas before the main earthquake mainly comes from deep fluid phase transitions, whereas the coseismic or postseismic gas may mainly come from the locally closed fluid in the brittle layer of the crust. However, due to the lack of any technical means to accurately identify the source and content of methane in the atmosphere before the earthquake, an in-depth discussion has not been conducted, and further studies on this issue may be needed. The gradient index can reflect the characteristics of gas emission to some extent.
for making available the AIRS dataset. The authors wish to thank the anonymous reviewers for their constructive comments that helped improve the scholarly quality of the paper.