Short-Term Monitoring of Geogenic Soil CO2 Flux in a Non-Volcanic and Seismically Inactive Emission Site, South Korea

Temporal changes of soil CO2 flux (FCO 2 ) and soil CO2 concentration ([CO 2 ] v ) were surveyed in a natural CO2 emission site to characterize the factors controlling the short-term temporal variation of geogenic FCO 2 in a non-volcanic and seismically inactive area. Due to a lack of long-term monitoring system, FCO 2 was discontinuously measured for three periods: Ⅰ, Ⅱ at a high FCO 2 point (M17) and Ⅲ about 30 cm away. Whereas [CO 2 ] v was investigated at a point (60 cm depth) for all periods. A 2.1 magnitude earthquake occurred 7.8 km away and 20 km deep approximately 12 h before the period Ⅱ. The negative correlation of FCO 2 with air pressure suggested the non-negligible advective transport of soil CO2. However, FCO 2 was significantly and positively related with air temperature as well, and [CO 2 ] v showed different temporal changes from FCO 2 . These results indicate the diffusive transport of soil CO 2 dominant in the vadose zone, while the advection near the surface. Meanwhile [CO 2 ] v rapidly decreased while an anomalous FCO 2 peak was observed during the period Ⅱ, and the CO2 emission enhanced by the earthquake was discussed as a possible reason for the synchronous decrease in [CO 2 ] v and increase in FCO 2 . In contrast, [CO 2 ] v increased to 56.8% during the period Ⅲ probably due to low gas diffusion at cold weather. In addition, FCO 2 was low during the period Ⅲ and showed different correlations with measurements compared to FCO 2 at M17, implying heterogeneous CO2 transport conditions at the centimeter scale. The abnormal FCO 2 observed after the earthquake in a seismically inactive area implies that the global natural CO2 emission may be higher than the previous estimation. The study result suggests a permanent FCO 2 monitoring station in tectonically stable regions to confirm the impact of geogenic CO2 to climate change and its relation with earthquakes.

Meanwhile, the temporal changes of geogenic FCO 2 and their controlling factors were relatively less studied and mostly in volcanic or seismic areas (e.g., Chiodini et al., 1998;Granieri et al., 2003;Carapezza et al., 2011;Camarda et al., 2016;Camarda et al., 2019;Oliveira et al., 2018;Morita et al., 2019;Chen et al., 2020). Morita et al. (2019) showed that barometric pressure, air temperature, soil temperature and humidity, and wind speed were decisive variables that could explain more than half of the variations in FCO 2 at 1 km southwest of the active crater of Aso volcano, while the residuals were explained using an increase in magmatic CO 2 flux. Repeated measurements by Chiodini et al. (1998) showed that FCO 2 was governed by the change in barometric pressure, while other meteorological parameters such as rain, soil and air temperature, and humidity also influenced FCO 2 in volcanic and geothermal areas. As for seismically active regions, Kerrick and Caldeira (1993), Kerrick and Caldeira (1994), Kerrick and Caldeira (1998), and Kerrick et al. (1995) studied metamorphic CO 2 degassing, including convective hydrothermal CO 2 emission. In addition, Chiodini et al. (1999) suggested the non-volcanic CO 2 derived from mantle degassing and/or metamorphic decarbonation in Central Italy. Lee et al. (2016) estimated 4 Mt/yr of mantle-derived CO 2 released along deep faults in the Magadi-Natron Basin at the border between Kenya and Tanzania. Ascione et al. (2018) introduced anomalously high FCO 2 resulting from the combination of 1) intense CO 2 generation from magmatic bodies causing decarbonation of carbonate rocks; 2) a very thin or absent top seal overlying the carbonate reservoirs; 3) the occurrence of a dense network of active fault segments at the tip of a major crustal fault zone.
Moreover, seismic activity has been considered as an endogenous cause of the temporal variation of geogenic FCO 2 (Camarda et al., 2016;Camarda et al., 2019;Fischer et al., 2017;Sciarra et al., 2017;Chen et al., 2020). Specifically, Camarda et al. (2019) found that, of the two anomalous FCO 2 periods (A and C), the period A had a seismic swarm (3,471 seismic events in 79 days; Ricci et al., 2015) and thus showed the higheramplitude anomalies than the period C. Camarda et al. (2016) showed the high spatial and temporal correlation between seismicity and FCO 2 in a district with continuous seismic activity, whereas FCO 2 varied independently in the districts with low and sporadic seismicity. According to Chen et al. (2020), seismic activity also can be responsible for the jumpily temporal variations of CO 2 concentration and flux in soil gas wells. Furthermore, vibro-stimulation was applied to increase the oil production based on the physics that the rate of degassing increases due to vibration energy (Kouznetsov et al., 1998;Kouznetsov et al., 2002).
However, studies have been rarely conducted about either the temporal variation of geogenic FCO 2 or the effect of small seismic events to FCO 2 in a geologically quiescent (e.g., non-volcanic or seismically inactive) environment. As an alternative, CO 2 -rich waters have been studied in tectonically stable regions, in particular as a natural analogue study of geologic carbon storage to understand CO 2 leakage (e.g., Chae et al., 2016), because non-volcanic CO 2 is discharged by high-CO 2 groundwater as well as by focused degassing (Chiodini et al., 1999). For instance, in South Korea, which has no active volcanoes and had been relatively safe from seismic activity until the 2016 Gyeongju earthquake (M 5.8) (Woo et al., 2019), CO 2 -rich waters have been studied to identify anomalously high soil CO 2 areas, and their origins (e.g., magmatic degassing, metamorphic devolatilization, oxidation of organic matter) and ascending pathways. According to Jeong et al. (2005), the CO 2 gas derived from a deep-seated source moves into the groundwater system along faults or geologic boundaries in South Korea. However, FCO 2 has been rarely studied, which motivated this study.
This study aimed 1) to characterize the temporal variation of FCO 2 in a non-volcanic and seismically inactive site ( Figure 1A) where soil CO 2 was suggested to have a deep-seated magmatic origin , and 2) to identify the factors controlling the temporal changes of geogenic FCO 2 . The time-variant CO 2 supply and the effect of a small earthquake were discussed. This study contributes to provide a new study direction of long-term FCO 2 monitoring to the atmosphere in tectonically stable regions, which is important to assess due to the impacts of CO 2 on climate change, whereas many existing studies have focused on volcanic or seismic regions.

STUDY AREA
The study area (Daepyeong; Lat. 36 o 29′01″N and Long. 127 o 20′21″E) is located in the central South Korea ( Figure 1A) and in the middle of a small basin (about 1.1 km 2 ). The small watershed is surrounded by mountains lower than 270 m above sea level, and the low and flat area of the basin is mostly used for rice cultivation. There are also farmhouses and gardens that cultivate vegetables, fruits, and pine saplings on a small scale.
The bedrock of the study area consists of Precambrian gneiss that was intruded by Jurassic granite ( Figure 1A). The gneiss and granite are overlain by Quaternary sediments at lower altitudes. It is noticeable that the study area is located at the geologic boundary between gneiss and granite, along which five CO 2rich groundwater wells and two CO 2 -rich springs occur ( Figure 1A; Jeong et al., 2001;Jeong et al., 2005;Chae et al., 2016). Chae et al. (2016) observed fractures (fissures and/or joints) and CO 2 bubbles from the fractures at a CO 2 -rich spring (s-1).  found a high FCO 2 point (M17 in Figure 1A) about 1.8 m away from a CO 2 -rich groundwater well (w-2) to release geogenic CO 2 through the soil layer among a total of 94 points within 1 km 2 . The well (w-2) has a depth of 80 m and a diameter of 150 mm, and CO 2 -rich water is irregularly taken at w-2 for domestic usage by countless residents. The contact of gneiss and granite is observed on a slope near the well w-2.
A fault has not been identified in the study area (MCT et al., 2006), whereas there are faults in a regional scale including the closest Gongju Fault approximately 13 km away from the study area ( Figure 1B). At the regional scale, the study area is located on the southwest of the NE/SW trending Ogcheon Belt (Ogcheon region). The Ogcheon Belt is a fold-and-thrust belt affected by several deformational phases, and the Ogcheon region is mainly composed of metamorphosed clastic and volcanic rocks (Kihm and Kim, 2003). In the Ogcheon region, CO 2 -rich waters occur in the NE-SW direction (Supplementary Figure S1), parallel to the Gongju Fault and Ogcheon Belt, which implies the relation of CO 2 -rich waters to faults or fractures, while no evidence has yet been found.
A total of 49 small earthquakes (≥magnitude 2.0) occurred within 30 km from the study area in the past 30 years ( Figure 1B; KMA, 2019), including the 2.1 magnitude (M) earthquake occurring on 03:34 November 19, 2018 at 7.8 km southwest of the study area and 20 km deep. The information of focal depths is available only for five earthquakes occurring after 2017 and in the range of 11-20 km. The distribution of earthquake epicenters and their magnitudes indicate that the study area is relatively free from seismic hazards, while there may be unidentified and buried fractures.
The annual average temperature of the study area was 12.2°C, while the annual average relative humidity (RH) was 70. 8% in 19678% in -20048% in (MCT et al., 2006. The atmospheric temperature (T aws ) varied from 2.5 to 17.8°C in the period Ⅰ (Table 1; KMA,  2019). There was no rainfall, while it rained a week before the period Ⅰ. The total rainfall amount was 35.3 mm from October 26 to 29, 2018. During the periods Ⅱ and Ⅲ, T aws ranged between −0.9 and 12.7°C and between −10.7 and 9.5°C (Table 1), and the total amount of precipitation was 6.5 and 1.0 mm (Supplementary Figure S2), respectively.

METHODOLOGY
This study was conducted around the M17 point in Figure 1A found by a preliminary study on the spatial variation of FCO 2 around the CO 2 -rich wells and springs with 50 -100 m spacing within 1 km 2 . Among a total of 94 points, M17 was the only point to show geogenic CO 2 outflow through the soil layer. FCO 2 was detected up to 546 g/m 2 /d, while the soil CO 2 concentration at a depth of 60 cm ([CO 2 ] v ) and its carbon isotope (δ 13 C [CO2]v ) were 36.0% and -5.7‰, respectively at M17 in summer, 2017 (Table 1), which were much higher than the values of biogenic origin (average FCO 2 of 44.9 g/m 2 /d, [CO 2 ] v of 0.7% and δ 13 C [CO2]v of −25.2‰) observed at 79 samples in the study area .

Three Periods
Field works were conducted through three periods Ⅰ (from November 2 to 5, 2018), Ⅱ (November 19, 2018 to January 30, 2019), and Ⅲ (December 2 to 8, 2019) ( Table 1; Supplementary Figure S2). Monitoring during the periods Ⅰ and Ⅲ was conducted to assess the background level of FCO 2 and CO 2 concentration in soil gas ([CO 2 ] v ), while FCO 2 and [CO 2 ] v were investigated during the period Ⅱ to assess the effect of a 2.1 M earthquake nearby (i.e., 7.8 km southwest and 20 km deep) on FCO 2 .
Note that there is no long-term automated FCO 2 measurement system in the study area unlike other seismic or volcanic areas (e.g., Chiodini et al., 2001;Camarda et al., 2016;Morita et al., 2019) because FCO 2 is not a big concern. Besides, the study area is a private land. Thus, data were missing between periods, and measurement frequency varied at each period (Supplementary Figure S2) depending on the situations in the field (e.g., accessibility, power supply). Specifically, FCO 2 measurement and soil gas sampling were conducted simultaneously every 2 h during the period Ⅰ. Atmospheric air samples were collected in an 8-h interval approximately 1 m above the surface. Two weeks after the period Ⅰ ended, a 2.1 M earthquake occurred. Thus, further investigations for FCO 2 and soil gas were conducted since about 12 h after the earthquake occurred (period Ⅱ). FCO 2 was measured three times a day (at 14:00, 15:00, and 16:00), while soil gas samples were taken once a day (around at 14:00) until November 25. Then FCO 2 and soil gas were monitored once a week between December 26, 2018 and January 30, 2019 (Table 1). Lastly, FCO 2 was frequently (every 30 min) measured for a week from 00:30 December 2 to 22:30 December 8, 2019 (period Ⅲ). Soil gas and atmospheric gas samples were collected once at 15:00 on 8 December, 2019 for comparison.
The FCO 2 measurement point during the period Ⅲ was not exactly the same as M17, and about 30 cm away from M17 (called M17-1 hereafter) because the flux measurement device (i.e., LI-COR) had to be reinstalled in December, 2019 and we did not expect that the small spatial distance (i.e., 30 cm separation) affected the FCO 2 measurement. In contrast, [CO 2 ] v was investigated from the same tube (M17v) installed in August 2017 by  for all three periods.

Sampling and Measurement
FCO 2 , compositions of soil gas and atmospheric air, and their stable carbon isotopes (δ 13 C CO2 ) were monitored. In addition, gas pressures were measured at a depth of 90 cm (P 90 ) and on the surface (P 0 ) during the period Ⅰ. Meteorological parameters were obtained from an automatic weather station (AWS) of the Korea Meteorological Administration (KMA) near the study area (about 10 km away). All the measurements and their devices were summarized in Supplementary Table S1. FCO 2 , P 0 , and P 90 measurements (CO 2 flux and Gas Pressure Measurement) and gas sampling and analysis (Ga Sampling and Analysis) were detailed below.

CO 2 flux and Gas Pressure Measurement
A PVC collar (height of 11.5 cm and inside diameter of about 20 cm) was implanted into the soil, on which a bottom-opened chamber was placed ( Figure 1C). Then pressure (P s ; hPa), temperature (T s ;°C), relative humidity (RH s ; %), CO 2 concentration ([CO 2 ] s ; ppm), and water vapor mole fraction ([H 2 O] s ; mmol/mol) in the soil chamber were measured for 2 min by a built-in infrared gas analyzer using LI-COR 8100A (LI-COR Inc. Lincoln, NE, USA). FCO 2 (g/m 2 /d) was calculated by Equation 1 as Jung et al. (2014): where k is a unit conversion factor (3.80 g·s/μmol/d), R is the universal gas constant (8.31 m 3 ·Pa/K/mol), S is the soil surface area (herein, 317.8 cm 2 for the about 20 cm diameter chamber), V is the system volume (i.e., the sum of the chamber volume and the extra volume by a offset), P, T,   Supplementary Table S1 for details. b weekly monitored. c P 90 values (thus ΔP P 90 -P 0 ) were missing between 8 PM on November 2 and 10 AM on November 3 (see Figure 3).
Frontiers in Earth Science | www.frontiersin.org January 2021 | Volume 8 | Article 599388 of P 90 and P 0 were used when the relationships with other measurements were assessed. In other words, P 90 and P 0 data (thus ΔP P 90 -P 0 ) were taken for 15 min before and after the FCO 2 measurement, respectively and then averaged for the 30 min.

Ga Sampling and Analysis
Soil gas samples were taken using a Teflon tube which had the outer diameter of 0.64 cm and was embedded down to 60 cm below the surface with the AMS Gas Vapor Probe (AMS, Inc., USA). Atmospheric air samples were collected 1 m above the surface. Soil gas and atmospheric air samples were purged for 5 min using a portable Masterflex E/S peristaltic pump (Cole-Parmer Instruments, USA) and then collected in a 1 L multilayered Tedlar bag (Restek © , USA). Soil gas duplicates were collected once a day (at 14:00; n 4) during the period Ⅰ to double check the results of δ 13 C [CO2]v analysis, with connecting the y-shaped adapter at the end of the sampling tube. The carbon isotope of CO 2 (δ 13 C CO2 ) in the gas samples were analyzed by the Picarro G2121-i isotope and gas analyzer (Picarro Inc., USA) at the Korea Institute of Geoscience and Mineral Resources (KIGAM). The Picarro cavity ring-down spectroscopy (CRDS) was calibrated by IAEA standard materials (δ 13 C CO2 2.492, −5.764, −47.321‰) before analyzing samples. All gas samples were purged for 10 min with laboratory air and then analyzed for 15 min to avoid the memory effect. Results were expressed relative to the international V-PDB standard. The soil gas duplicates (n 4) obtained during the period Ⅰ and a soil gas sample obtained during the period Ⅲ were analyzed by Thermo Fisher Delta V TM IRMS (isotope ratio mass spectrometer) at Beta Analytic Inc. (Miami, USA) for comparison.
Gas compositions (N 2 , O 2 , and CO 2 ) were determined by the Agilent 490 Micro Gas Chromatograph (GC) at KIGAM. Before the laboratory analysis, two columns in GC (i.e., CP-Molsieve 5A column for N 2 and O 2 and PoraPLOT U column for CO 2 ) were calibrated by three different standards (CO 2 49.98; 5.00; 0.04%: Rigas © , Korea). Each sample was analyzed at least three times, and the coefficient of variation was less than 0.9% for CO 2 .

Statistical Analysis and CO 2 Solubility Calculation
Simple statistical analyses were applied for each period because of discontinuous and short-term observations at different intervals. First, Pearson correlation coefficients (r) were calculated to evaluate the effect of environmental parameters to FCO 2 , and to find the relation varying depending on a measurement period. According to Camarda et al. (2019), the response of FCO 2 to exogenous parameter variations is dependent on the predominant process of CO 2 transport through the soil. At sites with diffusion-dominated CO 2 transport, FCO 2 was mainly affected by variation in the volumetric water content of the soil and the air temperature, while at sites with high fluxes and non-negligible advective components, FCO 2 was affected solely by variation in the atmospheric pressure (Camarda et al., 2019). The relationship between FCO 2 and exogenous parameters is also influenced by the amount of deep CO 2 supply. Then multiple regression was used to explain the relationship between FCO 2 and highly correlated environmental parameters and to estimate the effect of endogenous parameters. Regression analyses have been widely conducted (Granieri et al., 2003;Vodnik et al., 2009;Carapezza et al., 2011;Camarda et al., 2016;Oliveira et al., 2018;Morita et al., 2019) to distinguish the effect of each factor to the FCO 2 variation, including endogenous (e.g., seismicity) and environmental parameters. It should be noted that we used P s , T s , [CO 2 ] s and [H 2 O] s in the soil chamber measured for 2 min using LI-COR 8100A as environmental parameters (see Table 1 and Supplementary Table S1) and discussed their usefulness (Environmental Parameters), because they represent the mixture of soil efflux from the soil layer for 2 min and the air initially filling the chamber, and are different from the initial values used to calculate FCO 2 in Equation 1.
Second, autocorrelation of time series was assessed to find the periodicity for the data obtained during the periods Ⅰ and Ⅲ (Supplementary Figures S3, S4). Cross-correlation coefficients (R xy ) were evaluated for the data obtained during the period Ⅰ to characterize the lead-lag relationship between input variables (x; measurements in Table 1) and output variables (y; mainly FCO 2 in this study) and a time lag (Supplementary Figure S5): where C xy is the covariance between x and y, while σ x and σ y are the standard deviations of x and y in a lag time (t lag ), respectively. When t lag 0, r R xy (0). R statistical software was used for autocorrelation and cross-correlation analysis (R core team, 2019).
In addition, the amount of CO 2 degassing from a CO 2 -rich aquifer was estimated by calculating the variation in CO 2 solubility in groundwater due to the variation in pressure, salinity and temperature based on the method by Duan and Sun (2003).

Background Levels
Measurements during the periods Ⅰ and Ⅲ were compared to assess the background levels of FCO 2 and soil gas compositions ( Table 1) and their relations with environmental variables (Tables 2, 3). Table 4 showed R xy with a non-zero t lag . The other parameters showed a t lag of zero with FCO 2 , and thus R xy r in Table 2 when y FCO 2 .

CO 2 Flux
The average FCO 2 during the period Ⅰ was 564 g/m 2 /d and ranged from 449 to 674 g/m 2 /d, which was similar to 546 g/m 2 /d obtained at M17 in August 2017 (Table 1) and quite high compared to FCO 2 in the other 93 points (7.5-118 g/m 2 /d) in the study area measured by  and in typical normal soil systems (∼40 g/m 2 /d) suggested in Ascione et al. (2018). Meanwhile, FCO 2 during the period Ⅲ ranged between 159 and 315 g/m 2 /d (average of 228 g/m 2 /d), which was higher than those in the 93 points of  and in typical normal soil systems by Ascione et al. (2018), but much lower than   Figure S3), while the diurnal variation of [H 2 O] s was not clear during the period Ⅲ ( Figure 2D and Supplementary Figure S4), especially until December 6, 2019 when the temperature was lowered down to −10°C and the pressure exceeded 1,010 hPa ( Figures 2B,F). RH s (r −0.6) and RH aws (r −0.4) were negatively correlated with FCO 2 during the period Ⅰ ( Table 2), but not related with FCO 2 during the period Ⅲ ( Table 3).
Despite the different ranges in FCO 2 at each period, FCO 2 had significant correlations with T s , [H 2 O] s , T aws , P s and P aws at both periods (Tables 2, 3). FCO 2 increased with increasing [H 2 O] s , T s and T aws but decreased with P s and P aws . In addition, P 0 and P 90 were negatively correlated with FCO 2 during the period Ⅰ. The negative relation of FCO 2 with P aws was explained by the fact that a decrease in barometric pressure increases the pressure gradient of the ground, which subsequently enhances the viscous gas flux (Rogie et al., 2001;Granieri et al., 2003;Morita et al., 2019). The positive relation with T aws on the short time scale was explained as the effect of variations in soil gas diffusivity with air temperature as well as surficial biological productivity (Camarda et al., 2016 and references therein). The positive effect of [H 2 O] s can be explained by its positive relationship with T s and T aws and negative with P s and P aws in Tables 2, 3, and will be further discussed in Environmental Parameters regarding the usefulness of the environmental parameters obtained in the chamber.

Soil Gas and Air
CO 2 concentrations in the soil gas obtained at a depth of 60 cm ([CO 2 ] v ) ranged from 39.6 to 48.9% (average 43.8%) during the period Ⅰ ( Figure 4A), which were slightly higher than that measured in August, 2017 (36.0%) but lower than that during the period Ⅲ (56.8%) in Table 1. δ 13 C [CO2]v for the soil gas was between −10.1 and −5.4‰ (average −7.1‰) during the period Ⅰ and was −6.2‰ during the period Ⅲ, which were a little lower than the value (−5.7‰) obtained in August, 2017 ( Figure 4A; Table 1). However, the variations in δ 13 C [CO2]v were insignificant and the δ 13 C [CO2]v values were relatively high compared to the δ 13 C [CO2]v of biogenic origin in the study area between −32.0 and −13.0‰ (average −25.2‰) in .
[CO 2 ] v did not show a distinct diurnal variation during the period Ⅰ ( Figure 4A) similar to FCO 2 and [CO 2 ] s that showed the pattern out of the diurnal variation during both periods I and Ⅲ unlike T s , RH s or [H 2 O] s (Figure 2; Supplementary Figures S3, S4). Table 2, while positively with FCO 2 at the time lag of −16 h (R xy 0.5 in Table 4 Figure S5). [CO 2 ] v was negatively related with P 90 (r −0.6 in Table 2), indicating that the high P 90 caused [CO 2 ] v to decrease. The average N 2 ([N 2 ] v ) and O 2 concentrations ([O 2 ] v ) of soil gas obtained at a depth of 60 cm was different between the two periods: 42.9% (40.0-46.3%) and 11.4% (10.6-12.2%) respectively during the period Ⅰ, while 30.3% and 7.8% respectively during the period Ⅲ (Figure 5; Table 1), probably due to the high proportion of CO 2 during the period Ⅲ .
It is noticeable that both [CO 2 ] v and [CO 2 ] s showed a rapid increase during the period Ⅰ with the time lag of about 12 h. Specifically, [CO 2 ] v rapidly increased from 39.8 to 48.9% at 22:00 on November 2 ( Figure 4A), while [CO 2 ] s increased from 0.3 to 0.5% at 10:00 on November 3 ( Figure 2C). Consistently, the cross-correlation analysis showed that the correlation between [CO 2 ] v and [CO 2 ] s increased up to 0.55 at the time lag of -12 h in Supplementary Figure S5 from zero in Table 2, and suggested the transport rate of approximately 60 cm/ 12 h. [CO 2 ] s occasionally showed rapid increases during the period Ⅲ as well, in particular around December 4 (see the red arrow in Figure 2D), whereas [CO 2 ] v measurements were not available for comparison.
Meanwhile the CO 2 concentrations in the air samples ([CO 2 ] a ) measured during the periods Ⅰ (0.05-0.19%) and period Ⅲ (0.20%) in Figure 4B and Table 1 showed high values compared to a reported atmospheric CO 2 composition (0.04%) and the value previously detected in the study site (0.05%). [CO 2 ] a was negatively correlated with FCO 2 (r −0.8) and [CO 2 ] v (r −0.5), but not with [CO 2 ] s (r 0.0) during the period I ( Table 2). δ 13 C [CO2]a for the air samples was different between two periods. The average δ 13 C [CO2]a (−19.0 ± 2.8‰; n 10) during the period Ⅰ was much lower   After the Earthquake FCO 2 increased up to 1,073 g/m 2 /d by a factor of two approximately 12 h after the earthquake (Figure 6 and Supplementary Figure S2), which was much higher than the sum of mean (μ) and 2 times standard deviation (2σ) of FCO 2 during the period Ⅰ (μ + 2σ; 668 g/m 2 /d). Besides, relatively high FCO 2 was observed on November 25, 2018 (836 g/m 2 /d) and January 23, 2019 (802 g/m 2 /d) ( Figure 6). Those high FCO 2 values were not observed either during the period Ⅰ or in the previous study at M17 ( Table 1).
The high FCO 2 seemed to be related with [CO 2 ] s (Figure 7; r 0.9 in Table 5). Their high correlation was not observed in the other periods. Namely, [CO 2 ] s rapidly increased after the earthquake and decreased with FCO 2 , and high [CO 2 ] s was observed on both November 25, 2018 and January 23, 2019 (Figure 7), although the maximum [CO 2 ] s (5,451 ppm) was observed during the period Ⅰ (Table 1; Figure 2C). Unlike FCO 2 increasing abruptly, however, [CO 2 ] v dropped by about half after the earthquake, and then increased back but did not reach the pre-earthquake values until January 30, 2019, although a high value of [CO 2 ] v (56.8%) was observed in December, 2019 (Table 1). Thus, [CO 2 ] v was negatively related with FCO 2 during the period Ⅱ (r −0.5 in Table 5). The maximum [CO 2 ] s observed during the period Ⅰ and the increasing in [CO 2 ] s and FCO 2 but the decreasing in [CO 2 ] v after the earthquake will be discussed in relation to the earthquake in Possible Causes of a High CO 2 Emission During the Period Ⅱ. , and pressure (P s ; hPa) in the chamber and calculated CO 2 flux (FCO 2 ; g/m 2 /d) in (E,F) during the periods Ⅰ (left) and Ⅲ (right). In (E,F), the gray line addresses the FCO 2 estimated using T aws and P aws in Equations 3, 4, respectively. Note that [CO 2 ] s , [H 2 O] s , FCO 2 , and P s have different scales between two periods. The red arrow in (D) indicates the time of rapid increases in [CO 2 ] s . FIGURE 3 | Pressures at surface (P0; hPa) and at a depth of 90 cm (P90; hPa) measured by the BAT ® geosystem and the pressure difference between P 0 and P 90 (ΔP P 90 -P 0 ; hPa) during the period Ⅰ.
Frontiers in Earth Science | www.frontiersin.org January 2021 | Volume 8 | Article 599388 Besides, it was observed that the relative humidity (RH) in the chamber (RH s ) and from the AWS (RH aws ) were positively related with FCO 2 during the period Ⅱ (r 0.7 and 0.4 respectively in Table 5) unlike the other periods, probably due to the precipitation as well as earthquake (see Environmental Parameters).

Source of CO 2
Scatter plots of soil gas compositions revealed that the soil gas was the mixture of atmospheric air and geogenic CO 2 ( Figure 5). Specifically, the CO 2 vs. O 2 was plotted on the air-CO 2 mixing line ( Figure 5A), while the N 2 vs. CO 2 was not on the biological respiration or CH 4 oxidation ( Figure 5B). Besides, the consistent δ 13 C [CO2]v regardless of the season and the δ 13 C [CO2]v ranges ( Figure 5D) indicated a deep-seated CO 2 source despite the different proportion of CO 2 in soil gas at each period as the δ 13 C CO2 of geogenic CO 2 is reported to be −6‰ in Baines and Worden (2004) and −9.7 ∼ −2.7‰ (−6.5 ± 2.5‰) in Sano and Marty (1995). In contrast, the δ 13 C CO2 from the microbial decomposition of C3 plants generally ranges between −34 and −23‰ (Faure, 1998) and seasonally changes due to the change in biological activities (White and Corfield, 2006;Zhu et al., 2019). Figures S2-S4), and had a significantly negative relation with barometric pressure (Tables 2, 3, 5), which also indicates the geogenic CO 2 emission in the study site (Camarda et al., 2019). At the measurement site around M17, the geological FCO 2 seems to exceed the microbial respiratory FCO 2 by several orders of magnitude as in the sites with high CO 2 concentrations in Vodnik et al. (2009).

FCO 2 did not show the diurnal variation (Supplementary
During the period Ⅰ, P 90 was always higher than P 0 (Figure 3). The positive ΔP indicates that the atmospheric air did not intrude and dilute [CO 2 ] v at the measurement point (M17v). Instead, the mixing with air seemed to be diffusive at M17v due to concentration gradients with depth, which may be affected by the barometric pressure given that P 90 was synchronized with P 0 (Figure 3). Besides, the geogenic CO 2 uprising could be mixed with soil gas influenced by the atmospheric air in the vadose zone.
According to Massmann and Farrier (1992) and Chen et al. (2020), the changes in barometric pressure migrate air into the vadose zone. Air intrusion due to the barometric pressure fluctuation depends on soil characteristics, the thickness of vadose zone, and climate (Massmann and Farrier, 1992;Auer et al., 1996).
Meanwhile, δ 13 C [CO2]a was widely ranged particularly during the period Ⅰ ( Figure 4B) probably due to various sources of CO 2 in the air 1 m above the surface. A CO 2 source in the air seemed to be geogenic given high [CO 2 ] a values up to 0.2% (Table 1), the positive relation between [CO 2 ] a and δ 13 C [CO2]a (r 0.7 in Table 2; Figure 4B), and the significant relations between air compositions and FCO 2 ( Table 2).

Factors to Control Geogenic CO 2 Flux
The variation in geogenic FCO 2 depends on soil properties (e.g., air permeability and diffusion coefficient), prevailing mechanisms of CO 2 transport (e.g., advection and diffusion), environmental parameters (e.g., air pressure and temperature), and deep-seated CO 2 supply via endogenous processes (e.g., volcano and tectonic activity). We discussed the factors controlling the temporal variations of FCO 2 in relation to time-variant geogenic CO 2 supply due to exploitation of CO 2 -rich water (Time-Variant Supply of Geogenic CO 2 ), the effect of environmental variables including precipitation (Environmental Parameters), and the prevailing mechanisms of CO 2 transport (Prevailing Mechanisms of CO 2 Transport). Besides, the difference between the periods Ⅰ and Ⅲ was discussed with respect to the spatial variability of FCO 2 (Heterogenous Transport of Geogenic CO 2 ). Soil properties could not be investigated in this private land.

Time-Variant Supply of Geogenic CO 2
The geogenic CO 2 supply seemed to be variant with time given the irregular increases of [CO 2 ] s (i.e., CO 2 concentration in the chamber 2 min after closing the chamber in Figure 2) regardless of environmental variables (Tables 2, 3). According to , CO 2 gas in the study area forms by degassing from the water table of a CO 2 -rich aquifer and transports toward the M17 site and leaks at M17 via the voids surrounding the well w-2. The amount of CO 2 degassing is expected to be time-variant because the CO 2 -rich water is taken from w-2 by countless people, which changes the groundwater level and subsequently the fluid pressure, affecting the amount of CO 2 degassing (Duan and Sun, 2003 Figure S5E) and consequently FCO 2 with the time lag of −16 h (Table 4) during the period Ⅰ.   The impact of pressure variation on degassing was quantitatively compared to that of temperature and salinity in Table 6 because the saline fluid in the deep part can rise temperature and salinity as well as pressure of groundwater. The salinity was assumed to be low (100-150 mg/L) because the CO 2 -rich water in the study area was characterized by low pH and electrical conductivity (EC), probably due to short (<35 years) reaction times between water and rocks despite a large amount of CO 2 inflow into the aquifer (Kim et al., 2008;Chae et al., 2016). According to Chae et al. (2016), the  Frontiers in Earth Science | www.frontiersin.org January 2021 | Volume 8 | Article 599388 temperature and salinity (EC) of the CO 2 -rich water from w-2 varied between 11.6 and 15.8°C and between 108 and 175 μS/cm for 14 months, respectively. Table 6 shows that the change in CO 2 solubility (i.e., degassing) is more influential by the pressure variation than by the temperature or salinity in the study site. When the water is pumped and thus the water pressure drops, gas bubbles occurs in CO 2 -rich water. Subsequently the gas release ceases and the gas dissolved in the CO 2 -rich water is being released as a result of a slow diffusion process (Kouznetsov et al., 2002). The "bubbly" stage may be resumed at shaking (see Possible Causes of a High CO 2 Emission During the Period Ⅱ). The changes in the water table also affect FCO 2 in the vadose zone as in Schroder et al. (2017) who found the FCO 2 distribution shifted between two CO 2 release tests due to the changes in groundwater depth in wet and dry season. The low water table in dry season facilitates lateral CO 2 migration, reducing FCO 2 along a wellbore. The quantitative assessment of the effect of pressure variation caused by water usage on the CO 2 degassing (and subsequently on FCO 2 ) in the study area remains future work due to little information on the water table and the amount of water consumption at w-2.
In addition, Chae et al. (2016) speculated dry CO 2 flowing directly into the aquifer of w-2 (80 m depth) from the magmatic origin, since the study site is located at the geologic boundary between gneiss and granite ( Figure 1A) and unknown fractures may exist around the study area given the small earthquakes ( Figure 1B) and CO 2 -rich waters (Supplementary Figure S1), although none has been reported within 13 km around the study area. According to Kerrick and Caldeira (1998), the plutonicmetamorphic belt can be a source area to emit CO 2 to the atmosphere. In fact, Yu et al. (2015) observed the temporal variation of CO 2 supply to the aquifer for 48 h in a CO 2 -rich spring (s-2 in Figure 1A), which decreased pH and increased total dissolved inorganic carbon (TDIC) and δ 13 C TDIC and might further cause the temporal variation of [CO 2 ] v , [CO 2 ] s , or FCO 2 .

Environmental Parameters
We found the significant influence of air temperature (T aws , T s ) and pressure (P aws , P s ) to FCO 2 at all periods as the previous studies (Chiodini et al., 1998;Granieri et al., 2003;Carapezza et al., 2011;Camarda et al., 2016;Morita et al., 2019). The temperature was correlated with FCO 2 probably due to the effect of temperature to soil gas diffusivity near the M17 site, and not due to the biological effect in November (late fall during the period Ⅰ) and December (winter during the period Ⅲ) in South Korea (Camarda et al., 2016 and references therein). In particular, the high [CO 2 ] v at M17v during the period Ⅲ suggested the accumulation of geogenic CO 2 in the subsurface due to low soil gas diffusion at cold weather.
T s and P s were highly related with T aws and P aws respectively (r ≥ 0.8 at all periods), while RH s was significantly with RH aws (r 0.7 at all periods) despite the 10 km distance of AWS from the study area. These high correlations between measurements imply the usefulness of the environmental data from the chamber in LI-COR and the AWS data. A multiple regression line using T aws and P aws for FCO 2 during the periods Ⅰ and Ⅲ was obtained respectively as: which had r 2 0.65 and p 0.00 in Equation 3 and r 2 0.53 and p 0.00 in Equation 4, supporting the significant effects of T aws and P aws to FCO 2 . However, microclimates may greatly affect some environmental parameters. For instance, the wind speed (WS) from the AWS did not influence FCO 2 (Tables 2, 3, 5), probably because WS, which is sensitive to topography (Helbig et al., 2016), was not measured near the measurement point. Table 4 shows the negative correlation of FCO 2 with WS at the time lag of −18 h. According to Carapezza et al. (2011), the negative effect of WS reflects that the gas is confined underground under strong wind conditions. In addition to T s and P s , [H 2 O] s (i.e., water vapor mole fraction in the chamber 2 min after closing the chamber), showed significant correlations with FCO 2 (r ≥ 0.5) at all periods, while with [CO 2 ] s only during the period Ⅱ (r 0.8 in Table 5). [H 2 O] s increased with T s and T aws but decreased with P s and P aws at all periods as FCO 2 . Similarly, Schroder et al. (2016), Schroder et al. (2017 found some degree of spatial correlation between FCO 2 and surface H 2 O fluxes and suggested that water vapor can be used as a proxy for escaping gas in some leak scenarios. Moreover, Schroder et al. (2016) discussed the possible source of the water, e.g., the same CO 2 reservoir or a function of the CO 2 passage through the water table, and proposed that the water is sourced from the same CO 2 reservoir based on the slightly elevated temperature of the upwelled water compared with surrounding groundwater in the Qinghai research site, whereas the CO 2 efflux transported H 2 O from the top of the water table at or marginally above the release well at the Ginninderra controlled release facility, Australia (Schroder et al., 2017). Besides, Zhou et al. (2013) suggested that the released CO 2 not only depletes soil O 2 but also enhances evaporation and reduces the soil moisture. On the other hand, the relative humidity (RH) was positively related with FCO 2 during the period Ⅱ (e.g., r 0.7 of RH s and r 0.4 of RH aws in Table 5) but negatively with FCO 2 during the period Ⅰ (Table 2), while not related with FCO 2 during the period Ⅲ (Table 3). The different correlation between measurements suggests the different major mechanism for CO 2 transport, e.g., low gas diffusion at cold weather during the period Ⅲ and the earthquake during the period Ⅱ (see Possible Causes of a High CO 2 Emission During the Period Ⅱ). Similarly, Zhou et al. (2013) found the opposite relationship between the soil O 2 concentration and soil moisture for the during-release and post-release due to the pumping effect of the released CO 2 gas plume at the interface between the CO 2 plume and the soil liquid water. Besides, the period Ⅱ had precipitation. The precipitation may cause the CO 2 gas uprising, filling the pore with water (e.g., Johnson and Rostron, 2012;Garcia-Anton et al., 2014), although the opposite effect of precipitation is also possible, dissolving the CO 2 gas in infiltrating water (Annunziatellis et al., 2008) or reducing the gas permeability (e.g., Carapezza et al., 2011;Garcia-Anton et al., 2014).

Prevailing Mechanisms of CO 2 Transport
The high FCO 2 and its high correlation with P aws indicate that the study site has a non-negligible advective component (Hernández et al., 2001;Ascione et al., 2018). Similarly,  showed that the M17 site was located in the high-advection zone using the relationship between FCO 2 and [CO 2 ] s suggested by Jung et al. (2014). The negative correlation between P 90 and [CO 2 ] v ( Table 2) and the positive ΔP (1.0-1.3 hPa in Table 1; Figure 3) and thus the pressure gradient in the range of 1.1 and 1.4 hPa/m during the period Ⅰ also suggest the advective flow upward in the unsaturated soil. Takle et al. (2004) showed that pressure differences between −15 and 15 Pa at depths of 0-60 cm caused FCO 2 exceeding diffusional fluxes due to pressure pumping. According to Altevogt and Celia (2004), the leakage rate of 0.1 g/m s to 100 g/m s were reached at a CO 2 source with the vertical pressure gradient of 0.18-49.65 hPa/m adjacent to the source boundary.
However, FCO 2 was affected by air temperature and [H 2 O] s as well as air pressure in Tables 2, 3, 5, implying that the diffusive transport of CO 2 is also significant in the study site (Camarda et al., 2019). Accordingly, temperature gradient might affect the pressure gradient of 1.1 and 1.4 hPa/m during the period Ⅰ, while the gradient was not measured. In addition, [CO 2 ] v showed different temporal changes from FCO 2 during the period Ⅰ in Figures 2E, 4A and its correlation with FCO 2 increased at the time lag of −16 h in Table 4. Moreover, soil CO 2 seems to mainly form through degassing of a CO 2 -rich aquifer and the degassing may vary depending on the pressure variation in the aquifer for w-2 (Table 6).
Thus, it can be concluded that the advection is dominant at the near surface, while the transport of geogenic soil CO 2 is diffusivedominated in the vadose zone by the CO 2 gradient. Similarly, Kim et al. (2018) observed that CO 2 concentrations measured at 15 cm depth were significantly lower than those measured at 60 cm depth, as the CO 2 gas escaped quickly into the atmosphere at the ground surface due to the atmospheric pressure effect at an inject test, for which approximately 1.8 t CO 2 was injected at 2.5 m depth with a CO 2 release rate of 6 L/min. Rillard et al. (2015) showed that the advective flux between an injection point and the surface through a preferential path was the dominant gas transport process during the injection phase because it was difficult to avoid a slight overpressure at the injection point. Altevogt and Celia (2004) determined the diffusive flux as well as the slip and Darcy fluxes associated with natural CO 2 leakage into the vadose zone based on a two-dimensional numerical model, and showed that the mole fraction-driven flux played an important role in the development of the CO 2 plume even in situations where pressure-driven advection was the dominant flux mechanism.
Heterogenous Transport of Geogenic CO 2 FCO 2 was in different ranges ( Table 1) and had different correlations with measurements during the periods Ⅰ and Ⅲ (Tables 2, 3) probably because the measurement points were not exactly the same but at the 30 cm separation between M17 (period Ⅰ) and M17-1 (period Ⅲ), implying the spatial variability of FCO 2 , as reported by other researchers including Annunziatellis et al. (2008) and Ascione et al. (2018). Similar to this study result, Annunziatellis et al. (2008) mentioned a test showing a change of more than one order of magnitude over only 30 cm. However, the spatial difference in FCO 2 at the centimeter scale was not expected before the study area, and the extreme spatial variability of FCO 2 needs to be further studied in the study area.
Note that the different FCO 2 between November (period Ⅰ) and December (period Ⅲ) cannot be explained by the seasonal variation, given the similar FCO 2 between August (summer) in 2017 and November (fall) in 2018 despite distinct climate conditions. We acknowledge that the low FCO 2 might reflect the weather condition given the low temperature (T s ) and high pressure (P s ) during the period Ⅲ (Figure 2) and their positive and negative influence on FCO 2 respectively (Tables 2, 3). However, the different ranges in FCO 2 as well as in [H 2 O] s and [CO 2 ] s between the periods Ⅰ and Ⅲ in Figure 2 imply that the effect of heterogeneous CO 2 transport in the vadose zone is stronger than the weather effect. In addition, the FCO 2 estimated using Equation 3 was much higher than the measured FCO 2 during the period Ⅲ (Supplementary Figure S6).

Possible Causes of a High CO 2 Emission During the Period Ⅱ
The FCO 2 value of 1,073 g/m 2 /d was exceptionally high, exceeding the μ + 2σ of the period Ⅰ ( Figure 6). Besides, FCO 2 measurements were much higher than the estimations using Equation 3, in particular until November 25, 2018 ( Figures 7E,F), implying the effect of factors other than T aws and P aws . Meanwhile, [CO 2 ] v rapidly decreased after the earthquake. This anomalously high FCO 2 peak and sudden decreases in [CO 2 ] v during the period Ⅱ cannot be explained by the environmental parameters given similar climate conditions (Table 1), and imply a high CO 2 emission, causing that FCO 2 was negatively related with [CO 2 ] v (r −0.5 in Table 5). In addition, [CO 2 ] s was highly correlated with FCO 2 (r 0.9), and the residual FCO 2 filtered by T aws and P aws was also well correlated with [CO 2 ] s (Supplementary Figure S7; r 0.7). According to Camarda et al. (2019), the relationship between FCO 2 and environmental parameters depends on the amount of deep CO 2 supply as well as the prevalent process of CO 2 transport. Table 5 shows that the FCO 2 variations brought by [CO 2 ] s , RH s and RH aws increased after the earthquake.

Emission Scenario
We suggest that vibrations caused by the earthquake induced the soil gas to transport to the surface, with rapidly decreasing [CO 2 ] v and increasing [CO 2 ] s and FCO 2 since the study site is located between the geologic boundary ( Figure 1) and unknown fracture may exist, while CO 2 gas probably forms by degassing of a CO 2rich aquifer and leaks through the well (w-2) casing in the shallow (<80 m) subsurface given no high flux except M17 near w-2. Earthquake might increase air permeability by a change in site features or modification of the structural parts of the well (w-2) from which the CO 2 originates. Besides, mechanical processes, including vibrations induced by an earthquake, have been known to increase degassing of dissolved gas and to enhance the movement of gas bubbles in fractured aquifers (Kouznetsov et al., 1998;Toutain and Baubron, 1999;Manga et al., 2012), changing the physicochemical parameters of water (e.g., EC, pH, temperature, water level). According to Kouznetsov et al. (1998), degassing is related to local instantaneous ruptures in the formation fluid due to the effect of elastic waves. Nuclei of bubbles are formed in these ruptures, and gas diffusion from fluid into these bubbles takes place (Kouznetsov et al., 1998). Crews and Cooper (2014) also showed that seismic waves initiated bubble nucleation and growth in groundwater, which increased the water level in boreholes, reducing effective stress in critically loaded geologic faults, and consequently induced secondary earthquakes. Fischer et al. (2017) observed the CO 2 bubbles increasing in a CO 2 -rich well water 4 days after a 3.5 M earthquake occurring 9 km away. The high CO 2 bubble concentrations lasted for 150 days.
We acknowledge that it is difficult to determine whether the high FCO 2 and low [CO 2 ] v were caused by the earthquake because of the short-term and discontinuous data with only a small earthquake. Besides, this observation was opposite to the jumps of both FCO 2 and [CO 2 ] v in soil gas wells during the seismic activity in the active fault zones (Chen et al., 2020). We did not observe the CO 2rich water to support the emission scenario during this study. However, the impact of the small earthquake cannot be excluded for the high FCO 2 during the period Ⅱ, given that FCO 2 measured immediately after the earthquake was beyond the seasonal and diurnal variation of FCO 2 at M17 ( Table 1) and much higher than that estimated using T aws and P aws (Figure 7). Moreover, it should be noted that we began to measure FCO 2 12 h after the earthquake occurred, and we might miss higher values given a synchronous sharp increase of seismicity and FCO 2 in a seismically active area (Camarda et al., 2016) and the velocity of P (7-8 km/s) and S wave (4-5 km/s).

Suggestion of an Earthquake Precursor
We noted that [CO 2 ] v rapidly increased to be 48.9% on 22:00 November 2, 2018, and then the maximum [C O 2 ] s was observed 12 h later ( Figures 2C, 4A), approximately 16 days before the earthquake ( Figure 6). Besides, FCO 2 had an increasing trend during the period Ⅰ (r 2 0.3 in Figure 6), although the average FCO 2 (564 g/m 2 /d) was similar to 546 g/m 2 /d obtained at M17 in August 2017 (Table 1). These temporal variations might be a precursor of the earthquake. We acknowledge that many researches have been carried out on the precursors of earthquakes, while there is no general agreement among scientists on the earthquake precursors (Tsunogai and Wakita, 1995;Hernández et al., 2001;Pérez et al., 2008;Ingebritsen and Manga, 2014). Besides, the changes in soil CO 2 have been reported as a result of earthquakes or volcanic activities, rather than an earthquake precursor (Hernández et al., 2001;Troll et al., 2012). Moreover, only one earthquake case was observed in this study (Supplementary Figure S2).
However, many studies suggested the soil gas to be one of the most reliable tools to investigate earthquake precursory signals (Walia et al., 2010;Sciarra et al., 2017). For instance, Sciarra et al. (2017) suggested that crustal dilation linked to seismic activity favors the uprising of geogas toward the surface. Walia et al. (2010) showed that the spatial distribution of soil gases was useful in identifying tectonic systems since it showed a clear anomalous trend along the Hsinhua Fault. Chiodini et al. (2004) found in the Apennine that the anomalous FCO 2 suddenly disappeared in a narrow band with the seismicity concentrated, and suggested that the gas accumulates in crustal traps at depth, generating CO 2 overpressurized reservoirs, which induce seismicity.

CONCLUSION
Temporal variations of soil CO 2 flux (FCO 2 ) and soil CO 2 concentration ([CO 2 ] v ) were investigated for three periods to recognize the factors controlling the temporal variation of geogenic FCO 2 in a non-volcanic and seismically inactive area. The periods Ⅰ (November 2 to 5, 2018) and Ⅲ (December 2 to 8, 2019) were to assess the baseline, while the period Ⅱ (November 19, 2018 to January 30, 2019) was to survey the effect of a small (2.1 M) earthquake occurring 7.8 km away. The correlation coefficients indicated that the air pressure was the most significant controlling factor to FCO 2 regardless of the periods, and the air temperature was also noteworthy.
In contrast, some environmental parameters were significantly related with FCO 2 during one or two periods only, e.g., [CO 2 ] s and [CO 2 ] v during the period Ⅱ. In particular, the low [CO 2 ] v but high FCO 2 during the period Ⅱ implied the high emission of soil CO 2 after the small earthquake, which affected the relations between some environmental parameters (e.g., [CO 2 ] s , RH s and RH aw ) and FCO 2 . Meanwhile, the low FCO 2 during the period Ⅲ suggested the heterogenous subsurface conditions for CO 2 transport at the centimeter scale, while the high [CO 2 ] v implied the accumulation of soil CO 2 in the subsurface due to low soil gas diffusion at cold weather. Based on the high FCO 2 , its high correlation with air temperature as well as air pressure, and the different temporal changes of [CO 2 ] v from FCO 2 including the high [CO 2 ] v at cold weather, the study area seemed to have the diffusive transport of soil CO 2 dominant in the vadose zone, while the advection near the surface.
The merit of this study is to present the temporal variation of high FCO 2 of deep CO 2 origin in a non-volcanic and seismically inactive area and to discuss the controlling factors, given few studies on the temporal changes of geogenic CO 2 emissions in a geologically stable region, although CO 2 -rich water discharges. In particular, we found a rapid increase of FCO 2 after the small earthquake, which implies that the global natural CO 2 emission can be larger than the previous estimation. In addition, artificial vibrations (e.g., building construction and transportation vibration) may enhance natural CO 2 emissions, and thus CO 2 -rich waters or FCO 2 should be monitored to assess the effect of artificial and natural vibrations to CO 2 emissions. Besides, we provided the usefulness of data in the chamber and AWS data to understand the temporal variation in FCO 2 .
We acknowledge however that we only observed a period (Ⅱ) with respect to an earthquake because the study area has low and sporadic seismicity and no automated FCO 2 monitoring system. It was difficult to determine a cause for the high FCO 2 peak and decreases in [CO 2 ] v mostly because of short-term and discontinuous monitoring at different acquisition intervals. Thus our speculation about the effect of the small earthquake to abnormal increases in FCO 2 and the applicability of [CO 2 ] v as an earthquake precursor needs to be confirmed through a physics-based numerical modeling work and long-term monitoring data. A process-based understanding for the effects of earthquakes to FCO 2 and [CO 2 ] v remains future work. Besides, we could not clearly explain the irregular temporal variations of [CO 2 ] s and its high correlation with FCO 2 during the period Ⅱ. Thus, the carbon isotopic compositions of [CO 2 ] s is also needed to be investigated in the next study. With defining end-member properties, the proportion of geogenic CO 2 in [CO 2 ] s should be assessed to verify the temporal variation of geogenic CO 2 supply with [CO 2 ] s . Lastly, the low FCO 2 during the period Ⅲ implied the high spatial variability of FCO 2 . A spatially intensive FCO 2 investigation close to M17 will be conducted in the near future to address the reason for heterogeneity in the centimeter scale.
Based on the temporal changes in FCO 2 in this non-volcanic and seismically inactive study area, we suggest to install an automated FCO 2 monitoring system in natural emission sites to understand temporal increases in natural CO 2 emissions and their causes (e.g., earthquake) in geologically stable regions and consequently the global natural CO 2 emission based on long-term monitoring data. In particular, the FCO 2 monitoring should be complemented with the monitoring of degassing from groundwater to assess the impact of tectonic stresses because endogenous factors affect the physicochemical parameters of water as well, which subsequently changes CO 2 concentrations and FCO 2 .

DATA AVAILABILITY STATEMENT
The datasets presented in this study can be found in online repositories. The names of the repository/repositories and accession number(s) can be found below: https://zenodo.org/ record/3818108#.X0c3DOR7k6Y.

AUTHOR CONTRIBUTIONS
CK conducted the field work, wrote a first draft, and made figures. SY interpreted the data, wrote the manuscript, and made figures. Y-YO interpreted the temporal data and suggested the cross-correlation analysis. GC designed and initiated the field work, and acquired and interpreted the data. S-TY contributed to the conception and design of the work. YS supervised the field work and sample analysis. All authors contributed to discussions and revisions of the manuscript.