Effect of Pressure Perturbations on CO2 Degassing in a Mofette System: The Case of Hartoušov, Czech Republic

Mofettes are gas emission sites where high concentrations of CO2 ascend through conduits from as deep as the mantle to the Earth’s surface and as such provide direct windows to processes at depth. The Hartoušov mofette, located at the western margin of the Eger Graben, is a key site to study interactions between fluids and swarm earthquakes. The mofette field (10 mofettes within an area of 100 m × 500 m and three wells of 28, 108, and 239 m depth) is characterized by high CO2 emission rates (up to 100 t/d) and helium signatures with (He/He)c up to 5.8 Ra, indicating mantle origin. We compiled geological, geophysical, geochemical, and isotopic data to describe the mofette system. Fluids in the Cheb basin are mixtures between shallow groundwater and brine (>40 g/L at a depth of 235 m) located at the deepest parts of the basin fillings. Overpressured CO2-rich mineral waters are trapped below the mudstones and clays of the sealing Cypris formation. Drilling through this sealing layer led to blow-outs in different compartments of the basin. Pressure transients were observed related to natural disturbances as well as human activities. External (rain) and internal (earthquakes) events can cause pressure transients in the fluid system within hours or several days, lasting from days to years and leading to changes in gas flux rates. The 2014 earthquake swarm triggered an estimated excess release of 175,000 tons of CO2 during the following four years. Pressure oscillations were observed at a wellhead lasting 24 h with increasing amplitudes (from 10 to 40 kPa) and increasing frequencies reaching five cycles per hour. These oscillations are described for the first time as a potential natural analog to a two-phase pipe–relief valve system known from industrial applications.


Introduction
Mofettes are gas emission sites where high concentrations of carbon dioxide (CO 2 ) ascend together with other volatiles through conduits from as deep as the mantle to the Earth's surface. As such, they provide direct windows to metamorphic and magmatic processes at depth [1]. Despite the obvious potential, systematic studies of mofettes to monitor magmatic processes at depth are rare [2]. Often, observed anomalies and transients are difficult to decipher. For instance, transients in fluid properties of mofettes may occur long after seismo-magmatic events have been indicated by, e.g., earthquake swarms [3]. It is not really known how long it takes for CO 2 to ascend through the mantle and crust, and where models have been suggested, the propagation times vary extremely between them [4]. According to other reports, gas flow transients occur at the surface without relative to sub-continental lithospheric mantle (SCLM) composition. With respect to carbon, a limestone contribution cannot be excluded without analyzing rocks. Both the total gas discharge and the mantle contributions decrease with increasing distance away from the degassing center [18]. This study focuses on the Hartoušov mofette field, a key site to investigate the interaction between fluids and swarm earthquakes within the framework of the ICDP-Eger project. The objectives are related to the monitoring setup, the CO2 degassing occurrence, the hydrochemical properties of the fluids at this site, as well as anomalies detected during the first period of observations. We discuss their origin to establish a basic This study focuses on the Hartoušov mofette field, a key site to investigate the interaction between fluids and swarm earthquakes within the framework of the ICDP-Eger project. The objectives are related to the monitoring setup, the CO 2 degassing occurrence, the hydrochemical properties of the fluids at this site, as well as anomalies detected during the first period of observations. We discuss their origin to establish a basic understanding of mantle derived mofettes, providing a unique window to deep geodynamic processes.

Methods and Fluid Monitoring Setup
At Hartoušov, geochemical and geophysical fluid properties are monitored continuously in a set of adjacent boreholes at three different depth levels and at the surface ( Figure 2; for details, see Fischer et al. [23] and Woith et al. [24]): F1 (aka H1-31b), F2 (aka HJB-1), and F3 (aka HJB-2) were drilled to depths of 28.2, 108.5, and 239.3 m, respectively. This multi-level setup is designed to provide insights into the cause for temporal variations, possibly related to the admixture of crustal or meteoric fluids to an otherwise steady mantle degassing.
understanding of mantle derived mofettes, providing a unique window to deep geodynamic processes.

Methods and Fluid Monitoring Setup
At Hartoušov, geochemical and geophysical fluid properties are monitored continuously in a set of adjacent boreholes at three different depth levels and at the surface ( Figure 2; for details, see Fischer et al. [23] and Woith et al. [24]): F1 (aka H1-31b), F2 (aka HJB-1), and F3 (aka HJB-2) were drilled to depths of 28.2, 108.5, and 239.3 m, respectively. This multi-level setup is designed to provide insights into the cause for temporal variations, possibly related to the admixture of crustal or meteoric fluids to an otherwise steady mantle degassing. In addition to the continuous monitoring of fluid properties, discrete gas and water samples were systematically collected in situ. Gas samples were analyzed for their chemical (CO2, N2, O2, Ar, He, CH4, CxHy, and H2) and isotopic compositions (noble gases, CO2 and H2O). Details on sampling methods and analytical procedures along with the first results of the gas components are provided by [22]. Radon was measured directly in the field using a scintillation counter as described in [25].
Water temperature, specific electrical conductivity, pH, redox potential, and dissolved oxygen were measured on site. The measured redox value was converted to redox potential EH with respect to a standard hydrogen electrode. The HCO3 − concentration was determined by titration with HCl on site. Water samples for anion and cation determinations were filtered with cellulose-acetate filters (pore size 0.2 μm). Two aliquots of 50 mL each were collected; one was acidified with concentrated HNO3 (150 μL per 50 mL) to determine cations, and the other aliquot remained non-acidified to determine anions. Analyses of cation and anion concentrations were performed at the ElMiE-lab (Elements and Minerals of the Earth) at the GFZ, Potsdam. Samples were diluted using Bidest water and ultra-pure reagents. Cations were measured using a 5110 ICP OES (Agilent, Santa Clara, California, USA) in the acidified samples. Concentrations were calculated by external calibration using single element standards. Intermediate In addition to the continuous monitoring of fluid properties, discrete gas and water samples were systematically collected in situ. Gas samples were analyzed for their chemical (CO 2 , N 2 , O 2 , Ar, He, CH 4 , C x H y , and H 2 ) and isotopic compositions (noble gases, CO 2 and H 2 O). Details on sampling methods and analytical procedures along with the first results of the gas components are provided by [22]. Radon was measured directly in the field using a scintillation counter as described in [25].
Water temperature, specific electrical conductivity, pH, redox potential, and dissolved oxygen were measured on site. The measured redox value was converted to redox potential E H with respect to a standard hydrogen electrode. The HCO 3 − concentration was determined by titration with HCl on site. Water samples for anion and cation determinations were filtered with cellulose-acetate filters (pore size 0.2 µm). Two aliquots of 50 mL each were collected; one was acidified with concentrated HNO 3 (150 µL per 50 mL) to determine cations, and the other aliquot remained non-acidified to determine anions. Analyses of cation and anion concentrations were performed at the ElMiE-lab (Elements and Minerals of the Earth) at the GFZ, Potsdam. Samples were diluted using Bidest water and ultra-pure reagents. Cations were measured using a 5110 ICP OES (Agilent, Santa Clara, CA, USA) in the acidified samples. Concentrations were calculated by external calibration using single element standards. Intermediate measurement precision is generally within 5%, similar to repeatability, which is regularly tested on in-house standards. Anion concentrations were measured with a Dionex Ion Chromatograph (IC; Thermo Scientific, Waltham, MA, USA) in the non-acidified aliquots. Anions were eluted over a Dionex IonPac AG22 guard and column, which were calibrated using single element standards.
Mineral saturation indices (Table A1) were calculated using PHREEQC [26]. The saturation index (SI) indicates that a fluid is undersaturated (SI < 0), saturated (SI = 0), or oversaturated (SI > 0) with respect to a mineral phase. In other words, a negative SI of a given mineral in, e.g., groundwater will facilitate dissolution of that mineral. A neutral or positive SI indicates that a mineral may remain in equilibrium or precipitate, respectively.

Hydrogeology of Hartoušov Mofette and Cheb Basin
The Hartoušov mofette field is located in the eastern part of the Cheb basin. It comprises distinct wet and dry mofettes, and diffusive degassing within an area of about 0.35 km 2 , with gas flux hot spots of more than 50 kg/d/m 2 (compare red areas in Figure 3).
Geosciences 2022, 12, x FOR PEER REVIEW 5 of 23 measurement precision is generally within 5%, similar to repeatability, which is regularly tested on in-house standards. Anion concentrations were measured with a Dionex Ion Chromatograph (IC; Thermo Scientific, Waltham, Massachusetts, USA) in the nonacidified aliquots. Anions were eluted over a Dionex IonPac AG22 guard and column, which were calibrated using single element standards. Mineral saturation indices (Table A1) were calculated using PHREEQC [26]. The saturation index (SI) indicates that a fluid is undersaturated (SI < 0), saturated (SI = 0), or oversaturated (SI > 0) with respect to a mineral phase. In other words, a negative SI of a given mineral in, e.g., groundwater will facilitate dissolution of that mineral. A neutral or positive SI indicates that a mineral may remain in equilibrium or precipitate, respectively.

Hydrogeology of Hartoušov Mofette and Cheb Basin
The Hartoušov mofette field is located in the eastern part of the Cheb basin. It comprises distinct wet and dry mofettes, and diffusive degassing within an area of about 0.35 km 2 , with gas flux hot spots of more than 50 kg/d/m 2 (compare red areas in Figure 3).  [15], light gray colors indicate a CO2 flux above 25 g/d/m 2 ; red colors mark regions with a CO2 flux >50 kg/d/m 2 ), the location of the ICDP research boreholes (F1, F2, and F3), wet and dry mofettes according to [12], historic springs, boreholes (the latter two from GDO (geologically documented objects) database of the Czech Geological Survey), and shallow drillholes (from [27]). (A) Position of SW-NE and NW-SE profiles used for the stratigraphic and hydrogeological profiles shown in

Geophysical Investigations
During the preparation phase of the drillings, a series of geophysical investigations dedicated to obtaining information about the shallow underground beneath the Hartoušov mofette field were performed. Sandig et al. [28] combined soil gas measurements (CO2, Rn) along a 500 m long profile-parallel to the SW-NE profile shown in Figure 3A at a distance of about 20 m to the north of F3-with electrical resistivity tomography and self-potential methods. High CO2 concentrations were related to positive self-potential anomalies, low soil moistures, and high resistivity, as well as high δ 13 C values and increased Rn concentrations (see Figure 4A).  [15], light gray colors indicate a CO 2 flux above 25 g/d/m 2 ; red colors mark regions with a CO 2 flux >50 kg/d/m 2 ), the location of the ICDP research boreholes (F1, F2, and F3), wet and dry mofettes according to [12], historic springs, boreholes (the latter two from GDO (geologically documented objects) database of the Czech Geological Survey), and shallow drillholes (from [27]).

Geophysical Investigations
During the preparation phase of the drillings, a series of geophysical investigations dedicated to obtaining information about the shallow underground beneath the Hartoušov mofette field were performed. Sandig et al. [28] combined soil gas measurements (CO 2 , Rn) along a 500 m long profile-parallel to the SW-NE profile shown in Figure 3A at a distance of about 20 m to the north of F3-with electrical resistivity tomography and self-potential methods. High CO 2 concentrations were related to positive self-potential anomalies, low soil moistures, and high resistivity, as well as high δ 13 C values and increased Rn concentrations (see Figure 4A).   Table 2 and Figure 8a in Umlauft and Korn [30]).  [28]. Greenish colors indicate δ 13 C values typical for organic material, whereas red to brownish colors indicate mantle origin. (B) NW-SE profile; blue stars mark a channel-like structure obtained from matched field processing of seismic noise data (see Table 2 and Figure 8a in Umlauft and Korn [30]).
Sandig et al. [28] further noted that CO 2 ascends from deep geological sources via preferential pathways and accumulates in coarser sediments. Flechsig et al. [27] observed high CO 2 soil gas concentrations in areas where the shallow subsurface sediments display elevated contents of C org and pyrite.
Seismic noise measurements were applied to detect fluid channels below the Hartoušov mofette using matched field processing techniques [30,31]. Flores Estrella et al. [31] observed changes in the spatial degassing patterns from day to day. Umlauft and Korn [30] located channel-like structures below the known mofettes (an example from the northern mofette field is shown in Figure 4).

Hydrogeologic and Isotopic Inventory
The Hartoušov mofette field is located in the protection zone of natural healing resources of Františkovy Lázně, just 8 km to the west of Hartoušov village. The hydrogeological, hydrogeochemical, and geochemical inventory of the Hartoušov mofette field is summarized in Table 1 and Figures 4-6. Data from Soos Císařský pramen, Kyselecký hamr, and Františkovy Lázně are also presented here and used for comparison (chemical and isotopic data of Hartoušov mofette HM, Hartoušov F1, and Hartoušov F2 are taken from Daskalopoulou et al. [22]). Information about the analytical processes used for the determination of the chemical and isotopic gas components and the uncertainties for all the presented (published and unpublished) data can be found in [22].  [22]. The abbreviations n.a. and b.d.l. stand for "not analyzed" and "below detection limit", respectively. A pumping test performed immediately after drilling F2 revealed a Na-HCO 3 -SO 4type groundwater with a total mineralization of 6300 mg/L, a pH of 6.67, and a dissolved CO 2 concentration of 1628-1892 mg/L. Highly mineralized (40,900 mg/L) groundwater of Na-SO 4 -HCO 3 -Cl type-which is similar to waters of Františkovy Lázně-was discovered in the crystalline rocks in the F3 borehole. A dissolved CO 2 concentration of <90 mg/L was determined for F3, compared to 2580 mg/L for Františkovy Lázně.
For the Cheb basin, three groups of groundwater types can be distinguished on the basis of major ion concentrations ( Figure 5A). The first type (Ca-HCO 3 ) is represented by the waters of Hartoušov F1 and the natural mofettes HM, Bublák, and Smrčina. These waters have acidic pH values between 3.9 and 5.3 along with low conductivities (<400 µS/cm). Hartoušov F3, Soos Císařský pramen, and Soos mofette, as well as Hartoušov mofette H5 belong to the second type of waters (Na-Cl-SO 4 ) with a pH between 5.4 and 6.1, and conductivities between 1849 and 6950 µS/cm. F3 is characterized by high conductivities (43,000 µS/cm) and elevated temperatures of 20.5 • C. Hartoušov F2, Kyselecký hamr, and Wettinquelle (located in Bad Brambach) belong to the third type of waters (mixed Ca-Na-HCO 3 ) and are characterized by pH values between 5.8 and 6.7 and conductivities between 1880 and 7850 µS/cm. Immediately after the drilling of F2 in 2016, the sampled water was of Na-Cl-SO 4 type (see symbol labeled F2 in Figure 5A), while newer samples plot in the field of the third group. Figure 5B, where Na + is plotted vs. Cl − concentrations, reveals that the fluids of the Cheb basin plot on a mixing line between a shallow water end-member F1 and the brines of F3. The Cl − concentrations range from 10 mg/L to nearly 5000 mg/L (for comparison, seawater has typically 19,400 mg/L Cl − ). The water of mofette H5 has a specific electrical conductivity of 1849 µS/cm. This is too high for a classical wet mofette, which is usually defined as a site where CO 2 degasses through surface water (e.g., puddles, ponds, rivers, lakes). Further, as can be seen from the diagram shown in Figure 5B, the water of H5 is a mixture of a shallow-water end-member (like F1) and a deep-water end-member (F3). The same applies to the mofettes in Soos.  Evaporites (gypsum, anhydrite, halite) are undersaturated, which reflects their absence in the study area. In contrast, both dolomite and calcite are commonly found in the western Cheb basin [33]. The negative SI, indicating undersaturation with respect to carbonate minerals, can be attributed to the natural upwelling of CO2-rich subsurface fluids. Note that shortly after drilling, the waters of F2 and F3 were both oversaturated with respect to calcite and dolomite. Silica minerals (chalcedony and quartz) are oversaturated in these groundwaters, and hence play an important role in regulating the groundwater chemistry.
CO2 is the dominant gas species in the Hartoušov mofette field, with concentrations that exceed 99%. The remaining gas components comprise mostly N2 and O2, which are usually below 0.9%, while Ar, He, CH4, and CxHy are found in trace amounts. Hydrogen  [32] showing element ratios in mmol(eq)%, (B) Na + vs. Cl − concentrations plot on a mixing line between the end-members F1 and F3.
Evaporites (gypsum, anhydrite, halite) are undersaturated, which reflects their absence in the study area. In contrast, both dolomite and calcite are commonly found in the western Cheb basin [33]. The negative SI, indicating undersaturation with respect to carbonate minerals, can be attributed to the natural upwelling of CO 2 -rich subsurface fluids. Note that shortly after drilling, the waters of F2 and F3 were both oversaturated with respect to calcite and dolomite. Silica minerals (chalcedony and quartz) are oversaturated in these groundwaters, and hence play an important role in regulating the groundwater chemistry. CO 2 is the dominant gas species in the Hartoušov mofette field, with concentrations that exceed 99%. The remaining gas components comprise mostly N 2 and O 2 , which are usually below 0.9%, while Ar, He, CH 4 , and C x H y are found in trace amounts. Hydrogen is often below the detection limit [12,18,21,22]. It is worth noting that the amount of atmospheric gases varies in correspondence to the depth of the sampling site; the deeper the sampled gas is found, the lower the air content is.
The chemical and isotopic compositions mentioned above are considered background values as they occur in seismically quiescent periods. Variations in gas flow, water temperature, and seismic activity may induce transient changes in the chemical and isotopic composition of the gases. For instance, a significant increase in the gas flow was observed at Hartoušov after the 2014 earthquakes [34]. An increase in the 3 He/ 4 He ratio, and thus of the SCLM contribution from 38% in 1993 up to 89% in 2016, was also observed in the eastern Cheb basin [21]. Moreover, during the seismically quiescent drilling period of F3, gas flow changes at F1 resulted in fluctuations in the chemical composition of the gases (solubility differences) and may have led to higher δ 13 C CO2 values [22].
Higher CO 2 concentrations with respect to Hartoušov are observed at Soos Císařský pramen and Kyselecký hamr. These elevated CO 2 contents are mirrored by the depleted chemical composition of the minor components. For instance, He in Hartoušov mofette HM is 20.3 µmol/mol, while in Soos Císařský pramen, it is 0.50 µmol/mol (Table 1). Isotopically, gas within the Soos area has a lower mantle He contribution (~52% considering an SCLM end-member), while CO 2 indicates a mixed mantle-limestone origin. The sketch in Figure 6 refers to the situation before the F2 borehole had to be redesigned due to safety issues in 2019 (compare to the recent technical setup in Figure 2). Since the drilling of the F2 well in April 2016, the wellhead pressure remained rather constant around 500 kPa. In October 2016, an additional pressure sensor was installed at a depth of 94 m, showing a pressure of 900 kPa. From this, we concluded that the well was partly filled with water (40 m of water above the borehole sensor) and that the headspace The sketch in Figure 6 refers to the situation before the F2 borehole had to be redesigned due to safety issues in 2019 (compare to the recent technical setup in Figure 2). Since the drilling of the F2 well in April 2016, the wellhead pressure remained rather constant around 500 kPa. In October 2016, an additional pressure sensor was installed at a depth of 94 m, showing a pressure of 900 kPa. From this, we concluded that the well was partly filled with water (40 m of water above the borehole sensor) and that the headspace was occupied by a gas column of 50 m.

Effect of Natural Pressure Perturbations on CO 2 Degassing
In the following, we compare the impact of rain and earthquakes on the degassing of the mofette system. While rain is affecting the surface area and has presumably only short-term effects, it is crucial to distinguish external signals from those caused by the underground processes.

Rain-Induced Pressure Transients
Heavy rain on 24 May 2018 caused flooding of the Hartoušov meadow, resulting in a water level increase of 50 cm in the mofette HP (mofette with an iron pipe located about 10 m SSE of the wooden hut in Figure 7B) and an abrupt increase in the gas flow from 7 to 17 L/min in borehole F1. Contemporarily, the water level in the F1 well dropped by 10 cm ( Figure 7A). The F2 borehole reaction (not shown) was very weak compared to F1. High water level in the nearby meadow likely closed small escape paths for gas, causing almost three times higher measured gas flow through the well and a corresponding response of wellhead pressure (compare similar shapes of green and blue lines in Figure 7A). Two small negative anomalies on all three measured values cannot be resolved, limiting possible interpretations, due to the fast and turbulent flooding behavior that exceeds minimum sampling intervals of 10 min. A possible cause might be the closing and re-opening of gas leakage pathways.

Earthquake-Induced Gas Flow Transients
The most pronounced anomalies of gas flow in Hartoušov were observed during the 2014 earthquake swarm, when the CO2 flow in the F1 borehole (which was the only one existing at that time) increased from the pre-seismic rate of less than 10 L/min to more than 40 L/min four months after the beginning of the swarm [23,34]. The flow rate started to rise four days after the first swarm event, which was also the strongest one (ML 4.2) with a hypocenter depth of 9 km and an epicentral distance from Hartoušov of 10 km. After reaching a maximum, the flow slowly decreased for the following four years, reaching pre-seismic levels in 2018. This behavior is consistent with Sibson's fault valve model [35], suggesting that the earthquake rupture breached a sealing layer at the hypocenter depth, which released fluid from a relatively low-permeable lower crust [34]. Similar behavior was observed during the 2008 earthquake swarm, which occurred at the High water level in the nearby meadow likely closed small escape paths for gas, causing almost three times higher measured gas flow through the well and a corresponding response of wellhead pressure (compare similar shapes of green and blue lines in Figure 7A). Two small negative anomalies on all three measured values cannot be resolved, limiting possible interpretations, due to the fast and turbulent flooding behavior that exceeds minimum sampling intervals of 10 min. A possible cause might be the closing and reopening of gas leakage pathways.

Earthquake-Induced Gas Flow Transients
The most pronounced anomalies of gas flow in Hartoušov were observed during the 2014 earthquake swarm, when the CO 2 flow in the F1 borehole (which was the only one existing at that time) increased from the pre-seismic rate of less than 10 L/min to more than 40 L/min four months after the beginning of the swarm [23,34]. The flow rate started to rise four days after the first swarm event, which was also the strongest one (M L 4.2) with a hypocenter depth of 9 km and an epicentral distance from Hartoušov of 10 km. After reaching a maximum, the flow slowly decreased for the following four years, reaching pre-seismic levels in 2018. This behavior is consistent with Sibson's fault valve model [35], suggesting that the earthquake rupture breached a sealing layer at the hypocenter depth, which released fluid from a relatively low-permeable lower crust [34]. Similar behavior was observed during the 2008 earthquake swarm, which occurred at the same fault patch as in 2014. It is of interest to mention that the average flow rate for the 4-year decay period was about three times the pre-seismic level. Applying this to the whole Hartoušov mofette field and considering an average daily release of 60 tons of CO 2 [19], an excess of 175,000 tons of CO 2 released by the 2014 earthquake swarm from the lower crust during the four years decay period is estimated for this mofette field.
The fault valve hypothesis requires the opening of a valve in the hypocentral region and the subsequent diffusion of a pressure front through the crust. Another possible explanation for the increase in the gas flow would be a dynamically triggered response of the local hydrogeological system to the passing seismic waves. Since the observed increase in flow rate was not very sharp, it is difficult to determine the exact beginning of the pressure transient. We revisited the 2014 event and applied a statistical test to determine the breaking point(s) in the gas flow data using the code of Zeileis et al. [36]. The result is shown in Figure 8 together with the daily gas flow rate data from F1 and the seismic energy density for the site Hartoušov calculated according to Wang and Manga [37]. (1) re-adjustment of the head gradient due to increased permeability by removing bubbles from the pore space (see Roeloffs [38]), (2) advective overpressure due to rising of the gas bubbles [39], and (3) triggering of the degassing of local oversaturated fluids. All three options are feasible given the hydrogeological situation at Hartoušov mofette. Similar observations were described and modeled from an artesian well in Armenia, which responded repeatedly to distant earthquakes (see Woith et al. [40], Wang et al. [41]). Additionally, the shape of the anomalies including the recovery phase is similar to the ones described in Fischer et al. [34]. The differences in the time scales-several months vs. several years for the recovery-might be related to different permeabilities of both aquifers: the Armenian site is characterized by high permeability due to abundant fractured rocks, whereas Hartoušov is characterized by relatively low permeabilities within sedimentary and metamorphic rocks.
At this stage, we do not have a clear favorite. Both mechanisms, the fault valve and the dynamic interaction of seismic waves with the local aquifer, are feasible scenarios. Both of them need a certain threshold in terms of earthquake magnitude and distance. No earthquake with a magnitude above ML 4 occurred since the ML 4.2 event of 2014. (1) re-adjustment of the head gradient due to increased permeability by removing bubbles from the pore space (see Roeloffs [38]), (2) advective overpressure due to rising of the gas bubbles [39], and (3) triggering of the degassing of local oversaturated fluids. All three options are feasible given the hydrogeological situation at Hartoušov mofette. Similar observations were described and modeled from an artesian well in Armenia, which responded repeatedly to distant earthquakes (see Woith et al. [40], Wang et al. [41]). Additionally, the shape of the anomalies including the recovery phase is similar to the ones described in Fischer et al. [34]. The differences in the time scales-several months vs. several years for the recovery-might be related to different permeabilities of both aquifers: the Armenian site is characterized by high permeability due to abundant fractured rocks, whereas Hartoušov is characterized by relatively low permeabilities within sedimentary and metamorphic rocks.
At this stage, we do not have a clear favorite. Both mechanisms, the fault valve and the dynamic interaction of seismic waves with the local aquifer, are feasible scenarios. Both of them need a certain threshold in terms of earthquake magnitude and distance. No earthquake with a magnitude above M L 4 occurred since the M L 4.2 event of 2014.

Pressure Transients Related to Stimulation Measures of Well F3
Borehole F3 was drilled in August 2019 to a final depth of 239.2 m with PQ3 drill rods (hole size 122.6 mm, core diameter 83 mm). Borehole logging was carried out after the termination of the drilling. Thereafter, HRQ V-wall drill rods were installed as casing (OD 88.9 mm, ID 81.0 mm, at joints ID 77.8 mm) from 0 to 235 m. The total volume is about 1232 L, or 5.15 L/m. The casing was cemented from the bottom to the surface. Thus, it was necessary to perforate casing and cement at the target depth with explosives. A series of stimulation measures aimed to access a CO 2 -bearing horizon at depth have been performed since 2020, which included a first perforation, two draw-down tests, two airlifts to clean the well, and a second perforation. A complete timeline of the stimulations and the wellhead pressure evolution at F3 is shown in Figure 9.

Pressure Variations after Perforations
The first attempt to connect the well to the gas reservoir at depth was performed on 15 January 2020. In total, two rounds of perforation shots were fired (gun type TTG Link 45 mm, 8 g RDX DP, 20 spm). Shot #1.1 at a depth of 228.5 m took place at 08:29 UTC and consisted of 24 g explosives (three pieces of 8 g). A small fountain (some dm high) occurred at the wellhead. Well-logging was performed at 09:09 UTC (downwards) and 09:30 (upwards). Shot #1.2 was fired at 10:11 UTC again with 24 g explosives at a depth of 229 m. Well-logging was repeated at 11:07 (downwards) and 11:28 UTC (upwards).
The second attempt was performed on 19 January 2022. Before the perforation, the well depth and access were checked by MND company. The measured depth of the well was 232 m, which agrees with a depth of 232.5 m estimated from a camera inspection of the borehole in April 2021. In total, three perforation shots were taken at different depths; the wellhead pressure response is shown in Figure 10A. After each shot, the wellhead was closed with a slider valve, so that the pressure increase could be monitored with a pressure transducer mounted at the wellhead below the sliding valve.
Shot #2.1 consisted of 80 g explosives (10 pieces of 8 g), covering a depth span of 231.0-230.5 m, at 09:55:40 UTC (the exact time was obtained from the seismometer recordings at F1 and F2). A small geyser with an ~10 min delay occurred. The well was closed and the pressure raised up to 50 kPa. Thereafter, the well was inspected by a

Pressure Variations after Perforations
The first attempt to connect the well to the gas reservoir at depth was performed on 15 January 2020. In total, two rounds of perforation shots were fired (gun type TTG Link 45 mm, 8 g RDX DP, 20 spm). Shot #1.1 at a depth of 228.5 m took place at 08:29 UTC and consisted of 24 g explosives (three pieces of 8 g). A small fountain (some dm high) occurred at the wellhead. Well-logging was performed at 09:09 UTC (downwards) and 09:30 (upwards). Shot #1.2 was fired at 10:11 UTC again with 24 g explosives at a depth of 229 m. Well-logging was repeated at 11:07 (downwards) and 11:28 UTC (upwards).
The second attempt was performed on 19 January 2022. Before the perforation, the well depth and access were checked by MND company. The measured depth of the well was 232 m, which agrees with a depth of 232.5 m estimated from a camera inspection of the borehole in April 2021. In total, three perforation shots were taken at different depths; the wellhead pressure response is shown in Figure 10A. After each shot, the wellhead was closed with a slider valve, so that the pressure increase could be monitored with a pressure transducer mounted at the wellhead below the sliding valve. The maximum amplitude of the wellhead pressure changes is obviously related to the amount of used explosives as shown in Figure 10B. The explosive RDX, also called hexogen, has a normal gas volume of 927 L/kg. Thus, during the three shots, gas volumes of about 75, 150, and 300 L were generated inside the well casing. The pressure pulse traveled at velocities of the order of 0.3 m/s when correcting the perforation depth for the expanded volume of the explosive ( Table 2). The molecular formula of hexogen is C3H6N6O6, which might explain the presence of hydrocarbons in a gas sample that had been taken about one month after the perforation. The sample collected from F3 was analyzed in situ for its elemental composition with the use of a quadrupole mass spectrometer (QMS). The instrument was previously calibrated. The results showed that CO2 is the dominant gas species with 93.5%, followed by N2 and O2 with 2.3%. Methane and hydrogen displayed concentrations up to 1.9% and 2.1%, respectively, while Ar and He were present in minor amounts.
This chemical composition is quite abnormal considering that gas from F3 is expected to be the least air-contaminated of the three boreholes and the natural mofette (surface expression). In addition to this, CH4 and H2 concentrations are comparatively high. The enhanced amount of H2 may be related either to low temperature interactions between groundwater, rock fragments, and steel detritus from the casing or to H2 dissolution from the surrounding strata [42]. Moreover, CH4 might have been generated by anaerobic bacterial consumption of H2 in the borehole slurry and by oxidation of the steel casing [43]. The elevated content of CH4 may also be related to C3H6N6O6 decomposition [44].

Pressure Recovery in a Closed Hole Following Draw-down Tests
Following the first perforation, two draw-down tests were performed. The aim was to lower the water level in order to trigger degassing of dissolved CO2 due to the reduced  F1 and F2). A small geyser with an~10 min delay occurred. The well was closed and the pressure raised up to 50 kPa. Thereafter, the well was inspected by a camera-bubbles with foam were visible at the water table and down to about 80-100 m. Visibility in the well was good to~200 m; below that, the visibility was reduced due to whirling particles in the vicinity of the explosion. When going up with the camera, a weakened bubbling on the water level was noticed.
Shot #2.2 consisted of 160 g explosives (20 pieces of 8 g) covering a depth range of 207-206 m, at 12:27:10 UTC. After the shot, we waited for water to flow out, then the well was closed. Pressure raised up to 150 kPa, but then decreased quickly-also due to a leaking slider valve at the wellhead.
The last shot #2.3 consisted of 320 g explosives (40 pieces of 8 g), covering a span of 188.5-186.5 m, at 13:27:15 UTC. Water started to leak from the well almost immediately after the shot. The slider valve was closed and a wellhead lid was also attached for better sealing. Pressure increased to more than 250 kPa and stabilized at 200 kPa almost 2 h after the shot.
The maximum amplitude of the wellhead pressure changes is obviously related to the amount of used explosives as shown in Figure 10B. The explosive RDX, also called hexogen, has a normal gas volume of 927 L/kg. Thus, during the three shots, gas volumes of about 75, 150, and 300 L were generated inside the well casing. The pressure pulse traveled at velocities of the order of 0.3 m/s when correcting the perforation depth for the expanded volume of the explosive (Table 2). The molecular formula of hexogen is C 3 H 6 N 6 O 6 , which might explain the presence of hydrocarbons in a gas sample that had been taken about one month after the perforation. The sample collected from F3 was analyzed in situ for its elemental composition with the use of a quadrupole mass spectrometer (QMS). The instrument was previously calibrated. The results showed that CO 2 is the dominant gas species with 93.5%, followed by N 2 and O 2 with 2.3%. Methane and hydrogen displayed concentrations up to 1.9% and 2.1%, respectively, while Ar and He were present in minor amounts.
This chemical composition is quite abnormal considering that gas from F3 is expected to be the least air-contaminated of the three boreholes and the natural mofette (surface expression). In addition to this, CH 4 and H 2 concentrations are comparatively high. The enhanced amount of H 2 may be related either to low temperature interactions between groundwater, rock fragments, and steel detritus from the casing or to H 2 dissolution from the surrounding strata [42]. Moreover, CH 4 might have been generated by anaerobic bacterial consumption of H 2 in the borehole slurry and by oxidation of the steel casing [43]. The elevated content of CH 4 may also be related to C 3 H 6 N 6 O 6 decomposition [44].

Pressure Recovery in a Closed Hole following Draw-down Tests
Following the first perforation, two draw-down tests were performed. The aim was to lower the water level in order to trigger degassing of dissolved CO 2 due to the reduced water column. The first one (compare "draw-down #1" in Figure 9) was performed on 6 February 2020. A pump (Grundfos MP1, diameter 4.6 cm) was installed at a depth of 80 m. The water level could be lowered from 5 m to 15 m below surface (measured manually with a water level contact gauge) before the pump failed due to problems with fuse/power supply. A second draw-down test was made on 10 March 2020. Again, the pump was installed 80 m below the surface. The test lasted 33 min (from 16:05 to 16:38 local time), and the estimated pumped volume of water was about 386 L. The steel casing has an inner diameter of 81 mm, which gives a volume of 5.15 L/m of casing; water level was lowered from 5 m to 80 m, i.e., by 75 m. This converts to a pumping rate of about 11.7 L/min. A retry to pump water again after waiting half an hour was not successful, because water inflow into the well was too slow. After both draw-down tests, the wellhead was closed tightly, and the wellhead pressure monitored with a pressure transducer mounted at the wellhead increased from 0 to 68 and 126 kPa, respectively.
The observed pressure increases are likely not related to an increased CO 2 degassing, but simply due to the compression of the air in the headspace of the well while the water level was recovering to the pre-test level of 5 m below the surface as the following estimates suggest: When the wellhead was still closed, a pressure of 16 kPa was measured inside the headspace, and the barometric pressure was 98 kPa. Thus, the absolute initial pressure was p 0 = 114 kPa = 114 kJ/m 3 within an initial volume V 0 = 5 m * 5.15 L/m = 0.02575 m 3 , given p*V = constant = 2.941 kJ. During the test, the water level was at 15 m, i.e., V 1 = 0.07725 m 3 , and p 1 = 2.941 kJ/0.07725 m 3 = 38 kJ/m 3 = 38 kPa. This is calculated to a pressure increase of p = p 0 − p 1 = 114 kPa − 38 kPa = 76 kPa, which is close to the observed value of 68 kPa. Accordingly, the theoretical pressure increase for the second draw-down test is calculated to 115 kPa. Whether the difference of 11 kPa between prediction and observation is due to an additional pressure build-up from degassed CO 2 remains unclear.

Water Level Recovery in the Open Well following Airlift Operations
In order to clear the bottom of the borehole from drillmud and sediments which entered the borehole after the perforations, two airlift operations were conducted on 10 September 2020 and 3 February 2021. In both cases, slim steel tubing was installed inside the borehole casing to the bottom of the well. Compressed air was then pumped down inside the spaghetti tubes with a pressure of 2 MPa, pushing the water towards the surface. In the beginning, the water was muddy, containing minor amounts of drillmud and later phyllitic material. Finally, clear water was raised to the surface. During the airlift operation in 2021, pumping started at a depth of 185 m. Then, the depth was lowered to 203 m and finally to 227 m, where the jet became stuck. At the end of the first attempt, the rise in the water level was measured, indicating an inflow rate of about 3.4 L/min, compared to about 2.1 L/min during the 2020 airlift. Meanwhile, the drillers had removed, cleaned, and reinstalled the pipes. A second attempt started at a depth of 206 m, continued at 227 m, and finally at 230 m, producing clear water with phyllitic particles. From the recovery data (for details of the recovery tests, see Figure A1), rough estimates of the transmissivity T could be obtained using the Cooper-Jacob method [45], indicating values of T of the order of 5-9 × 10 −8 m 2 /s. For F2, a transmissivity of T = 8 × 10 −6 m 2 /s had been determined. Thus, the transmissivity of F2 is about two orders of magnitude higher than that of F3. For comparison, aquifers suitable for water exploitation have transmissivities greater than 0.015 m 2 /s [46].

Pressure Drops and Recovery after Drilling F2
The wellhead pressure at F2 was about 500 kPa since the installation of a pressu transducer (KELLER PR-33X, 10 bar, sampled at 1 min intervals) on 19 May 2016. In Ju 2016, the pressure repeatedly dropped to values below 100 kPa ( Figure 12A). On 3 Ju 2016, the pressure dropped within a few hours to 50 kPa and remained between 50 a 100 kPa for more than two days before returning to the pre-event pressure level. Su pressure excursions occurred again on 24 and 27 July 2016. The events do not correl with rainfall or local earthquakes as recorded by the Czech WebNet.
Details of the third event, which started on 27 July 2016, are shown in Figure 1 Within several hours, the pressure dropped to below 100 kPa, before slowly recovering almost 540 kPa on the morning of 29 July. The pressure started to oscillate with increasi amplitudes (from 10 to 40 kPa) and varying frequencies until the early morning of 30 Ju

Pressure Drops and Recovery after Drilling F2
The wellhead pressure at F2 was about 500 kPa since the installation of a pressure transducer (KELLER PR-33X, 10 bar, sampled at 1 min intervals) on 19 May 2016. In July 2016, the pressure repeatedly dropped to values below 100 kPa ( Figure 12A). On 3 July 2016, the pressure dropped within a few hours to 50 kPa and remained between 50 and 100 kPa for more than two days before returning to the pre-event pressure level. Such pressure excursions occurred again on 24 and 27 July 2016. The events do not correlate with rainfall or local earthquakes as recorded by the Czech WebNet.

Tentative Hydro-Mechanical Model
Within the Cheb basin, four main CO2 degassing centers can be distinguished: Soos, Bublák, Hartoušov, and the area SE of Františkovy Lázně. The existing data and models do not allow us to identify the position of the supply paths of CO2 from the lithospheric mantle [48]. However, analysis of the water chemistry revealed that the fluids of Františkovy Lázně, Hartoušov F2, Kyselecký hamr, and Soos are mixtures between shallow groundwater and brine located at the deepest parts (>230 m below surface) of the segmented basin fillings.
CO2 eruptions (blow-outs) related to drilling activities were repeatedly reported, e.g., several erupting wells NE of Františkovy Lázně (boreholes HJ-11 and HJ-9 [49], as well as borehole HJ-1). In particular, the documentation of the gas eruption at the HJ-1 well near Spálená (W of Nový Kostel) during the drilling of the CO2 outlet zone in the Cypris formation testifies to the fact that other tectonically based structures than just the abovementioned contacts in the basin floor and faults are responsible for the distribution of CO2. The eruption of the H-11 well near Horní Ves (E of Františkovy Lázně) in 1957 [49] occurred during the excavation of the well in the underlying clays of the layered zone, an eruption of a mixture of water and gas with a strong mechanical admixture (sand, coal) with an estimated total yield of both phases of approx. 250 L/s. A secondary accumulation of gasified water was hit by the well, and the eruption affected the regime of some mineral water sources in Františkovy Lázně. The impact of the eruption was mainly caused by the drop in CO2 pressure in the individual collectors. The difference in the reaction of the resources of the Františkovy Lázně outflow area to the eruption of the H-11 well near Horní Ves and to the eruption of the HJ-1 well near Spálená confirms the hypothesis about the hydraulic role of the Nebanice-Soos ridge in the basin floor.
In the following, we present a tentative hydro-mechanical model of the Hartoušov mofette system made of pipes and relief valves. The stratigraphic analysis of cores showed a layer of higher permeability below 80 m depth, which is sealed by 60 to 70 m clays and mudstones of the Cypris formation on top of it. The permeable layer consists of porous Details of the third event, which started on 27 July 2016, are shown in Figure 12B. Within several hours, the pressure dropped to below 100 kPa, before slowly recovering to almost 540 kPa on the morning of 29 July. The pressure started to oscillate with increasing amplitudes (from 10 to 40 kPa) and varying frequencies until the early morning of 30 July ( Figure 12C). Thereafter, the pressure reached again the pre-event level of 500 kPa.
The most striking feature of the observed pressures is the reliability with which the pressure is kept around the level of 500 kPa, particularly when compared to the three anomalies, when the pressure was lowered as much as below 100 kPa, only to return over 1-3 days to the previous level again. Particularly interesting, then, is the oscillatory behavior after the third anomaly. Qualitatively, similar oscillations can be observed in two-phase pipe-relief valve systems in industrial applications, e.g., in hydropower stations (see engineering application presented by Wang et al. [47]). Nevertheless, the oscillations at Hartoušov lasted almost 24 h with changing amplitudes and frequencies (see Figure 12C). In the first eight hours after the beginning of the oscillations, 2-3 cycles per hour (cph) were observed, while in the following six hours, the frequency dropped to about 1 cph before increasing to 5 cph until the oscillations finally stopped after 23 h.

Tentative Hydro-Mechanical Model
Within the Cheb basin, four main CO 2 degassing centers can be distinguished: Soos, Bublák, Hartoušov, and the area SE of Františkovy Lázně. The existing data and models do not allow us to identify the position of the supply paths of CO 2 from the lithospheric mantle [48]. However, analysis of the water chemistry revealed that the fluids of Františkovy Lázně, Hartoušov F2, Kyselecký hamr, and Soos are mixtures between shallow groundwater and brine located at the deepest parts (>230 m below surface) of the segmented basin fillings. CO 2 eruptions (blow-outs) related to drilling activities were repeatedly reported, e.g., several erupting wells NE of Františkovy Lázně (boreholes HJ-11 and HJ-9 [49], as well as borehole HJ-1). In particular, the documentation of the gas eruption at the HJ-1 well near Spálená (W of Nový Kostel) during the drilling of the CO 2 outlet zone in the Cypris formation testifies to the fact that other tectonically based structures than just the above-mentioned contacts in the basin floor and faults are responsible for the distribution of CO 2 . The eruption of the H-11 well near Horní Ves (E of Františkovy Lázně) in 1957 [49] occurred during the excavation of the well in the underlying clays of the layered zone, an eruption of a mixture of water and gas with a strong mechanical admixture (sand, coal) with an estimated total yield of both phases of approx. 250 L/s. A secondary accumulation of gasified water was hit by the well, and the eruption affected the regime of some mineral water sources in Františkovy Lázně. The impact of the eruption was mainly caused by the drop in CO 2 pressure in the individual collectors. The difference in the reaction of the resources of the Františkovy Lázně outflow area to the eruption of the H-11 well near Horní Ves and to the eruption of the HJ-1 well near Spálená confirms the hypothesis about the hydraulic role of the Nebanice-Soos ridge in the basin floor.
In the following, we present a tentative hydro-mechanical model of the Hartoušov mofette system made of pipes and relief valves. The stratigraphic analysis of cores showed a layer of higher permeability below 80 m depth, which is sealed by 60 to 70 m clays and mudstones of the Cypris formation on top of it. The permeable layer consists of porous sandy mudstones of the Main Seam Formation having a thickness of 10 m and 20 m in F2 and F3, respectively. During drilling of F2, a small gas blow-out occurred when the drilling reached the top of the Main Seam formation at a depth of about 80 m. The permeable layer consists of porous sand directly above the weathered phyllitic basement rock. During drilling, it was documented that the local aquifer was under high overpressure and fully saturated by gas and water. We assume a steady supply of CO 2 from depth by a gas reservoir below the local aquifer. The gas reservoir leads to continuous inflow of CO 2 into the saline water. The aquifer itself is therefore assumed to be oversaturated, and the gas is migrating upward through small channels as manifested by the mofettes in Hartoušov. Since gas migration in the sediment and soil column experiences resistance, the system can be described as an overpressure valve. Measurements indicate that the sediment/soil cover can resist an overpressure of about 500 kPa.
The F2 well was dressed with a PVC-U casing with a diameter of DN 114 mm with perforated sections at three depth intervals from 58.5 to 63.5 m, 68.5 to 83.5 m, and 88.5 to 103.5 m. The bottom of the borehole was equipped with a full casing with a length of 5 m. The natural gas migrates into the well to replace any water due to the lighter density. We assume that the water-gas interface in the drilling is at the first open section in about 58 m depth. All inflowing CO 2 escapes at the first open section into the sediments, and the measured overpressure at the well head is relatively constant at 500 kPa. The overpressure in the cased section of the well is controlled by the difference in water level in the drill hole and the hydraulically connected aquifer. From the 500 kPa, we estimate a depth difference of about 50 m. Since the uppermost perforated section is only a few meters below, at 58.5 m, it can be assumed that the local aquifer is hydraulically connected to the near surface ground water. In other words, the wellhead pressure depends mainly on geometrical parameters (e.g., ground water level minus open section of casing), but not on CO 2 flux from depth, which explains why the pressure has been rather stable over time.
In July 2016, just 2 months after the drilling had been completed, we observed a sudden pressure drop. We postulate that the pressure was too high, causing the relief valve to open. The relief valve likely was situated between the casing and the clays of the Cypris formation; the pressure dropped and the valve closed again. This process repeated twice before the valve, i.e., the ring space between casing and wall rock, was finally closed completely by the plastic clay. It is interesting to note that the cores of F2 show dike-like injection structures-also described as injectites [50]-cutting through the host rock at high angles. Bussert and co-workers [29,51] suggest that fluidized sediments were injected into overlying beds, probably caused by over-pressurized fluids.

Conclusions
We compiled geological, geophysical, geochemical, and isotopic data from the Cheb basin with special focus on the Hartoušov mofette field, which is a key site to study the interaction between fluids and swarm earthquakes. Several pressure transients were observed related to natural disturbances as well as human activities. The main findings can be highlighted as follows: -Fluids in the Cheb basin are mixtures between shallow groundwater and brine located at the deepest parts (>230 m below surface) of the segmented basin fillings. -Deepwater components were also found in two wet mofettes (Hartoušov H5 and Soos), which are usually defined as sites where CO 2 degasses through surface water (puddles, ponds, rivers, lakes). -Over-pressured CO 2 -rich mineral waters are trapped below the mudstones and clays of the sealing Cypris formation. Drilling through the sealing layer led to several blow-outs in different compartments of the basin. -External (rain) and internal (earthquakes) events can cause pressure and gas flowrate transients in the fluid system within hours or several days, lasting from days to years. - The transmissivity at F3 (235 m, T = 5-9 × 10 −8 m 2 /s) is about two orders of magnitude lower than at F2 (108 m, T = 8 × 10 −6 m 2 /s). This has implications for the response time to pressure transients at different depth levels of the mofette system. -Pressure oscillations at a wellhead are described for the first time as a potentially natural analog to a two-phase pipe-relief valve system known from industrial applications.
The collected data provide the basis for further research at this unique site to study the role of mantle fluids in the triggering of swarm earthquakes. This work is also part of the research projects "MoRe-Mofette Research" and "MoCa-Monitoring Carbon", which were funded by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation)-419880416, 461419881. We received co-funding from the Swedish National Research 783 infrastructure for scientific drilling Riksriggen at Lund University, Sweden. Data Availability Statement: Data will be made available in the data repository of GFZ at https: //dataservices.gfz-potsdam.de (accessed on 15 December 2022), see Heinicke and Woith [17]. 0 Airlift operations to clean the well F3 were conducted on 10 September 2020 and 3 February 2021. In both cases, the recovery of the water level was monitored manually. Thus, rough estimates of the transmissivity could be obtained using the Cooper-Jacob [44] method ( Figure A1), indicating values of T of the order of 5-9 × 10 −8 m 2 /s. For comparison, aquifers suitable for water exploitation have transmissivities greater than 0.015 m 2 /s [45]. Figure A1. Results of well recovery tests according to Cooper-Jacob after two airlift operations.