Subterranean Karst Environments as a Global Sink for Atmospheric Methane

The air in subterranean karst cavities is often depleted in methane (CH4) relative to the atmosphere. Karst is considered a potential sink for the atmospheric greenhouse gas CH4 because its subsurface drainage networks and solution-enlarged fractures facilitate atmospheric exchange. Karst landscapes cover about 14 % of earth’s continental surface, but observations of CH4 concentrations in cave air are limited to localized studies in Gibraltar, Spain, Indiana (USA), Vietnam, Australia, and by incomplete isotopic data. To test if karst is systematically acting as a global CH4 sink, we measured the CH4 concentrations, δCCH4, and δHCH4 values of cave air from 33 caves in the USA and three caves in New Zealand. We also measured CO2 concentrations, δCCO2, and radon (Rn) concentrations to support CH4 data interpretation by assessing cave air residence times and mixing processes. Among these caves, 35 exhibited subatmospheric CH4 concentrations in at least one location compared to their local atmospheric backgrounds. CH4 concentrations and δCCH4 and δHCH4 values suggest that microbial methanotrophy within caves is the primary CH4 consumption mechanism as the atmosphere exchanges with subsurface air. The pattern of δCCH4 and δHCH4 values along CH4 concentration gradients in cave air provides evidence for incomplete oxidation by methanotrophy. Only 5 locations from 3 caves showed elevated CH4 concentrations compared to the atmospheric background and could be ascribed to local CH4 sources from sewage and outgassing swamp water. Several associated δCCH4 and δHCH4 values point to carbonate reduction and acetate fermentation as biochemical pathways of limited methanogenesis in karst environments and suggest that these pathways occur in the environment over large spatial scales. Our data show that karst environments function as a global CH4 sink. Estimates of CH4 flux in karst landscapes are needed in order to include the subterranean CH4 sink in climate models.


Introduction
Atmospheric methane (CH4) is a greenhouse gas and its concentration is increasing in the atmosphere (Dlugokencky et al., 2011;Sussmann et al., 2012;Ciais et al., 2013). The increase in atmospheric CH4 is due to an imbalance between CH4 sources and sinks. Anthropogenic and natural sources combine to contribute about 680 Tg of CH4 to the atmosphere while reactions with hydroxyl (•OH) and chlorine radicals in the troposphere and stratosphere remove about 600 Tg a −1 (Kirschke et al., 2013). Methanotrophic consumption in soils removes 30 Tg a −1 (Kirschke et al., 2013). The present globally averaged CH4 concentration is 1.87 ppmv which is 2.5 times higher than preindustrial levels (Nisbet et al., 2016). Despite improvements in estimating individual sources and sinks of atmospheric CH4, the associated errors remain large (Kirschke et al., 2013). Recent studies suggest that caves may act as an additional CH4 sink (Mattey et  Caves and associated karst landscapes may be an important overlooked sink for atmospheric CH4 because they are estimated to cover as much as 10 to 20 % of the continental surface with the more precise estimates suggesting about 13.8 % (Palmer, 1991;Ford and Williams, 2007). Karst landscapes are frequently associated with the chemical dissolution of limestones, but can form in any soluble rock body. The resulting caves, solution-enlarged fractures, and internal drainage networks that function to transport mass from high elevations to low elevations also allow for subsurface-surface atmospheric exchange (Kowalczk and Froelich, 2010; Garcia-Anton et al., 2014). The total volume and surface area of karst conduits able to interact with the atmosphere is unknown, in part due to small fractures and the difficulty of imaging the subsurface with geophysical methods. Karst caves, due to their accessibility, provide opportunities for non-invasive, in-situ analyses and sampling.
Cave and karst landscapes form in two common ways, each of which influences karst's capacity to act as a CH4 sink. Epigenic karst forms through the interaction of limestone with carbonic acid derived from the dissolution of atmospheric and soil CO2 into surface waters, and hypogenic caves form when corrosive water from deep sources migrates into and dissolves limestone bedrock. Epigenic caves are more widespread, and atmospheric to subatmospheric CH4 concentrations of 1.8 ppmv to < 0.1 ppmv have been observed in these settings (Mattey et Lennon et al., 2017). In some hypogenic caves be contrast, elevated CH4 concentrations from 2 ppmv to 1 % have been observed in association with CH4-rich springs or seeps related to fluid migration from deep hydrocarbon-bearing sedimentary rocks, i.e. seepage processes that are widespread on Earth (Sarbu et al., 1996;Hutchens et al., 2004;Jones et al., 2012;Webster et al., 2017). The dominance of epigenic karst suggests that karst is likely functioning as a CH4 sink at the global scale, but more observations are needed.
Different hypotheses have been put forward to explain the low CH4 concentrations observed in epigenic cave air. The combination of subatmospheric CH4 concentrations and the stable carbon isotopic ratio of CH4 in the air of caves in Gibraltar led to the hypothesis that microorganisms were responsible for the removal of CH4 (Mattey et al., 2013). In turn, low CH4 concentrations in Spanish caves, in the presumed absence of CH4-consuming (methanotrophic) bacteria, led to the hypothesis that CH4 oxidation was induced by ions and •OH generated by the radioactive decay of radon and daughter nuclides ( should allow for the determination of cave air CH4 sources. The objective of the present work is to extend the karst CH4 dataset and test the hypothesis that karst systems act as a CH4 sink on a global scale. To this aim, we studied CH4 concentration, δ 13 CCH 4 , and δ 2 HCH 4 in cave air from 33 epigenic caves in the USA and three epigenic caves in New Zealand. CO2, δ 13 CCO 2 , and Rn were also measured to support CH4 data interpretation via assessing cave air residence times and mixing processes. Data analysis is focused on determining CH4 concentrations, origin, mixing processes and isotopic fractionations.

Sampling and analyses
Air samples from 16 limestone or dolostone caves in the Appalachian fold and thrust belt, 17 limestone caves in gently warped intracratonic basins of the USA, and 3 caves from the North Island of New Zealand were collected over a timespan of roughly four years ( Fig. 1; Table 1). Cave air was analyzed using in-situ methods and was further sampled for laboratory analysis. In-situ CH4, CO2, and Rn abundance analyses were carried out using a suite of instruments (Table 2). Discrete samples of cave air were collected in preevacuated 50-mL serum vials, in 1 to 3-L Tedlar ® bags, or in 4-L glass bottles. CH4 and CO2 concentrations of discrete samples were measured via gas chromatography.
We assessed cave air mixing processes through the following techniques. A qualitative estimate on cave air residence time was obtained by comparing CH4 to CO2 concentrations at individual locations in each cave. Additionally, we measured the Rn concentrations of caves 32 through 36 to assess the relationship between cave air residence time, CH4 concentrations, and CO2 concentrations. δ 13 CCO 2 data were used to assess the sources of CO2 and thus of air entering the caves. We also assessed the distance from each sampling location to cave entrances as another tool to understand the sources and sinks of CH4 in caves.
CH4 and CO2 concentrations from discrete air samples were measured at Indiana University using a Varian 450 gas chromatograph (GC) (Varian -Agilent Technologies, Palo Alto, California). The GC was fitted with a flame ionization detector (FID) for CH4 and a thermal conductivity detector (TCD) for CO2. Standard gas mixtures from Air Liquide America Specialty Gasses LLC (Plumsteadville, Pennsylvania) were used for 3-point calibration curves to convert signals measured on the GC to concentrations. CH4 standards measured on the GC had errors of ± 5 to ± 14 % of the reported concentrations. CH4 concentrations are reported with the uncertainty associated with the standard curve unless the calculated uncertainty was ≤ 0.1 ppmv. Samples with calculated uncertainties ≤ 0.1 ppmv were assigned uncertainties of 0.1 ppmv based on replicate measurements. The uncertainty associated with standard curves for CO2 concentrations varied from < ± 1 to 5 %. CO2 concentrations were assigned uncertainties based on their associated standard curve.
The stable carbon isotope ratios of CH4 and CO2 and hydrogen stable isotope ratios of CH4 were measured on a ThermoFinnigan Delta Plus XP mass spectrometer in the Stable Isotope Research Facility at Indiana University. Carbon stable isotope ratios are expressed as conventional δ 13 CCH 4 and δ 13 CCO 2 values in ‰ along the scale anchored to Vienna Pee Dee Belemnite (VPDB). Hydrogen stable isotope ratios are expressed as δ 2 HCH 4 values in ‰ along the scale anchored to Vienna Standard Mean Ocean Water (VSMOW). CH4 samples were measured in continuous-flow mode using CH4 preconcentration, cryofocusing (Miller et al., 2002), and a gas chromatography-oxidation/pyrolysis-isotope ratio mass spectrometer (GC-ox/pyr-IRMS) interface. Varying sample extraction times were used to isolate roughly 0.45 and 0.90 μmol of CH4 prior to the introduction of the sample to the GC-ox/pyr-IRMS for analysis of δ 13 CCH 4 and δ 2 HCH 4 values, respectively. In-house CH4 standards methane #3, methane #6, and methane ALM with δ 13  values were calculated using a standard curve that accounted for the peak size of the measurement. Analytical repeatability of internal standards ranged from 0.14 to 0.6 ‰ for δ 13 CCH 4 and from to 7 to 18 ‰ for δ 2 HCH 4 . δ 13 CCO 2 values were measured in continuous-flow mode using a GasBench II inlet (Tu et al., 2001). Measured 13 C/ 12 C ratios of CO2 from cave air were converted to the VPDB scale using a single isotopically characterized in-house standard that has a value of 12.0 ± 0.2 ‰.

Data elaboration and quality control
In-situ measurements were preferentially used when statistically analyzing gas concentration data. When in-situ measurements were not available, concentrations measured on the GC were used in the statistical analyses. Samples and data were screened for quality control by comparing the samples with insitu measurements and visual estimation of the volume of sample bags. If a sample bag had been shown to exhibit a leak for one analyte, data from that sample were discarded. CH4 and CO2 concentrations measured by both GC-FID and FTIR showed strong agreements (GC(CO2) = 0.92 ± 0.04 * FTIR(CO2) + 100 ± 300, r 2 = 0.99, p = 5*10 −19 ; GC(CH4) = 0.7 ± 0.2 * FTIR(CH4) + 0.2 ± 0.2, r 2 = 0.62), and the stable isotopic composition of the samples was not related to their storage time (δ 13 CCH 4 = 0.2 ± 0.3*day −47 ± 2, r 2 = 0.03, p = 0.32; δ 2 HCH 4 = 0.03 ± 0.14*day −96 ± 11, r 2 = 0.005, p = 0.72; δ 13 CCO 2 = −0.05 ± 0.16*day −19.9 ± 1.8, r 2 = 0.006, p = 0.57). In locations where more than one sample was taken with in-situ methods, the values of the samples were averaged. Two different modeling techniques were used to assess trace gas sources and sinks in the studied caves. Keeling plots were used to assess the possibility of a two end member mixing system affecting δ 13 CCO 2 . The stable isotopic composition of CO2 entering the caves (δ 13 Cs) was assessed through equation 1 where δ 13 Cm is the δ 13 CCO 2 of the sample, δ 13 Catm is the δ 13 CCO 2 of the atmosphere, and Fatm is the fraction of atmospheric CO2 in the CO2 concentration of the sample (Peyraube et al., 2016). We used values of −10 ‰ for δ 13 Catm and 400 ppmv for the concentration of atmospheric CO2. Rayleigh distillation models were used as the theoretical basis to examine changes in the stable isotopic composition of CH4 in cave air caused by methanotrophy or •OH oxidation. The δ-value of an isotope system in a chemical compound of interest (e.g., CH4) in cave air can be modeled as where δc is the instantaneous δ-value of a particular isotope system in cave air after partial consumption, δi is the initial δ-value of the isotope system in cave air, f is the fraction of the compound remaining, and α is the kinetic isotope fractionation factor (Mattey et al., 2013).

Results
Each of the 36 caves showed atmospheric to subatmospheric CH4 concentrations in at least one location. Only five locations from three different caves showed elevated CH4 concentrations relative to the surface atmosphere (Fig. 2, Supplemental Table 1). The CH4 concentration in the surface atmosphere at study sites ranged from 1.8 ± 0.3 to 2.8 ± 0.7 ppmv. CH4 concentrations in cave air ranged from ≤ 0.1 ± 0.1 ppmv to 5 ± 1 ppmv, and were generally observed to decrease with the distance from cave entrances (Fig.  3, Supplemental Table 1). Two thirds of the caves where three or more air measurements and distance data were recorded showed decreases in CH4 concentration from cave entrances to interiors. For example, caves 7, 8, and 9 from Kentucky all showed progressive decreases in CH4 concentration from about 2 ppmv at the entrance of the cave, down to zero or near zero ppmv in the more inner rooms (from 2 to 0 in caves 8 and 9 and from 1.9 to 0.3 ppmv in cave 7). Additionally, CH4 concentrations were negatively correlated with CO2 concentrations in cave air following an inverse power law relationship ([CH4] = 17.5[CO2] −0.41 , r 2 = 0.26) (Fig. 2) (Fig. 5). However, many points fell below and to the left of the line representing the theoretical incomplete oxidation of atmospheric CH4 (Figs. 5, 6). When δ 2 HCH 4 and δ 13 CCH 4 values were plotted against each other, many samples clustered tightly near the signature of atmospheric CH4 (Fig. 7). Some points, like those from caves 25 and 26, plotted near the expected trend of partial atmospheric CH4 oxidation by methanotrophy. Other points plotted below or above this trend and indicated that at least two additional sources of CH4 had to be entering the caves.

Subsurface-Surface Atmospheric Exchange
The concentrations and stable isotopic compositions of CH4, CO2, Rn in cave air overlapped and diverged from those of the surface atmosphere. This suggests that atmospheric and cave processes influenced the composition of cave air. The atmospheric CH4 concentration at many study locations was above the globally averaged atmospheric background concentrations (1.87 ppmv CH4; Ciais et al., 2013) probably because they were near roads or pastures with local CH4 sources (Gioli et al., 2012;Harper et al., 2014). Rn concentrations were positively correlated with CO2 concentrations in cave air which agrees with other observations that have shown that CO2 concentrations in cave air track cave air residence time. The majority of cave air samples were depleted in CH4 and enriched in CO2 relative to the atmosphere, pointing to processes like in-situ CH4 oxidation and diffusion of air from the epikarst to decrease CH4 and increase CO2 concentrations (Fig. 2). Additionally, cave air CH4 concentrations generally decreased as the distance from an entrance increased (Fig. 3), and departures from this trend can be explained by fast airflow (caves 13, 25), cave air flowing out of the entrance (cave 26), multiple entrances resulting in multiple flow paths (cave 24), distance scales that were too small to observe a decrease in CH4 concentration (cave 20), or internal CH4 sources (cave 9). The patterns of CH4, CO2, and Rn concentration indicate that CH4 concentrations decrease with cave air residence time.
The concentration and stable isotopic composition of CO2 in cave air show that in addition to an atmospheric end member, CO2 with δ 13 Cs values ranging from −28 to −20 ‰ were entering the caves and that the average source δ 13 CCO 2 value was −23.3 ± 0.5 ‰ (Fig. 4). These observations generally agree with a CO2 source from the soils and an atmospheric end-member. The δ 13  Subatmospheric CH4 concentrations in cave air suggest that CH4 from the outside atmosphere is consumed in caves, and the addition of these observations to the existing datasets shows that karst is behaving as a sink for atmospheric CH4 (Mattey et  In several caves, values of δ13CCH4 and δ2HCH4 show that non-atmospheric CH4 enters caves because they were more negative than those predicted by the oxidation of atmospheric CH4 via methanotrophy or reaction with •OH. Other studies have also shown microbially produced CH4 entering caves (Mattey et al., 2013;Webster et al., 2016). Our data point to the methanogenic sources of acetate fermentation and carbonate reduction (Fig. 7). Mixing between residual atmospheric CH4 after partial methanotophic oxidation and CH4 from acetate fermentation in the soil-epikarst-cave system will cause a decrease in the δ13CCH4 and δ2H CH4 values of cave air compared to the atmospheric oxidation curve. Caves 23, 24, and 26 all appear to be influenced by acetoclastic methanogenesis. CH4 produced from carbonate reduction is inferred to enter the caves based on samples that had δ2HCH4 values that were roughly equal to, or more positive than atmospheric values (Fig. 7). The stable isotopic compositions of CH4 from caves 5, 15, 25, and 27 can be explained by partial methanotrophic consumption of CH4 generated from carbonate reduction with the strongest source signal in cave 5 (Fig. 7). CO2 reduction is typically observed in lake sediments, but has been observed in oxidizing environments such as biological soil crusts in deserts after rain events (Angel et al., 2011). We hypothesize that karst environments, which are less oxidizing, exhibit similar behavior. Our data show that carbonate reduction and acetate fermentation can occur in similar environments over large spatial scales and are not limited to arctic environments.
Sites of methanogenesis in or near the studied cave systems may include waterlogged soils above caves, cave soils themselves, and the epikarst. It is possible that after rain events anoxic micro niches occur in soil, the epikarst, or caves themselves and that the generated CH4 is dissolved and later introduced into caves with drip waters. We confirmed that dissolved CH4 outgasses in drip water of cave 32, which was situated underneath a wetland, by placing our CH4 detecting probe near the water and measuring increased CH4 concentrations in its vicinity. We confirmed that in-situ CH4 production can take place in locally anoxic environments within caves by measuring CH4 concentrations close to a bat guano deposit in cave 25, Tennessee (site 4h; average CH4 concentration = 0.3 ± 0.5 ppmv). Time series measurements near the large bat guano deposit showed that CH4 concentrations oscillated between 0.5 ppmv and 0.1 ppmv over the course of seconds, presumably in response to episodic migration of CH4 bubbles through the moist guano (similar oscillations in ammonia, NH3, were also observed). Additionally, we observed circumstantial evidence for local in-situ CH4 production in cave 3 because measured CH4 concentrations upstream of a restroom in the cave were low, while measured CH4 concentrations downstream of the restroom were enhanced. Our data show that caves are capable of expressing elevated CH4 concentrations Preprints (www.preprints.org) | NOT PEER-REVIEWED | Posted: 3 August 2017 doi:10.20944/preprints201708.0009.v1 due to in-situ CH4 production when accumulations of organic matter, such as guano or plant material, foster methanogenesis or when dissolved CH4 from waters outgasses into cave air. We observed minor amounts of thermogenic CH4 entering in at least one cave. Locally elevated CH4 concentrations in cave 9 (Mammoth Cave, Kentucky) were associated with a known hydrocarbon seep that is also transporting sulfide (Olson, 2013). Elevated CH4 concentrations, thought to derive from thermogenic CH4, have also been observed at sulfidic springs in Cueva de Villa Luz (Webster et al., 2017). Some δ13CCH4 and δ2HCH4 values in cave air, i.e. two of the samples from cave 24, are compatible with thermogenic CH4 isotopic signatures (e.g., Schoell, 1988;Etiope et al., 2009 ; Fig. 7). It cannot be excluded that small fluxes of CH4 from shales or hydrocarbon deposits underlying the limestone are entering caves through natural fractures, but our present isotope data cannot confirm this source for cave 24. The Antes and Utica shales, which contain hydrocarbon gases, are stratigraphically below cave 24 (Coleman et al., 2014), and geologic faults and joints, which are often aligned with caves, may serve as conduits for the flow of hydrocarbons (Powell, 1969;Koša et al., 2003). A confirmation of hydrocarbons entering from deep sources may be obtained through measurements of 'radiocarbon-dead' CO2, Rn, and ethane.

Methane Oxidation Mechanisms
The combination of δ 13 CCH 4 and δ 2 HCH 4 values allow for inferences to be made about how CH4 oxidizing reactions in karst environments. We distinguish between two scenarios with distinct sets of assumptions. Many isotopic compositions of CH4 in this study cannot be accounted for in the first scenario in which we assume that (i) CH4 enters the caves from the atmosphere, through acetate fermentation, and through carbonate reduction, and that (ii) CH4 is removed from cave air through reactions involving the •OH. Conversely, in a second scenario, if it is assumed that (i) sources of CH4 in cave air include the atmosphere, acetate fermentation, and carbonate reduction, and that (ii) CH4 is removed from cave air by methanotrophy, all of the points fall within the plausibility envelope of the model, suggesting that methanotrophy is the mechanism responsible for removing CH4 from cave air (Fig. 7). Consideration of an additional source of thermogenic CH4 (natural gas) from deep geologic sources enlarges the plausibility fields of both prior scenarios to encompass all of the data. Our δ 13 CCH 4 and δ 2 HCH 4 data also agree with observations of δ 13 CCH 4 and δ 2 HCH 4 from a cave in Indiana where it appeared that CH4 from both acetate fermentation and carbonate reduction influenced cave air geochemistry (Webster et al., 2016). Additionally our data resemble an arctic system characterized by acetoclastic and hydrogenotrophic CH4 sources and methanotrophy (McCalley et al., 2014). Our isotopic evidence for in-situ microbial CH4 oxidation in caves is corroborated by recent results from in-situ mesocosm experiments in Vietnam where cave rocks with live microorganisms were shown to consume CH4 even in cases where surface soils were very thin to nonexistent Nguyễn-Thuỳ et al., 2017).

Conclusions
Subterranean karst air generally shows subatmospheric CH4 concentrations. CH4 and CO2 concentrations were negatively correlated in cave air showing that as the residence time of cave air increases the CH4 concentration of cave air decreases. The stable isotopic composition of CH4 in studied caves suggests that CH4 is being oxidized by microbial methanotrophy. This evidence adds to earlier reports that methanotrophy is the mechanism by which CH4 is removed in cave air (Mattey et al., 2013;McDonough et al., 2016;Webster et al., 2016;Lennon et al., 2017). The observations of sub-atmospheric CH4 concentrations in cave air from this study and other studies shows that karst is behaving as a global sink for CH4. CH4 flux data from cave and karst landscapes are needed to estimate the size of the karst sink.
The stable isotopic composition of CH4 in the studied caves suggests that, in addition to atmospheric CH4, at least two additional CH4 sources are present in some caves. We suggest that the sources include CH4 produced from acetate fermentation and from CO2 reduction. These data corroborate recent findings of partially oxidized CH4 entering cave air from acetate fermentation and CO2 reduction in Indiana (Webster et al., 2016). These observations of CH4 production by acetate fermentation and carbonate reduction suggest that both processes happen over a wide scale in the environment.
Cave air CH4 appears to have been isotopically shifted towards more positive δ 13 CCH 4 and δ 2 HCH 4 values due to methanotrophy in the soil-epikarst-cave system. Even though minor CH4 production appears to be taking place in karst environments, CH4 consumption is the dominant process.