Antarctic Permafrost Degassing Revealed By Extensive Soil Gas and Co 2 Flux Survey in Taylor Valley

McMurdo Dry Valleys comprise 10% of the ice-free soil surface areas in Antarctica. Permafrost stability plays an important role in C-cycle as it potentially stores considerable quantities of greenhouse gases. While the geomorphology of the Dry Valleys reects a long history of changing climate conditions, comparison with the rapidly warming Northern polar region suggests that future climate and ecosystems may change more rapidly from permafrost degradation. In Austral summer 2019/2020 a comprehensive sampling of soil gases and CO 2 ux measurements was undertaken in the Taylor Valley, with the aims to identify potential presence of soil gases in the active layer. The results obtained show high concentrations of CH 4 , CO 2 , He and an increasing CO 2 ux rate. We identify the likely source of the gas to be from dissolved gases in deep brine moving from inland (potentially underneath the Antarctic Ice Sheet) to the coast at depth beneath the permafrost layer. t. Our results emphasize that extensive surveys are necessary to properly evaluate greenhouse gas (GHG) emissions in regions with permafrost. We also established the rst extensive baseline maps that can be used to compare and monitor soil gas concentrations and CO 2 in the region. The collected data identied areas characterized by multigas where during where uids with main of the We suggest that the from the that rich in dissolved The supercial melting of the and presence allowed these to at different layers in the surcial environment. Typical average values of soil gas concentration of Ne, O 2 , N 2 , H 2 O, Ar, CO 2 , He, CH 4 , and H 2 in the Taylor Valley (this work), atmospheric air, atmosphere-soil interface and soil gas.


Introduction
Permafrost is any ground (soil or rock and any ice and organic material inclusions) that remains completely frozen (0°C or colder) for at least two years 1 . Its thaw and the microbial decomposition of previously frozen organic carbon is considered one of the most likely positive climate feedbacks from terrestrial ecosystems to the atmosphere in a gradually warming planet 2,3 . Permafrost is present in both hemispheres at high latitudes and its temperature, thickness, and continuity are controlled by the geographic setting and, to a large extent, by the surface energy balance and thus vary strongly with latitude and it is present in both hemispheres at high latitudes 4 . Climate warming effects are going to impact these regions in the upcoming decades 5,6 and all three types of permafrost below 1000 m elevation (dry, ice-cemented, and massive ice) may be susceptible to warming-related degradation depending on future emission pathways, for example through slumping of meltlubricated sediments and surface ablation by sublimation-driven ice removal 7,8 . Measurements of CO 2 and CH 4 soil concentrations and uxes are essential to understand the C cycle in terrestrial ecosystems, although less is known about controls over CO 2 ux (φCO 2 ) in ecosystems lacking vascular vegetation, including polar deserts (such as the McMurdo Dry Valleys, thereafter MDV) and some hot deserts, where autotrophic inputs are low and abiotic factors tend to dominate in determining φCO 2 9 .
In the Arctic and boreal regions, permafrost is found in Greenland, Alaska, Canada, Northern Europe, Russia and China 4 , and represent 22% of the exposed land surface 10 . Studies carried out on permafrost soils in these ecosystems have shown how these areas store almost twice the carbon currently present in the atmosphere 4,11 . These regions are rich in frozen organic matter, that would lead to an increase of the production of CO 2 and CH 4 by microbial activities in case of thawing 4 . Furthermore, part of the released carbon could easily dissolve in water and, through solar radiation, produce CO 2 by the photomineralization process 12 . Large methane deposits currently stored at high latitude regions are either frozen within permafrost or trapped below impermeable buffer zones 13 . Current and future warming will signi cantly affect polar and sub-polar regions triggering the melting of ice-bound sediments, releasing methane in to the Earth's atmosphere. Methane has a global warming potential 28 times higher than that of CO 2 on a 100-year time horizon 14 . It is therefore imperative to provide estimates of methane and other gases released from the highlatitude regions. In remote and scarcely monitored regions soil gas release can endure for decades or even centuries before being detected and quanti ed.
In the Southern Hemisphere, permafrost is found in the Sub-Antarctic islands, in the Antarctic Peninsula, at high elevations, and in the ice-free areas of the Antarctic region. Since Antarctica is much colder than the Arctic, it has limited organic content in soils 4 and its ice-free area represents only 0.35% of the continent 15 , the degradation of permafrost has not been widely studied 16 . Nevertheless, if the temperature warms in Antarctica, the potential total amount of carbon contribution could be signi cant even at a fraction of the Northern Hemisphere. Also, the role of bacteria was in the past underestimated, and is potentially more important in Antarctica than previously believed. Indeed, microbial activity affects the amount of total organic carbon and is more susceptible to weak temperature variations 4 . The MDV are the largest ice-free regions in Antarctica 17 ; their geomorphology reveals how the landscape is strongly controlled by climate processes 18 . Attempts to quantify CO 2 emissions in the Antarctic continent have been carried out in the MDV soils, highlighting that φCO 2 is driven primarily by physical factors such as soil temperature and moisture, indicating that future climate change may alter the soil C cycle 17,19−23 . The lack of mechanistic understanding makes it di cult to predict the contribution of soil φCO 2 to the C-cycle due to climate change in the polar deserts of Antarctica. In the MDV, φCO 2 has been used to characterize a variety of ecosystem processes and properties, including soil C turnover, the functional role of differing origins of organic matter supporting C cycling, and biotic distribution and activity [24][25][26][27][28] . In situ φCO 2 in MDV soils is low and spatially variable 29 , and it is therefore di cult to separate the biological processes (e.g. C-xation) from physical factors (e.g. carbonate dissolution). Parsons et al. 30 hypothesized that in extreme desert environments, abiotic factors, like temperature gradients, parent material and soil water dynamics, may have the same magnitude of the biological processes in uencing φCO 2 rates; on the contrary, in lowest latitude ecosystems the physical φCO 2 is negligible. Recent studies have revealed a diffuse subsurface brine system in the MDV area, occurring preferentially near the coast and under the surface sediments of the main valleys that could be carried from beneath the East Antarctic Ice Sheet 31 . The presence of this deep uid circulation could also promote the uprising of geogenic gases. Soil gas measurements in the MDV were performed by Gregorich et al. 29

and by
MacIntyre et al. 23 but both works provide few measurement points and were focused on biological process and temporal variability, respectively. In order to understand better the different mechanisms of production and migration of gas species in this environment, it is necessary to carry out a comprehensive survey. To date, no studies have been completed to investigate the soil gas spatial distribution in relation to possible fault and/or fracture systems and characterize seepage for both CO 2 and CH 4 in Antarctica. Soil gas geochemistry is an alternative powerful approach that is widely used to detect diffusive/advective gas emissions and identify preferential migration pathways such as buried faults and fractured areas [32][33][34][35][36][37][38] . Permafrost is generally a barrier to the migration and leakage of endogenous gaseous species. However, the presence of faults, fractures and the thawing, could allow surface migration of anomalous concentrations of endogenous gaseous species. The challenge and the goal of this research is to understand the greenhouse gas potential that is trapped by MDV permafrost and, therefore, how much of these greenhouse gases would be released during thawing events. The rate of carbon release from permafrost soils is highly uncertain 39 . More accurate estimates are crucial to predict the impact and timing of this carbon-cycle feedback effect, and thus how signi cant permafrost thaw will be for climate change this century and beyond. We report here the rst large scale soil gas survey in Antarctica targeting the Taylor Valley as ideal locality for such type of study (Fig. 1).
The sampling strategy was developed considering the logistical constraints and nalized to obtain the most representative results of the study area. Our work evaluates the magnitude and spatial distribution of the concentrations of some gases in the soil and of φCO 2 emission from permafrost and/or thawing shallow strata. The goal is to provide a rst total CO 2 emission estimate for the lower Taylor Valley that can be used for future monitoring surveys and extrapolated more broadly across the continent. The main statistics obtained for soil gas concentrations and φCO 2 are reported in Table 1. All gas species highlight broadly skewed distributions with the presence of few outliers (see SD and SK in Table 1). By comparing the mean and median values, the presence of outliers is particularly evident for H 2 and CH 4 (mean values > median values). The difference between the mean and median values also suggests a log-normal distribution for φCO 2 , CO 2 , CH 4 and H 2 .
To understand better the magnitude and the signi cance of the soil gas concentrations measured in Antarctica, calculated mean values are compared with the average concentrations of the same gaseous species present in the atmosphere, in the soil-atmosphere interface and in soil gases from the literature ( Table  2). In Taylor Valley, O 2 , N 2 and Ne concentrations are broadly equal to their atmospheric concentration. In contrast, H 2 , CO 2 and CH 4 concentrations are higher than atmospheric concentrations. CO 2 mean concentration is twice as high as those normally measured in the soil gas. He concentrations highlight a mean value lower than the atmospheric concentration, and as evidenced by the 90% percentile, and only less than 10% of the total samples shows higher concentrations than the atmosphere.

Spatial distribution of soil gas concentrations and φCO 2 values
The soil gas and φCO 2 distributions were investigated to detect potential permafrost, or to identify the possible presence of faults and fractures, which may provide gas migration pathways. NPPs highlighted the following anomaly threshold values: 5.4 ppmv for He, 18.8 ppmv for Ne, 4 ppmv for H 2 , 9 ppmv for CH 4

Discussions
In the MDV, very limited data about soil gas concentrations are available (only CO 2 and CH 4 ), while for the most part they concern CO 2 and CH 4 Table 1). The highest CH 4 value reported in Gregorich et al. 29 is in the same order of magnitude as found in this study, although more than three times lower than our maximum value (18,447 ppmv, Table 1). Both referenced studies, however, collected a limited number of measurements and focused on temporal variability rather than spatial variability. Worldwide, CO 2 and CH 4 soil gas data are numerous, and measured in different environments: in the Arctic Finnish Lapland, Voigt et al. 42  In contrast, φCO 2 has been measured in Antarctica since 1994 (see Table 3). The φCO 2 measurements in this study are up to 3 orders of magnitude higher than previous studies, therefore, the highest measured in the MDV hereto. φCO 2 measurements in this work in Taylor Valley are on average comparable with the values measured in the Arctic 25,44 , desert 25,45 and alpine 46 areas and lower than other areas of the globe 25,47 ( Table  3).
The distribution of the positive anomalies for CH 4 , CO 2 , H 2 and φCO 2 are consistent in the NE, NW and S sectors of the study area, while He shows a good correlation with CO 2 and CH 4 anomalies in the S and N sectors, respectively. He detected in shallow soil is generally indicative of deep sources 32,38 , and is typically associated with CO 2 and CH 4 emissions that act as carrier gas for trace gases (i.e., He, Rn) 35,48 . The source of these gas anomalies could be linked with shallow depth hypersaline uids 31,49 that during summer periods can release the dissolved gases after permafrost thawing. These gases can then easily migrate toward the surface through permeable layers, as well as local fractures and/or buried faults. These may act as preferential migration pathways thus resulting in the linear multigas anomalous zones observed in the study area 35,37,48 (Fig. 4). Figure 4 shows gas anomalies aligned in ENE-WSW direction in the S and N sectors of the study area, respectively. At the northern boundary, the anomalous zone does not appear as continuous as that occurring in the southern boundary of the area because of the presence of the Commonwealth Glacier and related wetlands and streams (Fig. 2), that most likely prevent gas upwelling.
As for the southern sector, there are also two physical factors that may increase permafrost degradation: solar radiation and soil albedo. Solar radiation is more intense on North facing slopes than South facing slopes and contribution to the soil system, but may be linked to a different intake of CO 2 , for example, geological and/or abiotic contributions 53 . These hypotheses can be con rmed only with isotope analyses of gas samples. In the literature, the origin of CO 2 in Antarctic soils is suggested to be linked to biological activity, favored by soil alkalinity and by shallow abiotic processes (CO 2 solubility), soil moisture content and soil temperature variations 19,21,23,29,30 . Shanhun et al. 9 and Risk et al. 54 suggested an abiotic origin of CO 2 based on the isotopic analyses, reporting very high positive φCO 2 values that cannot be explained by normal microbial activity. We suggest that the measured high anomaly values, originate from the subsurface brine system 31,49 , and are linked to permafrost cap discontinuities at the edges of the valley (fractures or buried local faults).
The spatial distribution of these anomalous zone could also be locally in uenced by shallow permafrost thawing.
Although the samples were collected at shallow depths (i.e. permafrost is often reached at 30 cm depth) and are potentially affected by atmospheric gas dilution 55,56 , the anomalous values of He, CH 4 , CO 2 and φCO 2 cannot entirely be explained by biological activity and/or super cial physical processes. Another aspect concerns the presence of soil moisture which prevents both sampling and gas rising 55,56 . In Taylor Valley, the wetlands represent an area where water permeates the pores of the soil but does not emerge on the surface. Within these areas (Fig. 2) we managed to complete some sampling stations, however it should be noted that the gas concentrations are certainly underestimated compared to those conducted in dry and ventilated soil conditions. In these areas, of about 2.1 km 2 (equal to 10% of the total area) 16 soil gas samples and φCO 2 were collected. Statistical analyses (Table S1) con rm the low values in these areas. Figure 6 shows the ranging and average values of φCO 2 measured over time in Antarctica. It should be noted that the reported measurements were conducted using various methods and in different environments, e.g.
there are measurements on dry soil, near and from lakes and from the Ross Sea. Data collected in Dec 2019 -Jan 2020 from Taylor Valley show that the average value is in the same order of magnitude of those reported from the Garwood Valley 29 . The maximum values measured in this work, instead, is at least one order of magnitude higher than those previously reported. Focusing on Taylor Valley, various measurements have been carried out over the years, both in the soils 9,25 and around the three lakes 30 . The comparison of these data shows that our values are two orders of magnitude greater than those previously measured.
The total CO 2 gas emission rates over the surveyed area (A = 21.6 km 2 ) have been computed following a statistical approach (see 2.3 Statistical and geostatistical analysis) and, the calculated average CO 2 output is 14.95 t d −1 (Table S2). The calculated total CO 2 emission considering the three summer months, is about 1,345.5 t. This value is also considered as the total annual CO 2 emitted, providing a conservative estimate for the warmest months expecting much lower emissions during the rest of the year. Then, the estimated emission factor in the study area is 62.3 t km −2 y −1 . Comparing this value, for example, with those estimated for the central Apennines in Italy, ranging from 350 to 1,050 t km −2 y −1 (57) , it is evident that the uxes reported herein are remarkably lower, even compared with others reported worldwide 58 . This discrepancy is due to both climate/environmental differences and organic content and biological activity usually present in the soils of the other continents. Although our emission values are low, forecasting that ice-free regions in Antarctica are likely to expand with gradual warming, this amount will tend to increase and should be counted in the global CO 2 budget estimations.
To conclude, we provide the rst spatial distribution maps of soil gas concentrations and φCO 2 in a large area (> 20 km 2 ) of the Taylor Valley, Antarctica. The calculated CO 2 emission output during the summer period is 1,345.5 t. Our results emphasize that extensive surveys are necessary to properly evaluate greenhouse gas (GHG) emissions in regions with permafrost. We also established the rst extensive baseline maps that can be used to compare and monitor soil gas concentrations and CO 2 emissions in the region. The collected data identi ed areas characterized by multigas anomalies where permafrost partial melting may occur during the summer period and where uids migrate to the surface through structures/fractures aligned with the main direction of the valley. We suggest that the gases originate from the subsurface brine system that is rich in dissolved gasses. The super cial melting of the permafrost and the presence of permeable zones inside it, allowed these gases to migrate to the surface.

Site description and sampling strategy
The MDV feature a mosaic of ice-covered lakes, ponds, ephemeral streams, valley glaciers and glacial, uvial, lacustrine and aeolian sediments. Mean annual air temperature in the valleys is −17°C, and annual precipitation (snow water equivalent) spans 3-50 mm 59 , making the MDV a cold, polar desert 60 . Continuous permafrost, by de nition, is a regional land surface with temperatures below 0°C on interannual timescales, and underlies 90-100% of the MDV. This permafrost is predominantly ice-cemented (ranging from icesaturated to weakly cemented), although overlying "dry-frozen" (ice-free) permafrost is common in the upper  (Fig. 1). Our study area is located in the eastern sector of the valley and extends for 6 km to the east of the Lake Fryxell bordering the southern part of the Commonwealth glacier (Fig. 2). The area is characterized by hummocky moraines, lacustrine deposits, and outwash fans where ephemeral streams and of the accumulation chamber, using the formula: (φCO 2 * (86400 * P * (V/A)) / (1000000 * R * T)) * M where φCO 2 is the soil ux expressed in ppm/sec; P is the pressure in mbar; V is volume (m 3 ) and A is surface area (m 2 ) of the accumulation chamber; T is the temperature in K; M is molecular weight; R = 0.08314472 in bar L (K mol) −1 that is used to calculate the volume in L of an ideal gas from its temperature in K, pressure in bar and mole number.

Statistical and geostatistical analysis.
Exploratory Data Analysis (EDA) (numerical and graphical techniques) was applied to elaborate soil gas data in terms of main statistical parameters, distribution type, background, and anomalous values. Normal probability plots (NPP) were interpreted according to the Sinclair method 63 in order to distinguish different populations and statistical anomaly threshold values for each gas species (see Fig. S1). Subsequently, geostatistical analysis (e.g., variogram analysis and kriging 35,43 ) was applied to construct contour maps to represent the spatial distribution of gas concentrations in the surveyed area. Furthermore, φCO 2 measurements were used to estimate the total output of CO 2       Summary diagram of φCO 2 measurements carried out since 1994 in Antarctica. The diagram shows CO 2 measurements performed in Antarctica by using various methods and in different environments.
Measurements on dry soil (in black), lakes (in light blue) and the Ross Sea (in blue) are reported.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.