Strong geologic methane emissions from discontinuous terrestrial permafrost in the Mackenzie Delta, Canada

Arctic permafrost caps vast amounts of old, geologic methane (CH4) in subsurface reservoirs. Thawing permafrost opens pathways for this CH4 to migrate to the surface. However, the occurrence of geologic emissions and their contribution to the CH4 budget in addition to recent, biogenic CH4 is uncertain. Here we present a high-resolution (100 m × 100 m) regional (10,000 km²) CH4 flux map of the Mackenzie Delta, Canada, based on airborne CH4 flux data from July 2012 and 2013. We identify strong, likely geologic emissions solely where the permafrost is discontinuous. These peaks are 13 times larger than typical biogenic emissions. Whereas microbial CH4 production largely depends on recent air and soil temperature, geologic CH4 was produced over millions of years and can be released year-round provided open pathways exist. Therefore, even though they only occur on about 1% of the area, geologic hotspots contribute 17% to the annual CH4 emission estimate of our study area. We suggest that this share may increase if ongoing permafrost thaw opens new pathways. We conclude that, due to permafrost thaw, hydrocarbon-rich areas, prevalent in the Arctic, may see increased emission of geologic CH4 in the future, in addition to enhanced microbial CH4 production.

The emission of biogenic methane (CH 4 ) from arctic permafrost landscapes caused by microbial decomposition of carbon is widely discussed [1][2][3][4][5][6] . The carbon pool of arctic permafrost soils is estimated at 1300 Pg (ref. 7). Emission estimates of biogenic CH 4 from permafrost regions vary between 32 and 112 Tg per year 8 . The production of biogenic CH 4 in wetlands is limited to soil temperatures allowing for microbial activity. This is the case during the warm season of the year, but also at soil temperatures around 0 °C, the "zero curtain" 9 .
In addition to the biogenic CH4 sources, oil and gas reservoirs in the Arctic contain vast amounts of old, geologic CH 4 (ref. 10) that can be of biogenic or thermogenic origin. Currently, in large areas this source is sealed by a cap of permafrost. Geologic CH 4 can reach the surface if pathways in this cap exist along faults (cf. ref. 11 for lower latitudes) or form due to permafrost thawing 12,13 . For subsea permafrost no consensus has been reached on the amount of geologic CH 4 released from e.g. the Siberian shelf 14,15 . In terrestrial permafrost landscapes geologic CH 4 can reach the surface through open taliks (=thaw bulbs) below lakes 13 , or in the unfrozen areas of discontinuous permafrost, comparable to what was shown for lower latitudes 11,16 . Until now, the occurrence of the entirety of these geologic sources in the Arctic and their contribution to the overall CH 4 emission is uncertain.
Among the largest amounts of oil and natural gas in the Arctic are likely stored in the Canadian Mackenzie Delta and its surroundings 10 . In this region, strong natural gas seeps have been described sporadically 17,18 , some of which are fed by thermogenic CH 4 (refs 17-19). The permafrost in the delta is of the non-Yedoma type, i.e. without the ice-rich, organic rich characteristics of Yedoma permafrost 20 . The permafrost state in our study area shows differences due to glaciation history and surface characteristics 21 . The delta itself is characterized by relatively thin (i.e. up to 100 m) and discontinuous permafrost as it was glaciated during the Pleistocene and waterbodies and channels are changing the surface and warming the ground 21 . The region east of the delta was unglaciated during the Pleistocene 22 resulting in a sharp increase in permafrost thickness between the delta and Richards Island (marked with black line in Fig. 1b). On Richards Island the permafrost is up to more than 500 m thick and continuous 21 (Fig. 1b). On the Yukon coastal plain the permafrost thickness reaches about 300 m. The soil organic carbon stocks in the study area are classified as medium to high soil carbon content with mainly 50-70 kg m −2 in the upper 1 m (ref. 7).
The overall aim of this study is to improve our understanding of geologic CH 4 emissions on a regional scale in the Mackenzie Delta region, Canada (Fig. 1). Therefore, the purpose of this study is twofold: (i) to map the CH 4 flux and its spatial variability at high spatial resolution and to derive the abundance of geologic CH 4 emission hotspots and (ii) to assess the relative contribution of biogenic and geologic sources to the annual CH 4 budget of the Mackenzie Delta region.

Results and Discussion
We conducted airborne eddy-covariance (EC) measurements (cf. ref. 23) covering large areas of the Mackenzie Delta region during two extensive campaigns in July 2012 and 2013 (Fig. 1b).
The CH 4 flux map (cf. ref. 24) resulting from our measurements covers 10,000 km² at a spatial resolution of 100 m × 100 m (Fig. 2a). This CH 4 flux map enables to detect spatial patterns in CH 4 emission and to identify and localize CH 4 emission hotspots in our entire study area. The median of all 2012 and 2013 CH 4 flux data was 1.1 mg m −2 h −1 , which corresponds to fluxes measured by the EC technique in similar ecosystems [25][26][27][28][29] . Most importantly, we found the largest emissions to be spatially stationary and temporally constant to <30% standard error, both within and between the two campaigns. A majority of high fluxes up to 14.7 mg CH 4 m −2 h −1 (with a standard error of <30%) occurred as clustered peaks in the northern part of the study area and only inside the delta (Fig. 2) where the permafrost is thin and discontinuous. Biogenic CH 4 emissions from arctic wetlands [25][26][27][28][29] and lakes 30 are typically lower and driven by changing meteorological and surface properties, making stationary, repeatedly observable emission peaks at this strength and spatial extent unlikely.
For lack of isotopic data, we used published biogenic CH 4 flux data from EC flux tower measurements in the Arctic north of 61 °N for a conservative, approximate separation between biogenic emissions and strong geologic CH 4 hotspots. The area of influence of tower based EC measurements, the so-called footprint, is of a comparable spatial scale as those of our airborne EC measurements, integrating over several hectares. Measurements with chambers and bubble traps, on the other hand, cover much smaller scales and do not readily compare with spatially integrated EC measurements. They were therefore not considered for the following threshold definition. The maximum daily biogenic CH 4 flux from permafrost landscapes found in the literature is roughly 5.0 mg m −2 h −1 (e.g. refs 25-29; max. 4.58 mg m −2 h −1 , ref. 28). At thermokarst margins of lakes with strong biogenic point sources, CH 4 emission can reach 5.0 mg m −2 h −1 as well 13 . Individual hotspots of biogenic CH 4 emission in lakes in carbon rich Yedoma type permafrost areas, can exceed that threshold 31 , but have an area of a few square metres, and thus are much smaller than the areas with CH 4 emission peaks that we observed. These high emissions from lakes do occur on a seep scale and their signal would disappear among neighbouring areas with less CH 4 emission, when we consider spatially integrated fluxes resulting from the EC method.
Therefore, we defined a flux of 5.0 mg m −2 h −1 as upper threshold for biogenic CH 4 fluxes from arctic permafrost landscapes. We thus assumed that reoccurring emissions exceeding 5 mg m −2 h −1 independently of atmospheric or surface conditions, day, time or year of flight, as found in the northern Mackenzie Delta, were not of recent biogenic origin. Instead, we attributed them to deeper geologic sources that release CH 4 through seeps, which can be related to taliks, faults or artificial pathways such as oil and gas exploration wells (Fig. 2). Combined, the footprints of our 2012 and 2013 measurements comprise 9,754 km² excluding areas with a standard error >30%. We find that about 1% of the mapped area (116 km 2 ), releases CH 4 at rates exceeding 5 mg m −2 h −1 . Single areas with these peak emissions are up to several square kilometres large. At this spatial extent, biogenic emissions from thermokarst lakes that exceeded this threshold have not been reported 30 . Single biogenic seeps in lakes would not result in an integrated signal that large.
We only found areas with high emissions within the northern part of the delta where the permafrost thickness is less than 100 m and the permafrost is discontinuous 21 and therefore permeable for gas from the subsurface. In contrast, on the adjacent coastal plain and Richards Island, with their continuous and thick permafrost of up to 300 m and more than 500 m, respectively, such high emissions were not observed. We attribute that absence of emission peaks, despite established natural gas and oil deposits in the region (Fig. 2), to the thick impermeable permafrost in these areas. These CH 4 reserves might eventually be emitted into the atmosphere if the permafrost cap becomes permeable due to thawing.
To estimate the contribution of geologic sources to the annual CH 4 emissions of the study region, we divided the year roughly into a season with and without biogenic emission. For the following calculations, we did not account for potential biogenic emission from hotspots along thermokarst lake margins during winter time 6, 32, 33 , as they are weaker than geologic hotspots as described above. Biogenic emissions into the atmosphere from permafrost wetlands and lakes are largely limited to the growing or ice-free season, respectively, and the "zero curtain" period with soil temperatures near 0 °C (cf. ref. 9). Once soil temperatures drop permanently further during the cold season, the CH 4 emissions decrease rapidly to close to 0 mg m −2 h −1 (ref. 9). We derived the period of potential biogenic CH 4 emissions from soil temperatures in 10 cm depth from the North American Regional Reanalysis data 34 . In the particularly warm year of 2013, soil temperatures in the Mackenzie Delta exceeded −2 °C on about 165 days from end of May to beginning of November. We used this duration for biogenic emissions in the following back-of-the-envelope calculations. In contrast, we assumed that the emission of CH 4 from strong geologic sources is constant throughout the year, as it is independent of recent microbial activity and causes sufficiently strong ebullition in water bodies to maintain pathways through the ice during winter 13 . Using the median fluxes of biogenic (0.84 mg m −2 h −1 ) and geologic (6.3 mg m −2 h −1 ) emissions, respectively, geologic CH 4 contributes 3 Gg to the overall 35 Gg CH 4 emitted within the covered area during the 165 days warm period. Annually, 38 Gg CH 4 are emitted within this area, including 6.4 Gg CH 4 from geologic sources. The geologic estimates are very conservative, first, because, fluxes <5 mg m −2 h −1 may also include a significant amount of CH 4 from weaker geologic sources and, secondly, because extrapolating July biogenic fluxes, which are at or near a possible seasonal peak, into the shoulder seasons might overestimate the contribution of biogenic sources. Relative to their areal coverage, strong geologic sources that occur in 1% of the area thus contribute a disproportionally large fraction of at least 17% of the emissions to the estimated annual budget, making geologic sources 20 times stronger CH 4 emitters than biogenic sources.

Conclusions
Our study suggests, first, that thinning permafrost in a warmer climate may not only result in the frequently reported and discussed increased emission of biogenic CH 4 , but also in increased emissions of geologic CH 4 , that is currently still trapped under thick, continuous permafrost, as new emission pathways open due to thawing permafrost. Secondly, in other arctic regions with natural gas and oil reservoirs that are currently capped under mainly continuous permafrost, e.g. the North Slope of Alaska or Siberia 10 , CH 4 contributions from those geologic sources might need to be included when addressing future CH 4 emissions under ongoing permafrost thaw.
Thirdly, our results indicate that geologic CH 4 emissions may contribute strongly to the permafrost-carbon-climate feedback, especially in permafrost areas vulnerable to thawing and therefore warrant much more attention.

Methods
Experimental Setup. This study was performed in the Mackenzie Delta, which is the second largest arctic delta (13,000 km²), along the adjacent Yukon coastal plain towards Herschel Island, and on Richards Island to the north east of the delta (67°26′N-69°33′N, 133°22′W-140°5′W). The study area extended about 320 km from west to east and had a north-south extension of 240 km. The study period was from July 4 th to July 10 th , 2012 (5 flight days, 44 flight tracks) and July 19 th to July 26 th , 2013 (7 flight days, 40 flight tracks).
Measurements were conducted aboard the research aircraft Polar 5 of the Alfred Wegener Institute Helmholtz Centre for Polar and Marine Sciences (AWI). A 3 m nose boom including a 5-hole probe for measuring the 3D wind vector was mounted to the front of the airplane. Sample air was drawn from an inlet above the cabin at about 9.7 l s −1 and analysed in an RMT-200 (Los Gatos Research Inc., Mountain View, California, USA) in 2012 (CH 4 concentration only) and a Fast Greenhouse Gas Analyser FGGA 24EP (Los Gatos Research Inc.) in 2013 (CH 4 , carbon dioxide and water vapour). Gas concentrations were recorded at 20 Hz, wind speed at 100 Hz. Additionally, the aircraft was equipped with an Inertial Navigation System (Type Laseref V, Honeywell International Inc., Morristown, New Jersey, USA), several Global Positioning Systems (NovAtel Inc., Calgary, Alberta, USA), a radar altimeter (KRA 405B/Honeywell International Inc., Morristown, New Jersey, USA) and a laser altimeter (LD90/RIEGL Laser Measurements Systems GmbH, Horn, Austria). Air temperature was measured with an open wire Pt100 in an unheated Rosemount housing, and air humidity with an HMT-330 (Vaisala, Helsinki, Finland) placed in a Rosemount housing. We derived the wind components u, v, and w with respect to the earth coordinate system using the method suggested by ref. 35. For the analysis only flux measurements derived after the warming up phase of the gas analysers (about 0.5 h) were used. Flight altitude was between 40-80 m above ground level, true airspeed was 60 ms −1 . A height dependency of CH 4 fluxes was not considered in our calculation. The atmospheric boundary layer (ABL) was determined with in-situ measurements of relative humidity and potential temperature during vertical profile flights at the beginning and end of each track. Flights above the ABL were excluded from the analysis.
Flux calculation. The data were analysed in GNU R version 2.15.3 using an early version of the eddy4R software 36 . The analysis is based on 20 Hz dry mole fraction data. We calculated the dry mole fraction by using the humidity information from the Vaisala HMT-330 in 2012, and that from the FGGA itself for the 2013 data. Spectroscopic correction of the data was done following ref. 37. CH 4 fluxes were calculated with a time-frequency-resolved version of the eddy covariance technique using vertical wind speed and CH 4 concentration data 23 . The result was an in-situ observed space-series of the surface-atmosphere exchange of CH 4 at 100 m spatial resolution. With a footprint model 38 the CH 4 emissions were related to the underlying surface. The data were quality controlled 39 and only data up to flag 6 (steady state and integral turbulence characteristics test ≤100%), the upper threshold for data that are considered eligible for long-term measurements, were included in the analysis. This left 66957 flux observations (=6696 km of flight tracks), more than 90% of the original data, as input data for our study. In this study we did not apply spectral, vertical flux divergence, and storage term corrections.

Flux topography calculation.
To derive a flux map including CH 4 fluxes from all flights, we used the flux values along the flight tracks and the corresponding footprints. The footprints provide (i) the information about the location, size and shape of the area influencing the flux measurement at the respective point, and (ii) the percentage contribution, i.e. footprint weight, from each grid cell to the combined flux signal observed by the aircraft. The weight of the contribution of each grid cell depends mainly on the distance from the measurement point, wind direction, and atmospheric stratification. That is, the footprint itself does not separate the strengths of individual sources. Instead, for each grid cell this is achieved by superimposing joint replicates of footprint weights and corresponding flux observations as the aircraft passes by. In result, the flux gradient around strong emitters appears continuous rather than representing the potentially discrete nature of geologic features. This last discretization step is achieved by applying the >5 mg CH 4 m −2 h −1 thresholding.
We used flux topographies 24 to derive the CH 4 flux in each 100 m × 100 m grid cell using the sum of the products of CH 4 fluxes and footprint weights divided by the sum of footprint weights covering that grid cell , with the number of flight tracks j … N, the number of footprint weights i … M that cover a grid cell during a given flight track, the CH 4 flux f and the footprint weight g.
Error calculation. In this study error sources include random measurement error, error in the wavelet calculation and in the footprint calculation. Standard error topographies were calculated following ref. 40. Grid cells with a standard error <30% (963843 grid cells; 57% of the data) were included in the analysis. The shape of the CH 4 flux map presented in Fig. 2  in wind speed and direction, areas not included within the footprints of our CH 4 flux measurements, and the exclusion of these grid cells with a standard error >30%.
Anthropogenic features. In the ArcGIS software ArcMap 10.1 by Esri we digitalized, georeferenced and compiled information from the geologic maps of refs 41-44 and marked locations of exploration gas and oil wells in Fig. 2.
Estimating annual geologic and biogenic CH 4 emissions. The upper threshold for biogenic fluxes was set to 5 mg CH 4 m −2 h −1 based on maximum fluxes reported in the literature (e.g. refs [25][26][27][28][29]. Emissions in grid cells exceeding this value were classified as geologic CH 4 . We derived the period of potential biogenic CH 4 emissions from soil temperatures in 10 cm depth from the North American Regional Reanalysis (NARR) data 34 . Microbial activity and CH 4 emission occurs in the warm period of the year and in the zero curtain period when soil temperatures fluctuate around 0 °C (ref. 9). CH 4 emissions decrease close to zero, once the soil temperature drops well below 0 °C for the duration of the cold season 9 . Therefore, we defined the period between spring and autumn with soil temperatures exceeding −2 °C as period with biogenic emissions. In our study region this period lasted for about 165 days from end of May to beginning of November. In contrast, strong geologic sources are assumed to release CH 4 to the surface continuously during 365 days a year if pathways exist. To calculate the emission from each of the two sources of CH 4 , we used the respective time periods, area size of occurrence and the median fluxes. We used the median flux of all measurements equal to or larger than 5 mg m −2 h −1 (6.3 mg m −2 h −1 ) for geologic emissions, and the median of those emissions smaller 5 mg m −2 h −1 (0.84 mg m −2 h −1 ) for biogenic emissions.