Interannual Variability in Methane and Nitrous Oxide Concentrations and Sea-Air Fluxes Across the North American Arctic Ocean (2015–2019)

Between 2015 and 2018, we collected approximately 2,000 water column measurements of methane (CH 4 ) and nitrous oxide (N 2 O) concentrations in the North American Arctic Ocean during summer and early fall. We also obtained 25 measurements of CH 4 and N 2 O concentrations in rivers along the Northwest Passage and Ellesmere Island in midsummer 2017–2019. Our results show that N 2 O is generated in the highly productive Bering and Chukchi Seas and transported northeastward, producing a persistent subsurface N 2 O peak in the Beaufort Sea. The Chukchi and Beaufort Sea sediments are a significant source of CH 4 to the water column. These sedimentary sources and associated water column consumption display significant spatial gradients and interannual variability. CH 4 isotope data demonstrate the importance of CH 4 oxidation across the study region. We find that rivers are not a significant source of CH 4 or N 2 O to the Arctic Ocean at the time of year sampled. The estimated annual sea-air flux across the study region (2.3 million km 2 ) had a median (first quartile, third quartile) of 0.009 (0.002, 0.023) Tg CH 4 y −1 and −0.003 (−0.013, 0.010) Tg N y −1 . These results suggest that the North American Arctic Ocean currently plays a negligible role in global CH 4 and N 2 O budgets. Our expansive data set, with observations at many repeat stations, provides a synopsis of present-day Arctic CH 4 and N 2 O distributions and their range of variability, as well as a benchmark against which future climate-dependent changes can be evaluated. 4 ) and nitrous oxide (N 2 O) are powerful greenhouse gases, and N 2 O contributes to the depletion of the ozone layer. Rapid warming could change Arctic emissions of CH 4 and N 2 O. To date, most Arctic studies of CH 4 and N 2 O have focused on small geographic regions and/or not involved repeat observations across multiple years. To address these limitations, we collected approximately 2,000 measurements of CH 4 and N 2 O across the North American Arctic Ocean between 2015 and 2018 in summer and early fall, including many repeat stations. We show that N 2 O is produced in the Bering and Chukchi Seas and transported eastward into the Beaufort Sea. The Chukchi and Beaufort Sea sediments are a considerable source of CH 4 to the water column. These sedimentary sources and the associated water column consumption display significant variability. We collected river measurements between 2017 and 2019, which demonstrated that rivers are not a substantial source of CH 4 or N 2 O to the region in midsummer. Across the North American Arctic Ocean, the calculated rate of exchange of CH 4 and N 2 O between the ocean and the atmosphere was low, suggesting that this region currently plays a minor role in regulating the global atmospheric concentrations of these greenhouse gases.

and altered hydrological cycles (Blunden & Arndt, 2019), may result in climate feedbacks through impacts on biogeochemical cycling and atmospheric emissions of the potent greenhouse gases methane (CH4) and nitrous oxide (N2O) (James et al., 2016). In addition to contributing to global warming, N2O is currently the most significant ozone-depleting substance being emitted to the atmosphere (Ravishankara et al., 2009).
Some previous work has suggested that ocean warming could to lead to the destabilization of subsea CH4 hydrates, as well as subsea and terrestrial permafrost, but potential CH4 emissions due to these processes remain highly uncertain (Archer et al., 2009;James et al., 2016;Lenton et al., 2008). Many studies have also shown elevated CH4 concentrations in bottom waters and surface sediments of some Arctic shelf regions (Coffin et al., 2013;Fenwick et al., 2017;Myhre et al., 2016;Sparrow et al., 2018). However, these studies have reported a wide range of sea-air CH4 flux estimates, suggesting significant variability in seafloor release rates and subsequent water column oxidation. Furthermore, the potential impact of increasing freshwater discharge (from rivers and groundwater) on CH4 emissions is currently poorly understood (Lamarche-Gagnon et al., 2019;Manning et al., 2020). To date, most studies of Arctic Ocean CH4 distributions have focused on a single region and/or have lacked repeat observations across multiple years (Lapham et al., 2017;Rogener et al., 2020), making extrapolations of fluxes to broader spatial and temporal scales challenging.
As with CH4, the importance of the Arctic Ocean in the global N2O budget is currently poorly constrained, with some regions likely acting as N2O sinks and others as net sources (Fenwick et al., 2017;Yang et al., 2020;Zhan et al., 2021;Zhang et al., 2015). Thawing terrestrial permafrost may act as a net source of N2O to the atmosphere, but there is no consensus on whether thawing subsea permafrost will be a net source or sink of N2O (Voigt et al., 2017b;Voigt et al., 2017a). Oceanic N2O production and consumption rates are controlled by a wide range of biogeochemical processes, e.g. nitrification and denitrification (Bange et al., 2010;Codispoti, 2010), and are likely to be impacted by ongoing Arctic change. Some studies have reported significant N2O supersaturation can occur in the Bering and Chukchi Seas (Hirota et al., 2009;Toyoda et al., 2021;Zhan et al., 2021). Fewer published N2O data exist in other North American Arctic Ocean regions, including the Northwest Passage, and Baffin Bay (Kitidis et al., 2010;Zhang et al., 2015). Fenwick et al. (2017) presented CH4 and N2O concentration profiles from ~50 stations across the North American Arctic Ocean during 2015. Their observations presented the first three-dimensional picture of CH4 and N2O across a ~10,000 km transect of the North American Arctic. Here we combine the existing 2015 dataset with new oceanographic observations from 2016-2018 to investigate broader patterns of spatial and interannual variability in the distributions and sea-air fluxes of these gases. We also use measurements of the isotopic composition of CH4 (δ 13 C-CH4) to investigate CH4 sources and sinks in the water column and provide an updated treatment of sea-air fluxes that accounts for time-variability in both wind speed and sea ice coverage. Finally, we present CH4 and N2O measurements from 19 Arctic rivers to investigate the potential summertime impact of rivers on the distribution of these gases in the Arctic Ocean.
Our study region spans a broad range of hydrographic and biological regimes (section 3.1), including both high and low productivity waters, as well as shelf and slope regions of strong and weak sedimentary CH4 sources. With these observations, we are able, for the first time, to characterize the persistent patterns of subsurface N2O supersaturation in the Beaufort Sea shelf and slope (observed in all four years of observations), and contrast these with the strong interannual variability of N2O in the shallow Bering and Chukchi Seas. We also demonstrate strong spatial gradients and inter-annual variability in subsurface CH4 distributions in the Chukchi and Beaufort Seas associated with variability in seafloor release and water column consumption. Our new multi-year dataset, based on measurements conducted in a single laboratory, enables confidence the identification of inter-annual variability. These new ocean and river measurements also provide an important benchmark of current CH4 and N2O distributions in the North American Arctic, enabling the detection of future changes under a rapidly warming climate.

Research expeditions
We collected oceanographic samples during annual research expeditions on three Canadian icebreakers between 2015 and 2018, during summer and early fall (early July to mid-October; see Table S1 for sampling dates). The Bering and Chukchi Sea samples came from the Distributed Biological Observatory (DBO) program conducted on the CCGS Sir Wilfrid Laurier (Grebmeier et al., 2019) in July of each year, while the Beaufort Sea samples primarily came from the Joint Ocean Ice Study (JOIS) program conducted on the CCGS Louis S. St-Laurent in September-October. These research programs involve annually repeated stations and transects. Samples from the Northwest Passage (Canadian Arctic Archipelago) primarily came from the ArcticNet and GEOTRACES sampling programs conducted on the CCGS Amundsen between July to September. Additional sampling from the CCGS Amundsen was conducted in Baffin Bay (2015, 2017, and 2018), the eastern Beaufort Sea (2016), and Hudson Bay, James Bay and Foxe Basin (2017). The research programs, priorities, and cruise tracks on the CCGS Amundsen vary from year to year, so the sampling stations are less consistent than those of the DBO and JOIS programs.
The full dataset includes a total of 220 water column profiles and 1977 measurements collected between 2015 and 2018. We sampled 129 unique stations, with 17, 12, 16, and 84 stations sampled in four, three, two and one years, respectively ( Figure 1). Additionally, we collected 25 measurements from rivers discharging into the North American Arctic Ocean during CCGS Amundsen expeditions between 2017 and 2019 (19 rivers were sampled in one year and three rivers were sampled in two years). River sites were accessed via helicopter and samples were collected at a single location near the mouth of each river. The data collected from oceanographic stations during the 2019 CCGS Amundsen cruise will be published separately. The solid black line designates a transect connecting many of the stations and/or repeat hydrography lines. The along-transect distance is indicated in 1000 km intervals from west to east. Colored diamonds, circles, triangles, and squares represent stations sampled one, two, three, or four times, respectively. White triangles represent off-transect stations (typically >200 km from the transect line) that were only sampled in one year. Data from the off-transect stations are not presented in this paper, but are included in the accompanying dataset. Dashed black lines indicate the boundaries between different oceanic regions/seas, defined based on the International Hydrographic Organization (1953).
To focus our analysis of regional and temporal trends in the dataset, we defined a nearly 8000 km-long transect line (starting in the Bering Sea) that intersected many of the repeat stations ( Figure 1). For analyses based on along-transect distance, station locations were interpolated to the nearest point on the transect. Data from stations located more than 200 km from the transect, as well as measurements collected directly over a hydrocarbon seep at Scott Inlet (northeastern coast of Baffin Island, 155 km from the transect), are not discussed in this manuscript. The Scott Inlet CH4 data are described in Cramm et al. (2021). These off-transect stations (14 out of 129 stations, indicated by white triangles in Figure 1) were only sampled in one of the four years and cannot be used to investigate interannual variability in gas distributions. In accordance with the principles of findable, accessible, interoperable and reusable (FAIR) data (Wilkinson et al., 2016), all station locations are shown in Figure 1, and all data have been archived in the open access repositories PANGAEA and MEMENTO (see Open Data section).

Sample collection and analysis
CH4 and N2O samples were collected from Niskin bottles into glass serum vials (60 mL for concentration and 200 mL for CH4 isotopic composition) and sealed with butyl rubber stoppers and aluminum crimp seals. Before sealing, saturated mercuric chloride solution was added as a preservative to each sample (50 or 100 μL for 60 mL samples, and 200 μL for 200 mL samples). Preserved samples were stored in the dark and analyzed in the laboratory within four months of sample collection.
In 2017, 2018, and 2019, CH4 and N2O samples were collected from rivers in the Canadian Arctic Archipelago. Sampling sites were chosen based on their proximity to the cruise track, and reflected the wide range of geological variability within the region (Alkire et al., 2017). All sampling sites were accessed by helicopter. River water was collected using Tygon tubing secured to an extendable pole and connected to a peristaltic pump on shore. A hand-held probe (Oakton Con5 or Cole-Parmer C100) was used to measure sample temperature and conductivity in freshly collected samples. The river water depth at the sampling location was typically less than 1 m. The elevation difference between the peristaltic pump head and the water intake was always less than 1 m to prevent sample degassing due to vertical suction. Gas samples were collected in glass serum vials and treated as described above for seawater depth profile measurements.
Gas concentrations were measured using a purge and trap system coupled to a gas chromatograph-mass spectrometer . Calibration was performed during every run using a Praxair certified standard (5% accuracy for CH4 and N2O), which was diluted with helium to prepare working standards of varying concentration. The Praxair standards were validated against an international standard prepared by NOAA and found to be equivalent to the certified values within the uncertainty of the test (~5%). As part of every run, air-equilibrated water samples were analyzed as an additional quality control measure. For concentration measurements, duplicate samples were analyzed, and the precision (standard deviation) is reported. For the entire dataset, the average standard deviation was 0.5 nmol kg −1 for both gases, representing 10% and 3% mean relative standard deviation for CH4 and N2O, respectively.
In the 2018 dataset, measured N2O concentrations in air-equilibrated water samples were 9 to 13% lower than the expected concentrations, and the N2O concentrations measured on deep water samples (where little interannual variability is expected) displayed a similar offset. The offset may have been due to deactivation of the Nafion tube used to dry the sample, an issue which was identified after the analysis was complete. To correct for this offset, samples collected on the CCGS Amundsen, CCGS Sir Wilfrid Laurier, and CCGS Louis S. St-Laurent cruises in 2018 were increased by 9.1, 13.4, and 13.2%, respectively, which is the mean offset observed for the air-equilibrated water samples analyzed along with the samples for each cruise. After applying this correction, deep water N2O measurements at repeat stations were consistent across all four years. The measured CH4 concentrations in air-equilibrated water samples from 2018 did not have a systematic offset, and therefore the 2018 CH4 concentrations were not adjusted.
Methane isotopic measurements ( 13 C-CH4) were conducted at the Alfred Wegener Institute using a Finnigan Delta XP Plus mass spectrometer following the method of Damm et al. (2015). The samples were pre-concentrated through a purge and trap system (Finnigan PreCon Trace Gas Pre-Concentrator). Methane carbon isotope ratios are reported as δ 13 C-CH4 = 13 Rsamp/ 13 Rstd -1, where 13 R is the ratio of 13 C/ 12 C in the sample or standard, respectively, and the standard is Vienna Pee Dee Belemnite (VPDB). A CO2 reference gas (Air Liquide) was inserted via the reference gas port in each acquisition to correct for effects that occur in the mass spectrometer ion source. The isotope ratios of all peaks were calculated against this CO2 working standard and then referenced to VPDB. For most depths, only a single sample was analyzed, but the typical reproducibility derived from duplicates by this method is 1-1.5‰.
All CTD rosette hydrographic data from each cruise has been archived with metadata documenting the calibration procedures, including the shipboard analysis of salinity samples with a salinometer and oxygen samples by Winkler titration (see Data Availability Statement).

Sea-air flux calculations
Sea-air fluxes of CH4 and N2O integrated over the mixed layer residence time of each gas were calculated, incorporating variability in wind speed, sea ice cover, and sea level atmospheric pressure. This approach contrasts with some previous studies based on Arctic Ocean seawater measurements that have calculated sea-air fluxes of CH4 and N2O using either instantaneous wind speeds (Heo et al., 2021;Hirota et al., 2009;Zhan et al., 2017), or wind speeds averaged over a fixed amount of time, e.g., one month Lorenson et al., 2016). As gas exchange is a nonlinear function of wind speed, high wind speeds have a disproportionately large impact on the total sea-air flux. As described in Supporting Text S1, the majority of gas transfer velocities we derived from instantaneous or monthly-averaged wind speeds were systematically biased low compared to mixed layer residence time-weighted sea-air fluxes (Wanninkhof et al., 2002). In addition, sea ice decreases the effective gas transfer velocity, and it is thus necessary to incorporate ice cover variability prior to sampling to accurately calculate the timeweighted sea-air flux (Ji et al., 2019;Ouyang et al., 2021).
Sea-air flux calculations were performed using the CCMP V2.0 ocean vector wind speed analysis product, which provides data every 6 hours with 0.25 o x 0.25 o resolution (https://www.remss.com, Atlas et al. (2011)). Sea ice concentration products with daily frequency and 10 km resolution from the EUMETSAT Ocean and Sea Ice Satellite Application Facility (https://osi-saf.eumetsat.int). Sea ice concentration product AMSR-2 (identifier OSI-408) was used in 2017-2018 and SSMIS (identifier OSI-401-b) was used in 2015-2016 (as product AMSR-2 was not available). Sea level atmospheric pressure data (6-hour temporal and 2.5 o x 2.5 o spatial resolution) were obtained from the NCEP/NCAR reanalysis product, which is provided by the NOAA-ESRL Physical Sciences Laboratory (https://psl.noaa.gov/data/gridded).
The gas transfer velocity integrated over the residence time of the gas in the mixed layer was calculated based on equations 5-7 of Teeter et al. (2018), which modify equation 6 of Reuer et al. (2007).
Here k is the gas transfer velocity calculated with the equation of Ho et al. (2006) with Schmidt numbers for CH4 following Jähne et al. (1987) and N2O following Wanninkhof (2014), based on the work of Hayduk and Laudie (1974) and Wilke and Chang (1955). In equations 1-3, t is the time step, with t = n representing the most recent gas transfer velocity, and t = 1 the first measurement (60 days prior to the measurement date). The variable fi is the fraction of the mixed layer ventilated at index i, ωi is the weighting coefficient at index i, MLD is the mixed layer depth, and Δt is the time interval between each index (6 h). To account for the effect of sea ice on gas exchange, the fraction of open water, fopen was incorporated into the method of Teeter et al. (2018), as shown in equation 1. We assumed that the gas transfer velocity scales linearly as a function of the fraction of open water (Butterworth & Miller, 2016;Manning et al., 2019).
Sea-air gas flux was calculated as Here [C]meas and [C]eq are the measured and equilibrium gas concentrations, respectively, with positive values indicating a net sea to air flux (i.e., ocean outgassing). Equilibrium concentrations were calculated from the measured temperature and salinity following Wiesenburg and Guinasso (1979) for CH4 and Weiss and Price (1980) for N2O. The atmospheric concentration for each gas was estimated based on the mean monthly flask measurements reported for Barrow, Alaska by the NOAA Earth System Research Laboratory Global Monitoring Division for June-October in each year (Dlugokencky et al., 1994;Hall et al., 2007). Dry concentrations were 328.25, 329.14, 330.11, and 330.96 ppb for N2O and 1919.64, 1933.67, 1934.92, and 1933.50 ppb for CH4 in 2015, respectively (Dlugokencky et al., 2020b, 2020a. These dry atmospheric concentrations were adjusted to local sea level pressure using the mean sea level pressure from the 30 days prior to each measurement and adjusted to wet concentrations by assuming 100% relative humidity.
The mixed layer depth was defined based on a potential density difference criterion of 0.125 kg m −3 relative to the density at 5 m depth, using CTD profiles binned to 1 m. The mixed layer depth was set to a minimum value of 5 m.
To estimate annual sea-air fluxes and the contribution of our study region to the global budgets of CH4 and N2O, seasonal variability in the gas transfer velocity, driven by changes in sea ice cover and wind speeds was incorporated into the calculations. The average sea-air flux was calculated over the calendar year in which the sample was collected, using 12 months of wind speed, sea ice fraction, and sea level pressure data at the sampling location. Since only one observation per year was available, we assumed a constant gas saturation anomaly (i.e., [C]meas − [C]eq) at each location over the 12-month period. We calculated fluxes for each of the five ocean regions (Bering Sea, Chukchi Sea, Beaufort Sea, Northwest Passage, and Baffin Bay/Davis Strait) by taking the median or average of all fluxes in each region within 200 km of the transect line (see section 2.1, Figure S1, and Table S2). Due to the sparse sampling, variability in stations sampled each year, and significant interannual variability among the stations, we did not perform any spatial or temporal interpolation when upscaling the fluxes to each region. We calculated the area of each ocean region using the ETOPO1 1 Arc-Minute Global Relief Model, including all locations with a seafloor depth of at least 1 m (Amante & Eakins, 2009). Although these annual flux estimates have uncertainty due to the lack of seasonally-resolved measurements and the limited number of observations, they are useful for comparison with other Arctic studies that have also estimated annual CH4 and N2O fluxes based on instantaneous measurements (Hirota et al., 2009;Thornton et al., 2016Thornton et al., , 2020, as well as with global emissions estimates (Ciais et al., 2013;Saunois et al., 2020;Yang et al., 2020).
Further details on the calculation methodology for the regional fluxes and the impact of the time weighting approach on the calculated sea-air fluxes are provided in the Supporting Information.

Circulation and biogeochemistry in the North American Arctic Ocean
To aid in the interpretation of our datasets, we summarize some key characteristics of Arctic Ocean circulation and productivity regimes (Mclaughlin et al., 2006;Timmermans & Marshall, 2020). The Arctic Ocean is primarily salinity-stratified, with a strong halocline inhibiting convection, thereby enabling the formation of sea ice in winter (Carmack, 2007;Padman, 1995;Rudels, 2015). Freshwater inputs driving this stratification are provided via river discharge, precipitation, and inflow of Pacific-derived water through the Bering Strait (Carmack et al., 2016;Haine et al., 2015). Liquid freshwater storage in the Arctic Ocean has significantly increased over the past 40 years and is expected to increase in the future due to increased supply from these sources, as well as reductions in sea ice volume (Carmack et al., 2016;Haine et al., 2015).
In addition to representing a significant freshwater source to the Arctic Ocean, Pacificderived waters also contain elevated nutrients, which support high rates of biological productivity in the northern Bering and Chukchi Seas (Grebmeier et al., 2006). A component of the dense and nutrient-rich Pacific-derived water flowing into the Chukchi Sea is transported eastward into the southern Beaufort Sea as a shelf break jet (von Appen & Pickart, 2012;Brugler et al., 2014), providing a subsurface source of nutrients to this region. High productivity conditions can lead to the production of N2O through water column and sedimentary nitrification and incomplete sedimentary denitrification, and the production of CH4 from sedimentary methanogenesis during organic matter decomposition (Bange et al., 2010;Valentine, 2011). In contrast to the Bering and Chukchi Seas, the Beaufort Sea is generally considered oligotrophic (Bergeron & Tremblay, 2014;Carmack et al., 2004).
Complex circulation patterns within the Northwest Passage (Canadian Arctic Archipelago) result from the interaction of topographic constrictions, tidal flows, and significant freshwater inputs from rivers and sea ice melt (Mclaughlin et al., 2006). These circulation patterns regulate the transport of water masses formed in the Bering/Chukchi Seas that contain unique biogeochemical signatures. In near-surface waters, there is net eastward transport toward Baffin Bay, a marginal sea located between Baffin Island and Greenland (Mclaughlin et al., 2006). At depth, westward transport of warmer, saltier Atlantic-derived waters occurs within the Northwest Passage, but the presence of sills such as Barrow Strait (average depth 175 m) effectively limits the lateral transport of these waters into the Beaufort Sea (Mclaughlin et al., 2006). Atlantic-derived waters primarily enter the central Arctic Ocean via Fram Strait, between Greenland and Svalbard, Norway (Rudels, 2015). This dense Atlantic water is not found on the shallow Bering and Chukchi shelves, but is present throughout much of the remainder of the North American Arctic Ocean.
In Baffin Bay, surface flows are cyclonic. Northward inflow of water occurs along the Greenland shelf via the West Greenland Current (Arctic origin), and West Greenland Slope Current (North Atlantic origin). Southward outflow occurs via the Baffin Island Current, including near-surface waters from the Northwest Passage, waters from the northern archipelago (Smith Sound), and recirculation of the inflowing waters (Curry et al., 2011;Münchow et al., 2015). The exchange and ventilation of the deep and bottom waters in Baffin Bay are restricted by a sill at Davis Strait, which marks the boundary between Baffin Bay and the Labrador Sea, as well as sills within the Canadian Arctic Archipelago (Barrow Strait and Smith Sound) (Top et al., 1980;Wallace, 1985). High primary productivity has been reported in the North Water Polynya region of northern Baffin Bay, although productivity may have decreased in recent years due to increased stratification (Bergeron & Tremblay, 2014;Tremblay et al., 2002).
Prior studies have used a variety of definitions and names for the water masses present in the North American Arctic Ocean (Brugler et al., 2014;Danielson et al., 2017;Gong & Pickart, 2015;Pickart et al., 2016). In this manuscript, we focus on the eastward transport of Pacificderived, nutrient-rich water masses. We follow the definitions of Danielson et al. (2017) where the Bering-Chukchi Winter Water is defined by temperatures between −2 and 0 o C and practical salinity between 30 and 33.5, and the Bering-Chukchi Summer Water (BCSW) is defined by temperatures between 0 and 7 o C and practical salinity between 30 and 33.5. The BCWW forms during cooling and brine rejection in winter and the BCSW primarily consists of BCWW that has subsequently warmed. A portion of the dense BCWW and BCSW is transported eastward to the Beaufort Sea along the shelf break and upper slope (von Appen & Pickart, 2012;Brugler et al., 2014;Rudels, 2015).

Regional profiles of CH4 and N2O
Figure 2 displays the median depth profiles of N2O and CH4 from each region, with all four years of data pooled based on potential density anomaly (σϴ), with a bin size of 0.4 kg m −3 . Median CH4 concentrations at all density surfaces in the Bering Sea ranged from 3.3 to 5.6 nmol kg −1 . In the Chukchi Sea, we observed an increase in CH4 concentrations in the densest waters, with peak median concentrations of 11.1 nmol kg −1 at a potential density anomaly (σϴ) of 26.2 kg m −3 . The Beaufort Sea displayed a CH4 peak at intermediate density surfaces (median concentrations were ~9.5 nmol kg −1 from 24.8 to 26.0 kg m −3 ), but significant variability existed within each density surface. The elevated CH4 concentrations and large variability in subsurface waters of the Chukchi and Beaufort Seas likely reflect the release of CH4 from sedimentary organic matter diagenesis, as well as spatially-heterogeneous sedimentary hydrate and permafrost deposits throughout parts of the Beaufort Sea and eastern Chukchi Sea (Coffin et al., 2013;Ruppel et al., 2016). The Northwest Passage had relatively low CH4 concentrations throughout the water column (median concentrations ranging from 3.2 to 8.3 nmol kg −1 ), reflecting the limited sedimentary sources. In Baffin Bay and Davis Strait, the median CH4 concentrations ranged from 1.3 to 5.1 nmol kg −1 . The CH4 undersaturation observed in the stagnant deep waters of Baffin Bay reflects in situ oxidation, as well as lower atmospheric concentrations at the time of ventilation (Punshon et al., 2014). In the Bering and Chukchi Seas, N2O concentrations were highest in the densest waters. In the Bering Sea, the potential density surface 25.8 kg m −3 had a median N2O concentration of 16.6 nmol kg −1 (interquartile range 16.2 to 17.0 nmol kg −1 ) and in Chukchi Sea, the potential density surface 26.2 kg m −3 had a median N2O concentration of 18.9 nmol kg −1 (interquartile range 16.8 to 19.8 nmol kg −1 ). The high annual primary production and relatively low zooplankton grazing lead to substantial export of organic matter to the seafloor and active benthic ecosystems and remineralization (Campbell et al., 2009;Grebmeier et al., 2006;. Previous studies have suggested that the elevated N2O concentrations observed in this region likely reflect production via incomplete denitrification in the sediments and/or nitrification in the water column or sediments (Fenwick et al., 2017;Hirota et al., 2009;Toyoda et al., 2021). The peak N2O concentrations occur within the densest waters of the BCSW and BCWW and reflect supersaturation relative to atmospheric equilibrium (~16 nmol kg −1 ).
Maximum N2O concentrations in the Beaufort Sea were associated with a potential density anomaly of 26.6 kg m −3 and showed remarkable consistency across the four-year dataset (median concentration 18.1 nmol kg −1 , interquartile range 17.8 to 18.6 nmol kg −1 ). N2O concentrations decreased in deeper waters of the Beaufort Sea, with a median concentration of 13.2 nmol kg −1 and in the densest waters sampled (σϴ = 28.2 kg m −3 ). This value is undersaturated relative to the present-day equilibrium concentration of ~16 nmol kg −1 , but close to equilibrium concentrations at the time of ventilation several hundred years ago, i.e. ~13 nmol kg −1 at 270 ppb N2O (Druffel et al., 2017;Macdonald & Carmack, 1991;MacFarling Meure et al., 2006;Östlund et al., 1987). In the Northwest Passage, Baffin Bay and Davis Strait, N2O concentrations were relatively homogeneous throughout the water column (median concentrations 14.0 to 16.6 nmol kg −1 ), apart from the deep waters (>1200 m depth) in stagnant Baffin Bay where N2O concentrations increased to ~20 nmol kg −1 . This deep N2O excess is generated locally via sedimentary denitrification (Lehmann et al., 2019).
The overall distribution of N2O across our survey region was thus characterized by production in the Bering and Chukchi Seas, eastward subsurface transport of this N2O supersaturation into the Beaufort Sea and significantly attenuation of the N2O supersaturation within the Northwest Passage. This attenuation likely reflects the mixing and dilution of Pacificderived waters with Atlantic waters associated with the narrow channels and shallow sills within the Northwest Passage.

CH4 and N2O signatures of different water masses based on salinity, temperature and O2 analysis
We characterized regional trends in CH4 and N2O distributions within discrete water masses using temperature-salinity plots, and by evaluating the relationship between N2O saturation anomaly and apparent oxygen utilization (Figure 3). The saturation anomaly is . Apparent oxygen utilization (AOU) is an estimate of the amount of O2 consumed by respiration and is calculated as AOU = [O2]eq -[O2]meas. Because nitrification generates N2O and consumes O2, water column nitrification can generate a positive correlation between AOU and N2O (Yoshinari, 1976).
The highest concentrations of CH4 occurred in the Chukchi and Beaufort Seas (maximum concentration of 53 nmol kg −1 , as compared to atmospheric equilibrium of ~4 nmol kg −1 ). Elevated CH4 concentrations in the Chukchi and Beaufort Seas frequently occurred in the subsurface, and were likely associated with sedimentary sources (including seafloor CH4 seeps, subseafloor permafrost, and CH4 hydrates, etc.), as suggested in previous studies (Coffin et al., 2013;Kvenvolden et al., 1993;Lapham et al., 2017;Li et al., 2017;Lorenson et al., 2016;Matveeva et al., 2015;Sparrow et al., 2018). In addition, high CH4 concentrations also occurred across a wide range of density surfaces and depths, suggesting the importance of mixing and circulation in transporting sedimentary CH4 signatures. However, most of the samples with the highest CH4 concentrations had temperature/salinity properties consistent with the BCWW and BCSW.
CH4 concentrations in the Bering Sea, Northwest Passage and Baffin Bay were comparatively low; all samples had concentrations ≤15 nmol kg −1 , except for measurements from station 177 off Baffin Island near Cape Dyer (683 m bottom depth), which displayed peak CH4 concentrations of 38 nmol kg −1 at 247 m depth (salinity 33.5 PSS and potential temperature −0.85 o C, Figure 3e). Seafloor CH4 seeps in the Cape Dyer region have been previously reported (Punshon et al., 2014(Punshon et al., , 2019, and the peak CH4 concentration occurring at mid-depth suggests the sample collection occurred offshore of the seeps.  Figure 1 for regional boundaries). Temperature/salinity plots (a-j) showing the CH4 and N2O concentrations of water masses, and ΔN2O versus apparent oxygen utilization plots (k-o) showing the potential density anomaly of different water masses. On (a-j), the dashed lines represent the potential density anomaly (σϴ, kg/m 3 ) and the upper and lower black rectangles denote the temperature/salinity properties of the BCSW and BCWW, respectively. The red lines indicate the potential density anomaly range 25.5 to 26.9 kg m −3 that is highlighted on Figures 2, 5, and 7. On panel (m), the regression line for all Beaufort Sea samples with σϴ ≤ 26.9 kg m −3 is shown with a black dashed line. Note panels (k-n) use a different AOU range than panel (o).
In the Bering, Chukchi and Beaufort Seas, samples with the highest N2O concentrations were typically associated with the BCWW as well as the coldest, most saline waters of the BCSW (potential density anomaly 25.5 to 26.9 kg m −3 ). This observation is consistent with trends reported observed in 2015 (Fenwick et al., 2017). The magnitude of the supersaturation of N2O within the BCWW was similar in the Chukchi and Beaufort Seas. This N2O supersaturation was partially attenuated in the Northwest Passage and strongly attenuated in Baffin Bay and Davis Strait, where the influence of higher-salinity Atlantic Water becomes substantial.
Data from the shallow Bering and Chukchi Seas showed no consistent relationship between ΔN2O and AOU (r 2 values for a linear regression between the two parameters were 0.05 and 0.04 in the Bering and Chukchi Seas, respectively). The lack of a correlation between N2O and O2 could be interpreted as evidence that water column nitrification is not the dominant source of N2O in these regions (Nevison et al., 2003;Yoshinari, 1976). However, because the whole water column is ventilated annually, physical dynamics may act to erase the biogeochemical signature of nitrification on the O2 and N2O distributions. For example, Reeve et al. (2019) found that water column ventilation in the Chukchi Sea leads to a negative bias in fixed N losses estimated from N2/Ar as compared to estimates from nutrient ratios (N:P), which are not affected by ventilation.
In the Beaufort Sea, there was a weakly positive correlation between ΔN2O and AOU at σϴ ≤ 26.9 kg m −3 (r 2 = 0.40, p < 0.0001). This relationship reflects mixing between the surface waters, which were near equilibrium, and subsurface waters from the Beaufort/Chukchi Seas (BCWW/BCSW). In the dense Atlantic Water of the Beaufort Sea (σϴ > 26.9 kg m −3 , salinity > 33.5 PSS), N2O concentrations were undersaturated, reflecting the lower atmospheric concentrations at the time of ventilation of these water masses, i.e. 270 ppb compared to presentday values of 330 ppb (Druffel et al., 2017;MacFarling Meure et al., 2006). Within the Northwest Passage, ΔN2O was consistently close to 0 and showed no relationship with AOU (r 2 = 0.0001).
In Baffin Bay and Davis Strait there was a weakly positive correlation between ΔN2O and AOU across the entire dataset (r 2 = 0.21, p < 0.0001). However, this correlation was primarily driven by the ΔN2O and AOU increase in the deep waters of Baffin Bay below ~1000 m (Figure 3j, σϴ ~27.7 kg m −3 potential temperature ~0 o C and salinity ~34.5 PSS; Figure 3o, AOU > 140 μmol kg −1 ) and was not present throughout the region. The ΔN2O/AOU correlation in deep Baffin Bay is likely not a nitrification signature. Lehmann et al.(2019) used N2O isotopomer data to demonstrate that the excess N2O in deep Baffin Bay was generated by incomplete sedimentary denitrification and upward diffusion, whereas the AOU signature was driven by aerobic respiration in the water column.

CH4 and N2O in rivers
Although the Arctic Ocean represents less than 1% of total ocean volume, it receives over 10% of global river discharge, and this strongly seasonal discharge is increasing due to global warming (Blunden & Arndt, 2019;Peterson, 2002). The sparse measurements of CH4 and N2O in Arctic rivers make it difficult to quantify current and predict future impacts of rivers on current Arctic Ocean greenhouse gas budgets.
We collected 25 river CH4 and N2O measurements from 19 different rivers between 2017-2019 (three rivers were sampled in both 2017 and 2018), at latitudes ranging from 68 to 80 o N (Figure 4). Sampling occurred between 31 July and 26 August each year. Overall, the results indicate that rivers were not a significant source of CH4 or N2O to the North American Arctic Ocean at this time of year. For example, the median river CH4 concentration was 7 nmol kg −1 and 12 of 25 measurements were within 2 nmol kg −1 of equilibrium. Additionally, 15 of the 25 measurements were undersaturated in N2O (concentration range 12-19 nmol kg −1 , median 15 nmol kg −1 ). All samples from Ellesmere Island were supersaturated in N2O and rivers south of Lancaster Sound were undersaturated or near equilibrium in N2O. The origin of the N2O undersaturation is not known and could reflect physically or biologically-induced N2O undersaturation in melting ice and snow, denitrification consuming N2O in sediments, and/or temperature changes that occur faster than the re-equilibration timescale (Quick et al., 2019;Randall et al., 2012).
CH4 concentrations displayed significant variability across rivers, but no consistent regional trends. For example, the highest observed CH4 concentration, 76 ± 4 nmol kg −1 , was found from a stream on the surface of marine-terminating Eugenie Glacier on Ellesmere Island, and the second highest CH4 concentration, 24 ± 2 nmol kg −1 , occurred in the Tingmeak River on mainland North America (Figure 4). Other studies have reported similarly elevated CH4 levels in glacial rivers and air in Greenland (Christiansen & Jørgensen, 2018;Lamarche-Gagnon et al., 2019). Seasonally-resolved measurements from a small river in the North American Arctic demonstrated a strong seasonality in riverine concentrations of CH4 (but not N2O), with the highest concentrations occurring during the early melt season (Manning et al., 2020). Our measurements in late July to late August may thus not reflect peak annual concentrations of CH4 in Arctic rivers. Observations of both gas concentrations and river discharge throughout the discharge season are needed to better constrain annual greenhouse gas emissions from Arctic rivers.

Interannual variability in CHand N2O distributions and sea-air fluxes
Our sampling program across four consecutive years enabled us to examine interannual trends in CH4 and N2O distributions. We first discuss CH4 (sections 4.1-4.4) and then N2O distributions (sections 4.4-4.7), and then consider the sea-air fluxes (section 4.8).

CH4 variability at repeat stations along the transect between the Bering Sea and Northwest Passage
Across all years and sampling depths, mean and median CH4 concentrations were consistently close to atmospheric equilibrium in the Bering Sea (mean 5.0 and median 4.6 nmol kg −1 ) and the Northwest Passage (mean 5.2 and median 4.6 nmol kg −1 ). In contrast, CH4 distributions displayed strong interannual variability and a wider range in concentrations in the Chukchi and Beaufort Seas, likely due to temporal variability in sedimentary CH4 sources, as well as heterogeneous production and consumption of CH4 within the water column (Coffin et al., 2013;Damm et al., 2005Damm et al., , 2015Fenwick et al., 2017;Kitidis et al., 2010). Figure 5 shows concentration profiles at six representative stations that were sampled in all four years. Many of the stations with the greatest interannual variability in CH4 were located in the Beaufort Sea along the continental shelf and slope (see also section 4.2), much of which lies within the gas hydrate stability zone (Lorenson et al., 2016). CH4 profiles from the Beaufort Sea frequently displayed one or more subsurface maxima. For example, at station BL4, in all four years, there was a subsurface CH4 peak at 50-70 m depth (σϴ = 24.7-25.8 kg m -3 ), but the peak CH4 concentration varied from 9.5-26.2 nmol kg −1 (Figure 5d). In 2017 and 2018 there was a second, deeper, CH4 peak at BL4 at σϴ ≈ 27.7 (12.0 nmol kg −1 at 280 m depth in 2017 and 11.7 nmol kg −1 at 160 m depth in 2018) that was not observed in 2015 and 2016. For both BL4 and MK3, the highest CH4 concentrations were found in 2016 (up to 26 nmol kg −1 , at σϴ ≈ 25.8 kg m −3 ). These trends suggest significant interannual variability in CH4 release from shelf and slope sediments, and potential variability in water column CH4 production and consumption. Compared to stations BL4 and MK3, stations SLIP-1 in the Bering Sea, and AG5 and 312 in the Northwest Passage, displayed limited interannual variability and a narrower range in CH4 concentrations. There were no apparent relationships between CH4 concentrations and hydrographic water column properties across sampling years, providing further evidence for the importance of intermittent sedimentary sources of variability.  Figure 1). Profiles of CH4 (b-g) and N2O (h-m) concentrations as a function of potential density anomaly (σϴ); station names and bottom depths are indicated above each column. The grey bar highlights the density surface σϴ = 25.5-26.9, which represents dense waters of the BCSW/BCWW that are transported into the Beaufort Sea. The black numbers represent distance along the transect in km.
Interannual variability in CH4 distributions was evident along a transect through the Bering and Chukchi Seas south of Barrow Canyon (Figure 6). In the Bering Sea, CH4 concentrations were consistently near equilibrium, apart from the bottom water at one station in the northern Bering Sea in 2018 (Figure 6j). The Chukchi Sea displayed higher CH4 concentrations than the Bering Sea, but the locations and magnitude of peak CH4 concentrations varied from year to year. In many cases, the highest CH4 concentrations in the Chukchi Sea occurred adjacent to the seafloor (Figure 2b), suggesting a sedimentary source, which is consistent with published measurements of CH4 in surficial sediments in this region and reports of subseafloor permafrost in the eastern Chukchi Sea (Collett et al., 2011;Matveeva et al., 2015). In 2016 and 2018, the highest CH4 concentrations occurred within the densest waters of the BCWW/BCSW (potential density ≥ 25.8 kg m −3 ), whereas the highest CH4 concentrations in 2015 and 2017 were found in less dense waters. These trends likely reflect the complex interplay of biogeochemical processes (sedimentary CH4 production and water column CH4 oxidation) with physical processes, such as water column stratification and sea ice formation and melt, which will affect the rate at which excess CH4 is ventilated to the atmosphere.

CH4 spatial and temporal variability on a cross-slope transect in the Beaufort Sea
Repeat measurements along the BL hydrographic line in the Beaufort Sea, which spans an onshore-offshore transect, also exhibited strong interannual variability in CH4 distributions (Figure 7). In 2016, a subsurface plume of CH4 (up to 26 nmol kg −1 ) appeared to diffuse from the sediments at ~200 m depth and be transported laterally offshore. Observations from 2017 and 2018 displayed lower peak CH4 concentrations at each station, and in the both years there was a local minimum in CH4 between 150-200 m depth at some of the off-shelf stations. This middepth CH4 minimum may reflect spatial and temporal variability in CH4 release along the Beaufort slope, which has been reported in other regions (Philip et al., 2016;Veloso-Alarcón et al., 2019). The CH4 distributions did not display a consistent correlation with depth, density, O2, or chlorophyll (Figure 7). Similar temporal and spatial variability in CH4 distributions was observed at other stations along the Beaufort Sea shelf and slope. . These observations support our hypothesis that the observed CH4 supersaturation on the Beaufort slope is driven by intermittent sedimentary release, rather than water column production.

CH4 distributions in the Northwest Passage, Baffin Bay, and Davis Strait
As previously discussed (section 3.2, Figure 2d-e), CH4 concentrations in the Northwest Passage, Baffin Bay, and Davis Strait were generally low. Some seafloor seeps are present along the coast of Baffin Island (Cramm et al., 2021;Punshon et al., 2019). For example, as reported in Cramm et al. (2021), we measured bottom water CH4 concentrations at the Scott Inlet methane seep ranging from 9 to 592 nmol kg −1 during five repeat measurements over 24 h in 2017, and video surveys with a remotely operated vehicle confirmed the presence of active ebullition from the seafloor at this site. Punshon et al. (2019) reported bottom-water CH4 concentrations of ~60 nmol kg −1 near Scott Inlet in 2012.
The sampling stations in the Northwest Passage, Baffin Bay, and Davis Strait varied significantly from year to year, making detection of interannual variability more challenging in these regions. Nonetheless, nearly all measured CH4 concentrations in these regions were less than 10 nmol kg −1 (median 3.8, and 0.05-to 0.95-quantile 1.0 to 8.6 nmol kg −1 ), suggesting limited variability, and in good agreement with previous observations (Kitidis et al., 2010;Punshon et al., 2014Punshon et al., , 2019.

CH4 isotopic measurements support occurrence of water column oxidation
Paired measurements of CH4 concentration and isotope ratios can provide information on CH4 sources and sinks (Whiticar, 1999). Dissolved CH4 in equilibrium with the atmosphere has a δ 13 C-CH4 of ≈ −46.7‰, and a concentration of 4 nmol kg −1 CH4 at S = 31 PSS, T = 0 o C. This δ 13 C-CH4 value is based on atmospheric measurements of −47.5‰ at Barrow, Alaska from 2015-2018 (White et al., 2018) and an 0.8‰ enrichment in 13 C in water relative to air due to kinetic isotope fractionation during air-water gas transfer (Happell et al., 1995;Knox et al., 1992).
Microbial oxidation of CH4 to CO2 (which occurs through the intermediate CH3OH) acts to enrich the residual CH4 in 13 C, resulting in an increase in δ 13 C-CH4 (Coleman et al., 1981). The effect of CH4 oxidation on the composition of CH4 can be approximated with the following modified Rayleigh fractionation equation (Coleman et al., 1981;Mariotti et al., 1981) C 13 = C 13 ,0 + ( Here f represents the fraction of reactant that is remaining, the subscripts R and P represent reactant and product, respectively, and the subscript 0 represents the initial value (at f = 1). The fractionation factor (α) is defined as: ( Other studies have fit CH4 concentration and isotope data to Rayleigh fractionation equations and interpreted the resulting patterns to determine the extent of water column CH4 oxidation (Damm et al., 2005(Damm et al., , 2007Li et al., 2021). Fenwick et al. (2017) showed that measurements of δ 13 C-CH4 and [CH4] from the Bering and Chukchi Seas in 2015 (water column depth ≤112 m) were consistent with a Rayleigh fractionation curve with initial [CH4] of 30 nmol kg −1 , initial δ 13 C-CH4 of −40‰ vs. VPDB and a fractionation factor 13 αCH4/CH3OH = 1.002. This fractionation factor was similar to results reported in other studies, which are typically in the range of 1.002 to 1.025 (Coleman et al., 1981;Damm et al., 2005Damm et al., , 2007Damm et al., , 2015Li et al., 2021;Whiticar, 1999;Whiticar & Faber, 1986). This result suggested that CH4 in the Bering and Chukchi Seas reflected a relatively homogeneous source (uniform initial concentration and isotopic composition) and that this CH4 underwent partial oxidation in the water column.
With our expanded multi-year dataset, we show that the Beaufort Sea data cannot be represented with a single Rayleigh relationship, and that some CH4 in the Beaufort Sea is more 13 C-depleted than CH4 in the Bering/Chukchi Seas. The broader range of δ 13 C-CH4 values in the Beaufort Sea reflects the wider range of water masses present in this region, including CH4-rich waters transported eastward from the Chukchi Sea, and CH4-depleted waters from the Atlantic, all of which are likely impacted by in situ oxidation. The range of δ 13 C-CH4 values also suggests that some of the endmember CH4 sources within the Beaufort Sea may have a different isotopic composition compared to the Bering/Chukchi Seas. Specifically, the highest CH4 sample in the Bering and Chukchi Seas had a CH4 concentration of 30 nmol kg −1 and δ 13 C-CH4 of −41‰. In contrast, the Beaufort Sea dataset from 2016 had four samples with CH4 ranging from 24 to 27 nmol kg −1 and δ 13 C from −51 to −47‰ (collected at stations MK1, MK4, and BL1). The Beaufort Sea isotope data also reflect the spatiotemporal variability of CH4 in the Beaufort Sea: in 2015, CH4 concentrations at station MK1 ranged from 2 to 9 nmol kg −1 and δ 13 C-CH4 ranged from −39 to −36‰, whereas in 2016, the maximum CH4 concentration at MK1 was 27 nmol kg −1 , with a δ 13 C-CH4 of −47‰ (Figure 8 c-d).
Figure 8. CH4 isotope data from 2015-2017. Relationship between CH4 and δ 13 C-CH4 for (a) all data and (b) Beaufort Sea data only. The white star represents CH4 at equilibrium with the atmosphere. The purple line in (a) is a Rayleigh fractionation curve fit to the Bering/Chukchi data (purple triangles) following Fenwick et al. (2017) and the blue line in (a-b) represents a Rayleigh fractionation curve fit to data from the Beaufort Sea with σϴ ≤ 26.9 kg m −3 and CH4 ≤ 11 nmol kg -1 . In (b) the samples with σϴ ≤ 26.9 kg m −3 are indicated with diamonds and samples with σϴ > 26.9 kg m −3 are indicated with squares with a black outline. The bottom row shows depth profiles of (c) CH4 and (d) δ 13 C-CH4 from stations MK1, MK2 and BL4 on the Beaufort Sea shelf (see Figure 1 for locations of the MK and BL transect lines). Note that the vertical sampling resolution is lower for the δ 13 C-CH4 measurements than for CH4.
Our δ 13 C-CH4 and [CH4] data in the Beaufort Sea are comparable with previous measurements of δ 13 C-CH4 from the water column and seafloor in this region. On the Beaufort Sea shelf near Prudhoe Bay, Alaska (water depth from 2-38 m), Sparrow et al. (2018) reported δ 13 C-CH4 of −72 to −49 ‰ at CH4 concentrations of 10 to 51 nmol kg −1 , and used radiocarbon analyses to demonstrate the CH4 was a mixture of modern and ancient sources. Hart et al. (2011) reported δ 13 C-CH4 of −55 to −59 ‰ in a methane hydrate core collected from the Beaufort Sea slope (2538 m depth). Our samples from the Beaufort Sea had δ 13 C-CH4 ranging from −52 to −33‰ at sample depths ranging from 5 to 1529 m (bottom depths from 60 to 2096 m). The observations from other research groups support our conclusion that most of our Beaufort Sea samples experienced some degree of water column CH4 oxidation, based on their enriched 13 C signatures relative to the methane hydrate samples, and samples from shallower shelf waters with shorter gas residence times. However, more data are needed to better characterize the spatial variability in δ 13 C-CH4 across the range of potential sedimentary CH4 sources affecting isotope signatures in the Arctic water column.
By limiting our analysis of the Beaufort Sea dataset to only CH4 concentrations ≤ 11 nmol kg −1 and density surfaces ≤ 26.9 kg m −3 (including the BCWW/BCSW and fresh surface waters, but not denser, Atlantic-derived waters), we were able to fit a Rayleigh curve with initial [CH4] of 11 nmol kg −1 , initial δ 13 C-CH4 of −42‰ vs. VPDB and a fractionation factor 13 αCH4/CH3OH = 1.006 (Figure 8b). This Rayleigh relationship may reflect a baseline CH4 source that was present throughout the lower density waters within the Beaufort Sea. A small number of samples from the Beaufort Sea had higher CH4 concentrations (≥15 nmol kg −1 ) and lower δ 13 C-CH4 values (≤−45‰), which may reflect sporadic sedimentary sources (e.g., 50 m depth at station MK1 in 2016, Figure 8c-d). High CH4 concentrations within the Beaufort Sea appear to be rapidly oxidized and/or mixed with lower CH4 waters, resulting in strong interannual variability in CH4, as shown in Figures 5 and 7. These processes resulted in localized effects on the CH4 concentration-isotope relationship.
In the Northwest Passage, Baffin Bay, and Davis Strait, δ 13 C-CH4 ranged from −43 to −32‰ at CH4 concentrations of 1-8 nmol kg −1 , with no clear relationship between concentration and isotopic composition (Figure 8a). All samples showed 13 C enrichment relative to present-day atmospheric equilibrium (−46.7‰), and many of the subsurface samples were undersaturated in CH4, supporting the occurrence of CH4 oxidation in the water column (Punshon et al., 2014). The lower CH4 concentrations observed in the Northwest Passage and Baffin Bay likely reflect the reduced importance of sedimentary CH4 inputs at the stations sampled in these regions, compared to the Beaufort and Chukchi Seas.
Isotopic measurements were also performed on CH4 samples from seven rivers draining into the Northwest Passage collected in 2017. The river with the highest CH4 concentration (15 nmol kg −1 ) had δ 13 C-CH4 of −48‰ and the two samples with the lowest CH4 concentration (4 nmol kg −1 ) both had δ 13 C-CH4 of −42‰ (Figure 8a). Given the large range of geological conditions across the rivers sampled, and likelihood of variability in δ 13 C-CH4 source signatures, more data from individual rivers (e.g., at multiple locations along a single river) would be needed to verify the importance of CH4 oxidation and characterize the source isotope signatures in these rivers.

N2O variability at repeat stations between the Bering Sea and Northwest Passage
N2O concentrations in the shallow Bering and Chukchi Seas showed significant interannual and spatial variability ( Figure 5 and 6). This variability may reflect heterogeneity in sedimentary or water column N2O production as suggested previously (Fenwick et al., 2017;Hirota et al., 2009;Toyoda et al., 2021;Zhang et al., 2015), and could also be influenced by interannual differences in physical processes, such as water column stratification, or sea ice cover. The highest N2O concentrations along our surveys were most frequently associated with the BCWW and BCSW (potential density anomaly σϴ > 25.5 kg m −3 , Figure 5). Elevated N2O concentrations can accumulate in these dense waters when they are isolated from the atmosphere due to water column stratification, with subsequent ventilation to the atmosphere when the water column is mixed. Fenwick et al. (2017) reported an increase in N2O concentrations along the northeast flow path of BCWW toward the Beaufort Sea. In contrast, we did not detect such a pattern in our fouryear data set ( Figure 6). Rather, our results were more indicative of localized, transient N2O accumulation and in the shallow Bering and Chukchi Seas, with regional and temporal differences in water column stratification and sea ice cover influencing ventilation of N2O at some stations and subsurface accumulation at others. A positive correlation between N2O and AOU is frequently used as evidence that water column nitrification is the dominant source of N2O to the water column (Nevison et al., 2003;Yoshinari, 1976). In our measurements from the Bering, Chukchi, and Beaufort Seas, we did not observe such a correlation ( Figure 3). As discussed in section 3.3, one possibility is that rapid ventilation of water in the shallow Bering and Chukchi Seas may have attenuated any N2O excess generated from nitrification, such that variability in observed N2O concentrations may reflect removal processes rather than source terms.

N2O variability on a cross-slope transect in the Beaufort Sea
We observed a persistent subsurface maximum in N2O associated with the BCSW/BCWW on the Beaufort Sea slope and in adjacent deeper waters. Figure 7 shows this feature along a cross-slope transect along the BL repeat hydrographic line in the western Beaufort Sea, and similar trends were present at other stations in the Beaufort Sea. The N2O peak concentration at each station was typically 18-19.5 nmol kg −1 (compared to equilibrium of ~16 nmol kg −1 ), and centered at a potential density anomaly σϴ of ~26.5 kg m −3 and a depth of ~150 m. The peak N2O concentration for each profile in the Beaufort Sea was similar to peak concentrations observed in the bottom waters of the Bering and Chukchi Seas (Figure 2g-h). In the deeper (Atlantic-derived) waters, N2O concentrations decreased and became undersaturated, reflecting lower atmospheric concentrations at the time of ventilation of the Atlantic Water (minimum N2O concentrations of ~12 nmol kg −1 ; equilibrium relative to present-day atmospheric concentration is ~15 nmol kg −1 ).

N2O variability in the Northwest Passage, Baffin Bay, and Davis Strait
Our sampling stations varied significantly from year to year within the Northwest Passage, Baffin Bay, and Davis Strait. However, N2O concentrations throughout these regions were usually close to equilibrium values (Figure 3), with the exception of the deepest waters of Baffin Bay where N2O supersaturation was associated with sedimentary denitrification and upward diffusion into the deep waters (Lehmann et al., 2019). The surface waters of the Northwest Passage, Baffin Bay/Davis Strait, and the Beaufort Sea were frequently undersaturated in N2O and this trend was likely driven by physical processes. Melting sea ice is a potential source of N2O-undersaturated water, as reported by Randall et al. (2012) from sea ice samples collected in the Beaufort Sea and Northwest Passage. Solubility effects, resulting from the strong temperature-dependence of N2O solubility (increasing by 4% for every 1 o C temperature decrease) could also lead to undersaturation of N2O. We conclude that these regions do not exhibit strong interannual variability in summer N2O distributions.

Sea-air fluxes of CH4 and N2O
We observed significant regional and interannual variability in CH4 and N2O fluxes (Table 1 and Figure 9). Across the four-year dataset, all regions acted as net sources of CH4. Median CH4 saturation anomalies ranged from 5 to 50%, while median short-term sea-air fluxes, integrated over the mixed layer residence time, ranged from 0.3 μmol m −2 d −1 in Baffin Bay to 2.1 μmol m −2 d −1 in the Chukchi Sea. N2O was close to equilibrium values in all regions (median saturation anomalies ranged from −3 to 3%), and short-term sea-air fluxes were lowest in the Beaufort Sea (median −1.0 μmol m −2 d −1 ) and highest in the Chukchi Sea (median 0.4 μmol m −2 d −1 ).
Of the five regions sampled, the Chukchi Sea displayed the highest median sea-air fluxes of both CH4 and N2O, and the strongest interannual variability (Table 1 and Figure 9). Our fouryear dataset suggests that the Bering and Chukchi Seas acted as a net N2O sink during 2015 but a net N2O source during 2016-2018, albeit with significant variability between stations (Figure 9). In this region, both CH4 and N2O sea-air fluxes were highest in 2016, when dense BCWW/BCSW was present near the surface at several stations ( Figure 6). In contrast, during conditions of strong water column stratification (due to the presence of sea ice and/or meltwater), the BCWW and BCSW can be isolated from the surface, limiting sea-air exchange and allowing elevated levels of CH4 and N2O to accumulate in subsurface layers. Subsequent water column mixing leads to the ventilation of these CH4 and N2O-rich waters to the atmosphere.
As observed in the Chukchi Sea, Bering Sea N2O fluxes were highest in 2016 (median 1.2 μmol m −2 d −1 ) and lowest in 2015 (median −0.7 μmol m -2 d -1 ) (Figure 9). The Bering Sea was a net source of CH4 in all four years, but median fluxes in this region were lower than in the Chukchi Sea, likely reflecting lower sedimentary CH4 sources. The Bering and Chukchi Sea sampling took place between mid to late July in each year, and annual variability in the timing and extent of ventilation of bottom waters could contribute to the observed interannual variability in sea-air fluxes. Other groups have also reported significant spatiotemporal variability in CH4 and N2O fluxes and distributions in the Chukchi and Bering Seas (Cline et al., 1986;Hirota et al., 2009;Zhang et al., 2015).
Across our four-year dataset, the Beaufort Sea, Northwest Passage, and Baffin Bay/Davis Strait all acted as net N2O sinks (median fluxes −0.2 to −0.9 μmol m −2 d −1 ), with the strongest ocean uptake occurring in the Beaufort Sea. These same regions acted as net CH4 sources (median fluxes 0.3 to 1.7 μmol m -2 d -1 ), with the largest sea-air flux occurring in the Northwest Passage. Most stations were supersaturated in CH4 at the surface, but the highest CH4 concentrations were typically observed in the subsurface. These trends may reflect competing production and loss terms, including release of CH4 from sediments, production in the upper water column and/or sea ice (Damm et al., 2005, sea-air gas exchange, and consumption via gas exchange. For example, upward transport of sediment-derived CH4 could potentially lead to elevated surface concentrations and sea-air fluxes at the shallowest stations where excess CH4 is not fully consumed below the mixed layer (Sparrow et al., 2018). Table 1. Short-term and annual CH4 and N2O sea-air fluxes (positive values represent a flux from the sea to the air), and surface saturation anomalies, based on pooled observations from 2015-2018. Median values include the 0.25-to 0.75-quantile range in square brackets, and mean values are reported with the standard deviation in parentheses. Short-term fluxes are calculated using a weighting scheme integrated over the residence time of the gas in the mixed layer, accounting for variability in wind speed and sea ice over this residence time (section 2.4). Annual fluxes are calculated using the wind speed, sea ice and sea level pressure over the calendar year in which each sample was collected (section 2.4). See Table S2 and Figure S1 for the boundaries for each region.  To situate our results in a broader context, we compared our CH4 and N2O flux estimates with other recently published values from our study region. In the Chukchi Sea, Heo et al. (2021) reported mean sea-air N2O fluxes of 2.3 ± 2.7 μmol m −2 d −1 in the southern Chukchi Sea and −1.3 ± 1.5 μmol m −2 d −1 in the Northern Chukchi Sea in summer 2017, based on instantaneous wind speeds at the time of sampling. Another study in the Chukchi Sea by Toyoda et al. (2021) reported sea-air N2O fluxes ranging from −4.2 to 2.6 μmol m −2 d −1 in September to October 2014 and 2015, based on instantaneous shipboard wind speeds. These results are comparable to our mean estimates of 0.5 ± 1.4 μmol m −2 d −1 in the Chukchi Sea in July of 2015-2018. In the Beaufort Sea, primarily in shallow shelf waters <100 m depth, Lorenson et al. (2016) estimated summertime CH4 sea-air fluxes in 1993-1995 and 2009 of ~16 μmol m −2 d −1 , assuming a mean surface CH4 of 12 nmol kg −1 . This flux rate is significantly higher than our estimates in this study, based on mean surface CH4 concentrations of 4.8 nmol kg −1 . We note that we sampled stations across a broader range of water depths in the Beaufort Sea, which likely resulted in the lower mean surface CH4 concentrations and sea-air fluxes compared to the study of Lorenson et al. (2016). In Davis Strait, Punshon et al. (2014) calculated mean sea-air fluxes of 1.6 μmol CH4 m -2 d -1 , while Gagné (2015) estimated N2O fluxes with mean of 5 (standard deviation 8) μmol N2O m −2 d −1 in the eastern Northwest Passage and Baffin Bay, and noted that different regions acted as sources and sinks.

CH4 fluxes
Overall, the estimated annual CH4 fluxes from our study region (representing 2.3 million km 2 or ~0.6% of the global ocean surface area) have a median value of 0.009 and mean of 0.015 Tg CH4 y −1 (Table 1). From this estimate, we conclude that the North American Arctic Ocean CH4 emissions represent a small fraction (~0.1%) of global oceanic emissions (mean 9 Tg CH4 y -1 , range 5-17 Tg CH4), as derived from the Global Methane Budget program (Saunois et al., 2020). Methane emissions from the North American Arctic Ocean are also significantly lower than the East Siberian Arctic Shelf, where sea-air fluxes of ~3 Tg CH4 y -1 have been estimated by Thornton et al. (2016Thornton et al. ( , 2020. Our results thus suggest that the North American Arctic Ocean currently plays a small role in the global ocean CH4 budget. The estimated annual N2O fluxes from our study region have a median of −0.003 and mean of 0 Tg N y −1 , with median surface concentrations across all regions within 3% of atmospheric equilibrium. A recent study estimated global oceanic N2O emissions are 4.2 ±1.0 Tg N y −1 (Yang et al., 2020), which is similar to previous estimates from the IPCC AR5 report of 3.8 (1.8-9.4) Tg N y -1 (Ciais et al., 2013). These results suggest that the North American Arctic Ocean also likely plays a small role in the global budget of N2O.
We note the potential for strong seasonality in near-surface gas saturation state and concentrations in the Arctic Ocean, for example due to accumulation of gases under ice during winter and rapid ventilation during ice melt (Kvenvolden et al., 1993;Manning et al., 2020). This seasonality and its impact on sea-air fluxes could not be assessed in this study, due to the timing of the field campaigns; however, this remains an important area for future work.
Based on extensive measurements between 2015-2019, we have characterized spatial and interannual variability in CH4 and N2O water column distributions and sea-air fluxes across a ~7000 km transect of North American Arctic Ocean. Our results show that N2O is generated within the highly productive Bering and Chukchi Seas and transported into the Beaufort Sea, where it forms a persistent subsurface maximum across years. CH4 distributions are more dynamic, with the highest concentrations and greatest variability occurring at repeat stations on the Beaufort Sea shelf and slope, likely under the influence of sedimentary sources. Calculated sea-air fluxes, accounting for variability in wind speed and sea ice coverage, demonstrate that the North American Arctic Ocean currently plays a small role in the global budgets of CH4 and N2O. Continued monitoring programs will enable scientists to determine how anthropogenic climate change is altering the cycling of CH4 and N2O in these regions.

Data Availability Statement
All CH4, N2O, temperature, salinity, pressure, O2 and δ 13 C-CH4 data, as well as derived sea-air flux estimates, have been submitted to PANGAEA, and a DOI will be included in the final published manuscript. For early access to the data, please contact the corresponding author (cmanning@alum.mit.edu).