Nitrous Oxide Distributions in the Oxygenated Water Column of the Sargasso Sea

ABSTRACT This study presents dissolved nitrous oxide (N2O) concentrations in the water column at the Bermuda Atlantic Time-series Study (BATS) station and uses a subset of these measurements to estimate air-to-sea flux for four specific time points between September 2018 and June 2019. N2O concentrations at BATS were in the range of 4.0 nmol L−1–16.9 nmol L−1, with vertical profiles which were the mirror inverse of dissolved oxygen. Regardless of season, N2O concentration maxima were found within the oxygen minimum zone (OMZ). The highest maximum N2O values were observed in November and lowest in October. As the water column at BATS remains consistently at dissolved oxygen concentrations greater than 140 µmol L−1, and therefore aerobic, we assume that the bulk of N2O production occurs through nitrification. A nitrification source is supported by a correlation between excess N2O (ΔN2O) below the mixed layer, apparent oxygen utilization (AOU) and nitrate concentrations. We estimate a pooled average yield of 0.027% to 0.038% N2O from nitrification at BATS. Finally, estimates of air–sea exchange of N2O using regional average monthly wind speeds indicated that this region acts as a weak source or a sink of atmospheric N2O, and varies between months.


Introduction
Nitrous oxide (N 2 O) is an important greenhouse gas, with an atmospheric warming potential 273 times that of carbon dioxide over a hundred-year time span in the troposphere (Eyring et al., 2021), and additional potential for ozonedepleting photochemical reactions in the stratosphere. Therefore, an increase of N 2 O can destabilize climate through alteration of the radiative heat balance and increased exposure of sensitive ecosystems to damaging UV-radiation (Ravishankara et al., 2009). The open ocean is generally accepted to contribute one quarter to one third of the nitrous oxide flux to the atmosphere, or 4.2 ± 1.0 Tg N yr −1 (Alley et al., 2007;Ehhalt et al., 2001;C. D. Nevison et al., 1995;Pachauri et al., 2014;Yang et al., 2020). Atmospheric N 2 O concentrations have increased by over 18% since time series measurements began at the Mauna Loa Observatory in 1978 (Aneja et al., 2019), to the global average levels of 330.90 ppb in 2018 when the data in this study were collected (Lan et al., 2022). As of 2022, the global average level is 334.31 ppb (Lan et al., 2022). In the Sargasso Sea, average atmospheric measurements of N 2 O from the Tudor Hill station in Bermuda between 2018-2020 place the current concentration at approximately 332 ppb (NOAA, 2020), which also corresponds with the latitudinal zonal average concentrations.
Microbially mediated reactions, which are modulated by reduction-oxidation potential, can produce and consume N 2 O in the marine environment. Nitrification, a process which produces nitrate from ammonia using oxygen as the electron acceptor, generally proceeds to completion via two groups of organisms: ammonia oxidizers, which convert ammonia to nitrite, and nitrite oxidizers, which convert nitrite to nitrate. Ammonia oxidizers can be further divided into ammonia-oxidizing archaea (AOA) and ammonia-oxidizing bacteria (AOB), which perform this reaction using distinct pathways. However, both groups generate N 2 O as a minor by-product (Santoro et al., 2011;Stein, 2019). Excess N 2 O concentrations (hereinafter referred to as ΔN 2 O; the difference between the measured concentration and the saturation concentration) correlate with apparent oxygen utilization, where nitrification is the primary contributor to the biogenic N 2 O pool (Bange, 2008). In AOB, N 2 O production can occur via 'nitrifier-denitrification', in which nitrite is reduced to N 2 O, or through hydroxylamine oxidation followed by nitric oxide oxidation (Kozlowski et al., 2016;Ward, 2013;Wrage-Mönnig et al., 2018). N 2 O production by nitrifiers is enhanced at low dissolved oxygen (DO) concentrations, presumably due to increased nitrifier-denitrification as a method of anaerobic respiration (Codispoti & Christensen, 1985;de Bie et al., 2002;Grundle et al., 2017;Keeling et al., 2010;Naqvi et al., 2010). The mechanism of N 2 O production by AOA has yet to be fully defined but may be a 'hybrid' mechanism that includes both hydroxylamine and nitrite, and nonenzymatic downstream reactions of these two compounds (Carini et al., 2018;Frey et al., 2020;Kozlowski et al., 2016;Stieglmeier et al., 2014). However, the possibility of direct biological production of N 2 O by AOA cannot be discounted. Interestingly, there is some evidence of AOA being able to produce N 2 O under suboxic or functionally anoxic conditions through an oxygen and dinitrogen production pathway, such that AOA may be important contributors to N 2 O accumulation in both oxic and anoxic waters (Kraft et al., 2022;Martens-Habbena & Qin, 2022;Trimmer et al., 2016).
Heterotrophic denitrification produces N 2 O as an intermediate during the reduction of NO 3 − to N 2 (NO 3 . The expression of the bulk of the enzymes involved (e.g. nitrite reductase, N 2 O reductase, nitrate reductase) is modulated in part by dissolved oxygen concentrations, such that denitrification is typically limited to environments in which DO is below 10 µmol L −1 (Capelle & Tortell, 2016;Fenwick & Tortell, 2018;Grundle et al., 2017;Jameson et al., 2021). Empirically, N 2 O production via denitrification occurs at broad oxic/suboxic interfaces in marine water columns (Voss et al., 2013). However, the potential of smaller scale, redox 'microniches' within particles or aggregates with low internal DO concentrations in otherwise aerobic water columns to contribute to the global budget as N 2 O sources is understudied (Stief et al., 2016), though these may sink rapidly out of the surface ocean. N 2 O consumption, previously thought to be primarily mediated by anaerobic denitrifiers (Devol, 2008), has been found to occur both in oxic and anoxic waters though observations of the former are more limited. The nosZ gene, which encodes for N 2 O reductase, is broadly distributed across these different oxygen regimes, and variants of it have been identified in both denitrifiers and non-denitrifiers (Bertagnolli et al., 2020;Rees et al., 2021;Spiro, 2012;Sun et al., 2017Sun et al., , 2021. There is abundant evidence that DO concentrations in the global oceans are decreasing as a direct result of anthropogenic climate change (Keeling et al., 2010;Levin, 2018;Matear & Hirst, 2003;Plattner et al., 2001;Sarmiento et al., 1998). While the direct impacts of climate change on N 2 O are still actively under research, declining DO in the ocean may increasing N 2 O production. Warming associated stratification may have an opposing effect (Landolfi et al., 2017). However, given the relationship between DO and N 2 O, research has largely focussed on N 2 O concentrations in the major low-oxygen ocean regions (Ji et al., 2015;Sun et al., 2017). Though systems with oxic water columns have been studied (e.g. Kock & Bange, 2015;Nevison et al., 2003;Yang et al., 2020) they are assumed to be unlikely to represent important sources of N 2 O to the atmosphere. However, as global DO levels decrease, oxic regions that are net neutral in regards to N 2 O flux or potential N 2 O sinks may transform into sources of N 2 O to the atmosphere. Here we report on updated N 2 O distributions, yield, and estimate air-sea fluxes of N 2 O at the Bermuda Atlantic Time-series Study (BATS) station in the persistently oxic water column of the Sargasso Sea (Lomas et al., 2013). Characterizing well-ventilated environments like the Sargasso Sea and updating measurements in previously studied locations is critical in order to increase our overall understanding of N 2 O distributions in the ocean and air-sea exchange, and to better validate our models of N 2 O flux in the future.
2 Methods a Sampling Regime N 2 O samples were collected at BATS (31°40 ′ N 64°10 ′ W) on five cruises conducted onboard the R/V Atlantic Explorer, taking place in September, October, and November 2018, and June 2019. Cruise dates are summarized in Table 1. Samples were collected from the surface, and 10,20,40,60,80,100,120,140,160,200,250,300,400,600,800,1000,1600,2000,2400,3000,3400, and 4000 m. We adjusted sampling depths where necessary to accommodate the locations of the deep chlorophyll maximum and the oxygen minimum zone. Continuous vertical profiles of salinity, temperature, and dissolved oxygen were obtained using a Seabird 9/11 Plus CTD (conductivity, temperature and depth) and auxiliary sensors attached to a rosette configuration of 24 × 12L Niskin bottles, which were used to collect seawater samples. CTD measurements were calibrated using bottle samples collected and processed by the BATS science team. We chose a full water-column sampling scheme in order to investigate the nitrous oxide pool in both the dynamic surface ocean with accompanying higher vertical resolution, and in the less variable deep ocean. Samples were taken on various days during the cruise period to accommodate BATS sampling objectives. Quality controlled nutrient data for the June 2019 was not available, so is not presented here nor used in metadata are outlined in Table 1. N 2 O data for an additional cruise in late October 2018 is available in the supplemental information ( Figure S1), as accompanying biogeochemical parameters were not available.
b Nutrients Samples for nitrite and nitrate were collected in acid-washed 15-mL Falcon tubes. Tubes were immediately frozen at −20˚C after collection and thawed just before analysis onshore. Samples were analysed using the method of Barwell-Clarke and Whitney (1996) on an Astoria Nutrient Autoanalyzer. Coefficient of variation for this method is 18.2%. The limit of detection (LoD) for nitrite was 0.05 µmol L −1 ; for nitrite and nitrate combined, the LoD was 0.2 µmol L −1 . Values below the LoD are represented as zero in relevant figures. All nitrite values were below LoD, so the nitrite and nitrate combined measurement is effectively nitrate for our observations. c Nitrous Oxide Measurements N 2 O samples were collected in triplicate into 20 mL glass serum bottles using tygon tubing. Bottles were overfilled three times to prevent contamination, and bubbles were avoided during filling. Bottles were filled to crowning and tightly stoppered with a butyl rubber stopper. The stopper was then sealed with an aluminum crimp cap, and samples were poisoned with 50 µL of saturated HgCl 2 using a syringe within one hour of sample collection to eliminate biological activity. Samples were then stored upside down and stored under dark and cool conditions until analysis ashore.
Analysis of N 2 O samples was performed by a combination of a purge-and-trap autosampler (model VSP4000, Mercury Instruments GmbH, Karlsfeld, Germany) with a small sample introduction module (SSIM) (model A0314, Picarro Inc., Santa Clara, CA, USA) and a cavity ring down spectrometer (CRDS) (model G5131i, Picarro Inc.). This method is described in detail by . In brief, samples were loaded onto the purge-and-trap system, where a dual needle system extracted the N 2 O with an N 2 carrier gas (Ultra High Purity 5.0 grade). N 2 O was then cryo-trapped using liquid nitrogen in the analytical trap, then heated for delivery to the SSIM. The SSIM coordinates sample delivery into the CRDS, as well as sample acquisition and the purge-and-trap activities of the VSP4000. Inside the CRDS, a mid-infrared laser alternates between on and off in order to measure the light dissipation rate, which correlates to the N 2 O concentration of the sample. The practical limit of detection as assessed by Ji and Grundle (2019) is 0.6 nmol L −1 , given through a blank measurement of 10 ± 2 ppb. Manufacturer specifications indicate a precision of <0.05 ppb. We utilized a 5-point standard curve for calibration. Standards were prepared in triplicate by flushing 20-mL vials with N 2 for 10 min to clear the vials of any contaminating N 2 O. Flushed vials were injected with a set volume (0, 25, 50, 75, 100 µL) of 80 ppm synthetic airbalance N 2 O reference gas (EMPA CB08976, Messer Schweiz AG, Lenzburg, Switzerland).
During analysis, we identified contaminated data by assessing if replicates were internally consistent or inconsistent with the cruise and depth trends. These data were removed for visualization and statistical analysis and are flagged in the supplementary data.

d Calculations
The mixed layer depth was identified by the first major changepoint in each density profile as identified by the MATLAB ischange function, which utilizes the pruned exact linear time method for changepoint detection (Killick et al., 2012). A temperature-salinity plot and temperature profile are shown in Fig. 1. Mixed layer depth and the temperature and salinity at the bottom of the mixed layer for all cruises are outlined in Table 2. Salinity is reported as absolute salinity, and calculated using the Gibbs Seawater Oceanographic Toolbox (McDougall & Barker, 2011), based on TEOS-10. Apparent oxygen utilization (AOU) with depth is shown for the full water column in Fig. 2. AOU was calculated using equation 1:   (Broecker & Peng, 2000) and thus, 270 ppb (Machida et al., 1995).
Ocean-atmosphere flux was calculated using equation 3: Where k N 2 O represents the gas transfer velocity of N 2 O in m h −1 , as estimated used the model of Wanninkhof (2014). k N 2 O depends on wind speed and Schmidt number. The Schmidt number was estimated as in Wanninkhof (2014) using the N 2 O diffusion coefficient measured by Wilke and Chang (1955) as adapted by Hayduk and Laudie (1974); and the dynamic viscosity of seawater as reported in Sharqawy et al. (2010). C W and C eq are the concentration of N 2 O in water and the surface equilibrium concentration with the atmosphere, respectively, in mol m −3 .

Results and discussion a Physical Characteristics and Hydrography
We observed the characteristic 18˚C mode water forming from 200-400 m on all cruises, with a salinity of approximately 36.7. The water column structure was similar between all cruises, though we observed a slight shoaling of the 27.2 σ θ isopyncal in both November and September   (Fig. 1). This may be caused by an encroaching Antarctic Intermediate Water mass (AAIW), reported to intermittently extend up to the latitudes of BATS (Machín & Pelegrí, 2009;Tsuchiya, 1989).
b Dissolved Oxygen and AOU In general, AOU declined from approximately zero at the surface to weakly negative near the base of the euphotic zone, presumably due to the influence of primary producers. Maximum AOU corresponded with the oxygen minimum zone for all cruises, then decreased to 65-70 µmol L −1 below 1500 m, after which concentrations were fairly consistent down to the bottom of the water column. In comparison to the subtropical North Pacific, AOU at intermediate depths of the subtropical North Atlantic is low and O 2 remains relatively high due the relative youth of the NADW in this region (Matsumoto, 2014). For each cruise, the DO concentrations were highest at depths below 2000m, ostensibly demarcating the top of the North Atlantic Deep Water. Surface values were also high and increased below the mixed layer, peaking between 60-120 m depth, corresponding approximately with the depth of the deep chlorophyll maximum. DO concentrations were lowest ( 150 µmol L −1 ) at the 27.2 σ θ isopycnal, corresponding to a depth of 800-890 m.
c Vertical Distribution of Nitrous Oxide N 2 O concentrations in the Sargasso Sea ranged from 4.0 ± 1.4-16.9 ± 1.9 nmol L −1 . The depth distribution of N 2 O concentrations were typically inverse to DO concentrations. Nitrous oxide profiles from each of our cruises show little variability between sampling dates, with concentrations increasing from the surface and reaching maximum concentrations at the depth of the OMZ (Fig. 3). Below the OMZ, all profiles decreased from maximum values to approximately 10 -12 nmol L −1 at 4000 m. The co-occurrence of highest N 2 O concentrations and lowest DO values has been well documented in many other oceanic regions (e.g. Butler et al., 1989;Cohen & Gordon, 1979;Grundle et al., 2012;Oudot et al., 1990;Walter et al., 2006). The elevated bottom water concentrations in the November profiles may be attributable to the shoaling of the 27.2 σ θ isopyncal, which we are considering AAIW. Walter et al. (2006) noted that in the 27.1-27.5 σ θ range in the North Atlantic, N 2 O concentrations were elevated, so increased intrusion of AAIW in November is a plausible cause for these observations. With the exception of North Atlantic low oxygen eddies (Grundle d N 2 O Saturation and Excess N 2 O in the mixed layer was largely undersaturated (Fig. 4), with the exception of September 2018 in which slight oversaturation was observed (Table 2). ΔN 2 O ranged from −1.6 ± 0.62 nmol/L to 10.5 ± 1.89 nmol/L throughout the water column. Average mixed layer ΔN 2 O values ranged from −1.6 ± 0.62 nmol/L to 1.7 ± 0.93 nmol L −1 , with a total observed average of −0.70 ± 0.90 nmol L −1 suggesting that in general, this region ranges from acting as a sink of N 2 O for the atmosphere to a weak source of N 2 O to the atmosphere. N 2 O saturation increased from the surface waters to the deep waters, with large seasonal variations in N 2 O saturation. As observed in N 2 O concentration profiles (Fig. 3), N 2 O saturation in the deep water was greatest in November. This may also be related to the intrusion of relatively N 2 O-rich AAIW apparent in November.
Below the 26 σ θ isopycnal, a significant positive linear correlation between ΔN 2 O and AOU was observed (pvalue < 0.01, adjusted R 2 : 0.51; Fig. 5a). Likewise, a positive correlation between ΔN 2 O and NO 3 concentrations was also observed (p-value <0.01, adjusted R 2 : 0.66; Fig. 5b.) Values above the 26 σ θ isopycnal, which is approximately 100 m water depth for these cruises, were not included in these calculations to avoid including depths in which oxygen is produced independently of N 2 O production/consumption reactions (i.e. photosynthesis). Further, nitrate and nitrite values in the surface ocean to at least 100 m were below the detection limit for the cruises analysed. For both relationships, regression analyses were performed on pooled data sets and individual cruises. Given the temporal difference between these data sets, we employed a Kruskal-Wallis H test to determine that there is no significant difference in medians between the September, October and November 2018 cruises for nitrate and nitrite concentrations, apparent oxygen utilization and ΔN 2 O, allowing us to pool these cruises for further comparison. Fig. 5 illustrates the pooled regressions and those from individual cruises. Separated data for individual cruises are available in the supplemental information ( Figures S2-S4). Positive linear correlations between ΔN 2 O, and both AOU and NO 3 − concentrations have been established as an indication that nitrification is primarily responsible for N 2 O formation (Bange, 2008;Grundle et al., 2012;C. D. Nevison et al., 1995;Walter et al., 2006). This is expected, given that, with the possible exception of hypoxic/anoxic microzones, the relatively highly oxygenated waters of the Sargasso Sea would not support an exponential increase in N 2 O yields from nitrification or lead to N 2 O production from canonical denitrification (Codispoti et al., 2001). Previously reported oceanic rates of DO consumption to N 2 O production ranged from 3000 to 33 000 moles of O 2 per mole N 2 O (Nevison et al., 1995). Based on the slope of the ΔN 2 O to AOU regression, we calculate that 15 100 moles of O 2 are consumed for each mole of N 2 O produced. Adhering to Redfield rationale in which 138 mol O 2 are used to remineralize 1 mol organic matter, and subsequently 24 mol O 2 are used for the oxidation of 16 mol NH 4 + , 17% of DO consumption can be attributed to NH 4 + oxidation (Grundle et al., 2012;Hsiao et al., 2014;Ward, 2008)  produce similar values. These yields are comparable to yields of N 2 O from nitrification estimated from data in the Eastern Tropical South Pacific and Eastern Tropical North Pacific (Ji et al., 2018), which ranged from 0.003-0.06% in oxic waters. Similarly, N 2 O yields in the Northeast Subarctic Pacific Ocean were estimated at 0.028-0.040% using similar methods (Grundle et al., 2012). Previous studies (e.g. Nevison et al., 2003) have noted that significant uncertainties exist when using simple linear regression to calculate N 2 O yield, especially when integrating over a large geographical region due to spatial variability in N 2 O yield and mixing effects. While our data were not spatially distributed, these issues are still important to address. Specifically, mixing between water mass end-members with different preformed nutrients, N 2 O and AOU compositions can influence the yield calculation. This mixing can provide the impression of a linear relationship due to N 2 O production (Nevison et al., 2003) when it is simply due to the qualities of distinct water masses. Nevison et al. (2003) recommend employing a twoend-member mixing model to avoid this isopycnal mixing effect. In order to do this effectively, we need a strong understanding of the influence of DO on N 2 O yields, and these estimates vary widely as discussed below. We therefore did not attempt to differentiate preformed N 2 O from in-situ production in our calculations. The ΔN 2 O relationships and yields we present should be interpretated as integrated values the span the history of the water masses, rather than strictly attributable to local N 2 O production. With these caveats, these linear relationships can still be a valuable metric for comparison of N 2 O yield. Fig. 6 shows the ΔN 2 O-AOU relationship defined in this study in the context of previous observations from a number of studies which are included in the MEMENTO database (Kock & Bange, 2015). ΔN 2 O vs. AOU regression lines from selected regions are illustrated in Fig. 6. Regression lines for further regions are shown in the Supplemental Information. The relationships observed in this study correspond well with the ΔN 2 O-AOU relationships for previous observations in the North Atlantic, Eastern Tropical North Atlantic, and Atlantic Meridional regions. These relationships are somewhat regionally constrained, with comparatively steeper slopes observed in the Equatorial Pacific, and a much shallower slope observed in the coastal region of Saanich Inlet, Canada (Capelle et al., 2018;Torres-Beltrán et al., 2017). We will briefly discuss some possible mechanisms for the variations in ΔN 2 O to AOU and NO 3 − relationships that we observe between the datasets within the MEMENTO database and our own data. Reported N 2 O yields from AOB in culture are in the range of 0.1% to 8%, depending on the organisms used, oxygen conditions, and growth supplements provided (Prosser et al., 2020 and references therein). AOB N 2 O production increases with decreasing DO. AOA cultures have N 2 O yields ranging from 0.04% to 0.3% (Prosser et al., 2020 and references therein). Field estimates of N 2 O production from AOA range from 0.003% to greater than 1% (Frey et al., 2020;Ji et al., 2018). The mechanisms of N 2 O production by AOA are not well elucidated and may involve the interplay of multiple separate pathways. Some of the complexities of our current understand of AOA and their interactions in the ocean have been reviewed in other recent works (e.g. Hutchins & Capone, 2022;Kim et al., 2021;Shafiee et al., 2021). Whether N 2 O from AOA is produced via nonenzymatic reactions of biological intermediates or directly from an as-yet-undefined enzymatic reaction, N 2 O production yield seems to negatively correlate with DO concentration as it does with AOB (Qin et al., 2017). In the ocean, AOA dominate the ammonia oxidizing community (Santoro et al., 2010(Santoro et al., , 2011Wuchter et al., 2006) and linked N 2 O production similarly holds this negative correlation (Ji et al., 2015(Ji et al., , 2018Trimmer et al., 2016). Thus, oxygen concentration may impact N 2 O production both by AOA and AOB. In coastal regions, the ratio of AOB to AOA tends to be higher than that of the open ocean, possibly due to differing metal availability as compared to the open ocean (Shafiee et al., 2021) and/or due to varying regional ammonium Fig. 6 Regional relationships between AOU and ΔN 2 O, compiled from the MEMENTO database (A. Kock and Bange, H.W. 2015). Data for this study (dark grey) are overlaid on all datapoints (light grey) from the MEMENTO database for which DO, N 2 O, temperature and salinity values existed. Saanich Inlet data provided through (Capelle et al., 2018;Torres-Beltrán et al., 2017). Linear regression fits for individual datasets are represented by the indicated lines. Cruises within a particular region share the same line style. AOU to ΔN 2 O relationships not pictured here but represented in the 'All Data' data points are illustrated in the Supplemental Information. availability, given that AOA generally have a higher affinity for ammonium than AOB (Kits et al., 2017). Given that AOA and AOB produce N 2 O via different mechanisms and have consequently different N 2 O yields. Further, low oxygen waters may also stimulate nitrifier-denitrification and/or partial denitrification, which would also lead to steeper ΔN 2 O to AOU slopes, such as observed in the oxygen minimum zones of the Eastern Tropical Pacific (Ji et al., 2015). Changing community composition between regions due to their biogeochemical characteristics and biological response to oxygen variations may contribute to the large variations in ΔN 2 O to AOU relationships between regions within the MEMENTO dataset and the data in this study.
e Air-Sea Exchange Due to undersaturation of the surface waters, the oceanatmosphere flux was negative for all sampled months excluding September. The largest negative N 2 O fluxes of −3.39 ± 0.68 mol km −2 d −1 and −2.19 ± 0.44 mol km −2 d −1 were observed in October 2018 and June 2019, respectively. November had a weakly negative flux of −1.08 ± 0.22 mol km −2 d −1 , and September was weakly positive (i.e. net N 2 O flux to the atmosphere), with a flux of 0.33 ± 0.07 mol km −2 d −1 . The uncertainty expressed here is due to the 20% uncertainty inherent in the Schmidt number estimation, as it is more than twice the average relative standard deviation of our N 2 O measurements (Wanninkhof, 2014). Estimates of N 2 O oceanatmosphere flux depend heavily on wind speed and surface saturation. Wind speeds vary both seasonally and over short time scales, and large wind events may contribute disproportionately to gas flux from the surface ocean. In order to best represent average monthly N 2 O flux, we calculated N 2 O flux using monthly average wind speeds as recorded at the L.F. Wade Bermuda International Airport (Bermuda Weather Service, n.d.). Though some discrepancy will exist due to the use of a land-based weather station, monthly averages will better account for large wind events that may otherwise be missed. The monthly average wind speeds used were comparable to those observed during the cruises for single timepoints at individual casts. Undersaturation in the surface ocean has been only sparsely recorded in the literature, and in many cases, has been attributed to the influence of sea-ice melt (e.g. Zhang et al., 2015): this is clearly not the case in the Sargasso Sea. In a similarly warm location, the Western Philippines Sea has been reported as slightly undersaturated at 90 ± 22%, with an estimated ocean-atmosphere flux of −1.7 ± 3.9 mol km −2 d −1 during the wet season (Tseng et al., 2016). Early models of N 2 O surface distributions (Suntharalingam & Sarmiento, 2000), based on observations by Weiss et al. (1992), also report N 2 O concentrations at equilibrium or undersaturated in some oceanic gyres. Notably, Yang et al. (2020) suggest undersaturation, or at most slight supersaturation, is expected in the subtropical gyres, which is consistent with our observations in the Sargasso Sea. The Sargasso Sea is susceptible to tropical rainfall events and the passage of notable cold fronts, especially during the volatile autumn season (Bermuda Weather Service, n.d.). We suspect that the low near-surface saturation we observed during most of these cruises may be due to sampling in close temporal proximity to one of these events, allowing us to reveal transient undersaturation of the surface waters due to a cooler-than-usual surface layer which would have a higher calculated equilibrium saturation. For example, sea surface temperature as measured by the Bermuda Weather Station dropped 2.5˚C over the course of the October 2018 cruise, where the lowest mixed layer N 2 O saturation was observed compared to the other cruises in this study. Yang et al. (2020) likewise indicated that temperature-dependent solubility and in some cases, primary productivity in overlying surface waters, are the main drivers of seasonal variability in N 2 O saturation in the Sargasso Sea region.

Conclusions
The current estimate of global N 2 O flux from the oceans is 4.2 ± 1.0 Tg N yr −1 (Yang et al., 2020) which translates to an approximate average flux of 1.03 mol N 2 O km −2 d −1 (Grundle et al., 2012). This global average, compared to the negative flux observed in three out of four months sampled, demonstrates that the Sargasso Sea region may be an important N 2 O sink. These data highlight the importance of highly oxygenated systems in the N 2 O budget. Further, the large variation we observe in mixed layer saturation of N 2 O is testament to the impact of short-term weather events that may cause rapid temperature changes of the surface ocean. Further work is needed to better constrain the seasonality of N 2 O fluxes in the area and the impact of transient weather events. We also estimated N 2 O yield from nitrification as 0.027-0.038% from ΔN 2 O to AOU and ΔN 2 O to NO 3 − relationships, which is on the order of those observed in previously sampled open ocean gyre systems.
Our work was spatially limited, rendering gyre-wide statements unwise. However, the demonstrated ease of implementation and reproducibility of the cavity-ring down spectrometry method utilized in this study  invites further studies along existing transects and time-series. This will help inform future models investigating the impacts of ocean deoxygenation trends on N 2 O production. This is greatly needed especially in locations that are currently N 2 O sinks but may become weaker sinks or even slight N 2 O sources as ocean deoxygenation continues through this century. also to Cordie Goodrich, Aleksander Gulkewicz, and Shea Wyatt for help with sampling, and to Qixing Ji for his technical assistance. Historical data used in this publication was drawn from the MEMENTO database. The MEMENTO database is administered by the Kiel Data Management Team at GEOMAR Helmholtz Centre for Ocean Research and supported by the German BMBF project SOPRAN (Surface Ocean Processes in the Anthropocene, http://sopran. pangaea.de). The database is accessible through the MEMENTO webpage: https://memento.geomar.de. Finally, this work was supported by a Canadian Associates of BIOS (CABIOS) internship award to ACSM.

Disclosure statement
No potential conflict of interest was reported by the author(s).

Funding
This work was supported by Bermuda Institute of Ocean Sciences: [Grant Number CABIOS].