The importance of alkyl nitrates and sea ice emissions to atmospheric NO x sources and cycling in the summertime Southern Ocean marine boundary layer

. Atmospheric nitrate originates from the oxidation of nitrogen oxides (NO x = NO + NO 2 ) and impacts both tropospheric chemistry and climate. NO x sources, cycling and NO x to nitrate formation pathways are poorly constrained in remote marine regions, especially the Southern Ocean, where pristine conditions serve as a useful proxy for the pre-industrial atmosphere. Here, we measured the isotopic composition ( δ 15 N and δ 18 O) of atmospheric nitrate in coarse-mode ( > 1 µm) aerosols collected in the summertime marine boundary layer of the Atlantic Southern Ocean from 34.5 to 70 ◦ S and across the northern edge of the Weddell Sea. The δ 15 N–NO − 3 decreased with latitude from − 2 . 7 ‰ to − 42 . 9 ‰. The decline in δ 15 N with latitude is attributed to changes in the dominant NO x sources: lightning at the low latitudes, oceanic alkyl nitrates at the mid-latitudes and photolysis of nitrate in snow at the high latitudes. There is no evidence of any inﬂuence from anthropogenic NO x sources or equilibrium isotope fractionation. Using air mass back trajectories and an isotope mixing model, we calculate that oceanic alkyl nitrate emissions have a δ 15 N signature of − 21 . 8 ± 7 . 6 ‰. Given that measurements of alkyl nitrate contributions to remote nitrogen budgets are scarce, this may be a useful tracer for detecting their contribution in other oceanic regions. The δ 18 O–NO − 3 was always less than 70 ‰, indicating that daytime processes involving OH are the dominant NO x oxidation pathway during summer. Unusually low δ 18 O–NO − 3 values (less than 31 ‰) were observed at the western edge of the Weddell Sea. The air mass history of these samples indicates extensive interaction with sea-ice-covered ocean, which is known to enhance peroxy radical production. The observed low δ 18 O–NO − 3 is therefore attributed to increased exchange of NO with peroxy radicals, which have a low δ 18 O, relative to ozone, which has a high δ 18 O. This study reveals that the mid- and high-latitude surface ocean may serve as a more important NO x source than previously thought and that the ice-covered surface ocean impacts the reactive nitrogen budget as well as the oxidative capacity of the marine boundary layer.

radiative heat budget (IPCC, 2013). It also plays a major role in the biogeochemical cycling of reactive nitrogen . NO − 3 aerosols originate from the oxidation of nitrogen oxides, collectively referred to as NO x (NO x = NO + NO 2 ). NO x cycling controls the chemical production 10 of tropospheric ozone (O 3 ), a greenhouse gas and pollutant (Finlayson-Pitts and Pitts, 2000), which in turn contributes to the oxidizing capacity of the atmosphere (Alexander and Mickley, 2015). Globally, fossil fuel combustion is the primary NO x source (van der A et al., 2008), which far exceeds natural emissions such as biomass burning (Finlayson-Pitts and Pitts, 2000), soil processes (Davidson and Kingerlee, 1997) and lightning (Schumann and Huntrieser, 2007).
Due to its remoteness, the summertime Southern Ocean marine boundary layer (MBL) can be representative of 20 pre-industrial-like atmospheric conditions (Hamilton et al., 2014). The chemical composition of the Southern Ocean MBL is characterized by low NO − 3 concentrations (Virkkula et al., 2006), representative of a background aerosol environment (i.e. minimal anthropogenic influence). Further-25 more, the South Atlantic sector of the Southern Ocean is primarily influenced by natural NO x sources. During summer, high lightning activity over South America and southern Africa results in NO x production between approximately 40 • S and the intertropical convergence zone (ITCZ) (Nes-30 bitt et al., 2000). As such, lightning is expected to be the dominant NO x source in the low-latitude MBL (Schumann and Huntrieser, 2007;van der A et al., 2008). Because of its pristine nature, the summertime Southern Ocean serves as a unique region in which to study atmospheric chemistry 35 and is a useful pre-industrial reference point for comparing the magnitude of anthropogenic aerosol impacts on climate (Haywood and Boucher, 2000;Hamilton et al., 2014).
The atmospheric chemistry of the polar MBL at the high southern latitudes differs from that of the mid-and low-40 latitude MBL. During summer, high levels of photochemistry result in the emission of reactive gases from sea ice and snow cover in the Antarctic. As a result, highly elevated concentrations of hydrogen oxide radicals (HO x = OH + peroxy radicals), halogens, nitrous acid (HONO), and NO x 45 have been observed during spring and summer in the polar regions (Brough et al., 2019). Furthermore, photochemical production of NO x within the surface snow of Antarctica and subsequent oxidation in the overlying atmosphere represents a significant NO − 3 source to the Antarctic tropo- 50 sphere (Jones et al., 2000(Jones et al., , 2001. NO − 3 photolysis near the surface-air interface of ice crystals produces NO 2 (Grannas et al. 2007;Jones et al., 2000), which can be released to the firn (i.e. the intermediate stage of ice between snow and glacial ice) air and escape the snowpack to the overlying at-55 mosphere (Erbland et al., 2013;Shi et al., 2015Shi et al., , 2018. During winter, additional NO x sources to the Antarctic atmosphere may include long-range-transported peroxyacetyl nitrates (PAN) and stratospheric inputs Lee et al., 2014;Walters et al., 2019). 60 Emission of alkyl nitrates (a group of nitrogen gases collectively referred to as RONO 2 ) from the surface ocean have been recently proposed as a potential NO x source to the MBL in remote regions (Williams et al., 2014;Fisher et al., 2018). Observations of elevated MBL alkyl nitrate con-65 centrations suggest that a direct oceanic source exists in both the tropics (Atlas et al., 1993;Blake et al., 2003) and the high-latitude Southern Ocean (Blake et al., 1999;Jones et al., 1999). Although the exact mechanism remains unclear, experimental evidence suggests that oceanic RONO 2 pro-70 duction occurs via photochemical processes involving the aqueous-phase reaction of RO 2 , derived from the photolysis of oceanic dissolved organic matter, and NO, derived from seawater nitrite photolysis (Dahl et al., 2003;Dahl and Saltzman, 2008). Supersaturated RONO 2 conditions in the surface 75 ultimately drive a net flux from the ocean to the atmosphere (Chuck et al., 2002;Dahl et al., 2005). The photolysis of emitted RONO 2 and subsequent OH oxidation in the overlying atmosphere leads to NO x formation (Fisher et al., 2018), and/or RONO 2 can form aerosol NO − 3 directly by hydrolysis 80 (Rindelaub et al., 2015).
Current global atmospheric models suggest that oceanic RONO 2 represents a significant source of nitrogen (N) to the Southern Ocean MBL, accounting for 20 % to 60 % of the reactive N pool at high latitudes (60 to 90 • S) (Fisher 85 et al., 2018). However, only one shipborne dataset with coincident ocean-atmosphere RONO 2 concentration measurements exists to substantiate this notion (Hughes et al., 2008). Additionally, the NO x source from RONO 2 degradation dominates relative to model-defined primary NO x emission 90 sources over the Southern Ocean, which include shipping, aircraft and lightning (Fisher et al., 2018). However, the lack of seawater observations available to constrain Southern Ocean RONO 2 distributions hampers the validation of model fluxes. Better understanding of the Southern Ocean 95 RONO 2 source is required to improve simulations and accurately evaluate its contribution to the Southern Ocean MBL NO x budget.

Natural abundance isotopes of atmospheric nitrate
Measurements of the oxygen (O) and N stable isotope ratios 100 of atmospheric NO − 3 can be used to constrain NO x sources, NO/NO 2 cycling and NO x to NO − 3 oxidation pathways, which are critical for understanding the reactive N budget in the atmosphere. This technique has been applied in polluted (Elliott et al., 2007;Zong et al., 2017), open-ocean (Hastings 105 et al., 2003;Morin et al., 2009;Kamezaki et al., 2019;Gobel et al., 2013;Altieri et al., 2013) and polar environments 3 (Morin et al., 2009;Walters et al., 2019). Stable isotope ratios are reported as a ratio of the heavy to light isotopologues of a sample relative to the constant isotopic ratio of a reference standard, using delta (δ) notation in units of per mil (‰) following Eq. (1): where R represents the ratio of 15  When NO x is converted to NO − 3 , the N atom is conserved. As such, it is generally expected that the N stable isotope ratio of atmospheric NO − 3 (δ 15 N-NO − 3 ) reflects the δ 15 N of the source NO x ,  plus any isotopic fractionation associated with NO/NO 2 cycling or NO x to NO − 3 con- 15 version. For example, the δ 15 N of lightning generated NO x is close to 0 ‰ (Hoering, 1957) and is distinct from stratospheric and snowpack NO x . Savarino et al. (2007) used the degree of N 2 O destruction in the stratosphere and the associated isotopic fractionation to derive an Antarctic strato-20 spheric δ 15 N-NO x source signature of 19 ± 3 ‰ . In contrast, snow-emitted NO x typically has a very low δ 15 N signature due to the large fractionation (  and the kinetic isotope effect (KIE) is the ratio of the rates with which the two isotopes of N are converted from reactant to product. If equilibrium isotope fractionation during NO/NO 2 cycling occurs, it results in the 15 N enrichment of NO 2 such that the NO − 3 formed from this NO 2 will have 30 a higher δ 15 N-NO − 3 than the initial NO x source (Freyer et al., 1993;. Equilibrium isotope fractionation during the transformation of NO x to NO − 3 also results in higher δ 15 N-NO − 3 compared to the original NO x source (Walters and Michalski, 2015).

35
In contrast to N, the O stable isotope ratio of atmospheric NO − 3 (δ 18 O-NO − 3 ) is reflective of the oxidants involved in NO x cycling prior to NO − 3 formation, as well as the dominant NO − 3 formation pathway (Hastings et al., 2003;Michalski et al., 2003;Alexander et al., 2020). The O atoms of NO x are 40 rapidly exchanged with oxidizing agents in the atmosphere to produce NO − 3 . Tropospheric NO x recycles rapidly with O 3 following the equations below: NO 3 can also react with hydrocarbons (HC) (e.g. dimethylsulfide, DMS) to form HNO 3 following Reaction (R7).

70
NO 2 + BrO → BrONO 2 (R8) Typically, aerosol δ 18 O-NO − 3 is interpreted as being determined by the dominant NO x oxidation pathways, Reaction (R3) vs. Reaction (R4) to Reaction (R9). If some com-75 bination of Reactions (R4)-(R9) occurs, then O 3 is the main oxidant, whereas during Reaction (R3), one of the O atoms originates from OH. The OH radical exchanges with H 2 O vapour in the troposphere; therefore the δ 18 O of OH is a function of the δ 18 O of H 2 O vapour, which generally ranges 80 from −27.5 ‰ to 0 ‰ in the subtropics and over the Southern Ocean (Michalski et al., 2012;Guilpart et al., 2017;Dar et al., 2020), and equilibrium isotope exchange between OH and H 2 O . In contrast, the δ 18 O of tropospheric O 3 is much higher, the most recent estimate 85 being 114.8 ± 10.4 ‰ (Vicars and Savarino, 2014). Therefore, a higher δ 18 O for atmospheric NO − 3 reflects the increased influence of O 3 on NO x to NO − 3 conversion (Reactions R4-R9), and the δ 18 O-NO − 3 is lower when Reaction (R3) is favoured, due to the lack of exchange of O atoms 90 with O 3 (Hastings et al., 2003;Fang et al., 2011;Altieri et al., 2013).
Here, we present the concentration and isotopic composition of coarse-mode (> 1 µm) atmospheric NO − 3 collected in the MBL of the Southern Ocean between Cape Town, South 95 Africa, and coastal Antarctica, as well as across the Weddell Sea gyre, during summer. Using air mass back trajectories, surface ocean nitrite measurements and the aerosol δ 15  lected is as follows: stage 1 (> 7 µm), stage 2 (3 to 7 µm), stage 3 (1.5 to 3 µm) and stage 4 (1 to 1.5 µm). A sector collector was used to restrict HV-AS activity to avoid contamination from ship stack emissions (Campbell Scientific Africa). The HV-AS only began operating if the 40 wind was blowing at an angle less than 75 • or greater than 180 • from the bow of the ship for a minimum of 10 min at a speed of at least 1 m s −1 . Filters were removed from the cascade impactor inside a laminar flow cabinet (Air Science), placed in individual zip-sealed plastic bags and stored at −20 • C until analysis.
Given that the MBL of the Southern Ocean is characterized by low atmospheric NO − 3 concentrations, an attempt was made to ensure that at least 24 h of in-sector sampling had passed before filters were removed from the cascade im- 50 pactor. However, this was not always possible as on occasion the filters had to be removed early to avoid contamination due to unusual ship manoeuvres or stagnant conditions. Therefore, sampling times ranged between 13 and 88 h across the three legs. The details of each cruise leg can be found in the 55 Supplement (Table S1).
During the research voyage, a field blank was collected by fitting the cascade impactor with a set of filters and walking the cascade impactor from the laboratory to the HV-AS in the same way that atmospheric samples were deployed. The cas-60 cade impactor was placed into the HV-AS and then immediately removed without the HV-AS turning on, after which the filters were removed from the cascade impactor and stored in the same manner as the atmospheric samples. All chemical analyses performed on samples were also performed on the 65 field blank filters to assess possible contamination during filter deployment or sample handling.

Sample analysis
Filter stages 1 to 4 were extracted using ultra-clean deionized water (DI; 18.2 M ) under a laminar flow cabinet (Air 70 Science). The extraction ratio was approximately 30 cm 2 of filter in 25 mL of DI. Extracts were immediately sonicated for 1 h and then stored at 4 • C for at least 12 h. Thereafter, extracts were filtered (0.2 µm) using an acid-washed syringe into a clean 30 mL HDPE bottle and stored at −20 • C until 75 analysis (Baker et al., 2010).
Aerosol nitrate concentrations ([NO − 3 ]) were determined using a Thermo Scientific Dionex Aquion ion chromatography (IC) system (precision of ±0.3 µmol L −1 ). The anion IC system contained an AG22 RFIC 4 × 50 mm guard column 80 and AG22 RFIC 4 × 250 mm analytical column. A six-point standard curve that encompassed the range of sample concentrations (extract [NO − 3 ]: 1.3 to 27.7 µmol L −1 ) was run on each day of analysis (Dionex Seven Anion-II Standard), and an R 2 value > 0.999 was required for sample analysis to  was corrected by subtracting the field blanks, which represented 35 % of the total [NO − 3 ] on average. Aerosol samples were also analysed for [NO − 3 ] using a Lachat QuikChem ® flow injection autoanalyzer (precision of ±0.8 µmol L −1 ). The average [NO − 3 ] measured us-90 ing the Lachat QuikChem ® flow injection autoanalyzer and the IC system is reported (Table S3).
Nitrogen and oxygen isotopic ratios were measured using the denitrifier method (Sigman et al., 2001;Casciotti et al., 2002). To determine the 15 N/ 14 N and 18 O/ 16 O of NO − 3 , 95 a natural strain of denitrifying bacteria, Pseudomonas aureofaciens, that lack the terminal nitrous oxide (N 2 O) reductase enzyme, was used to convert aqueous NO − 3 quantitatively to N 2 O gas. The product N 2 O was analysed by continuous flow isotope ratio mass spectrometry using a Delta 100 V Advantage isotope ratio mass spectrometer (IRMS) interfaced with an online N 2 O extraction and purification system. Individual analyses were referenced to injections of N 2 O from a pure gas cylinder and then standardized through comparison to the international reference materials of IAEA-N3 105 and USGS34 for δ 15 N-NO − 3 and IAEA-N3, USGS34 and USGS35 for δ 18 O-NO − 3 (Table S2) (Table S3).
Seawater samples were collected in triplicate every 2 h from the ship's underway system (position at depth approximately 5 m) for the analysis of surface ocean nitrite concen- was analysed using the colorimet-20 ric method of Grasshoff et al. (1983) using a Thermo Scientific Genesys 30 visible spectrophotometer (detection limit of 0.05 µmol L −1 ) (Table S4).

Air mass back trajectory analysis
To determine the air mass source region for each aerosol 25 sample, air mass back trajectories (AMBTs) were computed for each hour in which the HV-AS was operational for at least 45 min of that hour. Given that the ship was moving, a different date, time and starting location were used to compute each AMBT. An altitude of 20 m was chosen 30 to match the height of the HV-AS above sea level, and 72 h AMBTs were computed to account for the lifetime of NO − 3 in the atmosphere. All AMBTs were computed with NOAA's Hybrid Single-Particle Lagrangian Integrated Trajectory model (HYSPLIT v 4), using NCEP Global Data As-35 similation System (GDAS) output, which can be accessed at https://www.arl.noaa.gov/ready/hysplit4.html (last access: 12 January 2022) (NOAA Air Resources Laboratory, Silver Spring, Maryland) (Stein et al., 2015;Rolph, 2016).

40
The coarse-mode (> 1 µm in diameter) aerosol [NO − 3 ], computed by summing the [NO − 3 ] of stages 1 through 4, ranged from TS2 15.1 to 235.0 ng m −3 ( Fig. 1a and Table 1). The mass-weighted δ 15 N of coarse-mode aerosol NO − 3 ranged from −43.1 ‰ to −2.7 ‰ (Figs. 1b, 2 and Table 1). There  (Table S5). The highest nitrate concentrations occurred between 34 and 45 • S, and then they decreased with increasing latitude (Fig. 1a). Similarly, higher values characterized δ 15 N-NO − 3 50 between 34 and 45 • S (−4.9 ± 1.3 ‰), and then they de-creased with increasing latitude (Fig. 1b) 3 observed in the Weddell Sea (during Jan-60 uary to February) and δ 18 O-NO − 3 observed at corresponding latitudes (56 to 70 • S) during the early and late summer transects is statistically significant (p value = 0.009). The early and late summer cruise transects were similar spatially in that both took place along the same hydrographic 65 line (i.e. the Good Hope line), apart from the deviation to South Georgia during late summer (Fig. 2a, b). Even though the early and late summer cruise transects occurred in December and March, respectively, there is no statistically sig- .67) between them. Therefore, the early and late summer legs are discussed together and collectively referred to as the latitudinal transect.

75
Our observations reveal a latitudinal gradient in atmospheric NO − 3 concentration and δ 15 N-NO − 3 , which we hypothesize may be attributed to the varying contribution of the dominant NO x sources present between Cape Town and coastal Antarctica. In contrast, δ 18 O-NO − 3 depicts no latitudinal 80 trend; however, relatively low δ 18 O-NO − 3 values are observed in the Weddell Sea, which we hypothesize may be attributed to the influence of sea ice emissions on NO x cycling. Below, we first discuss the extent to which anthropogenic NO x sources may influence the observed atmospheric NO − 3 85 concentrations and δ 15 N signatures. Then we discuss the dominant NO x sources to low-, mid-and high-latitude Southern Ocean MBL NO − 3 , determined in part from 72 h AMBTs, as well as the role of various oxidants in NO/NO 2 cycling and NO 2 oxidation.   over the South Atlantic sector of the Southern Ocean ( Fig. 2a  and b), with no continental influence and limited opportunity for direct anthropogenic NO x emissions to contribute to aerosol NO − 3 , assuming NO − 3 has a lifetime of 72 h (Alexander et al., 2020). Furthermore, contamination from ship stack 5 emissions was avoided using a sector collector to restrict HV-AS activity to certain wind directions (Sect. 2.1). As such, the higher atmospheric NO − 3 concentrations observed near South  Fig. S1 in the Supplement. The white area represents the location of the sea ice determined using satellite-derived sea ice concentration data, obtained from passive microwave sensors AMSR2 (Advanced Microwave Scanning Radiometer 2; Spreen et al., 2008). Africa are best explained by greater lightning NO x production, which generally occurs between 40 • S and the ITCZ 10 during summer (Nesbitt et al., 2000;van der A et al., 2008). ing those sampled in the Weddell Sea; Fig. 1b). As discussed in Sect. 1.1, the δ 15 N-NO − 3 reflects the δ 15 N of the source NO x plus any isotopic fractionation imparted from NO/NO 2 cycling or NO x to NO − 3 conversion. Similar to previous studies, we surmise that NO x equilibrium fractionation 5 is unlikely to be relevant in our system, as NO x concentrations are significantly lower than O 3 concentrations (Elliott et al., 2007;Morin et al., 2009;Park et al., 2018). Typical O 3 concentrations observed at coastal sites in Antarctica are on the order of 20 ppbv (parts per bil-10 lion by volume) (Nadzir et al., 2018), whereas the sum of NO and NO 2 rarely exceeds 0.04 ppbv (Jones et al., 2000;Weller et al., 2002;Bauguitte et al., 2012). Under these conditions NO x isotopic exchange occurs at a much slower rate than Reactions (R1) and (R2), such that little to no equi-15 librium isotope fractionation is expressed, and the δ 15 N of the NO − 3 should reflect the δ 15 N of the NO x source (Walters et al., 2016). Additionally, equilibrium isotope effects are temperature-dependent (increasing with decreasing temperature), and here ambient temperatures decline with increas-20 ing latitude. Therefore, if equilibrium isotope fractionation were occurring during NO/NO 2 cycling and/or NO x to NO − 3 conversion, one would expect δ 15 N-NO − 3 to increase with latitude, as both fractionation processes produce NO − 3 with a higher δ 15 N than the source NO x . However, the opposite 25 trend is observed here, whereby δ 15 N-NO − 3 decreases with increasing latitude (Fig. 1b). Therefore, we discount the hypothesis that equilibrium isotope effects can explain the latitudinal gradient in δ 15 N-NO − 3 . NO − 3 in the Antarctic troposphere may also derive from 30 stratospheric denitrification, whereby HNO 3 is injected into the troposphere from the stratosphere via the subsidence and penetration of polar stratospheric clouds (PSCs). However, this phenomenon typically occurs in winter when the tropospheric barrier is weak, and the lower stratosphere is 35 cold enough for PSC formation Walters et al., 2019). Furthermore, δ 15 N-NO − 3 originating from stratospheric inputs is estimated to be 19 ± 3 ‰ , a value substantially greater than the atmospheric δ 15 N-NO − 3 observed here for high-latitude air masses; thus, 40 we discount a direct influence from stratospheric NO x . We propose that the observed variation in atmospheric δ 15 N-NO − 3 across the Southern Ocean is best explained by the changing contribution of three dominant NO x sources: lightning, surface ocean alkyl nitrate emissions and photochem-45 ical production on snow and ice, determined using AMBT analyses and typical NO x source signatures where possible, as discussed below.

High latitudes: photochemical NO x source
Aerosol δ 15 N-NO − 3 was relatively low in air masses from 50 the southern high latitudes, including in the Weddell Sea (average of −24.3 ‰; Figs. 1b and 2). The latitudinal gradient in lightning NO x production suggests that lightning NO x is greatly reduced at high latitudes (Nesbitt et al., 2000). Similar to other studies in the region Morin 55 et al., 2009), we suggest that photochemical NO x production on snow or ice accounts for the low aerosol δ 15 N-NO − 3 in high-latitude air masses, where high-latitude air mass samples are defined as those exposed to the Antarctic continent or the surrounding sea ice (with sea ice concentration being 60 at least 50 %). Antarctic estimates for isotopic fractionation associated with snow NO − 3 photolysis during summer range from −47.9 ‰ to −55.8 ‰ for laboratory and field experiments, respectively (Berhanu et al., 2014(Berhanu et al., , 2015, resulting in the emission of low δ 15 N NO x to the overlying atmosphere 65 Morin et al., 2009;Shi et al., 2018;Walters et al., 2019). Therefore, NO − 3 photolysis explains the very low δ 15 N-NO − 3 observed in high-latitude air masses in early and late summer that crossed snow-covered continental ice or sea ice before being sampled (Fig. 2a, b). During 70 early summer, air masses spent significantly more time over the snow-covered continent compared to late summer, and the sea ice extent was greater in early summer compared to late summer (Fig. 2a, b). Combined, these dynamics resulted in a much lower δ 15 N-NO − 3 for high-latitude air masses dur-75 ing early summer compared to late summer (minimum value of −42.9 ‰ vs. −25.6 ‰). Similarly low MBL δ 15 N-NO − values (< −30 ‰) were recently observed for the southern high latitudes of the Indian Ocean (Shi et al., 2021). Our data are also consistent with year-round studies of atmospheric NO − 3 at coastal Antarctica  and the South Pole (Walters et al., 2019), where δ 15 N-NO − 3 was re-5 ported to range from −46.9 ‰ to 10.8 ‰ and from −60.8 ‰ to 10.5 ‰, respectively. Both studies observed a seasonal cycle in δ 15 N-NO − 3 , whereby the lowest values occurred during sunlit periods (i.e. summer) due to snowpack NO x emissions, and the highest values occurred during dark periods 10 (i.e. winter) due to stratospheric inputs Walters et al., 2019).

Low latitudes to mid-latitudes: oceanic NO x source
At the northern extent of our transects, the low-latitude 15 aerosol samples, defined as those with air mass back trajectories originating from anywhere north of 43 • S in early summer and 41 • S in late summer (Fig. 2), had the highest average δ 15 N-NO − 3 signature (−4.9 ± 1.3 ‰; n = 5). These values can be attributed to lightning-generated NO x , which 20 has a δ 15 N signature close to 0 ‰ (Hoering, 1957). Lightning activity at low latitudes is also consistent with the higher atmospheric [NO − 3 ] observed (Fig. 1a) and is further supported by co-occurring high [NO − 3 ] and relatively high δ 15 N-NO − 3 (Fig. S2). An average atmospheric δ 15 N-NO − 3 signature of 25 −4 ‰ was previously reported for the low-latitude Atlantic Ocean, between 45 • S and 45 • N, and was similarly attributed to a combination of natural NO x sources including lightning (Morin et al., 2009). Aerosol samples across the mid-latitudes had an average 30 δ 15 N-NO − 3 of −13.2 ‰ (Figs. 1b and 2). Mid-latitude air masses are defined as those originating from anywhere south of 43 • S in early summer and south of 41 • S in late summer that made no contact with Antarctica or any surrounding sea ice (Fig. 2a, b). Furthermore 43 • S and 41 • S represent 35 the latitudes at which non-zero sea surface nitrite concentrations began to be observed in early and late summer (Fig. 4). Mid-latitude samples were therefore unlikely to be influenced by snow-emitted NO x with its light isotopic signature. However, the observed aerosol δ 15 N-NO − 3 was too low 40 (−14.5 ‰ to −11.2 ‰) to be explained solely by lightninggenerated NO x . In the absence of any signature of anthropogenic NO x emissions (Sect. 4.1), we argue that the dominant NO x source for the mid-latitude samples originates from seawater. 45 As mentioned in Sect. 1, the most likely mechanism for an oceanic NO x source is via the photolysis of surface ocean derived RONO 2 in the MBL. NO derived from seawater nitrite is thought to limit RONO 2 production (Dahl and Saltzman, 2008;Dahl et al., 2012), such that non-zero nitrite concen- 50 trations are required for RONO 2 production to occur. Here, surface ocean nitrite concentrations were relatively high, in particular from ∼ 41 to 50 • S (Fig. 4). Furthermore, the lat-itudinal extent of mid-latitude air masses with low δ 15 N-NO − 3 signatures corresponds well with the same latitudinal 55 extent in which non-zero surface ocean nitrite concentrations occurred (Fig. 4). As such, we suggest that in this region oceanic RONO 2 emission is the main source to the Southern Ocean MBL, ultimately resulting in the low δ 15 N-NO − 3 values observed for mid-latitude air masses. 60 No estimates exist for the δ 15 N of oceanic RONO 2 ; however RONO 2 photolysis may yield isotopically light NO x given that NO − 3 photolysis produces low δ 15 N products (e.g. Frey et al., 2009). Therefore, once oxidized in the overlying atmosphere, NO x derived from oceanic RONO 2 photol-65 ysis may form atmospheric NO − 3 with a low δ 15 N signature. Aerosol δ 15 N-NO − 3 values have been observed to range from −14.1 ‰ to −7.3 ‰ in the eastern equatorial Pacific (Kamezaki et al., 2019) and from −6 ‰ to ∼ 0 ‰ (average = −3.4 ‰) in the western equatorial Pacific (Shi et al., 2021). 70 Observed δ 15 N-NO − 3 is higher in the western compared to the eastern equatorial Pacific, which could be attributed to the proximity of the western equatorial Pacific to continental/anthropogenic NO x sources, resulting in NO − 3 having a higher δ 15 N signature. The low average δ 15 N-NO − 3 observed 75 for the mid-latitude air masses of the Southern Ocean MBL sampled in the present study (−14.5 ‰ to −11.2 ‰) is remarkably similar to that for the air masses observed in the eastern equatorial Pacific (Kamezaki et al., 2019). Kamezaki et al. (2019) also concluded that such low δ 15 N-NO − 3 val-80 ues cannot be explained solely by lightning NO x , and given the lack of considerable influence from any continental NO x sources, they invoked the contribution of oceanic N emissions in the form of ammonia (NH 3 ) and/or RONO 2 . However, NH 3 flux data for the summertime Atlantic Southern 85 Ocean derived from in situ ocean/atmosphere observations suggest that the ocean in this region is a net sink of NH 3 .
The latitudinal extent of our sampling campaign enabled us to estimate a range of likely values for the N isotopic 90 composition of NO − 3 derived from oceanic RONO 2 . We split the latitudinal transect into three regions, each characterized by the dominance of a different natural source of NO − 3 , i.e. lightning NO x at low latitudes (Fig. 5, light orange), oceanic RONO 2 emissions at mid-latitudes (Fig. 5, dark orange) and 95 snowpack emissions at high latitudes (Fig. 5, red).
Assuming that the dominant natural source of NO − 3 is the only source relevant in each latitudinal zone, we estimate the contribution of each source to total NO − 3 formation by ascertaining the amount of time air masses spent in each zone. We 100 further assume that atmospheric δ 15 N-NO − 3 reflects at most a combination of two sources based on the AMBTs of each sample, either lightning NO x and oceanic RONO 2 emissions near South Africa or oceanic RONO 2 emissions and snowpack NO x emissions near Antarctica (Fig. 5 and Table S6). 105 Using a two-endmember mixing model, the δ 15 N signature of the source NO − 3 derived from mid-latitude Southern Ocean RONO 2 emissions was calculated for all samples  where air masses from the mid-latitude region contributed at least 10 % (Table S6). This 10 % threshold was chosen as the isotopic endmember of oceanic RONO 2 is harder to determine with confidence when its contribution to total NO − 3 is less than 10 %. As an example, the AMBTs for sample ES 4 5 spent 3 % of the time in the low-latitude zone and 97 % in the mid-latitude zone.
Using the measured δ 15 N-NO − 3 for ES 4 of −14.5 ‰, and assuming lightning NO x has a δ 15 N signature of 0 ‰, we calculate the δ 15 N signature of the RONO 2 -derived NO − 3 10 to be −14.9 ‰. It is important to note that using this approach to estimate the δ 15 N-NO − 3 from oceanic RONO 2 emissions relies heavily on AMBTs generated using HYS-PLIT. While HYSPLIT is a frequently used tool for assessing air mass origin in the Southern Hemisphere and over 15 Antarctica (Morin et al., 2009;Walters et al., 2019;Shi et al., 2021), it is important to note that a spatial uncertainty of 15 % to 30 % of the trajectory path distance can be expected (Scarchilli et al., 2011). AMBTs also become increasingly uncertain the further back in time they are used (Sinclair et 20 al., 2013). Some of this uncertainty is alleviated by the fact that the AMBTs generated here are relatively short (< 5 d). Additionally, the spatial scale of the low-, mid-and highlatitude zones is large, such that some variation in sample AMBTs will not significantly alter the expected dominant 25 NO − 3 source. Using this approach for each filter deployment along the latitudinal transect, an average δ 15 N-NO − 3 from oceanic RONO 2 emissions of −21.8±7.6 ‰ was estimated. Furthermore, the contribution of RONO 2 emissions can explain the 30 lowering of δ 15 N from 0 ‰ for the low-latitude air mass samples. For example, the highest δ 15 N observed in the study was −2.7 ‰, and this sample has a < 5 % contribution from the mid-latitude zone. The other two low-latitude samples have 30 % to 40 % contribution from the mid-latitude zone, and their δ 15 N is lower (Table S3), as expected due to the influence of RONO 2 emissions.
The influence of low δ 15 N-NO − 3 from RONO 2 emissions is not limited to the Southern Ocean, and this estimate of the 5 N isotopic composition for the RONO 2 derived NO − 3 source may be useful to constrain the contribution of RONO 2 emissions to NO − 3 formation in other ocean regions with elevated surface ocean nitrite concentrations, such as the tropical Pacific. The corresponding δ 18 O values allow us to determine the pathways of NO − 3 formation from NO x . However, an assumption must first be made regarding the oxidation of NO to NO 2 . While the dominant oxidant of NO to NO 2 is O 3 (Re-15 action R1) in most of the troposphere, over the open ocean there can be a significant contribution via the reaction of NO with peroxy radicals (HO 2 and its organic homologues RO 2 ) (Alexander et al., 2020). Peroxy radicals compete with O 3 to convert NO into NO 2 via Reaction (R10).
We assume that 15 % of NO to NO 2 conversion occurs via 40 HO 2 /RO 2 oxidation and 85 % by O 3 oxidation, as is suggested by global models (Alexander et al., 2020), and use the minimum and maximum δ 18 O-H 2 O range of −27.5 ‰ to 0 ‰, the temperature-dependent equilibrium isotope exchange between OH and H 2 O  and the resulting minimum and maximum estimates for δ 18 O-OH of −67.4 ‰ to −41.0 ‰. Using these assumptions and Eqs. (3) and (4), the expected δ 18 O-NO − 3 for the daytime OH oxidation pathway (Reaction R3) is 46.5 ‰ to 71.4 ‰, and for the dark Reactions (R4)-(R6), it is 88.7 ‰ 50 to 113.5 ‰.
The observed δ 18 O-NO − 3 values were all less than 70 ‰ (Figs. 1c and 3), suggesting that NO x oxidation by OH (Reaction R3) was indeed the dominant pathway for atmospheric NO − 3 formation during summer. The low δ 18 O-NO − 3 values 55 observed suggest a minimal influence of O 3 in the oxidation chemistry, ruling out both the halogen-related (Reactions R8 to R9) and DMS-related (Reaction R7) NO − 3 formation pathways in addition to N 2 O 5 hydrolysis (Reactions R4 to R6). This is consistent with previous year-round studies of atmo-60 spheric NO − 3 at coastal Antarctica  and the South Pole (Walters et al., 2019), where δ 18 O-NO − 3 was at a minimum in summer (59.6 ‰ and 47.0 ‰, respectively). Both studies confirm the importance of HO x oxidation chemistry in summer when solar radiation enhances the production 65 of these oxidants, followed by a switch to O 3 -dominated oxidation chemistry in winter Ishino et al., 2017;Walters et al., 2019).
Interestingly, most aerosol samples have a δ 18 O-NO − 3 less than 46.5 ‰ (n = 19), the lower limit estimated above for 70 the OH pathway. This suggests that there is more NO to NO 2 conversion via HO 2 /RO 2 oxidation occurring than the global average. A maximum HO 2 /RO 2 contribution to NO oxidation of ∼ 63 % is required to explain the lowest δ 18 O-NO − 3 value, which was observed over the mid-latitudes dur-75 ing early summer. Increased RO 2 production over the midlatitudes could derive from RONO 2 photolysis in the MBL, which we hypothesize is happening in this region based on the δ 15 N-NO − 3 (Sect. 4.2.2). Although the lowest δ 18 O observation occurred in the mid-80 latitudes, the majority of low δ 18 O-NO − 3 values were observed in the Weddell Sea, away from the region of maximum RONO 2 emissions. Approximately half of the Weddell Sea samples have a δ 18 O-NO − 3 < 31 ‰, which would require a HO 2 /RO 2 contribution to NO oxidation upwards of 85 40 % (more than double the contribution estimated by global models; Alexander et al., 2020). These δ 18 O-NO − 3 observations are unusually low compared to previous observations for the same region in spring (Morin et al., 2009).
We hypothesize that the large contribution of HO 2 /RO 2 90 to NO/NO 2 oxidation (i.e. a decrease in f in Eq. 2) resulting in these low δ 18 O-NO − 3 values is due to the influence of sea ice emissions. The 72 h AMBTs for these low δ 18 O-NO − 3 Weddell Sea samples indicate that all the air masses either originated from, or spent a significant amount of time 95 recirculating, over the sea-ice-covered region of the western Weddell Sea (Fig. 6b). By contrast, aerosol samples from the Weddell Sea with δ 18 O-NO − 3 values greater than 31 ‰ have air masses that experienced significantly more oceanic influence (Fig. 6a). There is evidence that sea ice can lead to enhanced peroxy radical production (Brough et al., 2019). In that work, increased HO 2 + RO 2 concentrations were observed during spring at a coastal Antarctic site when air masses arrived from across a sea-ice-covered zone. This was attributed to the 5 oxidation of hydrocarbons by chlorine atoms, which leads to increased RO 2 concentrations via Reactions (R11) and (R12).
Cl atoms are much more reactive with hydrocarbons than OH (Monks, 2005) and can enhance hydrocarbon oxidation, even when present at low concentrations. Brough et al. (2019) suggest that air masses that traversed the sea ice 15 zone contained photolabile chlorine compounds that built up at night until photolysis occurred during the next day (Brough et al., 2019). Although our study was conducted in summer (the season of minimum sea ice extent), the sampling locations were uniquely positioned at the western edge 20 of the Weddell Sea gyre, where significant sea ice remained (Fig. 6). Therefore, we suggest that chlorine chemistry over the sea ice increased RO 2 concentrations at the time of our sampling, allowing the NO + RO 2 pathway to play a more significant role in the Weddell Sea and resulting in low δ 18 O-25 NO − 3 values. We note that the only other estimates of δ 18 O-NO − 3 from the Weddell Sea ranged from ∼ 50 ‰ to 110 ‰ during springtime, and these samples were associated with air masses that spent almost no time over the sea ice and therefore had limited potential for this peroxy radical chem-30 istry to drive down the δ 18 O-NO − 3 to the low values we observe (Morin et al., 2009).

Conclusions
Our observations across a large latitudinal gradient of the summertime Southern Ocean MBL suggest it is dominated 35 by natural NO x sources with distinct isotopic signatures. Aerosol NO − 3 was predominantly formed from lightninggenerated NO x with a δ 15 N of ∼ 0 ‰ at the lower latitudes, whereas snowpack NO x emissions with a δ 15 N ∼ −48 ‰ dominated the MBL inventory at higher latitudes. Over the 40 mid-latitudes, NO − 3 derived primarily from oceanic RONO 2 emissions, with an estimated δ 15 N signature of ∼ −22.0 ‰. Additional research is needed to improve our mechanistic and isotopic understanding of surface ocean RONO 2 formation, flux and conversion to aerosol nitrate in order to con-45 strain the contribution of oceanic RONO 2 emissions to NO − 3 formation in other ocean regions where this source has been invoked, such as the tropical Pacific (Kamezaki et al., 2019). The isotopic composition of NO − 3 observed here can further inform interpretations of Antarctic ice core NO − 3 isotope 50 records to understand aerosol climate forcing and controls on the atmospheric oxidation budget over millennia (Freyer et al., 1996;Jiang et al., 2019) -the interpretation of which relies on knowledge of the NO x isotopic source signatures in the polar atmosphere. 55 The δ 18 O-NO − 3 values were consistently lower than 70 ‰, which confirms NO x oxidation by OH (Reaction R3) to be the dominant pathway for atmospheric NO − 3 formation during summer. However, unusually low δ 18 O-NO − 3 values observed at the mid-latitudes and in the Weddell Sea indicate 60 the increased importance of peroxy radicals (and decreased importance of O 3 ) in NO oxidation to NO 2 in the MBL. At mid-latitudes, peroxy radicals (RO 2 ) may derive from RONO 2 photolysis, while in the Weddell Sea, sea ice appears to play an important role in the formation of this oxidant via 65 its influence on chlorine chemistry (Brough et al., 2019). This implies that snow-covered sea ice is not only a source of NO x but also other species that have the potential to change the composition of the atmosphere above the ice and impact NO x oxidation chemistry. These results also highlight the utility of δ 18 O-NO − 3 to identify the major oxidants in NO oxidation, as well as NO x to NO − 3 conversion. In particular, δ 18 O-NO − 3 can serve as a useful tool for testing our understanding of the relative importance of HO 2 /RO 2 in NO/NO 2 cycling, which can be difficult to constrain in some environments.
Our study challenges the traditional paradigm that considers the ocean to be a passive recipient of N deposition, as the Southern Ocean mid-latitude NO − 3 source may derive almost entirely from oceanic RONO 2 emissions. In the tropical equatorial Pacific atmosphere, Kamezaki et al. (2019) also suggested evidence for a low δ 15 N-NO − 3 source derived from the ocean. In the subtropical Atlantic Ocean MBL, Altieri et al. (2016) found that biogeochemical cycling in the surface ocean can directly influence the lower atmosphere, 15 serving as a source of aerosol organic N and ammonium. This study suggests that the surface waters of the Southern Ocean may also serve as a NO x source, ultimately resulting in NO − 3 aerosol formation. As such, the surface ocean may play a bigger role in atmospheric oxidative capacity over re-20 mote marine regions than previously thought. Data availability. Datasets for this research are available at https://doi.org/10.5281/zenodo.5006982 TS5 (Burger et al., 2021).

25
Author contributions. KEA designed the study and sampling campaign, acquired funding and supervised the research. KEA and JG provided financial and laboratory resources and assisted in data validation. KAMS and JMB conducted the sampling at sea, and JMB performed the laboratory analyses. MGH and EJ assisted in 30 data validation, reviewing and editing of the manuscript. JMB analysed the data and prepared the manuscript with contributions from all co-authors.
Competing interests. The contact author has declared that neither they nor their co-authors have any competing interests.

35
Disclaimer. Publisher's note: Copernicus Publications remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
istry Lab in the Oceanography Department at the University of Cape Town for their assistance in the field and laboratory. We also thank the Flotilla Foundation and the Weddell Sea Expedition (2019). We thank Lija Treibergs, Reide Jacksin and Peter Ruffino for their assistance in analysing the nitrate isotopes and Riesna Audh for her as- 45 sistance with satellite-derived sea ice concentration data. We thank Riesna Audh, Raquel Flynn and Shantelle Smith for nitrite concentration measurements and Raquel Flynn for quality controlling the nitrite concentration data.
Financial support. This research has been supported by a CA-50 REER award to Julie Granger from the U.S. National Science Foundation (OCE-1554474). It has also been supported by the South African National Research Foundation through a Competitive Support for Rated Researchers Grant to Katye E. Altieri (111716)