Department of Geological Sciences and Environmental Change Initiative, Brown University, Providence, Rhode Island, USA
Corresponding author: M. G. Hastings, Department of Geological Sciences and Environmental Change Initiative, Brown University, 324 Brook St., Box 1846, Providence, RI 02912, USA. (Meredith_Hastings@brown.edu)
 Identifying the dominant sources of atmospheric reactive nitrogen (Nr) is critical for determining the influence of anthropogenic emissions on Nr deposition, especially in marine ecosystems. To test the influence of anthropogenic versus marine air masses, samples were collected in Bermuda, where seasonal atmospheric circulation patterns lead to greater continental transport during the cool season. The 15N/14N of aerosol nitrate (NO3–) indicates changes in Nr sources and its 18O/16O indicates a seasonal shift in the relative strength of pathways of NO3– formation. The aerosol δ15N-NO3– was consistently lower than or equal to the rainwater from the same sampling period, the opposite trend of that observed in polluted systems. We propose that this is due to HNO3(g) uptake onto aerosol particles with a kinetic isotope effect, lowering the aerosol δ15N-NO3– relative to residual HNO3(g). The aerosol δ18O-NO3– was higher than that in rainwater during the cool season, but was not different during the warm season, which we tentatively attribute to the increased importance of heterogeneous halogen chemistry on the formation of NO3– during the cool season.
 The increasing deposition of anthropogenic reactive nitrogen (Nr) to the surface ocean has the potential to alter surface ocean biogeochemistry [Duce et al., 2008; Krishnamurthy et al., 2007]. The Sargasso Sea, in the low-nutrient core of the North Atlantic gyre, may be particularly sensitive to atmospheric inputs of Nr [Michaels et al., 1993]. Moreover, because it is located downwind of major industrial centers in the eastern United States, atmospheric deposition models predict that anthropogenic activities have already substantially altered the Nr deposition to the Sargasso Sea [Duce et al., 2008]. As a result, the region has been the subject of a number of studies of atmospheric deposition of Nr [e.g., Galloway et al., 1989; Knapp et al., 2010; Prospero et al., 1996], and especially of nitrate (NO3–), the primary sink of atmospheric NOx (NOx = NO + NO2) [e.g., Hastings et al., 2003; Jickells et al., 1982; Moody et al., 1989].
 During the daytime, atmospheric NOx undergoes rapid transformations via
 R1: NO + O3 → NO2 + O2
 R2: NO2 + hν → NO + O
 R3: O + O2 → O3 before conversion to HNO3 via
 R4: NO2 + OH + M → HNO3 + M, where M is an unreactive third body, usually N2. During the nighttime, NOx accumulates as NO2, allowing for both the heterogeneous formation of HNO3 via
 R5: NO2 + O3 → NO3 + O2
 R6: NO2 + NO3 + M ← → N2O5 + M
 R7: N2O5(g) + H2O(l, aerosol) → 2HNO3(aq), and the gas-phase reaction with hydrocarbons, especially dimethyl sulfide (DMS), in the marine boundary layer via
 Heterogeneous halogen chemistry, discussed in section 3.2, may also contribute to NO3– formation.
 The 15N/14N and 18O/16O ratios of NO3– have been used to distinguish NO3– sources and chemical formation pathways in both polluted [Elliott et al., 2007, 2009; Mara et al., 2009; Wankel et al., 2009] and remote environments [Altieri et al., 2013; Baker et al., 2007; Hastings et al., 2003, 2004; Morin et al., 2009]. The N atom is conserved during conversion from NOx to NO3– (R1–R8), so the δ15N (δ15N = [(15N/14N)sample/(15N/14N)reference − 1]*1000, where the reference is N2 in air) of the final NO3– is taken to indicate the δ15N of the NOx source [Altieri et al., 2013; Elliott et al., 2007, 2009; Hastings et al., 2003; Wankel et al., 2009], although isotopic fractionation during formation may also influence the δ15N of NO3– [Freyer, 1991; Vicars et al., 2013]. In contrast, the oxygen (O) atoms are exchanged with ozone (O3) in the nighttime pathways (R5–R8) and with both O3 and the hydroxyl radical (OH) in the daytime pathway (R1–R4). Because O3 has a much higher δ18O (δ18O = [(18O/16O)sample/(18O/16O)reference − 1]*1000, where the reference is Vienna Standard Mean Ocean Water) than that expected for OH (+90 to +122‰ and −6 to +2‰, i.e., H2O(v), respectively), the δ18O-NO3– can be used to distinguish among NO3– formed via the daytime (R4) and nighttime (R7–R8) pathways [Altieri et al., 2013; Elliott et al., 2009; Hastings et al., 2003; Wankel et al., 2009].
 The δ15N-NO3– in rainwater (δ15N-NO3–(aq)) at Bermuda is significantly lower during the cool season, from October to March (−5.9 ± 3.3‰; ± 1 SD unless otherwise noted), than during the warm season, from April to September (−2.1 ± 1.5‰), corresponding to two distinct atmospheric transport regimes [Hastings et al., 2003]. In the cool season, fast-moving fronts transport relatively polluted air masses from North America to Bermuda [Jickells et al., 1982; Miller and Harris, 1985] associated with tracers for anthropogenic activity (e.g., non-sea-salt sulfate, antimony, and selenium) and high concentrations of sea-salt aerosols (e.g., coarse-mode sodium, chlorine, and calcium) due to increased wind speeds [Arimoto et al., 1992; Huang et al., 1999; Wolff et al., 1986]. In the warm season, the Azores high pressure system develops over Bermuda, blocking most transport from North America [Jickells et al., 1982] and carrying dust plumes from the Sahara over the North Atlantic [Prospero et al., 1996] associated with increased mineral components (e.g., silica, aluminum, and non-sea-salt calcium) [Arimoto et al. 1992; Huang et al., 1999; Wolff et al., 1986]. Interestingly, for concurrent sampling campaigns conducted in 2000, the cool season δ15N-NO3–(aq) in Bermuda (−5.9‰) [Hastings et al., 2003] was much lower than that measured in the eastern United States (+0.1‰) [Elliott et al., 2007], the presumed source region, which could indicate an additional NOx source over the ocean or some isotopic transformation during transport from the U.S. to Bermuda.
 Aerosol NO3– (NO3–(p)) can contribute 20–65% of total NO3– deposition in marine environments [Baker et al., 2007] and to our knowledge has not been investigated in the marine atmosphere downwind of North America, a region of high anthropogenic NOx emissions. In addition, rainwater efficiently scavenges both particles and gases from the atmosphere. Therefore, the δ15N-NO3–(p) will both influence the δ15N-NO3–(aq) and illuminate the source and formation pathways of the NO3–(p) itself. This study determined the N and O isotopic composition of NO3–(p) to investigate the importance of anthropogenic contributions to NO3–(p) and to provide insight into the formation pathways of NO3–(p) in comparison to NO3–(aq).
 Aerosol samples were collected in March 2010 and from June to August 2010 on the island of Bermuda at the Tudor Hill Marine-Atmospheric Sampling Observatory (32.27°N, 64.87°W) using cassette-based samplers fitted with Whatman 41 cellulose substrates. The filters were frozen until extraction into aqueous solution following the methods of Chen et al.  and Wankel et al. . The extracts were subsequently analyzed for [NO3–], using reduction to NO followed by chemiluminescent detection of NO [Braman and Hendrix, 1989], and for 15N/14N and 18O/16O ratios, using the denitrifier method [Casciotti et al., 2002; Sigman et al., 2001]. Rainwater samples collected on an event basis from July 2009 through June 2011, including the dates of the aerosol sampling campaign, were analyzed using the same methods. NOAA's Hybrid Single-Particle Lagrangian Integrated Trajectory (HYSPLIT) model was used to calculate air mass back trajectories (AMBTs) for each aerosol sample in order to confirm that specific trajectories conformed to expected seasonal atmospheric circulation patterns (Figure S2). Further detail on methods is presented in the supporting information.
3 Results and Discussion
 HYSPLIT AMBTs support the assumption that the seasonal designations are effective proxies for air mass source, with cool season samples originating over North America and warm season samples originating over the subtropical North Atlantic (Figures S1 and S2), consistent with the longer time series reported by Hastings et al.  and Altieri et al. . Therefore, the aerosol results are considered representative of seasonal trends. Comparisons below were calculated by season, following the methods of Hastings et al. , and using a two-tailed t test for populations of unequal variance, where p < 0.05 indicates a statistically significant difference between populations.
 Previous studies [Arimoto et al., 1992; Huang et al., 1999; Wolff et al., 1986] and recent collections (Peters et al., unpublished data collected January 2007–June 2008) show that sea salt dominates the composition of Bermuda aerosols compared to mineral dust by over an order of magnitude year round. Infrequent strong dust events, usually during the summer, may cause mineral dust to reach levels comparable to sea salt concentrations [Arimoto et al., 1992; Huang et al., 1999; Peters et al., 2008, unpublished data]. For the samples in this study, however, AMBTs suggest little to no transport from North Africa, the main dust source to Bermuda. While we did not directly address bulk ionic composition of aerosols in this study, an (occasional) important contribution of mineral dust should not significantly affect the trends observed.
3.1 Aerosol and Rainwater δ15N-NO3–
 The δ15N-NO3–(p) was generally lower than the δ15N-NO3–(aq) for rain events concurrent with the aerosol sampling period. The difference (Δδ15N = δ15N-NO3–(aq)–δ15N-NO3–(p)) ranged from −2.9‰ to 6.0‰ and averaged 1.5‰ over the sampling period, although the populations were only significantly different during the warm season (Figure 1a). The isotopic difference between the δ15N-NO3–(p) and the δ15N-NO3–(aq) agrees with one other observation from the tropical North Atlantic [Baker et al., 2006] but contrasts with the trend reported in polluted regions, where the δ15N-NO3–(p) is consistently higher than the δ15N-NO3–(aq) [Elliott et al., 2009; Freyer, 1991; Mara et al., 2009]. This difference in δ15N between aerosol and rain NO3– is present during both the cool and warm seasons (Figure 1a). Thus, the observed difference is likely independent of the chemistry that converts NOx to NO3–, which has a strong seasonal dependence, as evidenced by the seasonal distribution in δ18O-NO3– (Figure 1b) and previous work on δ18O-NO3– [Hastings et al., 2003; Elliott et al., 2009; Wankel et al., 2009]. It also does not depend on AMBT, again suggesting little sensitivity to the bulk composition of the aerosol.
 In polluted regions, NO3–(p) occurs predominantly in the fine mode as NH4NO3 [Putaud et al., 2004], formed through the following reaction:
 R9: NH3(g) + HNO3(g) ← → NH4NO3(p)
Freyer  suggested that 15N is favored in the more stable solid phase, driving the δ15N-NO3–(p) higher than the δ15N-HNO3(g). The rainwater NO3–, comprising both the higher δ15N-NO3–(p) and the lower δ15N-HNO3(g), will therefore consistently have a lower δ15N than the aerosol NO3–, as is observed in polluted regions. This mechanism becomes insignificant, however, when aerosols are transported into the marine atmosphere and shift from fine mode NH4NO3 to coarse mode through an association between HNO3(g) and sea salt or mineral dust particles [Mara et al., 2009, and references therein]:
 Indeed, about 90% of aerosol NO3– is in the coarse mode in the North Atlantic marine atmosphere [Baker et al., 2006]. Conversion from fine mode to coarse mode tends to be a unidirectional process; therefore, coarse mode aerosols, unlike fine mode aerosols, are not in equilibrium with HNO3(g) [Keene and Savoie, 1998]. Instead, kinetic fractionation should preferentially form 14NO3–(p), leaving a 15N-enriched pool of HNO3(g). The rainwater NO3–, integrating the lower δ15N-NO3–(p) and the higher δ15N-HNO3(g), would yield a δ15N higher than that of the aerosol NO3–, as observed in the data presented here and by Baker et al. .
3.2 Aerosol and Rainwater δ18O-NO3–
 The δ18O-NO3–(p) was significantly higher than the δ18O-NO3–(aq) during the cool season but was not significantly different during the warm season (Figure 1b). Recent studies have shown the potential for halogens to play a significant role in NOx and NO3– chemistry, especially in the polluted marine boundary layer near the continents [Altieri et al., 2013; Osthoff et al., 2008; Thornton et al., 2010; Vicars et al., 2013]. There, NOx may be converted to NO3– through reaction with XO (X = Cl or Br) during the day to form XONO2 (R12–R14). First, HNO3(g) in polluted plumes reacts with NaCl in the marine boundary layer, releasing HCl and NaNO3(p) (R10) or the analogous reaction with NaBr to form HBr. The HX then reacts with OH to produce Cl or Br radicals, which quickly form XO by reacting with ozone. The XONO2 subsequently formed (R13) can then combine with sea-salt aerosol to form coarse mode NO3–(p) (R14).
 R12: X + O3 → XO + O2
 R13: NO2 + XO(g) + M → XONO2(g)
 R14: XONO2(g) + NaX(p) → X2(g) + NaNO3(p)
 NOx can also react with sea-salt aerosol to form XNO and NaNO3 (R15).
 R15: 2NO2 + NaX(p) → XNO(g) + NaNO3(p)
 Mineral dust is assumed to play a negligible role in these reactions for two reasons: the aerosol halogens participating in these reactions can reasonably be assumed to come from sea salt, and these reactions would occur predominantly in polluted, off-shore transport from North America, when mineral dust contributions to Bermuda would be minimal. Because the coarse mode NO3–(p) formed through these heterogeneous pathways derives all its O atoms from O3, it should have a higher δ18O than HNO3(aq) formed from N2O5 hydrolysis (R7). The δ18O-NO3–(aq) should reflect both the high δ18O-NaNO3(p) and the lower δ18O-HNO3(aq) formed from N2O5 hydrolysis and other NO3– formation pathways (R4–R7). During the warm season, however, the OH pathway (R4) dominates NO3– formation and should set the δ18O of both aerosols and rainwater. Thus, the direct formation of high δ18O-NO3–(p) in the polluted marine boundary layer could lead to the isotopic difference between cool season rain and aerosol NO3–.
3.3 Seasonal characteristics of δ15N and δ18O
 The range of concentrations for both NO3–(p) (8.1 to 42 nmol m−3; n = 12, Table S2) and NO3–(aq) (0.8 to 33.3 μM; n = 126, Table S3) was consistent with other analyses in the North Atlantic [Baker et al., 2006, 2007; Hastings et al., 2003]. On average, dry deposition was 30% of total N (wet + dry) deposition (following the calculations of Baker et al.  and assuming all NO3– was in the coarse mode). The concentrations of rainwater and aerosol NO3– did not vary significantly by season. However, the concentration-weighted average cool season δ15N-NO3–(p) was significantly lower (p < 0.01) than the warm season δ15N-NO3–(p) (−6.2 ± 2.0‰ and −2.6 ± 0.6‰, respectively). Consistent with the findings of Hastings et al. , the δ15N-NO3–(aq) was also significantly lower in the cool season than in the warm season (−4.3 ± 3.0‰ and −1.4 ± 3.1‰, respectively, p < 10−4). The concentration-weighted averages for cool season δ18O-NO3–(p) and δ18O-NO3–(aq) (+79.1 ± 1.7‰ and +74.7 ± 5.5‰, respectively) were significantly higher (p < 0.01 and p < 10−4, respectively) than those of the warm season (+69.7 ± 3.4‰ and +67.1 ± 4.6‰, respectively).
 The isotopic data together with the AMBT patterns suggest a seasonal variation in NOx source such that during the cool season, continental (and thus anthropogenic) NOx is the primary contributor to NO3–(p) deposition. However, during the cool season, both the δ15N-NO3–(aq) (−4.3‰) and the δ15N-NO3–(p) (−6.2‰) are distinctly lower in Bermuda than in the United States, the source region (+0.1 ± 0.2‰, winter average δ15N-NO3–(aq) [Elliott et al., 2007]; and −1.5‰, annual mean δ15N-NO3–(p), with winter values significantly higher than summer values, although they were not reported separately [Elliott et al., 2009]). The isotopic difference between NO3– deposition in the U.S. and Bermuda is unlikely to be driven by a significant source difference. AMBTs suggest that the air masses transported to Bermuda are representative of the regions sampled by Elliott et al. . Moreover, the difference was observed in the concurrent sampling events of Elliott et al.  and Hastings et al. , and the δ15N-NO3–(aq) observations of the latter study are comparable to those of this study.
 If the ultimate source of the NO3– deposited in Bermuda during the cool season is the same as that of the NO3– deposited in the U.S. as predicted by AMBTs, then the pool of NOx and its oxidation products must preferentially lose 15N between the U.S. and Bermuda. When NOx concentrations are high relative to O3, equilibrium fractionation between NO and NO2 results in 15N-enriched NO2 and 15N-depleted NO [Freyer et al., 1993]. By contrast, when O3 concentrations are greater than NOx concentrations, NOx tends toward NO2, such that the δ15N-NO2 is expected to equal δ15N-NOx. In Bermuda, O3 concentrations always exceed NOx concentrations [Prados et al., 1999 and references therein]; therefore, fractionation between NO and NO2 has not previously been considered important [Hastings et al., 2003]. In the atmospheric boundary layer over North America, however, NOx concentrations are comparable to O3 in many areas and exceed O3 in heavily industrialized zones [Liang et al., 1998]. Fractionation between NO and NO2 may therefore be important because 65% of NOx emitted in the U.S. is deposited over the continent in the winter [Liang et al., 1998], potentially altering the isotopic composition of the NOx exported from the continent. A simple numerical model described in the supporting information that assumes 65% loss of anthropogenic NOx with a starting δ15N of +6‰ and Rayleigh fractionation with an isotope effect of 1.022 [Freyer et al., 1993] results in a δ15N-NOx pool of −5.6‰, consistent with NO3– deposition at Bermuda (Figure 2; the supporting information also includes a discussion of the model's sensitivity to the assumed isotope effect). This mechanism could be tested by following a coherent plume of NOx and NO3–, as performed by Neuman et al.  and measuring the time-dependent evolution of both the concentrations and isotopic compositions of NO3– and NOx as NO3– is deposited during transport from the U.S. to Bermuda.
 Seasonal trends in the δ18O-NO3– similar to the results of this study have been observed in both aerosol [Elliott et al., 2009; Wankel et al., 2009] and rainwater [Elliott et al., 2009, Hastings et al., 2003] and are generally taken to indicate a change in the formation pathway, with the daytime reaction (R4) gaining importance during the warm season and heterogeneous nighttime reaction (R7) gaining importance during the cool season. This interpretation is consistent with the isotopic results of this study, the AMBTs, and modeled predictions of the importance of NO3– formation pathways [Alexander et al., 2009; Dentener and Crutzen, 1993]. Because of the uncertainty in the absolute values, seasonal trends, and spatial variability of the δ18O-O3 and δ18O-OH, the δ18O-NO3– alone gives us limited capacity to quantify the contributions of these pathways; nevertheless, a discussion of trends remains possible.
 The NO2 + OH pathway (R4) becomes relatively more important during the warm season: increased radiation from longer days and a decreased angle of solar incidence result in greater production of OH; and decreased atmospheric transport from the U.S. results in less competition with O3 in polluted air. The NO3 + DMS pathway (R8) should also contribute the most in absolute terms during the warm season, when the flux of DMS to the marine atmosphere peaks [Bates et al., 1992]. This pathway would tend to increase the δ18O-NO3– because NO3, the precursor, derives all its O atoms from O3. Since the average δ18O-NO3– decreases during the summer, we can infer that despite the likely increase in total NO3– formed by the NO3 + DMS pathway compared to the cool season, the relative strengthening of the NO2 + OH pathway drives the isotopic composition during the warm season.
 The N2O5 pathway (R7) becomes important under the opposite conditions that lead to increased importance of the NO2 + OH pathway: longer nights, increased angle of solar incidence, and increased pollution, all occurring during the cool season, lead to the increased relative contribution of O3. The heterogeneous halogen reactions (R12–R15) also become more important during the cool season because of the transport of halogens in polluted air and the increased salt load in the marine atmosphere due to higher wind speeds. Both halogen reactions and the N2O5 pathway would increase δ18O-NO3–, and thus both may contribute to the observed trend. Future studies should pursue the quantification of these pathways through simultaneous gas, aerosol, and rainwater sampling, and combined 18O and 17O measurements [Morin et al., 2009; Vicars et al., 2013].
 The N and O isotopic composition of NO3–(p) at Bermuda follows the same seasonal trends as NO3–(aq): lower δ15N and higher δ18O during the cool season than during the warm season. The seasonal change in δ15N-NO3–(p) is best interpreted as a seasonal change in NOx source associated with the shift in dominant transport patterns. The cool season δ15N-NO3– observed is much lower than that of NO3– observed in the presumed source region in North America, a difference best explained by the loss of 15N-enriched aerosols during transport of anthropogenic NOx from North America to Bermuda, although the mechanism of this loss requires further investigation. The seasonal change in δ18O-NO3–(p) is consistent with a cool-to-warm season shift in the relative importance of the O3-driven nighttime and halogen chemistry versus OH-driven daytime chemistry, respectively.
 Aerosol δ15N-NO3–(p) is generally lower than or equal to rainwater δ15N-NO3–(aq) collected concurrently. A study in the marine atmosphere showed the same trend [Baker et al., 2007], while studies over continents showed the opposite trend [Elliott et al., 2007, 2009; Freyer, 1991; Mara et al., 2009]. We propose that this isotopic difference is driven by kinetic fractionation during the formation of coarse mode aerosol NO3– from HNO3(g) due to acid displacement when HNO3(g) reacts with sea-salt and mineral dust particles. This mechanism explains the difference in δ15N between NO3–(p) and NO3–(aq) during both the cool and warm seasons, and why the offset over the continents is in the opposite direction. The difference between aerosol and rainwater δ18O-NO3– unique to the cool season could be driven by heterogeneous halogen chemistry in the polluted marine boundary layer leading to the formation of high δ18O-NO3–(p). This work demonstrates the utility of isotopes as a tracer of seasonal changes in NOx source and NO3– formation pathways, as well as differences in atmospheric chemistry between polluted and marine regions.
 This work was supported by NSF ATM-1044997. Postdoctoral research support for K.E.A. was also provided by the NOAA Climate and Global Change Fellowship. Operation of the Tudor Hill facility is supported by NSF OCE-1130395. Additional support for sample collection and analysis was provided by the Grand Challenges Program at Princeton University. We thank A. Marks and J. Rosset for sample collection assistance.
 The Editor thanks two anonymous reviewers for their assistance in evaluating this paper.