Ammonia in the summertime Arctic marine boundary layer : sources , sinks , and implications

Continuous hourly measurements of gas-phase ammonia (NH3(g))were taken from 13 July to 7 August 2014 on a research cruise throughout Baffin Bay and the eastern Canadian Arctic Archipelago. Concentrations ranged from 30 to 650 ng m (40–870 pptv) with the highest values recorded in Lancaster Sound (7413 N, 8400W). Simultaneous measurements of total ammonium ([NHx]), pH and temperature in the ocean and in melt ponds were used to compute the compensation point (χ), which is the ambient NH3(g) concentration at which surface–air fluxes change direction. Ambient NH3(g) was usually several orders of magnitude larger than both χocean and χMP (< 0.4–10 ng m ) indicating these surface pools are net sinks of NH3. Flux calculations estimate average net downward fluxes of 1.4 and 1.1 ng m s for the open ocean and melt ponds, respectively. Sufficient NH3(g) was present to neutralize non-seasalt sulfate (nss-SO 4 ) in the boundary layer during most of the study. This finding was corroborated with a historical data set of PM2.5 composition from Alert, Nunavut (8230 N, 6220W) wherein the median ratio of NH+4 /nssSO 4 equivalents was greater than 0.75 in June, July and August. The GEOS-Chem chemical transport model was employed to examine the impact of NH3(g) emissions from seabird guano on boundary-layer composition and nss-SO 4 neutralization. A GEOS-Chem simulation without seabird emissions underestimated boundary layer NH3(g) by several orders of magnitude and yielded highly acidic aerosol. A simulation that included seabird NH3 emissions was in better agreement with observations for both NH3(g) concentrations and nss-SO 4 neutralization. This is strong evidence that seabird colonies are significant sources of NH3 in the summertime Arctic, and are ubiquitous enough to impact atmospheric composition across the entire Baffin Bay region. Large wildfires in the Northwest Territories were likely an important source of NH3, but their influence was probably limited to the Central Canadian Arctic. Implications of seabird-derived N-deposition to terrestrial and aquatic ecosystems are also discussed.


Introduction
Ammonia (NH 3(g) ) is the dominant alkaline gas in the atmosphere and is an important component of the global nitrogen cycle. Its transport and deposition can have harmful effects for N-sensitive ecosystems such as eutrophication, loss of biodiversity and soil acidification (Krupa, 2003). The presence of NH 3(g) can impact climate by increasing 5 rates of new particle formation via stabilization of sulphuric acid clusters (Kirkby et al., 2011). Gas-phase NH 3 is also able to partition to acidic fine particulate matter (PM 2.5 ) to form particulate-phase ammonium (NH + 4(p) ), which alters various aerosol properties, such as scattering efficiency (Martin et al., 2004), hygroscopicity (Petters and Kreidenweis, 2007), ice nucleating ability (Abbatt et al., 2006) and heterogeneous chemistry 10 occurring on surfaces (Fickert et al., 1999).
As a result, the accurate quantification of the magnitude and location of NH 3(g) sources is important for chemical transport models (CTMs). The major anthropogenic source is agriculture (fertilization and animal husbandry) with biomass burning, transport and industry being minor contributors (Reis et al., 2009). Natural sources include 15 soils, vegetation, oceans and animal excreta (Sutton et al., 2013). Estimates for the annual global emissions of NH 3(g) range from 35-54 Tg N yr −1 ; however, large uncertainties exist for these values due to the area-wide nature and poor characterization of many sources. In remote marine environments, the ocean is thought to be the dominant source of NH 3(g) to the marine boundary layer and delivers an estimated Gibb et al., 1999;Johnson et al., 2008;Quinn et al., 1988Quinn et al., , 1990. Johnson et al. (2008) provided the most recent dataset and summarized the previous studies to show that the open ocean can be both a net source and a net sink of NH 3(g) , with sea surface temperature (SST) being a key determinant for the direction of flux. Colder SST reduces the emission potential due to increased solubility of NH 3 ; hence, at higher 10 latitudes the open ocean is more likely to act as a net sink (Johnson et al., 2008). Of the six previous studies, only Johnson et al. (2008) quantified NH 3 fluxes above the Arctic Circle (66 • 33' N) during a summer time study in the Norwegian Sea. Therefore additional measurements of sea-air NH 3 fluxes in the High Arctic are invaluable for improving constraints on oceanic NH 3 emissions. 15 During the summertime, freshwater melt ponds are a ubiquitous feature on top of melting Arctic sea ice and can comprise up to 80 % of the sea ice surface (Lüthje et al., 2006). These melt ponds form from melting sea ice and are anywhere from a few cm to over 1 m deep. They are chemically distinct from the bulk ocean owing to their low salinity and physical separation from the ocean mixed layer by sea ice or stratification. 20 To our knowledge, no studies to date have attempted to quantify melt pond-air NH 3(g) fluxes despite the abundant presence of melt ponds in the summertime Arctic.
Quantifying sea-air and melt pond-air NH 3 exchange in the Arctic will help elucidate the role these processes play as either sources or sinks in the Arctic nitrogen cycle. Many terrestrial Arctic ecosystems are N-limited and highly sensitive to perturbations 25 in N-input (Shaver and Chapin III, 1980), thus Arctic soils and vegetation are unlikely to represent important sources of atmospheric ammonia. Major sources at lower latitudes include agriculture, vegetation, transport and industry (Reis et al., 2009;Sutton et al., 2013) but these are expected to contribute minimally north of the Arctic Cir-Introduction cle. Since the lifetime of NH 3(g) is typically less than 24 h, long-range transport from lower latitudes is likely not important (Lefer et al., 1999). Substantial NH 3 emissions have been measured from both seabird guano (Blackall et al., 2007) and seal excreta (Theobald et al., 2006) so large colonies may be relevant point sources throughout the Arctic region. Biomass burning can also inject significant quantities of NH 3 into the free 5 troposphere and/or boundary layer (Bouwman et al., 1997). Although vegetation in the high Arctic is sparse, there can be large wildfires in boreal regions, and emissions may be transported poleward. The potential for the ocean and melt ponds to act as sources to the atmosphere will depend on the relative importance of sources and sinks within the atmosphere and the aqueous systems. 10 NH 3 emission to the atmosphere can affect the extent of non-sea salt sulphate (nss-SO 2− 4 ) neutralization, which has implications for N-transport (Lefer et al., 1999). Therefore, it is important to also consider the relative abundances of atmospheric NH x and nss-SO 2− 4 . The dominant source of the latter in the summertime Arctic is oxidation of dimethylsulphide (DMS) emitted from the Arctic Ocean (Leaitch et al., 2013;15 Sharma et al., 199915 Sharma et al., , 2012. Measurements of PM 2.5 composition in the summertime Arctic marine boundary layer are rare (e.g. Chang et al., 2011;Leck et al., 2001). Previous chemical transport model (CTM) studies with GEOS-Chem predict highly acidic aerosol (i.e. nss-SO 2− 4 NH x ) with negligible amounts of NH 3(g) throughout the summertime Arctic boundary layer (Breider et al., 2014). 20 The region for this study is the eastern Canadian Arctic Archipelago where shipbased atmospheric (NH 3(g) , NH + 4(p) , SO 2− 4(p) ) and oceanic ([NH x ], pH, SST) measurements were taken over a 4 week period in July and August 2014. To our knowledge, this study presents the first measurements of NH 3(g) in the Canadian Arctic. Motivated by a lack of atmospheric and oceanic measurements in the region, as well as sub-Introduction as possible to the inlet sampling point (Markovic et al., 2012). Ambient air is pulled through a PM 2.5 impactor to remove coarse (> 2.5 µm in diameter) particles at a flow of 3 L min −1 . Air then enters a parallel plate wet denuder where water-soluble gases are dissolved in a 2 mM H 2 O 2 solution which is continuously flowing across the denuder membranes. PM 2.5 particles with diameter larger than 2.5 µm have sufficient inertia to 5 pass through the denuder into a supersaturation chamber where they are collected as an aqueous solution via hygroscopic growth. These components were contained within an aluminum inlet box that was mounted to the hull near the bow of the ship. The height of the inlet was 1 m above the deck. The aqueous solutions collected in the inlet box were pulled down a 22 m sample line through a conduit leading to the IC systems 10 which were housed in a laboratory below deck. Half of each ∼ 10 mL aqueous aliquot (representing 1 h of sampling) was then separately injected onto both a cation IC and anion IC for quantification of water-soluble ions.
During the campaign, three zero air overflow experiments were performed to quantify the background signal of each analyte measured during AIM-IC ambient sampling. For each experiment the inlet was overflowed with high purity zero air (AI 0.0 UZ-T, PraxAir, Toronto, ON) at 4.5 L min −1 for 18 h. The average peak area during the final 25 8 h of each experiment was used as a background and subtracted from each ambient measurement. Detection limits were calculated by taking 3 times the standard deviation of each analyte peak area during the final 8 h of each zero air overflow. This value was then converted to either a mixing ratio or mass loading assuming standard temperature

Surface measurements
A total of 37 surface ocean and 9 melt pond samples were collected throughout the study. Melt pond samples were collected directly into a cooler jug using an electrical pump fixed on a telescopic arm. The water was sampled as far from the side as possible, between 1-2 m depending on the size of the melt pond. Temperature 20 was measured in situ with a VWR high precision thermometer and total aqueous [NH x ] was determined within 10 h of sampling using a fluorometric technique that has been optimized for low concentrations and complex matrices (Holmes et al., 1999 uously measured by a thermosalinograph (Seabird Electronics SBE 45) connected to the seawater inlet. For the purposes of flux calculations, the ocean pH and salinity were assumed to be 8.1 and 35 g kg −1 , respectively, which are representative for the region of interest (Takahashi et al., 2014). These assumptions have been made previously and were not found to be a major source of uncertainty when calculating sea-air NH 3 5 fluxes (Johnson et al., 2008). The melt pond pHs were measured using a pH-meter within four hours of sampling. A three point calibration of the pH probe (Orion ™ Model 91-72, Thermo Scientific) was performed using commercially available pH 4.01, 7.00 and 10.00 buffers. Salinity of the melt ponds were determined with a WTW Cond 330i handheld conductivity meter.

Flux calculations
The direction of sea-air NH 3 fluxes can be assessed by comparing ambient measurements of NH 3(g) to the atmospheric mixing ratio predicted from Henry's Law equilibrium calculations using seawater [NH x ] and surface temperature measurements (e.g. Asman et al., 1994;Johnson et al., 2008;Quinn et al., 1988Quinn et al., , 1996. This equilibrium NH 3 15 concentration signifies the ambient value at which the net flux changes direction, and is known as the compensation point (denoted χ ). In other words, one expects a net downwards flux if ambient NH 3(g) exceeds χ and a net upward flux if it is below χ . The magnitude of these fluxes are commonly computed using the "two-phase" model first developed by Liss and Slater (1974), which describes the sea-air transfer of gases 20 as being controlled by molecular diffusion on either side of the interface. The transfer of NH 3 across this interface is predominantly dictated by the air-side transfer velocity, given the relatively high water solubility of NH 3 (Liss, 1983). Hence, the equation to calculate sea-air NH 3 fluxes is: where F NH 3 is the sea-air flux of NH 3 (ng m −2 s −1 ), k g is the air-side transfer velocity (m s −1 ), NH 3(g) is the measured ammonia concentration (nmol m −3 ), χ is the compensation point (nmol m −3 ), and the molecular weight of 17.03 g mol −1 is to convert nmol to ng. Numerous parameterizations exist for k g with varying degrees of complexity (Johnson, 2010). Here we adopt the approach established by Duce et al. (1991): where u is the wind speed (m s −1 ) and MW is the molecular weight of the gas of interest (17.03 for NH 3 ). Although simple, this parameterization has been used previously to estimate sea-air NH 3 fluxes (e.g. Johnson et al., 2008) and has been shown to be in good agreement (within 20 %) with a more complex scheme, particularly at lower wind 10 speeds (Johnson, 2010). The following equation is used to calculate χ : where K H is the Henry's law constant (dimensionless) and [NH 3(sw) ] is the concentration of dissolved ammonia in the surface pool (nmol m −3 where T is the surface temperature (in K). The following equation is used to relate the NH 3(sw) to the concentration of total dissolved NH x ([NH x(sw) ]), which is the value actually measured by the procedure outlined in Sect. 2.3: 20 ACPD 15,2015 Ammonia in the summertime Arctic marine boundary layer where K a is the acid dissociation constant of NH + 4 . The pK a (≡ − log K a ) is calculated according to Bell et al. (2008), which provides an empirical correction for salinity (S, dimensionless) at a given temperature (T , in • C): Equations (2) and (4) closely follow that of Johnson et al. (2008) but are sufficiently 5 similar to analogous approaches for calculating K H and k g used in other sea-air NH 3 exchange studies (e.g. Asman et al., 1994;Gibb et al., 1999;Quinn et al., 1992). Johnson (2004) reported that fluxes calculated with these various schemes usually agree within 2 %. Melt pond-air exchange was also examined using Eqs.

10
The GEOS-Chem chemical transport model (www.geos-chem.org) is used to aid in the interpretation of the atmospheric measurements. We use GEOS-Chem version 9-02 at 2 • × 2.5 • resolution globally, and with 47 vertical layers between the surface and 0.01 hPa. The assimilated meteorology is taken from the NASA Global Modelling and Assimilation Office (GMAO) Goddard Earth Observing System version 5.11.0

15
(GEOS-FP) assimilated meteorology product. Boundary layer mixing uses the nonlocal scheme implemented by Lin and McElroy (2010). Our simulations use 2014 meteorology and allow a 2 month spin-up prior to the simulation.
The GEOS-Chem model includes a detailed oxidant-aerosol tropospheric chemistry mechanism as originally described by Bey et al. (2001). Simulated aerosol species 20 include sulphate-nitrate-ammonium (Park et al., 2004(Park et al., , 2006, carbonaceous aerosols (Park et al., 2003;Liao et al., 2007), dust (Fairlie et al., 2007(Fairlie et al., , 2010 and sea salt (Alexander et al., 2005). The sulphate-nitrate-ammonium chemistry uses the ISOR-ROPIA II thermodynamic model (Fountoukis and Nenes, 2007), which partitions ammonia and nitric acid between the gas and aerosol phases. For our simulations, the ACPD 15,2015 Ammonia in the summertime Arctic marine boundary layer GEOS-Chem simulates both wet and dry removal of aerosols and gases. Dry deposition follows a standard resistance in series scheme (Wesley, 1989) with an aerosol dry 10 deposition velocity of 0.03 cm s −1 over snow and ice (Fischer et al., 2011). Wet removal in GEOS-Chem takes place in large-scale clouds and convective updrafts (Liu et al., 2001). In-cloud scavenging of hydrophilic species takes place at temperatures warmer than 258 K, and hydrophobic black carbon and dust are also removed at temperatures colder than 258 K (Wang et al., 2011).

FLEXPART-WRF
FLEXPART-WRF (Brioude et al., 2013, website: flexpart.eu/wiki/FpLimitedareaWrf) is a Lagrangian particle dispersion model based on FLEXPART (Stohl et al., 2005) that is driven by meteorology from the Weather Research and Forecasting (WRF) Model (Skamarock et al., 2005). Here we use FLEXPART-WRF run in backward mode to study 20 the emissions source regions and transport pathways influencing ship-based ammonia measurements. A WRF simulation for the summer 2014 NETCARE campaign was performed using WRF 3.5.1 with initial and boundary conditions provided by the operational analysis (0.25 The relevant measurements needed to calculate χ for both the open ocean and melt ponds are listed in the Supplement (Tables S1 and S2, respectively). Only four unique co-ordinates are listed for the nine melt pond samples because multiple melt ponds were sampled at each location. Roughly half of the surface ocean samples had [NH x ] below the detection limit (20 nM) and in general values were significantly lower than in 5 the melt ponds. Open ocean samples ranged from < 20 to 380 nM whereas seven of the nine melt pond samples were between 640 to 1260 nM (with the other two below detection limit). These concentrations and their spatial variability are typical for the region during summer (Martin et al., 2010).
Parameters listed in Tables S1 and S2 were input into Eqs.
(3) to (6) to calculate χ 10 for both the surface ocean and melt pond samples. For samples with [NH x ] below the detection limit, a value of 10 nM (half of the detection limit) was assumed. A comparison of the calculated compensation points for the ocean (χ ocean ) and melt ponds (χ MP ) are shown in Fig. 2. Also shown is the range for the nearest valid measurement (see Sect. 2.2) of ambient NH 3(g) . The NH 3(g) concentration taken during the hour of surface 15 sampling could not be used since the ship remained stationary for up to 12 h while melt pond or ocean work was being conducted. Hence, the NH 3(g) measurement from several hours prior (as the ship approached the surface sampling site) had to be used. This approach should not significantly impact the analysis given that the ambient levels of NH 3(g) were observed to be fairly uniform from one hour to the next (i.e. no rapid 20 spikes of NH 3(g) were measured). Shown in lighter yellow are the ranges of NH 3(g) observed over the entire study (from ∼ 30-650 ng m −3 ). Figure 2 clearly shows that the ambient concentrations of NH 3(g) exceed both χ ocean and χ MP by several orders of magnitude throughout the entire region. This conclusively demonstrates that during the summertime, the ocean and melt ponds are net sinks of atmospheric NH 3(g) . This

Sulphate neutralization
The extent of neutralization of PM 2.5 influences aerosol properties as discussed pre- 15 viously. Figure 4 depicts the relative abundances (in neq m −3 ) of gas-phase ammonia and particulate-phase ammonium and sulphate. It is important to note that the value for sulphate is total PM 2.5 sulphate as opposed to non-sea salt sulphate (nss-SO 2− 4 ), which is commonly reported for marine boundary layer studies. High and variable backgrounds of Na + from the AIM-IC prevented the calculation of nss-SO 2− 4 , hence this 20 dataset provides an upper limit for nss-SO to numerous values near or below detection limit. On the other hand, after 25 July the nanoequivalents of NH 3(g) were substantially higher (NH x ≈ NH 3 ) than either NH + 4 or SO 2− 4 , which implies a nearly neutralized sulphate aerosol. It is important to note that a nearly neutralized aerosol does not equate to an aerosol with a pH of 7 since aerosol pH is highly sensitive to liquid water content as well as the precise NH + 4 : SO 2− 4 ratio. An 5 aerosol with NH + 4 : SO 2− 4 approaching 1 can still have an acidic pH. Figure 5 shows the distribution of the NH + 4 : nss-SO 2− 4 ratio (on a per equivalent basis) measured at Alert, Nunavut (82.50 • N, 62.33 • W) as a function of month from 1996-2011. Weekly-averaged PM 2.5 speciation measurements at Alert are made by Environment Canada and are available on-line (Environment Canada, 2014). The contribution from NO − 3 is minor and has not been included in this analysis. During July and August the nss-SO 2− 4 is, on average, completely neutralized by the NH + 4 in PM 2.5 as shown by a median neutralization ratio approaching 1 during these months. This implies there is sufficient NH 3(g) throughout the region to neutralize nss-SO 2− 4 produced from DMS oxidation which is consistent with the measurements shown in Fig. 4. 15 The AIM-IC and Alert measurements are both inconsistent with a previous study that used GEOS-Chem to predict a highly acidic aerosol and insignificant gas-phase ammonia (NH x ≈ NH + 4 ) throughout the summertime Arctic marine boundary layer (Breider et al., 2014). This inconsistency implies a missing process in a widely used CTM that we investigate further below.

Evidence for the importance of seabird guano
Observations collected on board the Amundsen and at Alert strongly suggest a significant source of NH 3 in the Baffin Bay region. Decomposition of uric acid in seabird guano (excreta) has been recognized as a significant source of NH 3 where large colonies exist (Blackall et al., 2007;Wilson et al., 2004). However, studies measur- 25 ing NH 3 from seabird colonies are limited due to the remoteness of most colonies and technical challenges in quantifying NH 3 in isolated locations (Blackall et al., 2007) few studies that have been done have focused on colonies located in the United Kingdom (Blackall et al., 2004;Wilson et al., 2004), Antarctica (e.g. Legrand et al., 1998;Zhu et al., 2011) and remote tropical islands (Riddick et al., 2014;Schmidt et al., 2010). Recently, Riddick et al. (2012) developed a global inventory to estimate the magnitude and spatial distribution of NH 3(g) from seabird guano. The authors employed a bioener-5 getics model, first developed by Wilson et al. (2004), to calculate the NH 3(g) emissions (in g bird −1 yr −1 ) for 323 different seabird species. After compiling a list detailing the populations and locations of 33 225 colonies, they were able to estimate global annual emissions between 97-442 Gg NH 3 yr −1 . Although this is less than 2 % of total global NH 3(g) emissions, it can be the dominant source in remote regions where seabird pop- 10 ulations are large and other sources are negligible. In order to assess the impact of seabird guano on NH 3 across the Baffin Bay region, seabird colony NH 3 emissions were implemented in the GEOS-Chem model, and the impact on monthly mean surface layer NH 3 was examined. The NH 3 emissions inventory used in the standard GEOS-Chem v9-02 (and in many other CTMs) is from Bouw-  Paulot et al. (2015). Scenario 3 was chosen since this represented the midpoint between the minimum and maximum emissions of scenario 1 and 2, respectively. Close inspection of this seabird inventory revealed that some large 20 seabird colonies in our study region were not accounted for. To investigate this, the spatial co-ordinates of northern colonies (> 50 • N) in the Riddick et al. (2012) inventory were cross-referenced against colonies in the on-line Circumpolar Seabird Data Portal (Seabird Information Network, 2015). Annual emissions for large colonies in the seabird data portal were calculated in the same manner as in Riddick et al. (2012). In 25 total, there were 42 colonies present in the seabird data portal but absent in the Riddick inventory in the region north of 50 • N. These additional emissions were added to the inventory we implemented in GEOS-Chem. These colonies totaled 7.5 Gg NH 3 yr −1 (approximately one quarter of the existing emissions north of 50 • N) and were primar-ACPD 15,2015 Ammonia in the summertime Arctic marine boundary layer ily in Siberia and western Alaska. The Riddick et al. (2012) bioenergetics model only counts emissions that occur during breeding season and while the seabirds are at the colony. Hence, the annual emission estimates (Gg NH 3 yr −1 ) per colony were temporally allocated evenly between the 15 May to 15 September. This period is when the majority of seabirds in the Baffin Bay region are nesting (e.g. Gaston et al., 2005;Mal-5 lory and Forbes, 2007;McLaren, 1982). One limitation to this approach is that it does not account for additional temporal variations in NH 3 emissions. For instance, moisture increases the rate of uric acid degradation, and fluxes of NH 3 from guano have been observed to increase 10-fold for up to a day after rain events (Riddick et al., 2014). Figure 6 shows the July mean output for surface layer NH 3 mixing ratio both with-10 out (Fig. 6a) and with (Fig. 6b) seabird emissions, along with the NH 3(g) measured by the AIM-IC denoted by circles in Fig. 6a. Comparing the top two panels reveals that seabird emissions make a substantial impact on modelled NH 3 levels in the boundary layer. Much better model-measurement agreement is achieved with the inclusion of the seabird colonies. Without the seabird emissions, NH 3 mixing ratios are underpredicted 15 by several orders of magnitude. Surface NH 3 is still underpredicted in Fig. 6b (with guano NH 3 emissions) which could be the result of modelled emissions being independent of rainfall, which can substantially increase NH 3 emissions. Episodic rainfall was persistent throughout the latter half of the campaign. Other contributing factors may include: challenges in representing boundary layer mixing, uncertainties in deposition 20 rates, comparing monthly averages (GEOS-Chem) to ambient hourly measurements, missing/underestimated bird colonies, and/or excreta from other fauna (e.g. seals, caribou, musk-ox) absent in the updated inventory. The bottom two panels ( Fig. 6c and d) show the influence of seabirds on the ammonium to non-sea salt sulphate ratio. Without seabirds (Fig. 6c) the ratio is less than 0.3 throughout most of the study region, 25 which is inconsistent with the abundance of NH 3 relative to SO 2− 4 measured by the AIM-IC. Adding the seabird emissions (Fig. 6d) increases the ratio to above 0.7 in most grid cells along the ship track. Although the high ratio (July average is ∼ 1) observed at Alert (denoted by the star in Fig. 6c and d) is underestimated in the GEOS-Chem ACPD 15,2015 Ammonia in the summertime Arctic marine boundary layer simulation, the bias is reduced by nearly a factor of 2 (from 0.32 to 0.57) when seabird emissions are included. Wildfires are also a source of NH 3 to the free troposphere and/or boundary layer. Particularly strong wildfire events were persistent in the Northwest Territories (NWT) during the study period. Blue circles in Fig. 7 show the location and average fire radia-5 tive power (representative of fire strength) of wildfires across the Arctic from 20-26 July. It was constructed using data from NASA's Fire Information Resource Management System (FIRMS) database (NASA, 2015). We used FLEXPART-WRF retro plumes to assess the importance of wildfire NH 3 emissions, as well as to further corroborate the influence of seabird guano. 10 The significant impact of seabird colonies on [NH 3(g) ] is supported by the analysis of FLEXPART-WRF retro plumes shown in Fig. 7. Periods of low [NH 3(g) ] (bottom panel in Fig. 7) correspond to air masses that spent at least the last 48 h over the ocean and/or aloft above the MBL (∼ 500 m) where NH 3 sources are negligible. This is clearly shown in Fig. 7a where the air mass sampled on 14 July 00:00 UTC spent the previous 96 h in 15 the MBL over Baffin Bay, consistent with low [NH 3(g) ]. In contrast, on 26 July 00:00 UTC (Fig. 7b) air had recently passed over seabird colonies (purple circles) surrounding Lancaster Sound as well as wildfires in the Northwest Territories (NWT) on mainland Canada (blue circles), coincident with the large increase in [NH 3(g) ]. A similar NH 3(g) peak occurs on 3 August that can also be examined by using a retro plume analysis. 20 Low NH 3(g) values observed on the morning of 2 August agree with Fig. 7c showing the air originating from the MBL over Baffin Bay. At 3 August 00:00 UTC (Fig. 7d) the air had spent the last 12 h in the boundary layer of Western Greenland where large seabird colonies exist. However, by 4 August 00:00 UTC (Fig. 7e) the retro plume shifted such that air is now originating from primarily above the boundary layer (altitude 25 plots not shown) leading to a decrease in NH 3(g) . In addition from 2-4 August the ship was north of 79 • N and in the Eastern Canadian Arctic, hence it is unlikely that this increase in NH 3 can be attributed to wildfires given how far removed this region is from wildfires in the NWT. While Fig. 7 only highlights five examples from the study period, ACPD 15,2015 Ammonia in the summertime Arctic marine boundary layer G. R. Wentworth et al. retro plumes throughout the entire campaign also support the hypothesis that NH 3(g) in the MBL originates primarily from seabird colonies (for the Eastern Canadian Arctic) with contributions from wildfires in some regions (central Canadian Arctic). All NH 3(g) spikes in the time series can be attributed to air that had recently passed over seabird colonies and/or wildfires, whereas low values coincide with air masses from either the 5 open ocean or free troposphere not influenced by wildfires.
To further investigate the potential influence of wildfires on NH 3 in the Arctic MBL, GEOS-Chem simulations were performed using a wildfire emissions inventory for 2014 (QFED2). Simulations with/without wildfires and with/without seabirds revealed that in Lancaster Sound (along 74 • N) roughly 40 and 55 % of the boundary layer NH 3 can be attributed to seabirds and wildfires, respectively. In other words, air sampled in Lancaster Sound (20 July to 27 July) was likely influenced by wildfires in NWT in addition to seabird guano. On the other hand, north of Lancaster Sound, contributions from seabirds and wildfires to surface layer NH 3 were approximately 95 and 5 %, respectively. Wildfires in the NWT are an important but episodic source of summertime NH 3 15 in the Canadian Arctic. This is due to periodic transport events associated with this source that is located remote to our study region. Whereas, seabird colonies are a local, and persistent source of NH 3 from May to September. Given the observation of consistently neutralized sulfate at Alert each summer, and the large interannual variability and episodic wildfire influence, emissions from migratory seabirds are likely to 20 be a significant contributor to NH 3 abundance in the Arctic marine boundary layer.

Implications for N-deposition to ecosystems
Previous studies have highlighted the important role that seabird-derived N can play in the nitrogen cycle of ecosystems adjacent to bird colonies due to large deposition rates of NH 3 and NH + 4 (e.g. Anderson and Polis, 1999;Lindeboom, 1984). However, 25 little attention has been paid to the effects of seabird-derived N on deposition at the regional scale. In this section, we consider the importance of seabird-derived nitrogen as an input of reactive N to Arctic ecosystems. These ecosystems tend to be N-limited 29993 Introduction during the summer and hence have a large sensitivity to N input (Shaver and Chapin III, 1980). In terrestrial ecosystems, soil N availability is a key factor in determining both plant community structure (McKane et al., 2002) and greenhouse gas emissions from soil (Stewart et al., 2012). Nitrogen (N 2 ) fixation via microbes is thought to be the primary N input to remote 5 Arctic terrestrial ecosystems (e.g. Cleveland et al., 1999;Hobara et al., 2006;Stewart et al., 2014). Numerous field studies have been conducted to estimate N 2 -fixation rates via the acetylene reduction technique (Hardy et al., 1968). The N 2 -fixation rates for most terrestrial Arctic sites fall within the range of 10 to 120 mg N m −2 yr −1 (Hobara et al., 2006). However, highly variable rates (due to spatial heterogeneity of microbial 10 populations) and assumptions in the acetylene reduction technique yield high degrees of uncertainty for N 2 -fixation rates (Stewart et al., 2014). Total atmospheric N-deposition (wet and dry) in the Arctic is thought to be smaller than fixation, with typical ranges from 8 to 56 mg N m −2 yr −1 (Van Cleve and Alexander, 1981). Only a few N 2 -fixation studies also quantify wet deposition, with dry deposition 15 being ignored altogether (e.g. Hobara et al., 2006). Nonetheless, in certain Arctic regions atmospheric deposition may exceed N 2 -fixation in soils (DeLuca et al., 2008). These processes are coupled since large inputs of NH + 4 have been shown to inhibit N 2 -fixation in certain microbial species and lichens (Chapin and Bledsoe, 1992). Figure 8 shows results from the GEOS-Chem simulation of total NH x deposition 20 (both wet and dry) for the months May to September (inclusive) both without (Fig. 8a) and with (Fig. 8b) seabird NH 3 emissions. The difference in total NH x deposition for birds and no birds is shown in Fig. 8c (absolute difference) and Fig. 8d (percent different). Areas near large colonies are heavily influenced by seabird guano with NH x deposition from seabirds exceeding 10 mg N m −2 yr −1 , particularly in western Green- useful for comparing N-deposition to N 2 -fixation since it captures deposition only during the growing season, and NH x is likely the dominant form of atmospheric reactive N in the summertime Arctic boundary layer. Furthermore, Fig. 8b provides information on regions where N-deposition rates could be comparable to input from terrestrial N 2 -fixation (> 10 mg N m −2 yr −1 ) which can help inform subsequent studies exploring 5 N-cycling in the region. According to Hobara et al. (2006), Arctic terrestrial N 2 -fixation only occurs from May-September (inclusive) and peaks in July, similar to migration patterns of Arctic seabirds. Estimates of N 2 -fixation rates in the Arctic Ocean mixed layer are even sparser than estimates for terrestrial ecosystems. To our knowledge, only Blais et al. (2012)  that nss-SO 2− 4 is, on average, completely neutralized by NH + 4 during July and August. These measurements imply strong regional source(s) of NH 3(g) in the eastern Canadian Arctic Archipelago that are sufficient to neutralize nss-SO 2− 4 produced from DMS oxidation. Our surface-air flux estimates show that the Arctic Ocean and melt ponds are not responsible for NH 3(g) in the marine boundary layer.
It is also noteworthy that even though these melt ponds have significantly higher [NH x ] than the open ocean (average of 670 vs. 55 nM), χ MP is only marginally higher. More acidic pHs and slightly lower temperatures mitigate the effect of higher [NH x ] on χ . Chemical transport models (CTMs) that explicitly account for bi-directional NH 3 exchange typically require χ as a predefined model input (e.g. Bash et al., 2013;Wichink 15 Kruit et al., 2012). Therefore, from a modelling standpoint, similar values of χ ocean and χ MP are convenient since they can be parameterized in a similar fashion which would remove the need for CTMs to resolve the spatial extent and temporal evolution of melt ponds to properly model surface-atmosphere NH 3 exchange in the summertime Arctic.
To investigate the impact of NH 3 emissions from seabird guano, we examined GEOS- 20 Chem simulations both with and without seabird colony NH 3 emissions. The seabird NH 3 emission inventory developed by Riddick et al. (2012) was updated for this study to include northern colonies (> 50 • N) that had been overlooked in the original inventory.
Without the seabirds, GEOS-Chem underestimated NH 3(g) by several orders of magnitude and predicted highly acidic aerosol at the surface in July, which is in direct con- 25 trast to our measurements. The inclusion of seabird emissions provided much better agreement with NH 3(g) observations and yielded more neutralized aerosol throughout most of the Baffin Bay region. The importance of seabird NH 3 emissions is also supported by analysis of FLEXPART-WRF retro plumes throughout the study period. Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | masses enriched in NH 3(g) had recently passed through regions with seabird colonies whereas periods of low NH 3(g) involved air masses originating from the open ocean or above the boundary layer. Together, these models provide strong evidence that seabird colonies are the dominant and persistent local source of NH 3(g) in the summertime Arctic. FLEXPART-WRF and GEOS-Chem were also used to assess the influence of wild-5 fires on NH 3 . Wildfires are an important but episodic source of NH 3 source to the Arctic due to ongoing changes in transport patterns and fire intensity. Further work should be done to examine the inter-annual influence of NH 3 emissions from wildfires in the NWT on other regions in the Arctic. Deposition estimates of NH x from GEOS-Chem during the seabird nesting season (May to September) exceed 10 mg N m −2 season −1 in grid cells close to large seabird colonies, which is on the lower end of microbial N 2 -fixation in Arctic tundra (Hobara et al., 2006). Hence, in some regions seabird-derived NH x could be a significant Ninput to terrestrial Arctic ecosystems which are typically very N-sensitive. Estimates of NH 3 fluxes into the open ocean are unlikely to be an important input of reactive-

15
N except for waters close to large seabird colonies; however, these fluxes may be important for the N-cycle in the much shallower melt ponds. There is strong evidence that seabird colonies are likely the dominant and persistent source of NH 3(g) to the summertime Arctic boundary layer. Emissions appear to be significant enough to at least partially neutralize nss-SO 2− 4 throughout most of the study 20 region, in contrast to previous model simulations that did not consider seabird colony emissions. Further research is required to better constrain the location, population, and NH 3 emissions of Arctic seabird colonies. It is also important to quantify meteorological effects (e.g. rainfall, wind speed) on seabird emissions. The NH 3 emissions inventory in CTMs should be updated to include seabird emissions with correct representation of 25 the breeding season so that emissions only occur when seabirds are nesting. Summertime measurements of atmospheric NH x elsewhere in the Arctic are needed to assess whether the impacts of seabirds observed in this study (substantial NH 3(g) , nss-SO 2− Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Bash, J. O., Cooter, E. J., Dennis, R. L., Walker, J. T., and Pleim, J. E.: Evaluation of a regional air-quality model with bidirectional NH 3 exchange coupled to an agroecosystem model, Biogeosciences, 10, 1635-1645, doi:10.5194/bg-10-1635-2013, 2013. Bell, T. G., Johnson, M. T., Jickells, T. D., and Liss, P. S.: Ammonia/ammonium dissociation coefficient in seawater: a significant numerical correction, Environ. Chem., 5, 183-186,  15,2015 Ammonia in the summertime Arctic marine boundary layer  Figure 8. GEOS-Chem simulation of for total NH x deposition (in mg N m −2 season −1 ) for the months May to September (inclusive). Panel (a) does not include seabird emissions, whereas panel (b) does. The difference in total NH x deposition between the two emissions scenarios (with birds minus without birds) is shown in panels (c) and (d) as an absolute amount and percentage increase, respectively.