Oxygen isotope mass balance of atmospheric nitrate at Dome C, East Antarctica, during the OPALE campaign

. Variations in the stable oxygen isotope composition of atmospheric nitrate act as novel tools for studying oxidative processes taking place in the troposphere. They provide both qualitative and quantitative constraints


Introduction
The search for ice core proxies to reconstruct past change of oxidative properties of the atmosphere is motivated by the need to simulate ozone and OH changes over preindustrialindustrial and glacial-interglacial timescales (Thompson, 1992;Wang and Jacob, 1998;Murray et al., 2014). Early ice core reconstructions of oxidants based on H 2 O 2 (Sigg and Neftel, 1991) and HCHO (Staffelbach et al., 1991) measurements were hampered by the occurrence of post-depositional alteration of H 2 O 2 and HCHO concentrations in the upper snowpack prior to preservation in the ice (Hutterli et Published by Copernicus Publications on behalf of the European Geosciences Union. al., 2003). More recently, the 17 O excess ( 17 O = δ 17 O-0.52 × δ 18 O) of nitrate, a unique isotopic signature inherited from ozone via bimolecular chemical reactions in the atmosphere, has shown promise as a conserved proxy for past oxidant concentrations (McCabe et al., 2005;Alexander et al., 2004). The 17 O signal of nitrate reflects the relative importance of NO x transformation mechanisms and recent studies suggest that the measurement of this isotopic signal in ice cores and ancient sediments may provide relevant information regarding the role of ozone in the overall oxidative capacity of the paleo-atmosphere (McCabe et al., 2007;Michalski et al., 2004Michalski et al., , 2003Savarino et al., 2007;Thiemens, 2006). Parallel studies of the nitrogen isotope ratios (δ 15 N) of nitrate in polar ice and snow suggest that this isotopic tracer may serve as a proxy for past variations in natural sources of atmospheric NO x (Hastings et al., , 2005Jarvis et al., 2008). However, the chemical and physical factors governing the oxygen and nitrogen isotopic composition of atmospheric nitrate and its nitrogen oxide precursors are not fully understood (Alexander et al., 2009;Morin et al., 2009). This is particularly true in the polar troposphere, where the UV photolysis of trace species present in the snowpack or marine aerosols initiates complex boundary layer oxidation processes involving reactive halogen species (Bloss et al., 2010;Grannas et al., 2007;Wang et al., 2007) and results in distinctive 17 O and δ 15 N signatures in atmospheric nitrate (Morin et al., , 2012. Here we present measurements of δ 15 N and δ 17 O and δ 18 O isotopic composition of atmospheric nitrate collected at Dome C between November 2011 and January 2012. These measurements were conducted within the framework of the OPALE project (Oxidant Production over Antarctic Land and its Export; Preunkert et al., 2012), which has provided an opportunity to combine nitrate isotopic observations with a wealth of meteorological and chemical observations, including measurements of species involved in nitrate production (NO x , O 3 , OH, HO 2 , RO 2 , etc.). The primary objective of this study was to reconcile observations of 17 O for atmospheric nitrate at Dome C with quantitative predictions based on nitrate isotope mass balance and atmospheric chemistry parameters, a unique opportunity offered by the OPALE campaign.

Site description and scientific context
Dome C is situated 3233 m above sea level on the East Antarctic Plateau (75 • 06 S, 123 • 23 E), approximately 1100 km from the coastal research station Dumont d'Urville and 560 km from the Vostok station. Deep ice cores were extracted at Dome C in the framework of the European Project for Ice Coring in Antarctica (EPICA) covering approximately 800 000 years (EPICA community members, 2004) and Vostok covering the last 420 000 years (Petit et al., 1999). In parallel, studies aiming to understand the meteorological, chemical, and physical factors governing the variability in trace constituents preserved in the ice were initiated .
Although the Antarctic Plateau is extraordinarily dry, cold, and far removed from sources of anthropogenic emissions, first atmospheric measurements of oxidants conducted in 1998-1999 during the ISCAT (Investigation of Sulfur Chemistry in the Antarctic Troposphere) field campaign revealed a high level of photochemical activity. For example, the average summertime OH concentration (2 × 10 6 cm −3 ) over the South Pole was found to be similar to that of the tropical marine boundary layer (MBL) . Unexpectedly high levels of nitric oxide (NO) were also detected, with concentrations 1 to 2 orders of magnitude higher than that typically observed in other remote regions . Model simulations revealed that the large OH concentrations observed at South Pole were a result of the elevated NO level, which catalyzes a rapid cycling of HO 2 to OH (Chen et al., 2004. The high concentrations of NO x were also inferred to drive in situ photochemical production of ozone during the ISCAT campaign (Crawford et al., 2001). Surface ozone and NO x measurements at Dome C suggest a similar level of enhanced oxidant production during November-January (Frey et al., , 2015Legrand et al., 2009).
The high levels of photochemical activity observed at South Pole and Dome C are now understood in terms of NO x release from the snowpack (Honrath et al., 2000(Honrath et al., , 1999Jones et al., 2001Jones et al., , 2000Zhou et al., 2001). This process is initiated by the photolysis of nitrate, which can lead to large fluxes of NO 2 , NO, and HONO from permanently sunlit snow (Anastasio and Chu, 2009;Grannas et al., 2007;Jacobi and Hilker, 2007;Legrand et al., 2014;Frey et al., 2013). Observed and modeled NO x production rates are largely capable of explaining the high levels of photochemical activity observed on the Antarctic Plateau during spring (France et al., 2011;Liao and Tan, 2008;Wang et al., 2007) although detailed and speciation of nitrogen oxides chemistry remain largely unknown in this NO x -rich/VOC-poor environment (Kukui et al., 2014;Frey et al., 2015;Legrand et al., 2014;Davis et al., 2008).

High-volume sampling
Bulk atmospheric samples were collected at Dome C on glass fiber filters using a high-volume air sampler (HVAS), which was installed on a platform 1 m above the ground. The HVAS was run by applying an average STP (standard temperature and pressure) flow rate of 0.7 m 3 min −1 that ensures the collection of a sufficient amount of nitrate for isotopic analysis. The atmospheric NO − 3 collected on glass fiber filters represents the sum of atmospheric particulate NO − 3 (pNO − 3 ) and gaseous nitric acid (HNO 3 )  . After each collection period, filters were removed from the HVAS and placed in clean 50 mL centrifuge tubes, which were sealed in plastic bags and stored at −20 • C. Upon arrival at our laboratory in Grenoble, atmospheric filter samples were extracted in 40 mL of ultra-pure water via centrifugation using Millipore Centricon ® filter units. Nitrate concentration was then determined for each filter extract solution using a colorimetric technique ).

Ozone collection
The nitrite-coated filter technique for ozone isotope analysis has been described in detail in Vicars et al. (2012) and Vicars and Savarino (2014). The principle of ozone collection underlying this technique is the filter-based chemical trapping of ozone via aqueous-phase reaction with nitrite (Adon et al., 2010;Geyh et al., 1997;Koutrakis et al., 1993;Krzyzanowski, 2004): By coupling this routine ozone measurement technique with recent analytical developments in the 17 O analysis of nanomole quantities of nitrate , the 17 Bhattacharya et al., 2008;Janssen and Tuzson, 2006), can be inferred from the oxygen isotopic composition of the nitrate produced via R1 within the coated filter matrix.
Ambient ozone collections were performed using an active air sampler consisting of 1/4 in. Teflon ® (PFA) tubing connecting three main sampler components: (i) a standard lowvolume vacuum pump (Welch ® , model 2522C-02) equipped with a volume counter and needle valve (or flow meter) for flow rate regulation, (ii) a closed PFA filter holder assembly (Savillex ® ) containing a pre-coated 47 mm glass fiber sampling substrate (Whatman ® , GF/A type), and (iii) an openfaced PFA filter holder assembly containing a 47 mm PTFE membrane filter (Zylon ® , 5 µm pore size) for the removal of particulate species upstream of the coated filter. Glass fiber sampling substrates were coated prior to sample collection with 1 mL of a nitrite-based ozone sampling solution (0.1 M NaNO 2 , 0.05 M K 2 CO 3 , 0.1 M glycerol) (Koutrakis et al., 1993), allowed to dry at 75 • C for approximately 10 min, and then stored frozen in the dark in individual plastic PetriSlide ™ containers (Millipore ® ). Samples were collected by loading pre-coated filters into the sampling filter holder, which was then connected to the prefilter (upstream) and needle valve/pump (downstream) and covered in aluminum foil to limit light exposure, a step that is necessary to limit the blank production rate during sample collection (Vicars et al., 2012). Air was then pumped through the sampling system at a target flow rate of 3.0 L min −1 .
Sampling was conducted during the OPALE campaign (December 2011-January 2012) from a climate-controlled shelter, and a total of 28 samples were collected. However, due to difficulties in the application of our analytical technique to the unique environmental conditions encountered in Antarctica, the results obtained from these samples were inconclusive (i.e., unrealistic variability) due to the lack of light protection of the filter holder (Vicars et al., 2012. Sampling and isotopic analysis of ozone was therefore repeated in 2012, and a complete annual record of 17 O(O 3 ) was obtained (n = 60). Procedural filter blanks were also collected at regular intervals and were subjected to all of the same preparation, storage, handling, and analytical procedures as field samples. After sample collection, filter samples and procedural blanks were returned to their containers, which were covered in aluminum foil and stored at −20 • C before processing and analysis.
Filter samples were extracted in 18 mL of deionized water (18.2 M , hereafter referred to as "MQ water"). In order to remove the excess (i.e., unreacted) nitrite reagent from the sample extracts, the solutions were treated with 1 mL of a 1 M sulfamic acid solution and then neutralized with a corresponding addition of high-purity sodium hydroxide (Granger and Sigman, 2009;Vicars et al., 2012). Extract solutions were then filtered via centrifugation using Millipore Centricon ® assemblies. The nitrate extracted from the coated filter samples was then subjected to isotopic analysis, as described in the following section.

Isotopic analysis
The comprehensive isotopic composition of nitrate ( 15 N / 14 N, 17 O / 16 O, 18 O / 16 O) was measured on a Finnigan ™ MAT253 isotope ratio mass spectrometer (IRMS), equipped with a GasBench II and coupled to an in-house-built nitrate interface . Nitrate in both the nitrite-coated filter and aerosol sample extracts was prepared for isotopic analysis by conversion to N 2 O via the bacterial denitrifier method (Casciotti et al., 2002;Kaiser et al., 2007;Michalski et al., 2002;Sigman et al., 2001). The detailed analytical procedure has been described elsewhere (see Morin et al., 2009) and is briefly presented here.
Denitrifying bacteria (Pseudomonas aureofaciens) were cultured in nitrate-amended soy broth and incubated for 5 days in stoppered glass bottles. Bacterial cultures, after concentration by centrifugation and resuspension, were dispensed as 2 mL aliquots into 20 mL glass vials, which were then crimped and purged with helium for 3 h. Approximately 100 nmol of sample nitrate was then injected into the purged vials and conversion of the sample nitrate to nitrous oxide (N 2 O) via bacterial denitrification was allowed to proceed overnight. The it was thermally decomposed at 900 • C into O 2 and N 2 . Following separation via gas chromatography, the O 2 and N 2 gas samples were directed into the ionization chamber of the IRMS. All analytical steps were identically performed on nitrate isotopic standards and their equimolar mixtures (International Atomic Energy Agency USGS 32, USGS 34, and USGS 35), which were prepared in an identical background matrix as the samples. Individual analyses were normalized through comparison with these three nitrate reference materials (Coplen, 2011;Werner and Brand, 2001). All isotopic enrichment values for nitrate are reported relative to VSMOW and air N 2 for oxygen and nitrogen, respectively. The overall accuracy of the method is estimated as the standard deviation of the residuals from the linear regression between the measured reference materials and their expected values. For the results reported here, the average uncertainties obtained for δ 18 O, 17 O, and δ 15 N were 1.6, 0.5, and 1.0 ‰, respectively.

Isotope ratios of ozone and atmospheric nitrate
Atmospheric nitrate concentrations observed at Dome C during the campaign are presented in Fig (France), as well as with measurements conducted along a latitudinal transect from 50 • S to 50 • N in the Atlantic Ocean (Vicars and Savarino, 2014). Although the 17 O(O 3 ) bulk seasonal cycle reveals some interesting features, like the winter maximum, probably in response of the permanent winter darkness and stratospheric air mass intrusions, a complete description is beyond the scope of the present paper. What should be kept in mind here is the quite stable 17 O(O 3 ) bulk value close to 26 ‰ that can be considered as representative of the OPALE campaign held in November-January.

Nitrate isotope mass balance
The availability of a large database of trace chemical species measurements at Dome C during a portion of the OPALE field campaign (December 2011)  nity to compare observed 17 O(NO − 3 ) values in the atmosphere to ones calculated from concurrent observations. As discussed at length in recent studies (for example, by Morin et al., 2011, andVicars et al., 2013), the 17 O-excess transfer functions associated with the various nitrate production pathways (i.e., 17 O(NO − 3 ) i values) can be estimated as a function of the 17 O of nitrate precursor gases (i.e., NO x , O 3 , OH, etc.) using mass balance calculations that trace the origin of oxygen atoms transferred during the chemical transformation of NO x in the atmosphere. All atmospheric nitrate production channels involve either NO 2 or a NO x reservoir species derived from NO 2 (e.g., N 2 O 5 ). The first step in determining the 17 O signature of each pathway is therefore a quantitative assessment of the steady-state 17 O value of NO 2 , which is typically calculated as a function of the 17 O value of O 3 and the reaction dynamics involved in the conversion of NO to NO 2 . As Dome C in summer is permanently under sunlight, photochemical inter-conversion of NO x continues: At photochemical steady state (i.e., R2-R4 being faster than NO 2 net sink reactions), an assumption that can be reasonably applied throughout the day at Dome C during summer, we have (Morin et al., 2011) where the term in bracket represents the laboratory-deduced anomaly transfer function of the NO + O 3 reaction , 17 O(O 3 ) bulk the 17 O excess of the bulk O 3 and α represents the fraction of the atmospheric NO 2 reservoir that has been produced through oxidation by O 3 rather than HO 2 / RO 2 at photochemical equilibrium (Alexander et al., 2009;Michalski et al., 2003;Morin et al., 2011;Röckmann et al., 2001):  (1) and (2), although established under the NO x steady-state approximation, are independent of NO 2 concentration, for which a bias in measurement cannot be ruled out. Indeed, as discussed by Frey et al. (2013Frey et al. ( , 2015, bias in NO 2 measurements is suspected partly because it remains difficult to explain the observed ratio of NO 2 / NO, which is systematically higher (up to a factor of 7) than predicted by calculations made by assuming photochemical steady state considering the NO 2 photolysis and reaction of NO with O 3 , HO 2 / RO 2 and BrO. Equation (2) also assumes that [HO 2 ] * is predominantly formed by the reaction H + O 2 and R + O 2 during the OPALE campaign (Kukui et al., 2014), resulting in the formation of [HO 2 ] * devoid of any significant 17 O excess (Morin et al., 2011). Using OPALE measurements of NO, O 3 , OH and HO 2 / RO 2 (Frey et al., 2015;Kukui et al., 2014), along with temperature dependent reaction kinetics data obtained from Atkinson et al. (2004), we have calculated the diurnally mass-averaged trend in α for the month of 2.5 a Production rates obtained from a 0-D box model (see Kukui et al., 2014, for details). b HONO production rate divided by a factor of 4 to balance the HO x radical budget (see Kukui et al., 2014, andLegrand et al., 2014, for justification via R3 (see also Table 1). In the absence of the α dilution effect introduced by the HO * 2 reaction, 17 O(NO 2 ) would equal 37 ‰, a value 8 ‰ lower than an estimation obtained from modeling only NO x -O 3 chemistry at standard temperature and pressure (Michalski et al., 2014). This difference is essentially explained by the use of different 17 O(O 3 ) bulk (32 ‰ in Michalski's simulation, 26 ‰ for our observations), which possibly corresponds to different conditions of the two studies.
By accounting for the origin of the oxygen atom transferred during the conversion of NO 2 to nitrate, the 17 O signature of the nitrate produced through different reaction mechanisms can be calculated. For summer conditions at Dome C, it is reasonable to assume that the dominant atmospheric nitrate formation pathway is the gas-phase association of NO 2 and the OH radical (Alexander et al., 2009): leading to the following 17 O-excess mass balance (Michalski et al., 2003;Morin et al., 2011): In order to predict the 17 O value of the nitrate produced through R5 by mass balance, the isotopic composition of tropospheric OH must be known. The OH radical participates in a rapid isotopic exchange with atmospheric water vapor, which represents a very large oxygen reservoir relative to OH, with a 17 O that is negligible compared to ozone or nitrate (Luz and Barkan, 2010). This exchange tends to erase the 17 O excess of OH under humidity and temperature conditions typical of the midlatitudes (Dubey et al., 1997); therefore, the 17 O of OH is normally assumed to be zero in modeling studies applied to these regions. As discussed by Morin et al. (2007), this assumption of 17 O(OH) = 0 is not valid under the low-humidity conditions encountered in the polar atmosphere. The degree of isotopic equilibration between OH and H 2 O can be determined as a function of the relative rates of the isotope exchange reaction and the main OH sink reactions: where L represents the total chemical loss rate of OH. β is the factor relating the initial 17 O transferred to OH upon its formation, denoted 17 O(OH) prod , to its steady-state 17 O value : In plain words, Eqs. (4)-(5) predict that when the isotopic exchange reaction dominates over OH chemical losses (i.e., β 1), the steady-state 17 O value of OH will be equal to that of water (i.e., 17 O ≈ 0 ‰). Conversely, when water vapor concentrations are low and the rate of chemical loss is large relative to the rate of the isotopic exchange, 17 O(OH) = 17 O(OH) prod . Kukui et al. (2014), using a Master Chemical Mechanism box model, constrained by the OPALE meteorological conditions and concurrent chemical observations, give the rate of the OH chemical sources and sinks. NO 2 as measured by Frey et al. (2015) represents at most only ca. 10 % (equivalent of ca. 1 ‰) of the total sink of OH, which is predominantly dominated by reactions with CO, CH 4 , aldehydes and to a lesser extent by reactions with O 3 , H 2 , and NO. Thus, the possible overestimation of NO 2 concentration has only a minor effect on β calculation and is well embedded within the total uncertainty of such calculation. To assess the value of 17 O(OH), we have computed β for the conditions found during the OPALE campaign using the same 0-D box model that is used to evaluate the budget of OH and RO 2 during the OPALE campaign (see Kukui et al., 2014, and Table 1) and used the exchange kinetic rates given in Dubey et al. (1997). The absolute water vapor concentration is deduced from relative humidity and temperature measurements using Bolton (1980) (i.e., P water = 6.112 × e , with P water in hPa and T in K). The results of this calculation (Fig. 6) indicate that β varies between 0.70 ± 0.10 (1σ ) and 0.30 ± 0.10 from midnight to noon for conditions prevailing during the OPALE campaign, suggesting that, on a daily average basis, approximately 43 % of the 17 O value originally present in OH is preserved from exchange with H 2 O, consistent with estimates for an Arctic site described by Morin et al. (2007).
The value of 17 O(OH) prod is more difficult to assess because of the interplay between HO and HO 2 , and the different sources involved in OH formation. In the NO xrich atmosphere at Dome C in summer, the O( 1 D) + H 2 O reaction forming OH is a minor reaction pathway. When multiple pathways are involved in the production of OH, 17 O(OH) prod can be estimated by a simple isotope mass balance equation where 17 O(OH) prod = i P i × 17 O i , with P i the relative production rate of the ith reaction pathway with respect to the total production rate and 17 O i its www.atmos-chem-phys.net/16/2659/2016/ Atmos. Chem. Phys., 16, 2659-2673, 2016 24.9 * HONO production rate divided by a factor of 4 to balance the HO x radical budget (see Kukui et al., 2014, andLegrand et al., 2014, for justification (Morin et al., 2011). Observations at Dome C during the OPALE campaign indicate that the photolysis of HONO and the HO 2 + NO reaction may represent the most significant sources of OH at Dome C during the period of seasonal snowpack emissions (Kukui et al., 2014). However, the measurement of HONO (around few 10 pmol mol −1 ) during OPALE, probably biased by HO 2 NO 2 interference (Legrand et al., 2014), is incompatible with the HO x (= OH + HO 2 / RO 2 ) radical budget. Best agreement is achieved when HONO at Dome C is assumed to originate from snow emissions with the emission strength evaluated by Legrand et al. (2014). Using a 1-D model, Kukui et al. (2014) show that the concentrations of HONO corresponding to about 20-30 % of measured HONO are consistent with those calculated from the budget analysis of OH radicals with the concentrations of NO 2 either calculated from NO measurements assuming PSS or observed by Frey et al. (2015). Therefore, the production of OH by HONO photolysis is consequently adjusted and the 0-D box model (Kukui et al., 2014) is used to calculate all other production rates of OH. Note that, even when lowering HONO to 20-30 % of the measured values, this species remains the major primary source of radicals at Dome C. Applying the isotope 17 O-excess transfer (Morin et al., 2011) and the OH prod isotope mass balance, 17 O(OH) prod on average equals 5 ± 2 ‰ (1σ ). Because the major process leading to the emission of HONO from the snowpack is the photolysis of nitrate, which possesses a 17 O value of approximately 32 ‰, both in the snow "skin layer"  and in the top 10 cm of snow   Fig. 2 and Table 2). The source of discrepancy between observed and modeled 17 O(NO − 3 ) during OPALE is presently unknown, but we note that such underestimation of the modeled 17 O(NO − 3 ) versus the observed 17 O(NO − 3 ) was also pointed out in 3-D modeling of the nitrate 17 O excess (Alexander et al., 2009). A critical evaluation may nevertheless offer some clues.

Alternative sources of NO 2
A possible explanation for the underestimation of 17 O(NO − 3 ) involves halogen chemistry in the troposphere over the Antarctic Plateau (Bloss et al., 2010;Morin et al., 2008). Reactive halogen oxides (XO = BrO, ClO, IO, etc.) are produced through the reaction of halogen radicals (X) with ozone, a pathway that plays an important role in the catalytic process responsible for ozone depletion events (ODEs) observed in the Arctic boundary layer since the Atmos. Chem. Phys., 16, 2659-2673 1980s (Fan and Jacob, 1992;Simpson et al., 2007): In terms of the chemical budget of NO x , the impact of XO can occur via two chemical mechanisms (see Sect. 4.2 for the second mechanism involving the formation of halogen nitrate, XONO 2 ). First, XO can oxidize NO to NO 2 , a pathway that competes with the NO + O 3 and NO + HO 2 / RO 2 reactions in terms of NO oxidation: For conditions typical of the Antarctic boundary layer, 1 pmol mol −1 of XO has roughly the same chemical activity as 4 nmol mol −1 of ozone in terms of NO oxidation (Atkinson et al., 2007). Therefore, when halogen oxides are present at relevant levels, R7 can result in concentrations of NO 2 that are higher than that predicted from the balance between NO 2 destruction via photolysis and production through the reaction of NO with O 3 or HO 2 / RO 2 (i.e., the extended Leighton mechanism): The involvement of XO in the NO x cycle at Dome C would have important implications for the 17 O of atmospheric nitrate. The production of halogen oxide radicals proceeds through a direct transfer of a terminal oxygen atom from ozone to the XO product (Zhang et al., 1997). Therefore, it is expected that the 17 O of XO is equal to 17 O(O 3 ) term , which means that the reaction of NO with XO is roughly equivalent to the NO + O 3 reaction in terms of 17 O transfer to NO 2 (note that the NO + XO transfer is greater than NO + O 3 as, in the later case, part of the central O 3 atom participates in the reaction). The participation of XO species in the oxidation of NO thus leads to a greater 17 O transfer to NO 2 by effectively increasing the value of α. However, on the Antarctic Plateau, BrO did not exceed 2-3 pmol mol −1 during the OPALE campaign (Frey et al., 2015). Including BrO chemistry would only increase α by 2 % (due to the specific form of α ≡ 1/(1 + x)), which is too low to significantly influence 17 O(NO 2 ) and ultimately 17 O(NO − 3 ). In the absence of measurements of other halogens we cannot completely rule out a role of the halogen chemistry there. However, even with α = 1, its maximum but unrealistic value due to the high concentration of HO 2 , 17 O(NO − 3 ) would reach the range of 23-25 ‰, in better agreement with the observations but still significantly lower. Similarly, in the event of a non-isotopic steady state of NO 2 (Michalski et al., 2014), it is very unlikely that 17 O(NO 2 ) could reach values greater than its primary snow nitrate source (i.e., 17 O(NO 2 ) > 17 O(NO − 3 ) snow = 30-35 ‰ in summer at Dome C; Erbland et al., 2013;Frey et al., 2009), still leaving the predicted 17 O(NO − 3 ) underestimated with respect to atmospheric observations.

Alternative oxidation pathways of NO 2
Considering R5 as the main source of HNO 3 , an alternative approach is to consider that OH bears a higher 17 O than the estimate calculated previously. Assuming a β of 1, which seems again unrealistic, will increase 17 O(NO − 3 ) by 1 ‰ at most (Table 2), still insufficient to explained atmospheric observations as NO + HO 2 remain a major source of OH, independent of the assumed β.
Alternatively, if measured HONO concentrations are considered instead of those assumed to constrain by the HO x budget (i.e., 4 times lower than measured), average 17 O(NO − 3 ) values of 23-24 ‰ are calculated (Table 2), again systematically lower than the observed range of 27-30 ‰. However, given the significant uncertainty surrounding the isotopic composition of HONO and its relative contribution to total OH production at Dome C, it is not possible to make a firm conclusion in this regard. Therefore, neither the common sources of NO 2 nor the daytime formation of HNO 3 seems to be able to explain the high 17 O(NO − 3 ) values of atmospheric nitrate observed at Dome C in summer. When this observation is taken together with the high NO 2 / NO ratio observed by Frey et al. (2013Frey et al. ( , 2015 during two summer seasons at Dome C, clearly our current understanding of the NO x chemistry on the Antarctic Plateau seems to be incomplete. There are several other processes that possibly account for the disagreement between the measurements and mass balance calculations. Indeed, in addition to its impact on NO x cycling through the R7 pathway, an increasing body of evidence points towards reactive halogen chemistry as a major NO x sink and source of nitrate via the production and subsequent hydrolysis of XNO 3 species (Sander et al., 1999;Savarino et al., 2013;Vogt et al., 1996): A critical analysis of the CHABLIS data led Bauguitte et al. (2012) to conclude that R8-R9 pathway exerted predominant control over the chemical loss rate of NO x during the campaign, despite the significant uncertainties involved in the parameterization of the uptake processes (Finlayson-Pitts, 2009). This implies that XNO 3 uptake may also represent a significant source of nitrate at Dome C should halogen oxide radicals (XO) be present at the required concentration. Experimental (Gane et al., 2001) and theoretical (Mc-Namara and Hillier, 2001) studies suggest that the oxygen atom initially associated with XO combines with the N atom of NO 2 to form nitrate, thus transferring the isotopic signature of both XO and NO 2 . The specific 17 O value induced by XNO 3 hydrolysis can thus be expressed as follows : efficiently bypassing the OH 17 O-excess budget. Through consideration of the increased 17 O transfer associated with R8, the observations of 17 O(NO − 3 ) during December can be reconciled with the values calculated by mass balance if approximately 10-20 % of total nitrate production is assumed to occur via XNO 3 hydrolysis. However, no sufficient halogen concentration has been observed on the Antarctic Plateau to sustain such a chemical pathway, but we note that chlorine chemistry has never been probed on the Antarctic Plateau.
There is increasing body of evidence that heterogeneous hydrolysis of NO 2 can be a possible source of HONO and HNO 3 in acidic conditions (Finlayson-Pitts, 2009), with the potential to explain the difference between the calculated and measured atmospheric 17 O(NO − 3 ) values. This mechanism would represent a source of nitrate with a 17 O value roughly equivalent to the nitrate originally present in the surface snow (i.e., 30-35 ‰), a signature significantly higher than that induced by R5. If this production mechanism is active at the air-snow interface at Dome C and results in the slow emission of nitrate to the atmosphere via physical release after its formation, it would act to increase the 17 O value of nitrate in the boundary layer relative to the local 17 O(NO − 3 ) R5 oxidation signature. However, considering the propensity of nitric acid to stick on snow (Crowley et al., 2010), the snowpack to act as a sink rather than a source of nitric acid (Dibb, 2004;Erbland et al., 2013) and the fast NO x recycling that should take place within the snowpack, it is very unlikely that 17 O(NO − 3 ) could be explained by a direct nitric acid emissions from snow, which has been ultimately shown to be limited (Slusher et al., 2010;Berhanu et al., 2014).

A critical analysis of 17 O(NO −
3 ) shows in fact that such high values correspond mainly to the nighttime chemistry of NO x (Michalski et al., 2003;Morin et al., 2008). Nighttime chemistry involves species like N 2 O 5 and NO 3 in the process of forming HNO 3 and again efficiently bypass the OH pathway. It is conceivable that below the photic zone, within the snowpack, N 2 O 5 and NO 3 could be produced when O 3 and NO 2 are transported at depth, but there is no reason to think that such dark NO x chemistry could in one way or another survive the photic zone transition and thus influence the overlying atmosphere.
Stratospheric nitrate deposited to the surface snow during winter, which has been observed to possess 17 O(NO − 3 ) values in the range of 35-41 ‰  and, possibly more, may act to buffer the 17 O of the atmospheric nitrate reservoir via evaporation late into the spring and summer. However, this seems again unlikely given the rapidity of NO x cycling and oxidative loss at Dome C during this time Legrand et al., 2009).
Alternatively, the discrepancy may originate from an systematic error in the 17 O values assigned to tropospheric ozone, specifically at Dome C (Fig. 3). In a 3-D global modeling exercise of 17 O(NO − 3 ), Alexander et al. (2009) were able to reconcile modeling and observation only by assuming a bulk composition of ozone at 17 O = 35 ‰ instead of the 25 ‰ generally assumed for the tropospheric ozone. While such high values would also solve our discrepancy, all observations and measurements published thus far are consistently closer to 25 ‰ than 35 ‰ (Vicars and Savarino, 2014;Johnston and Thiemens, 1997;Krankowsky et al., 1995). Given the low variability in the measurements observed at Dome C (Fig. 3) and elsewhere (Vicars and Savarino, 2014), if atmospheric measurements are underestimated, it should be by a systematic error common to both the liquid helium condensation or coated filter techniques. In the absence of such demonstration, we believe that the model-measurement discrepancy is likely due to false assumptions regarding NO x chemistry or 17 O excess transfer mechanisms. Equally, it is also possible that non-zero 17 O could be generated from know chemical reactions (i.e., well established) such as the CO + OH reaction which produces a positive 17 O in the remaining CO (Röckmann et al., 1998;Feilberg et al., 2005), although no evidence exists for this to occur. On the other hand, Table 2 shows that the variability in 17 O (but not the absolute values) is correctly caught by the model when α is constrained by the observations and 17 O(OH) by the observed HONO concentrations. This observation would favor the view that the chemistry and associated 17 O transfer are well understood and that a systematic error is likely at the origin of the discrepancy of the absolute values. However, this interpretation would be in contradiction with NO x -HO x chemistry observations showing that in fact such chemistry is not very well understood above the Antarctic ice sheet (Frey et al., 2015;Legrand et al., 2014;Slusher et al., 2010 3 ) values observed during spring and early summer could be associated with snowpack emissions of NO x . Considering only samples collected at Dome C between October and December, both those reported here and those collected in 2009 and described by Erbland et al. (2013), a strong anticorrelation (r = −0.90) is observed between the δ 15 N and 17 O values of atmospheric nitrate (Fig. 8). In other words, the atmospheric nitrate sampled in early spring, which is heavily depleted in 15 N due its formation from NO x emissions arising from a winter-nitrified snowpack, possessed consistently higher 17 O values than the nitrate sampled directly after this period. Conversely, in summer, atmospheric nitrate possesses a lower 17 O and is 15 N enriched with respect to early spring values as it is formed from a snowpack that has suffered several stages of denitrification, leading to 15 N enrichment of the snowpack and the emitted NO x . This finding suggests that the mechanism producing enhanced 17 O(NO − 3 ) values observed during early spring is tightly coupled in time and space with the intensity of NO x emissions from the snowpack, an observation very similar to that of Morin et al. (2012), who detected a similar relationship between δ 15 N and 17 O for atmospheric nitrate in the springtime boundary layer over Barrow, Alaska (71 • N). The authors of that study attributed the observed correlation to the coupling of snowpack NO x emissions and reactive halogen chemistry, suggesting that these two processes were interrelated and mutually strengthening. In the case of the OPALE 2011-2012 data, the correlation between δ 15 N and 17 O could arise from any of the potential pathways previously discussed. For example, as proposed by Morin et al. (2012), the R8 and R9 pathways may be enhanced during the period of polar sunrise. Alternatively, a correlation could result from an increased contribution to total OH production from the photolysis of HONO, which is co-emitted with NO x via nitrate photochemistry (Grannas et al., 2007) and may induce a larger 17 O excess in OH as compared to the conventional O( 1 D) + H 2 O pathway. Furthermore, the hydrolysis of NO 2 in snow, should it contribute significantly to nitrate production at Dome C, is likely amplified during periods when concentrations NO 2 are high in the snowpack interstitial air due to nitrate/nitrite photochemistry. Therefore, while the processes responsible for driving the formation of atmospheric nitrate at Dome C during summer cannot be unambiguously identified, the isotopic results presented here clearly indicate that snowpack emissions result in enhanced 17 O transfer to nitrate. Our understanding of NO x chemistry above the snow surface at Dome C is therefore incomplete.

Conclusions
Constraining the propagation of ozone's 17 O-excess signature within the NO x cycle is critical in polar areas where the opportunity is offered to extend atmospheric investigations based on 17 O measurements to the glacial-interglacial timescale using deep ice core records of nitrate. However, the factors governing the present-day isotopic composition of atmospheric nitrate over the Antarctic Plateau remain poorly understood, primarily due to the complex nature of the boundary layer photochemistry initiated during spring by NO x emissions from the snowpack. An isotopic mass balance performed for atmospheric nitrate during December 2011, informed by in situ oxidant concentration measurements conducted within the framework of the OPALE field study, suggests the existence of an unexpected process bypassing the commonly accepted daytime chemistry of NO 2 (i.e., NO 2 + OH) that contributes significantly to the atmospheric nitrate budget over Dome C. The strong negative correlation observed between the δ 15 N and 17 O values of nitrate between October and December suggests that this unknown process scales with the intensity of the snowpack photochemistry. Potential explanations for this observation include (i) an increased 17 O transfer from OH due to its formation from the photolysis of HONO released from the snowpack, (ii) heterogeneous hydrolysis of NO 2 due to the high concentrations of NO 2 in the snowpack interstitial air, and (iii) the co-emission of reactive halogen species that act as an intermediate in the transfer of 17 O from ozone to nitrate. Systematic measurement error and/or false assumptions regarding 17 O isotopic transfer functions cannot be completely excluded as potential causes for the observed discrepancy between the observed and modeled data. We encourage additional laboratory experiments to further reduce these uncertainties. However, given the other lines of evidence presented here, we hypothesize that this discrepancy is due to an unknown or misunderstood component of the NO x photochemical cycle over the Antarctic Plateau. Further research is needed to solve the many inconsistencies (e.g., high NO 2 / NO ratio, high concentration of NO 2 , unresolved HONO atmospheric concentration, interference such as HO 2 NO 2 , isotope mass balance) observed during the OPALE experiments.
The Supplement related to this article is available online at doi:10.5194/acp-16-2659-2016-supplement.