Atmospheric nitrogen oxides ( NO and NO 2 ) at Dome C , East Antarctica , during the OPALE campaign

Mixing ratios of the atmospheric nitrogen oxides NO and NO2 were measured as part of the OPALE (Oxidant Production in Antarctic Lands & Export) campaign at Dome C, East Antarctica (75.1 S, 123.3 E, 3233 m), during December 2011 to January 2012. Profiles of NOx mixing ratios of the lower 100 m of the atmosphere confirm that, in contrast to the South Pole, air chemistry at Dome C is strongly influenced by large diurnal cycles in solar irradiance and a sudden collapse of the atmospheric boundary layer in the early evening. Depth profiles of mixing ratios in firn air suggest that the upper snowpack at Dome C holds a significant reservoir of photolytically produced NO2 and is a sink of gasphase ozone (O3). First-time observations of bromine oxide (BrO) at Dome C show that mixing ratios of BrO near the ground are low, certainly less than 5 pptv, with higher levels in the free troposphere. Assuming steady state, observed mixing ratios of BrO and RO2 radicals are too low to explain the large NO2 : NO ratios found in ambient air, possibly indicating the existence of an unknown process contributing to the atmospheric chemistry of reactive nitrogen above the Antarctic Plateau. During 2011–2012, NOx mixing ratios and flux were larger than in 2009–2010, consistent with also larger surface O3 mixing ratios resulting from increased net O3 production. Large NOx mixing ratios at Dome C arise from a combination of continuous sunlight, shallow mixing height and significant NOx emissions by surface snow (FNOx ). During 23 December 2011–12 January 2012, median FNOx was twice that during the same period in 2009– 2010 due to significantly larger atmospheric turbulence and a slightly stronger snowpack source. A tripling of FNOx in December 2011 was largely due to changes in snowpack source strength caused primarily by changes in NO−3 concentrations in the snow skin layer, and only to a secondary order by decrease of total column O3 and associated increase in NO−3 photolysis rates. A source of uncertainty in model estimates of FNOx is the quantum yield of NO − 3 photolysis in natural snow, which may change over time as the snow ages.

Depth profiles of mixing ratios in firn air suggest that the upper snowpack at Dome C holds a significant reservoir of photolytically produced NO 2 and is a sink of gas phase ozone (O 3 ).First-time observations of BrO at Dome C show that mixing ratios of BrO near the ground are low, certainly less than 5 pptv, with higher levels in the free troposphere.Assuming steady-state, observed mixing ratios of BrO and RO 2 radicals are too low to explain the large NO 2 : NO ratios found in ambient air, possibly indicating the existence of an unknown process contributing to the atmospheric chemistry of reactive nitrogen above the Antarctic Plateau.During 2011-2012 NO x mixing ratios and flux were larger than in 2009-2010 consistent with also larger surface O 3 mixing ratios resulting from increased net O 3 production.Large NO x mixing ratios at Dome C arise from a combination of continuous sun light, shallow mixing height and significant NO x emissions by surface snow (F NOx ).During 23 December 2011-12 January 2012 median F NOx was twice that during the same period in 2009-2010 30 due to significantly larger atmospheric turbulence and a slightly stronger snowpack source.A tripling of F NOx in December 2011 was largely due to changes in snow pack source strength caused primarily by changes in NO 3 concentrations in the snow skin layer, and only to 35 a secondary order by decrease of total column O 3 and associated increase in NO 3 photolysis rates.A source of uncertainty in model estimates of F NOx is the quantum yield of NO 3 photolysis in natural snow, which may change over time as the snow ages.The nitrogen oxides NO and NO 2 (NO x = NO + NO 2 ) play a key role in the polar troposphere in determining its oxidation capacity, defined here as the sum of O 3 , HO x radicals, and hydrogen peroxide (H 2 O 2 ).The influence is achieved via photolysis of NO 2 , the only source for in situ production of tropospheric O 3 , through shifting HO x radical partitioning towards the hydroxyl radical (OH) via the reaction NO + HO 2 !NO 2 + OH, and finally through reactions with peroxyradicals NO+HO 2 2 M .M .F r e y e t a l .: A t m o s p h e r i c n i t r o g e n o x i d e s d u r i n g O P A L E (or RO 2 ) which compete with the formation of peroxides (H 2 O 2 and ROOH).
Atmospheric mixing ratios of NO x in the atmospheric boundary layer of coastal Antarctica are small, with average NO x values in summer not exceeding 30 pptv (Bauguitte et al., 2012).The build up of large mixing ratios is prevented by gas-phase formation of halogen nitrates (e.g.BrNO 3 , INO 3 ) followed by their heterogeneous loss (Bauguitte et al., 2012).Conversely, mixing ratios of NO x on the East Antarctic Plateau are unusually large, similar to those from the mid-latitudes (Davis et al., 2008;Slusher et al., 2010;Frey et al., 2013).Such large mixing ratios of NO x were found to arise from a combination of several factors: continuous sunlight, location at the bottom of a large air drainage basin, low temperatures leading to low primary production rates of HO x radicals, significant emissions of NO x from surface snow, and a shallow boundary layer (Davis et al., 2008;Frey et al., 2013, and refs. therein).
Snow emissions of NO x , observed at several polar locations (e.g.Jones et al., 2001;Honrath et al., 2000b), are driven by UV-photolysis of nitrate (NO 3 ) in snow (Honrath et al., 2000b;Simpson et al., 2002) and are now considered to be an essential component of airsnow cycling of oxidised nitrogen species above the polar ice sheets (Davis et al., 2008;Frey et al., 2009b) and likely also above mid-latitude snow packs (Honrath et al., 2000a;Fisher et al., 2005).Atmospheric dynamics, i.e. vertical mixing strength and mixing height, can explain some of the observed temporal variability and sitespecific chemical composition of the lower troposphere at South Pole and Summit, Greenland (Neff et al., 2008;Van Dam et al., 2013).Recently, the very strong diurnal cycle of mixing ratios of NO x observed at Dome C, East Antarctic Plateau, during summer was shown to result from the interplay between boundary layer mixing and emissions from the photochemical snow source; during calm periods a minimum of NO x mixing ratios occurred around local noon and a maximum in the early evening coinciding with the development and collapse of a convective boundary layer (Frey et al., 2013).A key parameter of the physical atmospheric processes at play is the turbulent diffusivity of the atmosphere, which controls the mixing height, h z , of the atmospheric boundary layer and contributes to the magnitude of the flux of trace chemical species emitted by the snow (e.g.Frey et al., 2013).
The impact of NO x emissions from snow on the oxidation capacity of the lower troposphere in summer can be significant.For example, NO x snow emissions can result in net O 3 production as observed in the interior of Antarctica (Crawford et al., 2001;Legrand et al., 2009;Slusher et al., 2010) as well as unusually large mixing ratios of hydroxyl radicals as detected at South Pole (Davis et al., 2008, and refs. therein).Furthermore, in Antarctica the gas phase production of hydrogen peroxide (H 2 O 2 ), the only major atmospheric oxidant preserved in ice cores, is sensitive to NO released by the surface snowpack (e.g.Frey et al., 2005Frey et al., , 2009a)).A steadystate analysis of ratios of NO 2 : NO at Dome C sug-110 gested that mixing ratios of peroxy radicals (not measured at the time) are possibly larger at Dome C than any previous observations in air above polar snow (Frey et al., 2013).The quantitative understanding of emissions of NO x from snow remains incomplete, but it is 115 a research priority to be able to parameterise global models to assess for example global impacts of chemical air-snow exchange on tropospheric O 3 (e.g.Zatko et al., 2013).Emissions of NO x from snow at Dome C are among the largest observed above either polar ice 120 sheet, but are typically underestimated by models, especially at large solar zenith angles (Frey et al., 2013).
The study presented here was part of the comprehensive atmospheric chemistry campaign OPALE (Oxidant Production and its Export from Antarctic Lands) 125 in East Antarctica (Preunkert et al., 2012) and provided the opportunity to measure NO x mixing ratios and flux during a second summer season, after a previous campaign in 2009-2010 (Frey et al., 2013).The study objectives were firstly to extend the existing data set with 130 mixing ratio profiles of the lower atmosphere and the firn air (interstitial air) column of the upper snow pack.Secondly, to investigate if observed NO 2 : NO ratios are consistent with measurements of hydroxyl and halogen radicals.And thirdly, to analyse the main drivers of the 135 atmospheric NO x emission flux from snow.

Methods
The measurement campaign of 50 days took place at Dome C (75.1 S, 123.3 E, 3233 m) from 23 November 2011 to 12 January 2012.Similar to the 2009-2010 140 campaign atmospheric sampling was performed from an electrically heated lab shelter (Weatherhaven tent) located in the designated clean-air sector 0.7 km upwind (South) of Concordia station (Frey et al., 2013, Fig.1a).All times are given as local time (LT), equivalent to 145 UTC + 8 h, and during the study period the sun always remained above the horizon.

NO
x concentration measurements and uncertainties Three 20 m-long intake lines (Fluoroline 4200 high pu-150 rity PFA, I.D. 4.0 mm) were attached to a mast located at 15 m from the lab shelter into the prevailing wind to continuously sample air at 0.01, 1.00 and 4.00 m above the natural snow pack.The intake lines were away from the influence of the drifted snow around the lab shelter.
On 9 January 2012 vertical profiles of the lower atmosphere were sampled by attaching a 100 m-long intake line to a helium-filled weather balloon, which was then manually raised and lowered.During selected time periods firn air was sampled, to depths 5-100 cm, by means of a custom built probe.The probe consisted of a tube (10 cm diameter) which was lowered vertically into a precored hole to the chosen snow depth, passing through a disc (1 m diameter) resting on the snow surface.The disk had a lip of 10 cm protruding into the snow.The lip and disk minimised preferential pumping of ambient air along the tube walls.The air intake was mounted so that only air from the bottom and sides could enter, using small horizontal holes at 0-10 cm above the open bottom end of the vertical tube.All probe components were made from UV-transparent plastic (Plexiglas Sunactive GS 2458).Furthermore, 2⇥3 m sheets of UVopaque (Acrylite OP-3) and UV-transparent (Acrylite OP-4) plexiglass, mounted on aluminium frames at 1 m above the snow surface, were used to deduce the effect of UV radiation on the mixing ratio of NO x in the interstitial air and avoid at the same time any temperature effect altering the snow surface.
To measure NO x the same 2-channel chemiluminescence detector (CLD) and experimental set up as during the 2009-2010 campaign were used (Frey et al., 2013, Fig.1b).Channel one of the CLD measured atmospheric mixing ratios of NO whereas the other channel determined the sum of the mixing ratios of NO and NO originating from the quantitative photolytic conversion of NO 2 .The difference between the two channels was used to calculate atmospheric mixing ratios of NO 2 .The three sample inlets were connected inside the lab shelter to a valve box, which automatically switched the CLD between sampling heights on a 90 s duty cycle.As described below, the 10-minute average concentration difference NO x between the 0.01 and 1.0 m inlets is used to estimate flux.Therefore, 10-minute mean NO x values are calculated on average from two sets of two subsequent 90 s intervals, separated by a 90 s interval during which the 4.0 m inlet was measured.Baseline count rates were determined by adding excess O 3 to sample air in a pre-chamber so that all electronically excited NO 2 has returned to ground state when reaching the reaction chamber.The baseline was measured for 60 s every 13.5 min alternating between all three inlets.The NO sensitivity of the CLDs was determined every 14 h by standard addition to the sample air matrix of a 1 ppm NO/NO 2 mixture (UK National Physics Laboratory traceable BOC certified), which is further diluted to 4 ppbv of NO.During standard runs also the conversion efficiency (CE) of the photolytic converter was determined by addition of a known mole fraction of NO 2 .This was achieved by gas phase titration of the NO/NO 2 mixture to NO 2 by O 3 generated from a pen-ray lamp, and monitoring the un-titrated NO mole fraction.The instrument artefact originating from NO x producing surface reactions in inlets and reaction cells was determined by overflowing the instrument inlet with scrubbed ambient air supplied by a pure air generator (Eco-Physics 215 PAG003).The artefact was measured every 14 h, offset by 7 h to the calibration runs.The CLD performance, e.g.sensitivity, random error and precision, was similar to that during 2009-10 (Frey et al., 2013, Table 1).
The mean wind direction during the measurement 220 period was from S (176 ) with an average speed of 4.0 m s 1 (Fig. 1b).During 2.5% of the time winds came from the direction of Concordia station, i.e. the 355-15 sector (Frey et al., 2013, Fig.1a), potentially carrying polluted air from the station power generator to the 225 measurement site.For example, during Period III.winds rotated 4 times through northerly directions (Fig. 1b).
Pollution spikes in the raw 1-s data typically exceeded 10 ppbv of NO x and were effectively removed before computing the 1-min averages by applying a moving 1-230 min standard deviation ( ) filter.Observations were rejected when 1-of NO and NO 2 mixing ratios within a 1-min window exceeded 24 and 90 pptv, respectively.The CLD employed also converts nitrous acid (HONO) to NO in the photolytic converter and thus 235 HONO sampled by the CLD is an interferent, as discussed previously (Frey et al., 2013).Average mixing ratios of HONO at 1 m above the snowpack measured with the LOPAP (Long Path Absorption Photometer) technique were ⇠ 35 pptv (Legrand et al., 2014).The corre-240 sponding downward correction for NO 2 at 1 m above the snowpack is ⇠ 5%.However the LOPAP technique may overestimate the mixing ratio of HONO owing to an interference by pernitric acid (HO 2 NO 2 ) (Legrand et al., 2014).True corrections of NO 2 inferred from modelled 245 HONO mixing ratios (Legrand et al., 2014) are more likely to be on the order of < 1.5%.Due to the uncertainty in absolute mixing ratios of HONO, no correction of NO x values for the HONO interference was applied.
The thermal decomposition of HO 2 NO 2 in the sam- tween 3 h (12:00 LT) and 7 h (00:00 LT), whereas that of HONO, ⌧ HONO , ranged between 4.5 min (12:00 LT) and 24 min (00:00 LT) (Legrand et al., 2014).The life time of HONO is comparable to the typical transport times of ⇠ 10 min between the surface and 1 m at Dome C in 290 summer (Frey et al., 2013).Hence, HONO : NO x ratios as well as corresponding corrections required for NO 2 are not constant with height above the snow surface.
No gradients of HONO mixing ratios were measured but modelled values were 18.8 and 10.2 pptv at noon, and 295 15.3 and 12 pptv at midnight, at 0.1 and 1.0 m, respectively (Legrand et al., 2014).Corresponding corrections of mean NO 2 mixing ratios for HONO are 1.3-1.5 % with a maximum difference of 0.2% between 0.1 and 1.0 m.Thus, at Dome C a strong gradient in the mixing ratios 300 of HONO was a negligible effect on the mixing ratios of NO x measured at 0.1 and 1.0 m and thus a negligible effect on the estimated NO x flux.

NO
x flux estimates The turbulent flux of NO x , F NOx , was estimated us-305 ing the integrated flux gradient method (e.g.Lenschow, 1995) and mixing ratios of NO x measured at 0.01 and 1.0 m.F NOx in the surface layer is parameterised according to the Monin-Obukhov similarity theory (MOST) whose predictions of flux-profile relationships at Halley, 310 an Antarctic coastal site of the same latitude as Dome C, agree well with observations (Anderson and Neff, 2008, and references therein): with the von Karman constant  (set to 0.40), friction 315 velocity u ⇤ , measurement height z, concentration gra-dient @c/@z, and h ( z L ) an empirically determined stability function for heat with L as the Monin-Obukhov length.Assuming constant flux across the layer between the two measurement heights z 1 and z 2 allows the inte-320 gration to be solved and yields: Stability functions h used are given in Frey et al. (2013), while their integrated forms can be found in Jacobson (1999).Friction velocity u ⇤ and L were com-325 puted from the three-dimensional wind components (u, v, w) and temperature measured at 25 Hz by a sonic anemometer (Metek USA-1) mounted next to the uppermost NO x intake line, at 4 m above the snow surface.Processing of raw sonic data in 10 min blocks included 330 temperature cross-wind correction and a double coordinate rotation to force mean w to zero (Kaimal and Finnigan, 1994;Van Dijk et al., 2006).Equation ( 2) implies that a positive flux is in upward direction, equivalent to snow pack emissions and a negative flux is in 335 downward direction, equivalent to deposition.
The application of MOST requires the following conditions to be met: (a) flux is constant between measurement heights z 1 and z 2 , (b) the lower inlet height z 1 is well above the aerodynamic roughness length of the sur-340 face, (c) the upper inlet height z 2 is within the surface layer, i.e. below 10 % of the boundary layer height h z (Stull, 1988), and (d) z 1 and z 2 are far enough apart to allow for detection of a significant concentration difference [c(z 2 ) c(z 1 )].

345
Condition (a) is met in the surface layer if the chemical lifetime ⌧ chem of NO x is much longer than the turbulent transport time scale ⌧ trans .Based on observed OH and HO 2 the ⌧ chem for NO x is estimated to be 3 h at 1200 LT and 7 h at 0000 LT during OPALE (Legrand 350 et al., 2014).Estimating ⌧ trans following the approach described previously (Frey et al., 2013, Eq. 6 and 7) yields 0.6, 1.7 and 2.5 min during the day (0900-1700 LT), the typical time of BL collapse (1700-1900 LT) and during the night (1900-0900 LT), respectively.Thus, 355 ⌧ chem exceeds ⌧ trans by at least a factor 100, confirming that vertical mixing always dominates over the gas phase photochemical sink and flux can be assumed constant between the two inlets.Condition (b) is met as discussed in Frey et al. (2013).For (c) the upper inlet height of 1 m is compared to estimates of mixing height h z from the MAR model (Gallée et al., 2015).The MAR model has been validated previously over the Antarctic Plateau, focusing on Dome C, during winter (Gallée and Gorodetskaya, 2010) and now also during summer (Gal-365 lée et al., 2015).Calculated flux values of NO x were removed when h z < 10 m resulting in the removal of 22 % (773 values) of all available 10 min flux averages.Flux es-timates are removed specifically during the evening and night, when the BL is shallow.Hence, fluxes during night time are less well constrained, but nevertheless support a significant diurnal cycle (Frey et al., 2013, Fig. 6b,g and Fig. 9).For (d) 10 min averages of [c(z 2 ) c(z 1 )] not significantly different from zero, i.e. smaller than their respective 1-standard error, were not included in the calculation of the flux of NO x .The 1-standard error in [c(z 2 ) c(z 1 )] was determined by error propagation of the 1-standard error of NO x mixing ratios.A total of 8 % (303 values) of all available 10 min flux averages were not significantly different from zero and thus removed.
In summary, the restrictions imposed by MOST and NO x measurement uncertainty justify placing inlets at 0.01 and 1.0 m and lead to the removal of 30 % (1076 values) of all available flux estimates.The total uncertainty of the 10 min NO x flux values due to random error in [c(z 2 ) c(z 1 )] (31 %), u ⇤ (3 % after Bauguitte et al., 2012) and measurement height (error in ln(z 2 /z 1 ) of ⇠ 7 %) amounts to 32 %.

Analysis of NO 3 concentrations in snow
During this study NO 3 concentrations in snow were measured every 2-3 days in the surface skin layer, i.e. in the top 0.5 cm of the snowpack, as well as in shallow snow pits within the clean-air sector.Snow NO 3 concentrations were determined using clean sampling procedures and a continuous flow analysis technique (e.g.Frey et al., 2009b).Samples were stored together with the additional snow samples discussed in Berhanu et al. (2014) and then analysed for NO 3 in batches by the same operator.The precision is 5% based on replicate standard measurements.Due to a systematic shift in the NO 3 standard response in between individual batch runs due to a calibration issue (Berhanu et al., 2014) results are less accurate than before.The overall accuracy including systematic errors in calibration and collection of just the top few mm of snow is of the order of 20%, and is therefore comparable to the spatial variability of NO 3 in surface snow at Dome C (France et al., 2011).
For the discussion below it should be borne in mind that temporal changes of NO 3 concentrations observed in surface snow are >50% (Fig. 7b) and therefore significantly larger than the measurement accuracy.

MAX-DOAS observations
Scattered sunlight was observed by a ground-based UVvisible spectrometer, in order to retrieve bromine oxide (BrO) column amounts.The instrument was contained in a small temperature-controlled box, which was mounted onto a tripod at 1 m above the snow surface.
An external gearbox and motor scanned the box in el-evation (so-called Multiple Axis).Spectra were anal-420 ysed by Differential Optical Absorption Spectroscopy (DOAS), the combination being known as the MAX-DOAS technique.See Roscoe et al. (2014) for more details of apparatus and analysis.Briefly, the observed spectrum contains Fraunhofer lines from the Sun's at-425 mosphere, which interfere with absorption lines in the Earth's atmosphere and are removed by dividing by a reference spectrum.The amounts of absorbers in the Earth's atmosphere are found by fitting laboratory cross-sections to the ratio of observed to reference spec-430 tra, after applying a high-pass filter in wavelength (the DOAS technique).
In our case the spectral fit was from 341 to 356 nm, and the interfering gases O 3 , O 4 (oxygen dimer) and NO 2 were included with BrO.The analysis was done 435 with two reference spectra, one from near the start of the campaign in December, the other following the addition of a snow excluder in January, necessary because it also contained a blue glass filter with very different spectral shape.The analysis was restricted to cloud-free 440 days or part-days.In MAX-DOAS geometry, the stratospheric light path is almost identical in low-elevation and zenith views, so stratospheric absorption is removed by subtracting simultaneous zenith amounts from lowelevation slant amounts, important for BrO as there is 445 much in the stratosphere.
To find the vertical amounts of BrO radicals the MAX-DOAS measurements were evaluated as follows: we divided by the ratio of the slant path length to the vertical (the Air Mass Factor, AMF), calculated by ra-450 diative transfer code (Mayer and Kylling, 2005), assuming all the BrO was in the lowest 200 m.

Ancillary measurements and data
Other co-located atmospheric measurements included mixing ratios of OH radicals and the sum of peroxy 455 radicals (RO 2 ) at 3 m using chemical ionisation mass spectrometry (Kukui et al., 2014) and mixing ratios of O 3 at 1 m with a UV absorption monitor (Thermo Electron Corporation model 49I, Franklin, Massachusetts).Photolysis rate coefficients, J, were determined based 460 on actinic flux, I, measured at ⇠ 3.50 m above the snow surface using a Met-Con 2⇡ spectral radiometer equipped with a CCD detector and a spectral range from 285 to 700 nm (further details in Kukui et al., 2014).Total column O 3 above Dome C was taken from ground based SAOZ (Système d'Analyse par Observation Zenitale) observations (http://saoz.obs.uvsq.fr/SAOZ_consol_v2.html).Standard meteorology was available from an automatic weather station (AWS) at 0.5 km distance and included air temperature (Vaisala 470 PT100 DTS12 at 1.6 m), relative humidity (at 1.6 m), wind speed and direction (Vaisala WAA 15A at 3.3 m).

The mixing height h
z of the atmospheric boundary layer was calculated from simulations with the MAR model as the height where the turbulent kinetic energy decreases below 5 % of the value of the lowest layer of the model (Gallée et al., 2015).

Modelling NO
3 photolysis The flux of NO 2 , F NO2 , from the snowpack owing to photolysis of the NO 3 anion in the snowpack can be estimated as the depth-integrated photolysis rate of NO 3 where J z (NO 3 ) is the photolysis rate coefficient of reaction NO 3 +h⌫ !NO 2 +O at depth, z, in the snowpack.
[NO 3 ] z is the amount of NO 3 per unit volume of snow at depth, z, in the snowpack.J z (NO 3 ) is calculated as described in France et al. (2010) using a radiative transfer model, TUV-snow (Lee-Taylor and Madronich, 2002), to calculate irradiances within the snowpack as a function of depth.The optical properties and detailed description of the Dome C snowpack are reported in France et al. (2011).Values of depth-integrated flux were calculated as a function of solar zenith angle and scaled by values of J(NO 3 ) measured by the Met-Con 2⇡ spectral radiometer described above to account for changing sky conditions.Scaling by a measured value of J(NO 3 ) is more accurate than previous efforts of scaling with a broad band UV instrument (e.g.France et al., 2011).The quantum yield and the absorption spectrum for NO 3 photolysis in snow were taken from Chu and Anastasio (2003).For the discussion below it should be borne in mind that the calculated F NO2 is a potential emission flux assuming that NO 2 is vented immediately after release from the snow grain to the air above the snow pack without undergoing any secondary reactions.

NO
x observations in ambient and firn air In summer 2011-2012 atmospheric mixing ratios of NO x with strong diurnal variability were observed (Fig. 1c), similar to the 2009-2010 season, and showed maximum median levels in firn air of ⇠ 3837 pptv, which rapidly decreased to 319 pptv at 0.01 m and 213 pptv at 1.0 m (Table 1).In the following we focus on measurements at 0.01 and 1.0 m, but statistics from all three measurement heights are reported in Table 1 and 4 m measurements were discussed for summer 2009-10 in Frey et al. (2013).
As seen previously at Dome C and other locations, NO x mixing ratios were weakly but significantly anti-correlated with wind speed (at 1.0 m R = 0.37, p < 520 0.001), especially when only the time period of the daily collapse of the convective boundary layer, i.e. 1700-1900 LT, was considered (R = 0.45, p < 0.001), and their diurnal cycle was dampened during storms (Fig. 1b-c).
The two main differences between summer 2011-2012 followed by a strong increase of daily averages from 300 to 1200 pptv at the beginning of Period III.(9-22 December 2011) (Fig. 1c).After that NO x mixing ratios gradually dropped over 10 days (Period III.-IV.) to median concentrations of ⇠120 pptv, slightly lower than 535 observed in late November (Fig. 1c, Table 2).During Period III. the median concentration of NO x at 1.0 m was 451 pptv, about 2.5 times that during the same time period in 2009, but similar thereafter (Fig. 1c, Table 2).
The NO x fluxes, F NOx , between 0.01 and 1.0 m were 540 mostly emissions from the snow surface, with a median of 1.6 ⇥10 13 molecule m 2 s 1 .Median values of F NOx at midnight and at noon were 0.4 and 2.9 ⇥10 13 molecule m 2 s 1 , respectively (Table 1).During Period III.F NOx showed an increase by a factor 3, ap-545 proximately around the same time when atmospheric mixing ratios of NO x increased (Fig. 1d, elled mixing heights h z of 200-550 m and observed turbulent diffusion coefficients of heat K h of ⇠ 0.1 m 2 s 1 (Fig. 2).However, in the late afternoon K h values decreased gradually over a few hours to reach in the evening levels half those during the day thereby giv-565 ing evidence of strongly reduced vertical mixing.Furthermore, around 18:30 LT modelled h z values decreased within minutes from 550 to < 15 m height (Fig. 2a) illustrating the collapse of the convective boundary layer typically observed at Dome C in the early evening dur-570 ing summer (King et al., 2006).At Dome C rapid cooling of the surface in the evening results in a strong shallow surface inversion (e.g.Frey et al., 2013), and is illustrated by a decrease in downward long-wave radiation and a negative heat flux, as observed in the evening of 9 January 2012 (Argentini et al., 2014, Fig.4).It follows that NO x snow emissions are trapped near the surface, which then leads to a significant increase in NO x mixing ratios below 15 m height measured almost immediately after collapse of the boundary layer (Fig. 2).During 22:20-22:40 LT a small increase in K h , due to the nightly increase in wind shear (see Frey et al., 2013), was sufficient to cause upward mixing of NO x accumulated near the surface to ⇠ 35 m height (Fig. 2).The vertical balloon soundings further underline the unique geographical setting of Dome C or other sites of similar latitude on the East Antarctic Plateau where air chemistry is dominated by strong diurnal cycles, both in down-welling solar radiation and atmospheric stability, contrasting South Pole where diurnal changes are absent and changes are more due to synoptic variability (Neff et al., 2008).
A vertical profile of mixing ratios of NO x and O 3 in firn air was measured on 12 January 2012 between 10:00 and 18:00 LT, for which depths were sampled in random order for 30-60 min each.Mixing ratio maxima of NO and NO 2 were ⇠ 1 and 4 ppbv, respectively, about one order of magnitude above ambient air levels (Table 1), and occurred at 10-15 cm depth, slightly below the typical e-folding depth of 10 cm of wind pack snow at Dome C (France et al., 2011) (Fig. 3a).NO dropped off quickly with depth, reaching 55 pptv at 85 cm, whereas NO 2 decreased asymptotically approaching ⇠ 2 ppbv (Fig. 3a).NO 3 concentrations in snow under the firn air probe did not follow the exponential decrease with depth typically observed at Dome C (e.g.Erbland et al., 2013).The firn air probe was installed onto untouched snow, and only removed after the end of the atmospheric sampling period.Thus contamination due to local activity appears unlikely, but a local anomaly remains a possibility as snow pits 5 m next to the lab shelter showed a similar increase of concentration with depth (data not shown).But NO 3 values within one e-folding depth were still in the range measured further away (Profiles P1-P3 in Fig. 3a), justifying a discussion of vertical profiles of mixing ratios.
O 3 mixing ratios in firn air were always below ambient air levels, suggesting the snow pack to be an O 3 sink as observed previously for the snowpack on the Greenland ice sheet (Peterson and Honrath, 2001), and showed a significant anti-correlation with NO 2 (R = 0.84, p < 0.001).This is further evidence for significant release of NO x by the snow matrix into the interstitial air, which then titrates O 3 through the reaction NO+O 3 !NO 2 +O 2 (Fig. 3).In particular, the drop of O 3 mixing ratios by >10 ppbv at 45 cm depth was not an outlier since collocated NO 2 mixing ratios were also significantly elevated compared to adjacent snow layers (Fig. 3a).However, no snow NO 3 measurements were available to further investigate the origin of the NO 2 630 peak.The observed vertical trends in NO x suggest that below a few e-folding depths the open pore space of the upper snowpack holds a significant reservoir of NO 2 produced photolytically above, as hypothesized previously (Frey et al., 2013).In contrast, NO disappears at depths 635 devoid of UV irradiance as it reacts with O 3 .

Response to UV irradiance
Changes in surface downwelling UV irradiance lead to a quick response of mixing ratios and speciation of NO x in ambient and firn air as observed during a partial solar 640 eclipse and during a shading experiment (Fig. 4).The solar eclipse occurred early in the season, on 25 November 2011, and caused a decrease in ambient NO mixing ratios at 1.0 m by about 10 pptv or 10 %, whereas NO 2 mixing ratios did not change significantly (Fig. 4a and 645 b).The NO gas phase source, UV photolysis of NO 2 , is reduced during the solar eclipse.But the sink of NO, the fast titration with O 3 , is unaffected by the reduction in UV irradiance.During the shading experiment on 11 January 2012 plastic sheets were placed at 1 m 650 above the snow surface, alternating in 30 min intervals between UV-opaque and UV-transparent materials.The impact of blocking incident UV irradiance (wavelengths < 380 nm) on firn air mixing ratios at 10 cm snow depth was up to 300 pptv or 30 % decrease in mixing ratios 655 of NO, whereas mixing ratios of NO 2 increased at the same time by ⇠ 150 pptv or 5 %, although often not statistically significant (Fig. 4c and d).Similar to the solar eclipse, the behavior of NO x mixing ratios in firn air is in accordance with a disruption of the fast gas phase inter-660 conversion of NO x species.Decrease of NO and increase of NO 2 mixing ratios are consistent with the suppression of NO 2 photolysis, which is both a NO source and a NO 2 sink.
Most importantly varying incident UV irradiance in 665 the wavelength region of NO 3 absorption (action spectrum maximum at 320 nm) over half-hourly time scales does not cause a depletion of NO 2 in firn air even though NO 2 is the main product of NO 3 photolysis in the snowpack.A dampened UV response of NO 2 mixing ratios 670 suggests that the NO x reservoir present in the open pore space of the upper snow pack discussed above must be large as it is not depleted during 30 min filter changes at the sample pump rates used.One implication is that the impact of changes in incident UV irradiance on the  2).A previous steady-state analysis indicated that high peroxy and possibly halogen radical levels must be present to explain deviations from the simple Leighton steady-state (Frey et al., 2013).The OPALE campaign provided observations needed to further investigate the NO 2 : NO ratios at Dome C.During summer 2011-2012 median concentrations of RO 2 radicals at 3 m, thought to consist mainly of HO 2 and CH 3 O 2 , were 9.9⇥10 7 molecule cm 3 (Kukui et al., 2014).
Figure 5 shows the BrO results, where the apparent vertical amounts at 15 are much larger than those at lower elevations -this shows that the vertical profile of BrO used to calculate AMFs, whereby all the BrO is in the boundary layer, must be incorrect.And interestingly, as at Halley in 2007 (Roscoe et al., 2014), much of the BrO must be in the free troposphere.The average of BrO at the three elevations is about 0.8 ⇥10 13 molecule cm 2 , with a slight decrease during the campaign.The average at Halley in 2007 was about 2.5 ⇥10 13 molecule cm 2 , so mixing ratios of BrO at Dome C are about a third those at Halley.The Dome C data were not inverted to determine the mixing ratio near the surface, but the changes in slant column with elevation angle are similar to those at Halley in 2007 (Roscoe et al., 2014).Based on the similarity of relative changes of slant BrO with elevation angles to those of Halley in 2007, and the approximate ratio of the slant columns at Halley in 2007 to those at Dome C of 3, we decided to divide the Halley inversion results by a factor 3 to arrive at a first estimate for Dome C of 2-3 pptv of BrO near the surface.Higher levels prevailing in the free troposphere possibly originate from a sea ice source in coastal Antarctica (Theys et al., 2011) or from stratospheric descent (Salawitch et al., 2010).
Assuming steady-state the total radical concentration with XO=BrO, can be calculated based on observed NO 2 : NO ratios and J(NO 2 ) (Ridley et al., 2000).Repeating the calculation as described in Frey et al. (2013) for 19 December 2011 to 9 January 2012 yields a median [OX] of 2.2⇥10 9 molecule cm 3 or 116 pptv.However, during the same period observations showed a median concentration of 9.9⇥10 7 molecule cm 3 or 5 pptv of [RO 2 ]+[HO 2 ] ( Kukui et al., 2014) and approximately 3 pptv of BrO, yielding a total radical concentration [OX] of 11 pptv.Hence, [OX] deduced from measured NO 2 : NO ratios exceeds available observations by a factor 10.3.NO 2 mixing ratios were then corrected for a potential interference with HO 2 NO 2 , assuming ambient levels of 130 pptv.It is found that the median steady-state esti-mate of total oxidant concentrations is still a factor 9.6 larger than the sum of observed radical mixing ratios.Hence, the large NO 2 : NO ratios observed at Dome C are either the result of an unknown measurement bias 735 or of an unidentified mechanism in boundary layer oxidation chemistry.A similar conclusion was reached in companion papers on the OPALE project (e.g.Legrand et al., 2014;Kukui et al., 2014;Savarino et al., 2015).

Drivers of seasonal NO
x variability 740 On diurnal time scales NO x mixing ratios at Dome C are controlled by the interplay between snowpack source strength and atmospheric physical properties, i.e. turbulent diffusion of heat K h and mixing height h z of the boundary layer.The median diurnal cycles of NO x mix-745 ing ratios in 2011-12 show with the exception of Period II.previously described behaviour (Frey et al., 2013), that is a strong increase around 1800 LT to maximum values, which last into the night time hours (Fig. 6a).
Night-time peaks of NO x are plausible if the weakening 750 of snow emissions is offset by a corresponding decrease of the chemical sink of NO x , i.e. the NO 2 +OH reaction, assuming no significant change in h z .This is consistent to a first order taking into account that observed OH concentrations (Kukui et al., 2014) and F NOx vary in a 755 similar way, by up to a factor 5 between local noon and midnight.During Period III.noon time values are similar to Period II.but the increase in the evening has a larger amplitude and generally larger mixing ratios prevail during 760 night time (Fig. 6a).Increased NO x mixing ratios during Period III. are consistent with the observed NO x emission flux F NOx , which always peaked at local noon, but also showed during Period III. a strong increase at all times of the day with a near doubling of the noon 765 time median (Fig. 6b).During Period IV. the diurnal cycles of both NO x mixing ratios and F NOx returned to low values and small diurnal amplitudes (Fig. 6a-b).
Below we evaluate potential causes of the unusual variability in NO x mixing ratios and flux observed on 770 seasonal time scales.

Atmospheric mixing vs. snow source strength
Similar to explaining diurnal NO x cycles at Dome C the seasonal variability of daily mean NO x mixing ratios during the first half of December 2011 can be attributed 775 to a combination of changes in F NOx and h z (Fig. 1).The strong increase of NO x around 11 December 2011 falls into a Period when F NOx almost tripled, while wind speeds slightly decreased and shallow boundary layer heights prevailed (Fig. 1, Table 2).For example, on 12 tively, while sodar observations yielded 10-150 m and 5-125 m, respectively (Gallée et al., 2015).After 13 December 2011 F NOx remained at high values, thus, the decrease of NO x mixing ratios appears to be primarily caused by stronger upward mixing into a larger volume, i.e. wind speeds increased and daily h z maxima grew, exceeding 600 m on 18 December 2011 (Fig. 1).After 23 December 2011 NO x mixing ratios drop to low levels, due to smaller F NOx and a deep boundary layer (Fig. 1).
F NOx depends on atmospheric turbulence (K h ) and concentration difference ( NO x ), which in turn is determined by the strength of the photolytic snow pack source at a given K h (Eq.1-2).However, the relative importance of K h and snow pack source strength can vary.For example, during Period IV. the median F NOx was 1.3⇥10 13 molecule m 2 s 1 , about twice that observed during the same period in 2009-2010 (Fig. 6g; Table 2).The inter-seasonal difference can be explained by both, significantly larger atmospheric turbulence and more negative NO x during all times of the day in 2011-2012 (Fig. 6h and i).Median K h was 0.08 m 2 s 1 , double that in 2009-2010, and median NO x was 51 pptv compared to 32 pptv in 2009-2010 (Table 2).
In contrast, during 2011-2012 the observed intraseasonal variability of F NOx is dominated by changes in the snow pack source strength.During Period III.median K h values (⇠ 0.05 m 2 s 1 ) and diurnal cycles were smaller than thereafter (Fig. 6c; Table 2), while NO x values were among the largest observed so far at Dome C, about three times those during the rest of the season, and therefore primarily caused the tripling of F NOx (Fig. 6d and i).In section 3.5.2we'll discuss underlying causes of changes in the strength of the snow source.
Previously, non-linear HO x -NO x chemistry and the associated increase in NO x lifetime were suggested to be an additional factor needed to explain large increases in NO x mixing ratios observed at South Pole (Davis et al., 2008, and references therein).In order to assess the relevance of this factor at Dome C we apply a simple box model to estimate net NO x production rates as done previously (Frey et al., 2013).It is assumed that mixing is uniform and instantaneous, that the snow emission flux F NOx is the main NO x source and the reaction with the OH radical is the dominant NO x sink and where k is the respective reaction rate coefficient.In 2009-10 no OH observations were available at Dome C and average values from South Pole were used instead.In 2009-10 estimated net production rates of NO x at night were on the order of 100 pptv h 1 and therefore explained the average increase in NO x from 110 to 300 pptv observed from 1700 to 1900 LT (Frey et al., 2013).In 2011-12 the same analysis is repeated using OH measurements available for most of Period IV. (Kukui et al., 2014) as well as h z calculated with the MAR model (Gallée et al., 2015).Resulting night time values of net NO x production rates are with about 40 pptv h 1 smaller than in 2009-10 but again to a first 840 order consistent with a smaller observed increase in NO x mixing ratios in the evening hours; i.e. during Period IV. median NO x increased between 1630 and 1930 LT from 114 to 242 pptv ((Fig.6a,f).The above model is oversimplified as the likely presence of HO 2 NO 2 will modulate 845 the diurnal variability of NO x sinks and sources with an impact on NO x lifetime as suggested by Davis et al. (2008).However without any information on the diurnal cycle of HO 2 NO 2 at Dome C further modelling is not warranted.850

Snow source strength
The NO x flux observed above polar snow is on the order of 10 12 to 10 13 molecule m 2 s 1 and contributes significantly to the NO x budget in the polar boundary layer.At the lower end of the range are F NOx observa-855 tions at Summit, Greenland (Honrath et al., 2002) and at Neumayer in coastal Antarctica (Jones et al., 2001) with 2.5⇥10 12 molecule m 2 s 1 , whereas on the Antarctic Plateau F NOx values are up to ten times larger.For example, the average F NOx at South Pole during 860 26-30 November 2000 was 3.9⇥10 12 molecule m 2 s 1 (Oncley et al., 2004), whereas at Dome C observed fluxes are 2-6 times larger, with seasonal averages of 8-25⇥10 12 molecule m 2 s 1 (Frey et al., 2013, this work).Due to the uncertainties in the processes leading to NO x 865 production it had been difficult to explain inter-site differences, e.g. by simply scaling F NOx with UV irradiance and NO 3 in the surface snow pack (Davis et al., 2004).Some of the variability in flux values may be due to differences in experimental set up or in the employed 870 flux estimation method (e.g.Davis et al., 2004;Frey et al., 2013).For example, the F NOx estimates for South Pole are based on measured NO gradients only, inferring NO x from photochemical equilibrium and using the Bowen ratio method (Oncley et al., 2004), whereas the 875 F NOx estimates for Dome C are based on observations of both atmospheric nitrogen oxides (NO and NO 2 ) and the flux-gradient method (Frey et al., 2013).
Model predictions of F NOx show in general a low bias on the Antarctic Plateau when compared to ob-880 servations.A first 3-D model study for Antarctica included NO x snow emissions parameterised as a function of temperature and wind speed to match the observed F NOx at South Pole (Wang et al., 2007).However, the model under-predicts NO mixing ratios ob-885 served above the wider Antarctic Plateau highlighting that the model lacks detail regarding the processes driving the emission flux (Wang et al., 2007).The first model study to calculate F NOx based on NO 3 photolysis in snow, as described in this work, reports 1-890 1.5⇥10 12 molecule m 2 s 1 for South Pole in summer (Wolff et al., 2002), about a factor 4 smaller than the observations by Oncley et al. (2004) and up to 16 times smaller than what is needed to explain rapid increases in NO x mixing ratios over a few hours (Davis et al.,895 2008, and references therein).Recent model improvements reduced the mismatch with the South Pole flux observations and included the use of updated absorption cross sections and quantum yield of the NO 3 anion, as well as e-folding depths measured in surface snow on the 900 Antarctic Plateau, and resulted in a factor 3 increase of flux calculated for South Pole (France et al., 2011).In light of major remaining uncertainties, which include the spatial variability of NO 3 in snow and the quantum yield of NO 3 photolysis (Frey et al., 2013), we discuss 905 below the variability of F NOx observed at Dome C.
A number of factors may contribute to changes in snow source strength of NO x .One possibility to explain increases in F NOx is that the NO 2 reservoir in the open pore space of the upper snowpack discussed above 910 may undergo venting upon changes in atmospheric pressure.However, no statistically significant relationship between F NOx and atmospheric pressure is found (data not shown).The main cause of large F NOx values appears rather to be related to changes in snow produc-915 tion rates of NO x from NO 3 photolysis, which depend on the NO 3 photolysis rate coefficient J NO 3 and the NO 3 concentration in the photic zone of the snow pack (Eq.3).
Trends in down-welling UV irradiance due to strato-920 spheric O 3 depletion were suggested previously to drive J NO 3 and therefore F NOx and the associated increase in net production of surface O 3 observed at South Pole in summer since the 1990's (Jones and Wolff, 2003).At Dome C the observed increase in F NOx and strongly neg-925 ative NO x values coincided with a period when total column O 3 declined from > 300 to about 250 DU (Fig. 7a  and c).During Period III. the median column O 3 was about 8 % lower than during the time periods before and after (Table 2).However, associated changes in J NO 3 on 930 the order of ⇠ 10 % are too small to account alone for the observed tripling in F NOx (Fig. 6e; Table 2).
Instead changes in F NOx can be linked to the temporal variability of NO 3 present in the snow skin layer.During the end of Period II. and beginning of Period III.skin 935 layer NO 3 concentrations were up to two times larger than before and after (Fig. 7b).F NOx is high during the end of Period II. and beginning of Period III., however drops off one week after the decrease of nitrate concentrations in surface snow (Fig. 7c).To confirm the link 940 between NO x emissions and NO 3 in snow F NO2 values were modelled (Eq. 3) based on observed J NO 3 , daily sampling of skin layer NO 3 and two depth profiles, at 100 m (P1) and 5 km (P2) distance from the lab shelter, in order to account for spatial and temporal variabil-945 ity of NO 3 in snow.Modelled F NO2 capture some of the temporal trends in observational estimates of F NOx confirming the link with J NO 3 and NO 3 concentrations (Fig. 7c).However, median ratios of observed F NOx and modelled F NO2 values are 30-50 during Period III. and 950 15-30 during Period IV. (Fig. 7c).
Disagreement between model and observations was previously attributed to the poorly constrained quantum yield of NO 3 photolysis in natural snow (Frey et al., 2013).The model employed here uses a constant quan-955 tum yield, i.e. its value at the mean ambient temperature at Dome C ( 30 C) of 0.0019 (Chu and Anastasio, 2003).However, quantum yield may vary with time, as the same lab study reports a positive relationship between quantum yield and temperature (Chu 960 and Anastasio, 2003).Comparison of time periods before and after 18 December 2011 shows an increase of mean air temperature from 34.2 C to 27.7 C and a decrease of its mean diurnal amplitude from 13 to 9.7 K (Fig. 1a).However, observations of F NOx showed 965 behaviour opposite to that expected from a temperature driven quantum yield, i.e.F NOx values decreased as air temperature increased (Fig. 1a and d).Yet, the large diurnal amplitude of air temperature at Dome C could explain diurnal changes of F NOx by a factor 1.5-1.75.970 However, contributions from the temperature effect are small when compared to the up to 20-fold change between night and day observed in F NOx .A recent lab study found that the quantum yield of photolytic loss of NO 3 from snow samples collected at Dome C decreased 975 from 0.44 to 0.003 within what corresponds to a few days of UV exposure in Antarctica (Meusinger et al., 2014).The authors argue that the observed decrease in quantum yield is due to NO 3 being made of a photolabile and a photo-stable fraction, confirming a previous 980 hypothesis that the range of quantum yields reflects the location of NO 3 within the snow grain and therefore availability to photolysis (Davis et al., 2008;Frey et al., 2013).Thus, the F NOx values observed at Dome C fall well within the range of predictions based on quantum 985 yield values measured in snow samples from the same site, which exceed that used in the current model by a factor 2-200.A systematic decrease in quantum yield due to depletion of photo-labile NO 3 in surface snow may have contributed to the observed decrease in F NOx 990 after 22 December 2011.However, a lack of information on snow grain morphology or NO 3 location within the snow grain limits further exploration of the impact of a time variable quantum yield on F NOx .It should be noted that during 2009-2010 large skin layer NO 3 val-995 ues did not result in F NOx values comparable to those in 2011-2012 which may be due to a different partitioning between photo-labile and photo-stable NO 3 in surface snow (Fig. 7b and c; Table 2).
The consequences of large NO x fluxes consist not only in contributing to high NO x mixing ratios but also in influencing local O 3 production, as suggested by significantly higher surface O 3 mixing ratios (> 30 ppbv) during 9-22 December in 2011-2012 (Period III.) compared to 25 ppbv in 2009-2010 (Fig. 7d).

Conclusions
Measurements of NO x mixing ratios and flux carried out as part of the OPALE campaign at Dome C in 2011-2012 allowed to extend the existing data set from a previous campaign in 2009-2010.
Vertical profiles of the lower 100 m of the atmosphere confirm that at Dome C large diurnal cycles in solar irradiance and a sudden collapse of the atmospheric boundary layer in the early evening control the variability of NO x mixing ratios and flux.In contrast, at South Pole diurnal cycles are absent and changes more due to synoptic variability (Neff et al., 2008).Understanding atmospheric composition and air-snow interactions in inner Antarctica requires studies at both sites as they together encompass the spectrum of diurnal variability expected across the East Antarctic Plateau (King et al., 2006;Frey et al., 2013).Large mixing ratios of NO x at Dome C arise from a combination of several factors: continuous sunlight, large NO x emissions from surface snow and shallow mixing depths after the evening collapse of the convective boundary layer.Unlike at South Pole it is not necessary to invoke non-linear HO x -NO x chemistry to explain increases in NO x mixing ratios.However, uncertainties remain regarding atmospheric levels of HO 2 NO 2 and its impact on NO x life time being a temporary NO x reservoir.
Firn air profiles suggest that the upper snow pack at Dome C is an O 3 sink and holds below a few e-folding depths a significant reservoir of NO 2 produced photolytically above, whereas NO disappears at depths devoid of UV as it reacts with O 3 .Shading experiments showed that the presence of such a NO 2 reservoir dampens the response of NO x mixing ratios above or within the snowpack due to changes in down-welling UV irradiance on hourly time scales.Thus, systematic changes in NO x mixing ratios and flux due to the impact of UV on the snow source are only observable on diurnal and seasonal time scales.
First-time observations of BrO at Dome C suggest that mixing ratios of BrO near the ground are low, certainly less than 5 pptv.Assuming steady-state observed mixing ratios of BrO and RO 2 radicals are about a factor ten too low to explain the NO 2 : NO ratios measured in ambient air.A potential interference of HO 2 NO 2 with the NO 2 measurements explains only a small part of 1050 this inconsistency.Hence, the large NO 2 : NO ratios observed at Dome C are either the result of an unknown measurement bias or of a yet unidentified mechanism in boundary layer oxidation chemistry, as similarly concluded in OPALE companion papers (e.g.Legrand et al., 1055 2014; Kukui et al., 2014;Savarino et al., 2015).
During 2011-2012 NO x mixing ratios and flux were larger than in 2009-2010 consistent with also larger surface O 3 mixing ratios resulting from increased net O 3 production.Large NO x mixing ratios and signifi-

525
and summer 2009-2010 are a strong intra-seasonal variability and larger atmospheric mixing ratios.A significant increase of NO x mixing ratios at 1.0 m from low values in Period I. (23-30 November 2011) occurred in two steps: a small rise in Period II.(1-8 December 2011), 530

675
snow source and thus NO x flux and mixing ratios is only observable on diurnal and seasonal time scales.M .M .F r e y e t a l .: A t m o s p h e r i c n i t r o g e n o x i d e s d u r i n g O P A L E 3.4 NO 2 : NO ratios, peroxy and halogen radicals In 2011-2012 the NO 2 : NO ratios at 1.0 m were up to 3 times larger than in 2009-2010 (Table

1060
cant variability during December 2011 were attributed to a combination of changes in mixing height and NO x snow emission flux F NOx .Trends in F NOx were found to be controlled by atmospheric turbulence and the strength of the photolytic snowpack source, of which 1065 the relative importance may vary in time.Larger median F NOx values in 2011-2012 than those during the same period in 2009-2010 can be explained by both, significantly larger atmospheric turbulence and a slightly stronger snowpack source.However, the tripling of F NOx 1070 in December 2011 was largely due to changes in snow pack source strength driven primarily by changes in NO 3 concentrations in the snow skin layer, and only to a secondary order by the decrease of total column O 3 and the associated increase in NO 3 photolysis rates.1075 Median ratios of observed F NOx and modelled F NO2 values ranged from 15 to 50 using the quantum yield of NO 3 photolysis reported by Chu and Anastasio (2003).Model predictions based on quantum yield values measured in a recent lab study on Dome C snow samples 1080 (Meusinger et al., 2014) yield 2-200 fold larger F NO2 values encompassing observed F NOx .In particular, a decrease in quantum yield due to depletion of photo-labile NO 3 in surface snow may have contributed to the observed decrease in F NOx after 22 December 2011.Yet in 1085 2009-2010 large skin layer NO 3 values did not result in elevated F NOx values as seen in 2011-2012 possibly due to different partitioning of NO 3 between a photo-labile and photo-stable fraction.In summary the seasonal variability of NO x snow 1090 emissions important to understand atmospheric composition above the East Antarctic Plateau depends not only on atmospheric mixing but also critically on NO 3 concentration and availability to photolysis in surface snow, as well as incident UV irradiance.However, the 1095 boundary layer chemistry of reactive nitrogen is not fully understood yet.Future studies on the Antarctic Plateau need to reduce uncertainties in NO 2 and HONO measurements, obtain also observations of HO 2 NO 2 and assess how quantum yield of NO 3 photolysis in snow varies 1100 as a function of snow chemical and physical properties.This is important to be able to close the mass budget of reactive nitrogen species between atmosphere and snow above Antarctica.Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper |

b
Based on concentrations at 1.0 and 0.01 m above the snow surface.

Figure 1 .
Figure 1.Meterorology and NO x observations at Dome C in summer 2011-2012 (highlighted periods I.-IV.as referred to in text and Table2): (a) air temperature (T ) at 1.6 m and modeled mixing height (h z )(Gallée et al., 2015), (b) wind speed (wspd) and direction (wdir) at 3.3 m (c), 1 min averages of NO x mixing ratios at 1 m (red line is 1 day running mean) and (d) 10 min averages of observational estimates of NO x flux (F NOx ) between 0.01 and 1 m (red line is 14 day running mean).

Figure 2 .Figure 3 .
Figure 2. Balloon profiles (vertical dashed lines) from 9 January 2012: (a) modelled mixing height h z (10 min running mean) and observed turbulent diffusion coefficient of heat K h at 1 m (symbols: 10 min averages; black line: 30 min running mean).(b) interpolated vertical profiles of NO x mixing ratios with contour lines representing 60 pptv intervals.The lower 100 m appear well mixed during the day, while after collapse of the convective boundary layer in the early evening snow emissions of NO x are trapped near the surface causing a strong increase in mixing ratios near the ground.

Figure 4 .
Figure 4.The impact of rapid changes in incident solar radiation on atmospheric NO x mixing ratios (1 min values).(a-b) ambient concentrations at 1 m during a partial solar eclipse on 25 November 2011 (shaded area) with black lines representing the 10 min running mean.(c-d) firn air concentrations at 10 cm depth during a shading experiment using UV-filters on 11 January 2012.Square symbols and error bars represent interval averages and standard deviation, respectively.Shaded areas and filled squares indicate time periods when the UV filter was in place.

Figure 7 .
Figure 7. (a) Total column O 3 above Dome C. (b) NO 3 concentrations in the skin layer of surface snow (top 0.5 cm).(c) observational estimates of NO x flux (F NOx ) between 0.01 and 1 m (10 min averages) and modelled F NO2 (multiplied by 10) based on NO 3 in the skin layer and depth profiles observed at 100 m (P1) and 5 km (P2) distance from the lab shelter (see Fig. 3a); the 1 day running mean of F NOx during 2009-2010 is shown for comparison (from Frey et al., 2013) (d) atmospheric O 3 mixing ratios.Highlighted periods I.-IV.as referred to in text and Table 2.
Table 2).The median flux of NO x during Period III.reached 3.1 ⇥ 10 13 molecule m 2 s 1 , almost 5 times the season median from 2009-2010.During Period IV. (23 Decem-January 2012 a total of 12 vertical atmospheric profiles of NO x mixing ratios were measured between 11:30 and 23:30 LT.The lower 100 m of the atmosphere appear well mixed throughout the afternoon, with mod- 555 3.2 The lower atmosphere-firn air profile On 9

Table 2 .
Seasonal evolution of median NO x mixing ratios and flux along with relevant environmental parameters at Dome C in summer 2011-2012 (time periods I.-IV.highlighted in Fig. and 7) and comparison to summer 2009-2010 (from Frey et al., 2013).
a At 1 m above the snow surface.