Large mixing ratios of atmospheric nitrous acid (HONO) at Concordia (East Antarctic Plateau) in summer: a strong source from surface snow?

. During the austral summer 2011/2012 atmospheric nitrous acid (HONO) was investigated for the second time at the Concordia site (75 ◦ 06 0 S, 123 ◦ 33 0 E), located on the East Antarctic Plateau, by deploying a long-path absorption photometer (LOPAP). Hourly mixing ratios of HONO measured in December 2011/January 2012 (35 ± 5.0 pptv) were similar to those measured in December 2010/January 2011 (30.4 ± 3.5 pptv). The large value of the HONO mixing ratio at the remote Concordia site suggests a local source of HONO in addition to weak production from oxidation of NO by the OH radical. Laboratory experiments demonstrate that surface snow removed from Concordia can produce gas-phase HONO at mixing ratios half that of the NO x mixing ratio produced in the same experiment at typical temperatures encountered at Concordia in summer. Using these lab data and the emission ﬂux of NO x from snow estimated from the vertical gradient of atmospheric concentrations measured during the campaign, a mean diurnal HONO snow emission ranging between 0.5 and 0.8 × 10 9 molecules cm − 2 s − 1 is calculated. Model calculations indicate that, in addition to around 1.2 pptv of HONO produced by the NO oxidation, these HONO snow emissions can only explain 6.5 to 10.5 pptv of HONO in the atmosphere at Concordia. To explain the difference between observed and simulated HONO mixing ratios, tests were done both in the ﬁeld and at lab to explore the possibility that the presence of HNO 4 had biased the measurements of HONO.


Introduction
The existence of an oxidizing boundary layer over the Antarctic continent was first highlighted by measurements carried out at the South Pole, where a mean concentration of 2.5 × 10 6 OH radicals cm −3 was observed (Mauldin III et al., 2001a), making the South Pole atmospheric boundary layer as oxidative as the remote tropical marine boundary layer (Mauldin III et al., 2001b). Chen et al. (2001) and  showed that the presence of high concentrations of NO x produced by the photolysis of nitrate present in surface snow permits the required efficient recycling of HO 2 into OH. Aside from snow photochemical emission of NO x that acts as a secondary source of OH, the role of atmospheric nitrous acid (HONO) as a primary source of OH remains unclear. Using a mist chamber followed by ion chromatography analysis of nitrite, Dibb et al. (2004) reported a median HONO mixing ratio close to 30 pptv at the South Pole. However, follow-up measurements by laser-induced fluorescence (LIF) indicated lower mixing ratios (6 pptv on average), and an interference with HNO 4 has been suspected (Liao et al., 2006). Furthermore, as discussed by Chen et al. (2004) the consideration of 30 pptv of HONO in the lower atmosphere over the South Pole leads to an OH over-prediction by gasphase photochemical models by a factor of 3 to 5. The authors questioned whether the discrepancy between observed and simulated concentrations of OH at the South Pole was due to measurements of HONO suffering from overestimation due to chemical interferences or if the mechanisms of the model missed HO x and NO x losses.
Even at the level of a few parts per trillion by volume (pptv), the presence of HONO requires a source other than the gas-phase reaction of NO with OH, and many studies measuring HONO in atmospheres overlying snow-covered regions have suspected HONO to be emitted from the surface snow in addition to NO x (see Grannas et al., 2007 for a review). It has to be emphasized that most of the studies of HONO have concerned high (Arctic, Greenland) and middle (Colorado and Alps) northern latitudes where, in relation to the chemical composition of snow, the involved HONO production processes would be very different compared to the case of Antarctica. Concerning Antarctic snow, following the pioneering shading experiment done by Jones et al. (2000) on snow from the coastal Antarctic site of Neumayer, numerous studies have investigated the release of NO x from the snow , but only two studies have reported on HONO snow emissions, and none of them examined together HONO and NO x emissions. Beine et al. (2006) reported small HONO fluxes (3 × 10 7 molecule cm −2 s −1 ) above the Browning Pass (coastal Antarctic) snowpack. However, the snow chemical composition at that site is very atypical, with a large presence of calcium (up to 4 ppm) attributed to the presence of a lot of rock outcrops at the site. As a consequence, even if nitrate is abundant (typically 200 ppb in fresh snow and more than 1 ppm in aged snow), the snow from that site appears to be weakly acidic and sometimes alkaline. Finally a few investigations of the vertical distribution of HONO were made at the South Pole (Dibb et al., 2004), but no fluxes were calculated. These previous Antarctic studies of HONO were carried out using either mist chambers  or high-performance liquid chromatography techniques (Beine et al., 2006). These "wet chemical instruments" sample HONO on humid or aqueous surfaces followed by analysis of the nitrite ion. However, it is well known that many heterogeneous reactions lead to the formation of nitrite on similar surfaces (Gutzwiller et al., 2002;Liao et al., 2006). In addition to these chemical interferences, it is also known that HONO can decompose or be formed on various surfaces (Chan et al., 1976). That may have affected data when sampling lines of up to 30 m length were used for polar measurements (see, e.g., Beine et al., 2006).
Motivated by a strong need to extend investigations of the oxidation capacity of the lower atmosphere at the scale of the whole Antarctic continent, the OPALE (Oxidant Production over Antarctic Land and its Export) project was initiated at the end of 2010 in East Antarctica. The first OPALE campaign was conducted during austral summer 2010/2011 at the coastal site of Dumont D'Urville  and focused on OH and RO 2 measurements . During this first campaign, preliminary investigations of HONO were performed at the continental station of Concordia located at Dome C (denoted DC, 3233 m above sea level). In spite of the use of a long-path absorption photometer (LOPAP), thought to avoid all known artefacts, high mixing ratios of HONO were observed (from 5 to 59 pptv; Kerbrat et al., 2012). In the framework of the OPALE project, a second summer campaign (2011)(2012) was conducted at Concordia with simultaneous measurements of HONO, NO, NO 2 , OH, and RO 2 , which are discussed in a set of companion papers of which this is one.
The paper presented here focuses on HONO data gained during the second campaign at Concordia. It also reports on snow irradiation experiments conducted in the laboratory at the British Antarctic Survey (BAS) on surface snow samples collected at DC in order to quantify a possible photochemical snow source of HONO. This was done by measuring simultaneously HONO with the LOPAP, NO, and NO 2 with a two-channel chemiluminescence detector. From these data we crudely estimate the amount of HONO released from snow within the lower atmosphere at DC on the basis of the NO x snow emissions derived from the vertical gradient of atmospheric concentrations measured during the campaign by Frey et al. (2014). The derived values of the HONO flux were used in 1-D modelling calculations to evaluate the contribution of this snow source to the large HONO mixing ratios observed at Concordia. Finally, to evaluate a suspected possible interference of HNO 4 on the HONO mixing ratio measured by the LOPAP, field experiments were conducted by heating sampled air prior to its introduction in the LOPAP device, heating being a convenient way to destroy HNO 4 . The selectivity to HNO 4 and the response of the LOPAP during the heating events was also investigated in laboratory by mass spectrometry at the Paul Scherrer Institute (PSI).

HONO measurement method
HONO was measured using a LOPAP which has been described in detail elsewhere (Heland et al., 2001;Kleffmann et al., 2002). In brief, after being sampled into a temperaturecontrolled stripping coil containing a mixture of sulfanilamide in a 1N HCl solution, HONO is derivatized into a coloured azo dye. The light absorption by the azo dye is measured in a long-path absorption tube by a spectrometer at 550 nm using an optical path length of 5 m. The LOPAP did not have long sampling lines or inlet. The stripping coil was placed directly in the atmosphere being sampled. The Atmos. Chem. Phys., 14, 9963-9976, 2014 www.atmos-chem-phys.net/14/9963/2014/ LOPAP has two stripping coils connected in series to correct interferences. In the first coil (channel 1), HONO is trapped quantitatively together with a small amount of the interfering substances. Assuming that these interfering species are trapped in a similar amount in the second coil (channel 2), the difference between the signals resulting from stripping in each coil provides an interference-free HONO signal (Heland et al., 2001). Air was sampled at a flow rate of 1 L min −1 , and the flow rate of the stripping solution was 0.17 mL min −1 . Calibrations were performed every 5 days. Relative deviations of the calibration signal were of 3 and 9 % at 3σ for channel 1 and 2, respectively. The quantification limit of the LOPAP instrument used in this study was as low as 1.5 pptv (taken as 10σ of all zero measurements done by sampling pure N 2 ) with a time resolution of 9 min. More details on the set-up of the LOPAP device in the fields can be found in Kerbrat et al. (2012). Similarly to the first campaign, the amount of interferences in the second coil was on average 9 ± 7 % of total signal (instead of 10 ± 5 % found by Kerbrat et al. (2012) in 2010/2011). The LOPAP was tested for numerous possible interfering species, including NO, NO 2 , HNO 3 , and alkylnitrates. It was concluded that, when significant, the two-channel approach was able to well correct the HONO data (Kleffmann and Wiesen, 2008). However, it has to to be emphasized that no tests have been conducted for HNO 4 . During the field campaign, HONO was occasionally sampled in the snow interstitial air by pumping air through a perfluoroalkoxy (PFA) tube (5 m long, 4 mm internal diameter) at a flow rate of 1 L min −1 . In addition, to evaluate a possible influence of HNO 4 on HONO measurements, field experiments were undertaken by heating air sampled through a 9 m long PFA tube. Tests were performed to evaluate potential loss or formation of HONO in the PFA tubes by running the LOPAP for 30 min with and without a tube connected to the inlet of the LOPAP, sampling air at the same height. In order to account for possible fast natural change of HONO mixing ratios, the test was repeated three times successively. The tests were carried out with ambient mixing ratios of 20 pptv as encountered at midday on 23 December and 40 pptv on the morning of 28 December. In the two cases losses of around 4 and 7 pptv were observed when using the 5 and 9 m long PFA tube, respectively. These losses will be considered in discussing HONO mixing ratios in interstitial air (see Sect. 3) or the interference of HNO 4 (see Sect. 6).

Field atmospheric measurements and snow samplings
The second OPALE field campaign took place at Concordia located over the high East Antarctic Plateau from late November 2011 to mid-January 2012. Nitrous acid was measured 1 m above ground level, about 900 m south-southwest from the main Concordia station. Measurements that started 4 December were interrupted from 9 to 15 December, 16 to 18 December, and 28 to 30 December afternoon due to problems on the LOPAP device. On 1 and 2 January and from 10 to 13 January air measurements were stopped to measure HONO in snow interstitial air. During the measurement campaign, the main wind direction was from the southeast to southwest. Several episodes with wind blowing from the north (10 • W to 60 • E sector), i.e. from the direction of the station, were encountered (see the red points in Fig. 1). During some of these pollution events (31 December around 22:00 for instance), sharp peaks of HONO mixing ratios exceeding 100 pptv were observed. These events were also detected in the NO x time series , with sharp peaks in the range of 100 ppbv or more (120 ppbv on 31 December at around 22:00 for instance). The ratio of excess of HONO to excess of NO x during these events is close to 10 −3 . The ratios of HONO / NO x reported by measurements made in traffic tunnels range from 3 × 10 −3 (Kirchstetter et  , 1996) to 8 × 10 −3 . When compared to ratios observed in tunnels, the lower ratio seen in the plume of the Concordia station when it reaches the sampling line is likely due the rapid photolytic destruction of HONO, whose lifetime is still as short as 20 min at the high solar zenith angles prevailing at DC around 22:00 in summer. In the following the data corresponding to red points reported in Fig. 1 were removed from the HONO data set. Concurrent measurements of chemical species that are relevant for discussion include those of ozone, NO, NO 2 , OH, and RO 2 . Surface ozone was monitored simultaneously to HONO using UV absorption monitors (Thermo Electron Corporation model 49I) deployed at Concordia since 2007 (Legrand et al., 2009). Nitrogen oxides were determined by deploying a two-channel chemiluminescence detector (Bauguitte et al., 2012;Frey et al., 2013Frey et al., , 2014. The chemiluminescence detector measured NO in one channel and the sum of NO, and NO originating from the photolytic conversion of NO 2 in the other channel. As discussed by Frey et al. (2013), among various nitrogen oxides able to interfere on the photolytic conversion channel, only HONO has to be considered, leading to an overestimation of NO 2 levels by less than 5 %. The radicals (OH and RO 2 ) were measured using chemical ionization mass spectrometry . During the campaign the photolysis rate of HONO was documented using a Met-Con 2π spectral radiometer equipped with a CCD detector and a spectral range from 285 to 700 nm (see details in Kukui et al., 2014).
Different surface snow samples were collected at DC and returned to the UK to be used in irradiation experiments (see Sects. 2.3 and 4). First, the upper 12 cm of snow was collected in December 2010. Second, the upper centimetre of snow corresponding to freshly drifted snow was collected 6 December 2011. The samples were characterized by their specific surface area (SSA). Measurements were performed using an Alpine snowpack specific surface area profiler, an instrument similar to that one described by Arnaud et al. (2011) based on the infrared reflectance technique. Briefly, a laser diode at 1310 nm illuminates the snow sample at nadir incidence angle, and the reflected hemispherical radiance is measured. The hemispherical reflectance at 1310 nm is related to the SSA using the analytical relationship proposed by Khokanovsky and Zege (2004). The SSA of the drifting snow is close to 26 m 2 kg −1 , and the upper 12 cm is 17 m 2 kg −1 . Such values appear to be close to typical DC values reported in the literature , suggesting that lab experiments conducted on these snow samples (see Sect. 4) may be relevant to discuss at least qualitatively natural processes occurring at DC.
The upper surface snow (from 0 to 1 cm, and from 0 to 12 cm) at DC were also sampled and analysed for major anions and cations following working conditions reported in Legrand et al. (2013). For cations (Na + , K + , Mg 2+ , Ca 2+ , and NH + 4 ), a Dionex 500 chromatograph equipped with a CS12 separator column was used. For anions, a Dionex 600 equipped with an AS11 separator column was run with a quaternary gradient of eluents (H 2 O, NaOH at 2.5 and 100 mM, and CH 3 OH), allowing for the determination of inorganic species (Cl − , NO − 3 , and SO 2− 4 ) as well as methanesulfonate (CH 3 SO − 3 ). The acidity of samples can be evaluated by the ionic balance between anions and cations, with concentrations expressed in micro-equivalents per litre (µEq L −1 ):

Snow irradiation experiments conducted at BAS
As discussed in Sect. 5, model simulations indicate that the production of HONO from the reaction of OH with NO is far too weak to explain observations at Concordia and that an additional light-driven HONO source is needed. To quantify a possible photochemical snow source of HONO, lab experiments were conducted at BAS by irradiating snow collected at DC and measuring gas-phase evolution of NO and NO 2 with a two-channel chemiluminescence detector (Bauguitte et al., 2012) as deployed at Concordia (Frey et al., 2013 and HONO with the LOPAP that ran at Concordia during the 2010/2011 and 2011/2012 campaigns. A 20 cm long cylinder (6 cm inner diameter) was filled with ∼ 120 g of snow inside an airtight glass reaction chamber (total length of 40 cm, 6 cm inner diameter) and put in a freezer of which the temperature was varied between −5 and −35 • C. Further details on the characteristics of the reaction chamber can be found in Meusinger et al. (2014). The reaction chamber is maintained vertically in a freezer, and a 1000 W xenon arc lamp was put above the freezer. The snow was irradiated by directing the light axially along the tube through a quartz window, which makes up the top surface of the chamber. Chemically pure air was supplied to the chamber from a pure air generator (Ecophysics, PAG003) in which air is dried at −15 • C. To match the relative humidity of the snow under investigation and limit metamorphism, this chemically pure air was passed through a cold trap at the temperature of the experiment. Note that, with this system and for temperatures above −30 • C, no condensation trace was observed in the tube outflow of the chamber. The flow rate of zero air was 4.3 L min −1 , while the detection systems sampled processed air at a rate of 2.0 L min −1 for NO x and 1.0 L min −1 for HONO. The overflow of 1.3 L min −1 was diverted through a flow meter to check for potential leaks. While the inlet line between the reaction chamber and the NO x analyser was several metres long, the length between the outlet of the reaction chamber and the LOPAP inlet was kept as short as possible (i.e. 25 cm). To do so the inlet of the LOPAP was arranged in the freezer. The wavelength range of the 1000 W xenon arc lamp (Oriel Instruments) was 200-2500 nm, modulated using filters with various cut-on points. The short residence time of NO 2 (∼ 4 s) in our small chamber prevents significant photolysis of NO 2 to occur during the experiments. Indeed, the J NO 2 of 2×10 −2 s −1 measured by Cotter et al. (2003) for a 1000 W xenon arc lamp, as also used in the present study, leads to a lifetime of NO 2 with respect to photolysis of 50 s at the front of the snow block.

Experiments performed at PSI to investigate a possible HNO 4 interference on HONO measurements
As will be discussed in Sect. 6, it may be difficult to reconcile typical mixing ratios of HONO measured 1 m above surface snow at Concordia with a reasonable estimate of the mixing ratio of HONO owing to emissions from snow due to snowpack photochemistry. It was suspected that HNO 4 was detected and measured as HONO by the LOPAP instrument. As briefly reported below, a few experiments conducted at PSI indicate that the LOPAP instrument does have an interference for HNO 4 . Mixing ratios of HNO 4 were not measured at Concordia, so the aim of the experiments described below was not to quantify the interference to enable correction of the Concordia HONO data, but to demonstrate that such an interference exits. The result of an experiment conducted under specific conditions is reported. A full characterization of the interference on HONO at various mixing ratios of HNO 4 in the presence or absence of other trace gases present at DC is beyond the scope of this paper. The interference of the LOPAP device was examined at PSI, where a gas-phase synthesis of HNO 4 has been developed by irradiating a mixture of NO 2 / H 2 O / CO / O 2 / N 2 at 172 nm (Bartels-Rausch et al., 2011). By-products of the synthesis are HONO, HNO 3 , and H 2 O 2 . The synthesis gas was fed into the sampling unit of the LOPAP, and the resulting LOPAP signals in the presence and absence of HNO 4 were compared. Heating the synthesis gas to a temperature of 100 • C prior to sampling by the LOPAP allowed selective removal of HNO 4 from the gas mixture. The mixing ratios of HONO, NO 2 , H 2 O 2 , and O 3 that are present in the synthesis gas were independently monitored with a chemical ionization mass spectrometer (CIMS), which was calibrated by using several analysers as detailed in Ulrich et al. (2012). An example of the mixing ratios of HNO 4 and HONO measured by CIMS and of the corresponding LOPAP signals in channel 1 and 2 is shown in Fig. 2. The relative amount of HONO (780 pptv) and HNO 4 (1000 pptv) observed in the synthetized mixture (prior heating) is typical for this synthesis (Bartels-Rausch et al., 2011). The experiment shows the response of the signals when the heating trap used to decompose HNO 4 is applied. As seen in Fig. 2, the mixing ratios of HONO, NO 2 , H 2 O 2 , or O 3 that may influence the response of the LOPAP instrument did not change upon the thermal decomposition of HNO 4 . A decrease of the LOPAP signal in channel 1 is observed during the heating event, indicating that 1 ppbv of HNO 4 corresponds to a signal in the LOPAP of 150 pptv. Examination of the signals of the two LOPAP channels (Fig. 2) suggests that HNO 4 has been efficiently sampled in the first channel. It is well known that HNO 4 efficiently decomposes to NO − 2 in acidic solutions (Regimbal and Mozurkewich, 1997), just like HONO does in the LOPAP sample unit. Based on the identical hydrolysis products, one might thus expect a rather large interference. The high sampling efficiency of HONO and potentially HNO 4 , both of which have similar partitioning coefficients to acidic solutions, is driven by the fast reaction of their hydrolysis product (NO − 2 ) with the reagents in the sampling solution of the LOPAP instrument. A full characterization of the interference by HNO 4 (its behaviour and quantification over a large range of concentrations, in the presence or absence of other gases) is needed to improve the use of the LOPAP in very cold atmospheres. We suggest a detailed investigation of LOPAP instrument response to different compositions of test gas mixture (i.e. with larger mixing ratios of H 2 O 2 ), and with an investigation of the potentially complex (non-linear) chemistry of sampled gases. At this stage we can only exclude an oxidation of the dye used in the LOPAP instrument by HNO 4 , as careful inspection of the absorption spectrum of the LOPAP dye reveals no significant change during heating. Assuming the interference of HONO signal by HNO 4 to be linear, one would expect an interference of ∼ 15 pptv in the HONO signal due to a mixing ratio of 100 pptv of HNO 4 . Given the absence of measurements of the mixing ratio of HNO 4 at Concordia, further experiments were conducted in the field at Concordia to directly estimate this interference as detailed in Sect. 6.

HONO observations at Concordia
Removing data suspected to have been impacted by pollution from station activities (see Sect. 2.2), a 1 min average mixing ratio of 35 ± 14 pptv is observed in December 2011/January 2012 compared to 28 ± 12 pptv measured by  for December 2010/January 2011.
The mean diurnal cycles of surface ozone, HONO, air temperature, and the planetary boundary layer (PBL) height simulated by the regional atmospheric MAR model (Modèle Atmosphérique Régional) are reported and compared for the two summers in Fig. 3. In the polar region, the strong static stability of the atmosphere often inhibits vertical mixing of surface emissions between the surface boundary layer and the rest of the atmosphere. At DC, the surface absorbs solar radiations during the day, heats the lower atmosphere, and generates positive buoyancy that is responsible for an increase of turbulent kinetic energy and the subsequent increase of the boundary layer height seen in Fig. 3. This boundary layer is referred to as the "daytime boundary layer". The surface cooling after 17:00 generates negative buoyancy near the surface. A new boundary layer referred as the "night-time boundary layer" develops but remains less active than the previous daytime boundary layer. The collapse of the boundary layer after 17:00 seen in Fig. 3 is in fact the representation of the transition between the daytime and night-time boundary layer.
The two mean summer ozone records indicate a drop of 1 to 2 ppbv around midday compared to early morning and evening values (Fig. 3). This small surface ozone change over the course of the day at DC has already been observed by Legrand et al. (2009), who attributed it to the increase of the PBL height in the afternoon that counteracts a local photochemical production of O 3 in the range of 0.2 ppbv h −1 during daytime.
Consistently with the previous 2010/2011 measurements from Kerbrat et al. (2012), the HONO mixing ratios exhibit a well-marked diurnal variation characterized by morning (around 05:00-07:00) and evening (around 20:00) maxima Atmos. Chem. Phys., 14, 9963-9976, 2014 www.atmos-chem-phys.net/14/9963/2014/ exceeding midday values by some 10 pptv. Therefore, in addition to an expected more efficient photolysis of HONO during the day, the increase of the daytime boundary layer may also accounts for the observed decreased HONO mixing ratios during the day in spite of a more active snow source (see discussions in Sect. 5). Such a diurnal variability characterized by noon minimum was also observed for NO x by Frey et al. (2013) and attributed to the interplay between photochemical snow source and boundary layer dynamics. As shown in Fig. 3, the larger HONO mixing ratios calculated for 2011/2012 (diurnal mean of 35 ± 5.0 pptv) with respect to the 2010/2011 ones (diurnal mean of 30.5 ± 3.5 pptv) concern both the midday minimum and the morning/evening maxima. The difference between the two summers, however, is reduced when the first week of measurements undertaken in December 2011 is removed with a lower diurnal mean (31.7 ± 4.3 instead of 35 ± 5 pptv over the entire measurement period; see the blue points in Fig. 3). The case of the beginning of December 2011 with respect to the rest of the summer 2011/2012 is highlighted in Fig. 3. It can be seen that the far thinner PBL height of early December (maximum of 145 instead of 350 m over the entire period) may have led to a more confined HONO production (see violet points in Fig. 3). Note also the relatively high ozone mixing ratios at that time (33 ± 4 ppbv in early December instead of 26 ± 1 ppbv over the entire period). Conversely, at the end of the period the PBL became thicker (maximum of 570 m) and the mixing ratios of ozone (24 ± 1 ppbv) and nitrous acid (31 ± 4 pptv) were lower than on average (see red points in Fig. 3). Finally, in early December 2011 the highest daily average mixing ratio of HONO observed, on 7 and 8 December (56 pptv, Fig. 1), corresponds not only to a thin PBL but also to lowest value of total ozone column (260 Dobson units (DU) instead of 296 ± 20 DU on average) measured by the SAOZ (Système d'Analyse par Observation Zénitale) at Concordia. Similarly, during the 2010/2011 campaign the highest values reported at the end of the campaign (44 pptv from 15 to 18 January) by Kerbrat et al. (2012) correspond to the lowest value of total ozone column (270 instead of 303 ± 17 DU on average). It therefore seems that HONO mixing ratios measured at 1 m at DC are also sensitive to the UV actinic flux reaching the surface. This link between stratospheric ozone and photochemistry of snow at the ground is discussed in more detail by Frey et al. (2014).
It therefore seems that one of the main causes of the difference between the 2011/2012 and 2010/2011 mean summer values is the slightly different atmospheric vertical stability conditions experienced over the different sampling times of the two summers, with an earlier HONO sampling in December 2011 than in December 2010 leading to higher HONO mixing ratios in a very thin and stable boundary layer. In conclusion, this second study of HONO confirms the abundance of this species in the lower atmosphere at DC with a typical mean mixing ratio of 30 pptv from mid-December to mid-January. As already discussed by Kerbrat et al. (2012) (see also Sect. 5), the existence of a large photochemical source of HONO in the snowpack is needed to explain these large mixing ratios of HONO measured above the snowpack. Measurements of the mixing ratio of HONO were therefore performed in snow interstitial air at different depths. From the top few centimetres of the snowpack down to 75 cm depth, mixing ratios of HONO in snowpack interstitial air tended to exceed those in the air above the snowpack, supporting the existence of a snow source of HONO (Fig. 4). However, given the interference of HNO 4 on HONO mixing ratio data as discussed in Sect. 6, it is difficult to use the observed vertical gradient of HONO mixing ratio to derive an estimate of emission of HONO from the snowpack. Indeed, typical values of HNO 4 mixing ratios are available in lower atmosphere of the Antarctic Plateau (Sect. 6) but not yet in snow interstitial air. Also it remains difficult to accurately estimate the production rate of HNO 4 in snow interstitial air from the reaction of NO 2 with HO 2 versus its uptake on natural ice surface.
To confirm the snowpack as a source of HONO (and as detailed in the following section), we carried out a laboratory experiment to evaluate the ratio of HONO to NO x released from natural surface snows collected at DC under controlled laboratory conditions (i.e. wavelength of light, temperature, snow specific area) to estimate the HONO snow emission flux relative to the snow emission flux of NO x for the same snowpack as derived from the atmospheric concentration vertical gradient measured during the campaign by Frey et al. (2014). Table 1. Results of irradiation experiments performed at the BAS laboratory on three different surface snows collected at DC. S1 and S2 are upper surface snows collected between 0 and 1 cm; S3 is the surface snow collected between 0 and 12 cm depth. The acidity is calculated by checking the balance between anions and cations (see Sect. 2.2). DL refers to detection limit, and N.C. means non-calculated value. . NO x and HONO are produced when snow is irradiated. Several laboratory experiments were conducted to investigate the wavelength, temperature, and snow chemical composition dependence of HONO release from snow. Similar to previous laboratory experiments conducted by Cotter et al. (2003) on surface snows collected in coastal Antarctica, the NO x release is found to halve when the optical filter in the front of the irradiation lamp (cut off for < 295 nm) is replaced by a cut-off filter for illumination wavelength smaller than 320 nm (Table 1). Cotter et al. (2003) demonstrated no measurable emission of NO x from the snow when illuminated with a lamp with wavelengths shaded below 345 nm, which is consistent with NO − 3 photolysis. Figure 5 illustrates the wavelength dependence of HONO release, showing the effect of insertion of a filter with different cut-off points. Similarly to the NO x , the HONO release is decreased by a factor of 2 when inserting the filter at 320 nm and becomes insignificant at 385 nm (Table 1).
While the observed wavelength dependency of the NO x release supports the hypothesis that the photolysis of nitrate present in snow is the major source of released NO x (via its major channel: NO − 3 +hν → NO 2 + O − ), for HONO it is still unclear whether the nitrate photolysis efficiently produces directly HONO from hydrolysis of NO − 2 produced by the second channel of the nitrate photolysis (NO − 3 + hν → NO − 2 + O) or HONO is secondarily produced from NO 2 (Villena et al., 2011). Indeed, lab experiments conducted on nitratedoped ice suggest that the first channel is a factor of 8-9 more efficient than the second one. It is suspected that the HONO production may be significantly higher than it is when considering this second channel since the NO 2 produced by the first channel may subsequently act as a precursor of HONO. The wavelength dependency of HONO release observed during previous experiments does not, however, help to separate the primary and secondary source of HONO during irradia-  Table 1) and inserting filters with cut-off points at 295, 320, and 385 nm on the xenon arc lamp (see Sect. 4). Vertical grey bands correspond to periods over which the lamp was switched off. tion since they were done with chemically pure air and, when placing the cut-off filter at 385 nm, we suppress the primary source of HONO as well as NO 2 that is needed for secondary HONO production. Among possible secondary productions it is generally accepted that the reduction of NO 2 on photo-sensitized organic material like humic acid (George et al., 2005;Bartels-Rausch et al., 2010) would proceed more efficiently than the disproportionation reaction of NO 2 (2 NO 2 + H 2 O → HONO + HNO 3 ) (Finlayson-Pitts et al., 2003). As discussed by Grannas et al. (2007), the relevance of this secondary production was supported even for Antarctica by the significant presence of dissolved fulvic acid reported for Antarctic snow (26-46 ppb C) by Calace et al. (2005). However, the previously assumed ubiquitous presence of organics in polar snow that is needed to reduce NO 2 into HONO was recently reviewed by Legrand et al. (2013), who found that organics (and humic acids) are far less abundant in Antarctica compared to Greenland or midlatitude glaciers like the Alps. For instance, the typical dissolved organic content of Atmos. Chem. Phys., 14, 9963-9976, 2014 www.atmos-chem-phys.net/14/9963/2014/ summer surface snow is only 10-27 ppb C at DC (Legrand et al., 2013) as opposed to 110 ± 45 ppb C at Summit and 300 ppb C in the Alps. Furthermore, recent HULIS (humiclike substance) measurements of surface snows collected at DC do not confirm the previously observed abundance (2 instead of 26-46 ppb C). From lab experiments conducted by irradiating ice films containing humic acid in the presence of NO 2 , Bartels-Rausch et al. (2010) derived production rates of HONO from NO 2 . From that, the authors roughly estimated light-driven HONO fluxes of 10 13 molecules m −2 s −1 from snow-covered surface area, assuming the presence of 100 pptv NO 2 in the snow interstitial air and a concentration of 10 ppb C of humic acid in snow. Keeping in mind uncertainties in extrapolating lab experiments to conditions relevant to the lower atmosphere at DC, with typical NO 2 mixing ratios of 1 to 10 ppbv in interstitial air at 10 cm below the surface at DC , the presence of 2 ppb C of HULIS in snow may still lead to a significant HONO production from NO 2 at the site. If HULIS are located at the surface of snow grains, much more than 2 ppb C of HULIS would be available to react with NO 2 present in interstitial air of the snowpack to produce HONO. Irradiation experiments with insertion of the filter at 295 nm were conducted at temperatures ranging from 240 to 260 K. As seen in Table 1, whereas the NO x release was found to be temperature independent (as previously shown by Cotter et al., 2003), a large dependence is found for HONO with an increase by a factor of 2.2 when the temperature of snow is increased from 240 to 260 K. A temperature dependence of the HONO emissions is expected since the partition coefficient of HONO between ice and air increases by a factor of 5.8 between 240 and 260 K (Crowley et al., 2010). As a consequence the HONO to NOx release is smaller at 240 K than at 260 K. For the example of the surface snow reported in Table 1, this ratio steadily increases from 0.3 at 240 K, to 0.5 at 250 K, to 0.8 at 260 K.
In Table 1 we report experiments with DC snow containing 160 to 1400 ppb of nitrate. As expected higher nitrate content leads to higher snow release of NO x and HONO, but the increase of HONO is larger than NO x . For instance, at a temperature close to −20 • C, the first upper centimetre of surface snow releases almost twice as much HONO compared NO x as the snow collected from the surface to 12 cm depth. The more acidic character of the snow collected in the upper first centimetre compared to that collected down to 12 cm below the surface (see Table 1) may favour the release of a weak acid species like HONO.

Model calculations
Observed atmospheric mixing ratios were compared with steady-state calculations made by considering major gasphase sources and sinks of HONO. The major sink of HONO is its photolysis. The photolysis rate constant (J HONO ) was measured with a 2π spectroradiometer (see Sect. 2.2). The value of J HONO was calculated for light from 4π steradians from the downwelling value of J HONO measured over 2π steradians by assuming a surface albedo of 0.95, a typical value for regions covered by dry snow and with a wavelength shorter than 400 nm (Hudson et al., 2006;France et al., 2011). The main gas-phase production of HONO is the reaction of NO with OH radicals. Steady-state calculations indicate that under noon conditions encountered at DC (a J HONO value of 3.7 × 10 −3 s −1 , 5 × 10 6 OH rad cm −3 , and 50 pptv of NO ) a HONO mixing ratio of 1 pptv is expected. The steady-state calculated diurnal HONO profile (Fig. 6) suggests a HONO maximum of 2.5 pptv at 19:00 due to the presence of a maximum of 120 pptv of NO . Another gas-phase source of HONO was recently proposed by Li et al. (2014) via reaction of the HO 2 (H 2 O) for possible fast natural change of HONO mixing ratios, the test was repeated three times successively. A systematic drop of HONO values was observed. Given the applied air sampling flow rate of 1.78 L min −1 (1 L STP min −1 ), the residence time of the air in the tube is 3.3 s. If attributed to the thermal decomposition of HNO 4 during the heating (64 % under these working conditions), the mean observed drop of 5.5 pptv of HONO would correspond to an HNO 4 artefact of around 9 pptv.
This indirect estimation of an overestimation of HONO measurements due to the presence of HNO 4 is consistent with experiments conducted at PSI if the presence of 50-100 pptv of HNO 4 is assumed at Concordia. On the other hand, the difference between observed and simulated HONO mixing ratios presented in Sect. 5 suggests a mean diurnal overestimation close to 20 pptv (ranging from 17 pptv around noon to 22 pptv during the night). In their discussions of the observed levels of HO x radicals, Kukui et al. (2014) found that the consideration of 30 pptv of HONO is inconsistent with radical observations leading to about 2 times overestimation of RO 2 and OH concentrations. Conversely, neglecting the OH production from HONO leads to an underestimation of radical levels by a factor of 2. Kukui et al. (2014) showed that a quite fair agreement with OH measurements is achieved with HONO mixing ratios derived from the 1-D modelling, with a HONO snow emission flux equal to about 30 % of that of NO x . Finally, though being slightly higher, the best guess of HONO mixing ratios derived in Sect. 5 for DC (8 to 12 pptv) are in the range of mixing ratios measurements made at the South Pole using laser-induced fluorescence (6 pptv; Liao et al., 2006).

Conclusions
This second study of HONO conducted in the atmosphere of the East Antarctic Plateau by deploying a LOPAP confirms unexpectedly high mixing ratios close to 30 pptv. A mixing ratio of 8-12 pptv can be rationalized based on emissions of HONO from snow of 0.5-0.8 × 10 9 molecules cm −2 s −1 derived from studies of the irradiation experiments surface snow collected at DC and scaled down to the NO x emissions derived from observations made at Concordia by Frey et al. (2014). Experiments conducted in the field and in the lab to identify the cause of such a discrepancy point to a possible overestimation of HONO in the range of 10 to 20 pptv due to the important presence of HNO 4 in this cold atmosphere. An accurate correction of the HONO data from the presence of HNO 4 is not yet possible. Further work -both in the lab, to quantify the interference at different levels of HNO 4 , and in the presence of various other species and in the field at Concordia, to obtain mixing ratios of HONO and HNO 4 at the same time -is needed.