Atmospheric Analysis of multi-year near-surface ozone observations at the WMO/GAW “ Concordia ” station (75°06 ′ S, 123°20 ′ E, 3280 m a.s.l. – Antarctica)

This work focuses on the near-surface O 3 variability over the eastern Antarctic Plateau. In particular, eight years (2006 – 2013) of continuous observations at the WMO/GAW contributing station “ Concordia ” (Dome C – DMC: 75°06 ′ S, 123°20 ′ E, 3280m) are presented, in the framework of the Italian Antarctic Research Programme (PNRA). First, the characterization of seasonal and diurnal O 3 variability at DMC is provided. Then, for the period of highest data coverage (2008 – 2013), we investigated the role of speci ﬁ c atmospheric processes in a ﬀ ecting near-surface summer O 3 variability, when O 3 enhancement events (OEEs) aresystematically observed at DMC (average monthly frequency peaking up to 60% in December). As deduced by a statistical selection methodology, these OEEs are a ﬀ ected by a signi ﬁ cant interannual variability, both in their average O 3 values and in their frequency. To explain part of this variability, we analyzed OEEs as a function of speci ﬁ c atmospheric variables and processes: (i) total column of O 3 (TCO) and UV-A irradiance, (ii) long-range transport of air masses over the Antarctic Plateau (by Lagrangian back-trajectory analysis – LAGRANTO), (iii) occurrence of “ deep ” stratospheric intrusion events (by using the Lagrangian tool STLEFLUX). The overall near-surface O 3 variability at DMC is controlled by a day-to-day pattern, which strongly points towards a dominating in ﬂ uence of processes occurring at “ synoptic ” scales rather than “ local ” processes. Even ifprevious studies suggested an inverse relationship between OEEs and TCO, we found a slight tendency for the annual frequency of OEEs to be higher when TCO values are higher over DMC. The annual occurrence of OEEs at DMC seems related to the total time spent by air masses over the Antarctic plateau before their arrival to DMC, suggesting the accumulation of photochemically-produced O 3 during the transport, rather than a more e ﬃ cient local production. Moreover, the identi ﬁ cation of recent (i.e., 4-day old) stratospheric in- trusion events by STEFLUX suggested only a minor in ﬂ uence (up to 3% of the period, in November) of “ deep ” events on the variability of near-surface summer O 3 at DMC. The available instruments included a piezoelectric sensor for atmospheric pressure, a capacitor device for relative humidity, a thermo-couple for air temperature, and a heated cup anemometer for wind speed and direction measurements. Meteorological data were recorded at 4 m above the snowpack for wind speed and direction, at 2 m for air temperature and relative humidity, and at 1.5 m for atmospheric pressure. Data were recorded every minute and validated on 1-h interval.


Introduction
Tropospheric ozone (O 3 ) is a greenhouse gas and a driver for atmospheric oxidation capacity (Schultz et al., 2015). It is recognized as a "short-lived climate forcer" (SLCF, see UNEP & WMO, 2011), and its mean global concentration has at least doubled since the pre-industrial age. In contrast to other "well-mixed" greenhouse gases, its short atmospheric residence time, due to the high reactivity, still makes the quantification of O 3 long-term trends a challenging task (Cooper et al., 2014;Monks et al., 2015;Schultz et al., 2015).
Near-surface measurements carried out at remote locations like Antarctica provide the opportunity to investigate O 3 without a significant anthropogenic influence, thus giving information on O 3 background variability. However, O 3 in Antarctica presents some peculiarities that need to be accurately taken into account. Different processes have been recognized to affect O 3 in this region, including: intrusions of stratospheric air masses (Gruzdev & Sitnov, 1993;Roscoe, 2004), transport of air masses from lower latitudes (Murayama et al., 1992), and depletion processes involving reactive halogen atoms (Barrie et al., 1988;Roscoe et al., 2001). Moreover, summer episodes of sudden O 3 increases were observed within Antarctica interior (e.g., Crawford et al., 2001;Legrand et al., 2009) mass transport from the interior of the continent (e.g., Cristofanelli et al., 2011). These events were attributed to the photo-denitrification of the summer snowpack, resulting in NO X emissions to the atmosphere and subsequent photochemical O 3 production (Jones et al., 1999(Jones et al., , 2000. These processes are capable of driving the seasonality of nearsurface O 3 over the Antarctic Plateau (Crawford et al., 2001;Legrand et al., 2009), thus potentially providing a significant input of O 3 to the whole Antarctic region (Legrand et al., 2016). Indeed, as shown in Cristofanelli et al. (2008) and Legrand et al. (2016), due to air mass transport, the photochemically-produced O 3 in the planetary boundary layer (PBL) over the Antarctic plateau can affect the O 3 variability also thousands of km far from the emission area. Previous investigations suggested that the O 3 production in the Antarctic PBL can be affected by several global/regional climate-related changes, i.e.: changes in UV fluxes due to total O 3 variability over Antarctica (Jones and Wolff, 2003;Frey et al., 2015), variability in long-range air mass transport patterns (Legrand et al., 2016), and depth of the continental mixing layer. Recently, Traversi et al. (2014Traversi et al. ( , 2017 suggested that the variability of air mass transport from the stratosphere to the Antarctic plateau can affect nitrate content in the low troposphere and in the snowpack. Besides, the direct transport of air masses enriched in O 3 would represent another process by which stratosphere-to-troposphere transport (STT) events can potentially affect O 3 in the Antarctic troposphere.
In this work, we present the multi-year data series of near-surface O 3 at the Concordia World Meteorological Organization/Global Atmosphere Watch (WMO/GAW) contributing station (GAW ID: DMC) over January 2006-December 2013. The observations were carried out in the framework of different research projects funded by the Italian National Antarctic Programme (PNRA). This near-surface O 3 record was already successfully used to verify the MACC/MACC-2/CAMS-Copernicus global atmospheric composition forecast system (Wagner et al., 2015), and to investigate the variability of atmospheric Hg (Dommergue et al., 2012).
The main purpose of this paper is to provide a systematic assessment of the specific processes which affect near-surface summer O 3 interannual variability over an East Antarctic plateau area: (i) TCO and UV irradiance variability, (ii) synoptic-scale air mass transport within Antarctic interior, and (iii) "deep" STT transport. To reach these objectives, we analyzed the 2008-2013 measurement period, for which the highest data coverage was achieved at DMC and simultaneous irradiance and TCO measurements were available. A better understanding of the role played by these different processes in affecting the regional tropospheric O 3 variability would also help in evaluating the future response of Antarctic tropospheric O 3 to the ongoing global/regional changes. Moreover, since the mixing ratio of O 3 in the surface layer can also affect the chemical composition of the snowpack, a better knowledge of these processes would be very valuable for the interpretation of the ice core records retrieved at DMC (see, e.g., Helsen et al., 2007), which represent a reference for the investigation of past global climate and atmospheric composition variability.

Measurement site
The French-Italian "Concordia" station is located at Dome C (East Antarctic plateau, 75°06′S, 123°20′E, 3280 m a.s.l.); about 1200 km inland from the Ross Sea, and about 1100 km South of the coast of Adelie Land (Fig. 1). The station is part of the WMO/GAW network, and identified by the DMC acronym. In agreement with a previous experimental campaign (Petenko et al., 2004), the January 2006-December 2012 dataset showed that, during summer (November-January), the hourly temperature values ranged between −61.7°C and −18.7°C, while during winter (March-July), with permanent darkness, the temperature varied between −78.9°C and −35.5°C. In contrast to other Antarctic locations, and because of the flat orography and the lack of local drainage flows, DMC is not strongly affected by katabatic winds: historical wind records (Petenko et al., 2004;Argentini et al., 2003) and current observations (Fig. 2) showed a prevalent SW direction (average intensity of about 3 m s −1 , with the highest values around 20 m s −1 ), without a clear annual cycle.

Near-surface O 3 measurements
During the measurement period (2006)(2007)(2008)(2009)(2010)(2011)(2012)(2013), near-surface O 3 measurements were performed by using different UV absorption analyzers (Thermo Tei 49C, Thermo Tei 49i, and Dasibi 1108). The analyzers did not work in parallel, but they were changed and intercompared during the summer maintenance campaigns, usually occurring from December to January. To assure the intercomparability of measurements, before and after the installation at DMC, the O 3 analyzers were intercompared and calibrated against transfer O 3 calibrators (Dasibi 1003, from 2006to 2010, and Thermo Tei49iPS, from 2010, hosted at the WMO/GAW Mt. Cimone global station (Italy, GAW ID: CMN), and referred (2006)(2007)(2008) to the national standard IMGC-O3SRP (hosted at the Italian National Institute for Metrological Research) and to the SRP#15 calibration scale (from 2008 onward), hosted at the WMO/GAW World Calibration Center for surface O 3 , CO and CH 4 at EMPA (see Schultz et al., 2015). For the different analyzers, and over the whole measurement period, the total expanded (coverage factor k = 2) uncertainty has been assessed to be lower than 2 ppb. As a further QA/QC procedure, at DMC, every 24 h the instruments were challenged with a span value of about 100 ppb (obtained by an internal span source) and zero air obtained by flushing ambient air throughout a cartridge of activated charcoal with purafil © .
The O 3 observations were carried out at the "CARO-Clean Air Research Observatory" shelter, located 700 m SW with respect to the main base and upwind to the prevalent wind direction (Fig. 1). Air was collected from a 2 m stainless steel inlet placed on the roof of the shelter, at approximately 5 m above the snow surface. As indicated by Frey et al. (2015), the boundary layer height is lower than 10 m for a notable fraction of time (especially during the night) at DMC during summer. Pietroni et al. (2012), using the temperature profiles measured with a passive microwave radiometer, found a stable layer ranging from 10 to 150 m during winter and summer nights. Thus, it can be possible that, in some cases, our air inlet can be very close to the inversion layer, sampling air masses from the layer above the surface layer. The inlet was made inert to O 3 by passivation with UV exposure before the installation at DMC. A regular sampling flux was guaranteed by a turbo blower, while the sampling room was heated to avoid too low internal temperatures. The near-surface O 3 data were collected at 10-s intervals and stored as 1-and 30-min averages, together with the control parameters provided by the O 3 analyzer. Finally, to prevent and/or point out possible technical problems, the O 3 analyzer and the sampling line were regularly inspected by the personnel working in situ. This permitted to achieve a data coverage of 80% from January 2006 to December 2013. The first two years of observations (January 2006-December 2008) were characterized by a lower data coverage (57%), due to the occurrence of technical problems that prevented the execution of measurements during spring-winter periods. The near-surface O 3 record can be accessed from the WCDCGG web site (http://ds.data.jma. go.jp/gmd/wdcgg/cgi-bin/wdcgg/accessdata.cgi?index=DCC775S00-ISAC&orgn=ISAC&para=O3&select=inventory).

Surface meteorological measurements, total column of ozone (TCO) and UV irradiance
The meteorological variables used in this work were recorded by using an automated weather station (Vaisala, Mod. MILOS 520) located at about 100 m SW from the CARO shelter and managed by the "Osservatorio Meteo-Climatologico Antartico" (www.climantartide.it).
The available instruments included a piezoelectric sensor for atmospheric pressure, a capacitor device for relative humidity, a thermocouple for air temperature, and a heated cup anemometer for wind speed and direction measurements. Meteorological data were recorded at 4 m above the snowpack for wind speed and direction, at 2 m for air temperature and relative humidity, and at 1.5 m for atmospheric pressure. Data were recorded every minute and validated on 1-h interval.
Since 2008, the total column of O 3 (TCO) is obtained from irradiance values observed by a narrowband filter radiometer UV-RAD, operating as part of the Baseline Surface Radiation Network (BSRN, see Ohmura et al., 1998). The UV-RAD is equipped with 7 spectral channels centered at peak wavelengths ranging between 0.300 and 0.364 μm, with each channel being defined by a set of pass-band filters determining an overall bandwidth of about 1 nm and transmittance varying between 12% and 27%. A comparison with OMI measurements showed an agreement on TCO measurements with mean uncertainty around ± 3% (Petkov et al., 2016).

LAGRANTO back-trajectories
To evaluate the synoptic-scale variability of the air masses reaching DMC, a set of 5-days back-trajectories starting at the measurement site was computed every 6 h (at 0:00, 6:00, 12:00 and 18:00 UTC), by using the Lagrangian analysis tool LAGRANTO (Wernli and Davies, 1997;Sprenger and Wernli, 2015). The model calculations were based on the ERA-Interim reanalysis dataset of the ECMWF (Dee et al., 2011). Each set comprised of 20 back-trajectories, with starting points shifted by ± 1°in latitude/longitude and in a vertical range of 50 hPa with respect to real DMC location, to partially compensate for uncertainties related to the absence of subgrid scale processes in LAGRANTO (e.g., convection and turbulent diffusion). The temporal resolution for the back-trajectories is one point every 2 h.

Identification of "deep" STT events
Stratosphere-to-troposphere transport (STT) events are well-known processes, able to influence tropospheric O 3 variability and near-surface O 3 within PBL of high-altitude regions (Škerlak et al., 2014). To assess the possible contribution of STT to near-surface O 3 variability in the DMC region, we identified the measurement periods possibly affected by "deep" STT events (i.e., stratospheric air masses transported down to the lower troposphere), by using the STEFLUX tool (Putero et al., 2016). STEFLUX is an algorithm for detecting the occurrence of STT events over specific target regions, by analyzing a pre-computed global stratosphere-to-troposphere exchange (STE) climatology produced by the IAC-ETH, Zurich (Škerlak et al., 2014). This climatology makes use of the ERA-Interim reanalysis dataset from the ECMWF (Dee et al., 2011), as well as a refined version of the well-developed Lagrangian methodology presented in Wernli and Bourqui (2002) for STE calculation. Basically, from a large set of global trajectories available each day, the ones that cross the tropopause (defined as the surface 2 pvu/ 380 K) within the first 24 h are retained. These are then extended backward and forward in time for additional 4 days, creating a 9-day trajectory. The application of a 3-D labelling algorithm is used to distinguish points of stratospheric and tropospheric nature, as well as potential anomalies (see Škerlak et al., 2014; for more details). STE-FLUX detects the passage of these stratospheric trajectories within a 3-D target box on the region of interest. For this work, we set the top lid of the box at 590 hPa, and the following geographical boundaries: 74-77°S, and 122-125°E. A "deep" STT event at DMC was defined if at least 1 stratospheric trajectory crossed the 3-D target box.
To remove trajectories representing transient exchanges, a two-way minimum residence time criterion is applied in the Škerlak et al. (2014) STE climatology. This guarantees that only "robust" (i.e., not transient) STE events are considered by STEFLUX. A caveat for the Lagrangian approach on which STEFLUX is based is that it relies on trajectories calculated from resolved wind fields, which are characterized by a poor representation of subgrid scale convective and turbulent processes (Fueglistaler et al., 2004). Moreover, the trajectory calculations also do not account for stretching and mixing of the air parcels along the transport. Thus, STT trajectory-based studies provide a useful quantitative information only under the assumption that most of that exchange is associated with the large scale flow rather than subgrid processes. However, several studies (e.g., Trickl et al., 2016, and references therein) were successful (at least for the middle latitudes) in identifying stratospheric intrusions basing on kinematic trajectories, giving a certain confidence that the vertical winds in these analysis and reanalysis datasets are reasonable to diagnose STE.  (2013), with values ranging from 32.6 ppb to 35.0 ppb. This behavior (already pointed out by Legrand et al., 2009Legrand et al., , 2016 clearly reflects the remoteness of Antarctica, with  net O 3 accumulation during the darkness period and destruction when sunlight starts to reach the region, due to photolysis and chemical loss by OH and HO 2 in a low NO X environment (Oltmans, 1981;Monks, 2000). However, in summer (i.e., November-January), episodes of sudden O 3 increase were systematically observed at DMC, with daily O 3 values approaching (or exceeding) those observed in winter. This feature appeared to be similar to what recorded at the other permanent continental station in Antarctica (the "Amundsen-Scott" station at South Pole), and can be attributed to the occurrence of O 3 enhancement events (OEEs) during the period November-January (Oltmans et al., 2008;Helmig et al., 2007). Despite their sporadic occurrence, as shown by the monthly O 3 values at DMC, these events are able to define the shape of seasonal near-surface O 3 cycle with monthly mean values in November and December exceeding 25 ppb (Fig. 4).

Near-surface O 3 variability
To characterize the typical monthly O 3 diurnal variations at DMC, we calculated the hourly deviations from the daily means for the period of investigation (Fig. 5). While no evident diurnal variations were recorded from March to October, a small reverse diurnal variation (∼amplitude 0.7-1.0 ppb) was pointed out from November to February. For these months, the diurnal O 3 minima were observed around 6:00 UTC (i.e., 14:00 local time, LT), while the highest values were recorded around 12:00 UTC (i.e., 20:00 LT). To explain this behavior, near-surface wind speed was used as a proxy to evaluate the development of the convective boundary layer at DMC. Indeed, as reported in Argentini et al. (2005), during summer systematic wind speed diurnal cycles can be related to the presence of strong stability conditions (typically from 18:00 to 6:00 LT) and to the development of convective activity eroding the surface-based inversion and transferring momentum to the surface (typically from 7:00 to 17:00 LT). As reported in 5, the wind speed maximum (4-5 m s −1 ) usually occurred before 14:00 LT, when the O 3 minima were observed from November to February. As also suggested by Legrand et al. (2016) and Mastrantonio et al. (1999), the diurnal development of the convective boundary layer favors the dilution of chemical species related to surface emissions. For near-surface O 3 this partially masks the role of local photochemical O 3 production related to NO X release from snowpack. Based on the observed mixing ratios of NO and NO 2 , and considering the loss reaction with OH, HO 2 and NO, Legrand et al. (2016), estimated a net averaged O 3 photochemical production of 0.3 ppb h −1 during summer light hours at DMC. Our observed O 3 average diurnal increase was 0.16 ppb h −1 from November to February, with significant variability of the 95% confidence interval (Table 1). In particular, the upper boundary of the 95th confidence level of the near-surface O 3 diurnal increase (in December and January) was well comparable with the estimated net photochemical O 3 production rates reported by Legrand et al. (2016) for specific case studies.

Identification of OEEs
The analysis of OEEs was restricted to years 2008-2013, in which simultaneous measurements of UV irradiance and TCO values were available at DMC. Our method to select the days characterized by OEEs is based on a two-steps procedure. The first step consists in fitting the annual cycle of O 3 mean daily values with a sinusoidal curve. This represents an "undisturbed" O 3 annual cycle, not affected by the occurrence of summer O 3 events. In the second step, the probability density function (PDF) of the deviations from the sinusoidal fit is computed, considering all of the daily data. In the Supplementary Material, Fig. S1 (2006)(2007)(2008)(2009)(2010)(2011)(2012)(2013). For each month, the average mean value is reported together with the 95% condense interval.  Fig. S2 reports the PDF of the calculated deviations. Then, a Gaussian fit was applied to the obtained PDF. As reported by Giostra et al. (2011), the Gaussian distribution corresponds to a well-mixed state, given the hypothesis that instrumental errors and natural background variability follow a Gaussian distribution. The deviations from the Gaussian distribution (calculated by the Sigmaplot © statistical tool) would indicate observations affected by non-background variability. To obtain a threshold value for selecting non-background O 3 daily values possibly affected by "anomalous" O 3 enhancements, we calculated a further Gaussian fit for the PDF points falling above 1 σ (standard deviation) of the Gaussian PDF, and we considered the intersection between the two curves as our threshold value (i.e., 6 ppb). In total, 246 days (11.6% of the dataset) were selected as affected by anomalous OEEs: 25.7% in November, 39.3% in December, 27.7% in January and 5.1% in February (Fig. 6, blue bars). Fig. 6 also reports (green bars), for each single month, the averaged occurrence of the anomalous events: the highest fraction of anomalous days was recorded in December, when more than half of the days were affected by the events, and slightly less values characterized November (35.5%) and October (37.1%); a non-negligible fraction of anomalous days was also recorded in February (7.6%).
Despite this clear annual cycle in the frequency of OEEs at DMC, we also observed significant interannual variability both in the O 3 mixing ratio and in the monthly frequency of events. In particular, the monthly frequency of OEEs ranged from 20% to 50%, as a function of the different years (Fig. 7).
To provide a quantification of the relative contribution of diurnal versus day-to-day O 3 variability during the OEEs, we compared the standard deviations of hourly and daily averaged datasets, as suggested by Chevalier et al. (2007). Indeed, the comparison between hourly and daily O 3 variability (expressed as standard deviations) allows to discriminate the relative contribution to the overall O 3 variability due to PBL photochemistry (and other processes occurring at diurnal scales) from processes occurring at longer time scales (i.e., weather variability, synoptic air mass transport, etc.). The variability based on hourly data presented a maximum in December (4.5 ppb) and a minimum in February (1.7 ppb): this clearly underpins the role played by diurnal-scale processes (i.e., PBL photochemistry and dynamics) in enhancing O 3 variability during summer (as suggested before). However, even in December, the ratio between daily and hourly standard deviations suggested that 84% of O 3 variability at DMC can be associated with processes occurring at longer temporal scales. With the purpose of attributing part of this variability, we analyzed several processes that, based on previous investigations, were indicated as likely to trigger the occurrence of summer O 3 enhancements over the Antarctic plateau. In particular, we investigated the possible role of: (i) TCO variability, (ii) synoptic-scale transport variability, and (iii) "deep" stratospheric intrusions events variability. Jones and Wolff (2003) proposed that TCO, and the related UV radiation variability at the surface, would affect near-surface O 3 by enhancing NO X snow emissions and affecting O 3 photochemical production in the South Pole PBL over long time scales. For this reason, and to assess whether TCO variability could have a role in modulating the interannual O 3 summer concentrations at DMC, we analyzed TCO data observed at DMC. For the months on which the occurrence of OEEs was most likely (i.e., December-January), and for each different year (2008)(2009)(2010)(2011)(2012)(2013), Fig. 7 shows the percentiles of hourly O 3 values during the OEEs (a-c), the OEEs frequency (d-f), together with percentiles of hourly TCO values (g-i) for the whole period of data availability. Although this analysis is somewhat qualitative and extended over a limited time period (6 years), we did not find clear anti-correlation between TCO and O 3 values during OEEs. On the other hand, by limiting the dataset to the days in which simultaneous data were available for all parameters (i.e., near-surface O 3 and irradiance measurements), we observed a positive (statistically significant at the 95% confidence level) correlation (r 2 = 0.75) between OEEs frequency and average TCO for November months (Fig. S3 in the Supplementary Material). A tendency for higher monthly occurrence of OEEs and higher monthly TCO can be also noticed for January ( Fig. 7 and Fig. S3 in the Supplementary Material). This not systematic relationship between O 3 levels, OEEs frequency and TCO supports the finding by Frey et al. (2015) that the flux of NO X in the surface atmospheric layer (and thus the related O 3 production) only depends on a secondary order from TCO variability. Frey et al. (2015) pointed out that 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 eclipse and during a shading experiment. However, they also stressed that, due to the presence of reservoir in the open pore space of the upper snowpack, the impact of changes in incident UV irradiance on the snow source, and thus NO X flux and mixing ratios, can be observable on diurnal and seasonal timescales only. To assess whether the variability of UV-A irradiance could affect near-surface O 3 interannual variability throughout impact on NO X flux from snowpack, we analyzed the monthly occurrences of OEEs and related near-surface O 3 values as a function of total dose of UV-A irradiance (315-400 nm) measured at DMC. UV-A has been considered since it covers the wavelength region of − NO 3 absorption (action spectrum maximum at 320 nm). As for TCO, it is difficult to find a clear relationship between monthly integrated UV-A dose and OEEs frequency, or average O 3 values during OEEs. A statistical significant (at the 90% confidence level) positive correlation for OEEs frequency and UV-A dose can be pointed out only for November (r 2 = 0.45), while a negative "tendency" (r 2 = 0.60) can be observed for January (Fig. S4 in the Supplementary Material).

Role of synoptic-scale air mass transport
As suggested by Legrand et al. (2016), the highest near-surface O 3 summer values were observed within air masses spending a significant fraction of time over the highest part of Antarctic plateau during the last 5 days before arriving at DMC. To investigate the possible influence of synoptic-scale air mass circulation to the occurrence of OEEs in November-January, we analyzed the 5-day LAGRANTO back-trajectories (Fig. 8). In particular, we used the potential source contribution function (PSCF) to calculate the conditional probabilities for identifying geographical regions related to the occurrence of OEEs at DMC (e.g. Hopke et al., 1995;Brattich et al., 2017). The residence time over each To obtain a robust PSCF spatial field, only domain cells visited by at least 45 back-trajectories were retained. The results clearly highlight the connection between the occurrence of OEEs at DMC and air masses which traveled over the eastern Antarctic Plateau: PSCF values ranging from 40% to 50% were calculated for this region, while air masses advected from the western Plateau are characterized by very low PSCF values. Interestingly, regions with PSCF values exceeding 50% were pointed out not far from the coastline (around 70°S, 60°E) and over the Ocean (60°S, 145°E). With the purpose of attributing these features, we calculated the altitude (Z) above ground level (a.g.l.) of the back-trajectory points during OEEs. While back-trajectories moving over the eastern Plateau were characterized by Z values mostly below 500 m a.g.l. (which support the role of emissions from snowpack as a source of near-surface O 3 ), the regions with high PSCF values along the coastlines and over the ocean were characterized by relatively high Z (above 3000 m a.g.l.). This would indicate that these air masses traveled at "free tropospheric" altitudes from outside the continent to the Antarctic Plateau. As already suggested by Fiebig et al. (2014), elevated nearsurface O 3 values in central Antarctica can be also related to the transport of free-tropospheric air mixed with aged pollution plumes originated at lower latitudes.
For each month, we computed the total time that air masses spent over the plateau (i.e., the area within the boundaries 90-70°S and 60-150°E) before reaching the measurement site (Fig. 7). Both considering the December-January period, as well as the single months, it is evident that an increase in the time spent by air masses over the plateau leads to an increase of enhancement event frequency at DMC. This correlation (Fig. 9) is statistically significant (at the 95% confidence level) for January (r 2 = 0.62) as well as for the whole December-January period (r 2 = 0.69). In agreement with Legrand et al. (2016), together with the analysis of hourly variability, this result points out that the synoptic-scale air mass transport plays a pivotal role in triggering the occurrence of enhancement near-surface O 3 events at DMC. If we (roughly) assume that the diurnal O 3 increase observed at DMC can be representative for the whole plateau, and if we assume that the air masses travelling over the plateau to DMC picked up "all" the Fig. 7. For each month (November, December, January; each month is represented by a different column) we report the yearly percentiles (barbs and whiskers denote 10th, 25th, median, 75th and 90th percentiles, outliers are indicated as points) of hourly O 3 during the OEEs (a-c), the frequency of OEEs (d-f), the percentiles of hourly TCO (g-i) as well as the total number of hours that LAGRANTO back-trajectories spent over the Antarctic plateau (l-n).
produced O 3 during a 5-day transport, an average O 3 accumulation on the air masses of about ∼5 ppb can be calculated. This value is comparable with the day-to-day variability of near-surface O 3 observed at DMC, thus supporting the hypothesis that snowpack NO X emissions may occur along the plateau on the daily scale. This would confirm that the synoptic-scale transport can favor the photochemical production and the accumulation of O 3 by air masses travelling over this region before their arrival to DMC.
3.5. Role of "deep" stratospheric transport The possible role of air mass transport from the stratosphere in affecting the variability of near-surface O 3 and tropospheric air-chemistry in the Antarctic coastal region was investigated in previous studies (i.e., Murayama et al., 1992;Roscoe, 2004;Mihalikova and Kirkwood, 2013). By using a well-known Lagrangian approach based on the FLEXPART dispersion model, Stohl and Sodemann (2010) assessed that the direct influence of stratospheric air masses over the Antarctic Plateau is only about 2% per year. At the Troll station (72.0°S, 2.5°E, 1275 m a.s.l.), by analyzing in situ data from an atmospheric radar and ECMWF meteorological analysis, Mihalikova and Kirkwood (2013) estimated a significant higher occurrence of tropospheric folds, with a winter maximum (15-26% occurrence rate) and a summer minimum (6-7%). Traversi et al. (2014Traversi et al. ( , 2017, described two possible case studies of stratospheric air mass transport at DMC, related to the advection of air masses linked to tropospheric folds, which occurred along the edge of the steep Antarctic coast, where air masses are orographically lifted up to 400 hPa. To provide a systematic assessment of the possible influence of "deep" STT events to the near-surface O 3 variability at DMC, we used the STEFLUX tool, proposed by Putero et al. (2016) as a relatively fast-computing Lagrangian algorithm to detect the occurrence of stratospheric intrusions over receptor regions in the troposphere, or in the boundary layer. Fig. 10 reports the average monthly frequency of "deep" stratospheric intrusions over the DMC 3-D target box (see Sect. 2.5). Although it is difficult to depict a clear seasonal cycle, due to the low frequency of "deep" STT events, our results are in agreement with Stohl and Sodemann (2010) (i.e., occurrence of STT lower than 2% on a monthly basis, without seasonal cycle). According to our STEFLUX outputs, the highest frequency of "deep" STT events was observed in November (2.8%). The frequency of occurrence of "deep" STT events identified by STEFLUX at DMC is about one order of magnitude lower than the occurrence of OEEs. Thus, a direct link of STT with OEEs interannual variability is unlikely. However, it should be noted that, according to Traversi et al. (2014Traversi et al. ( , 2017, STT can represent a source of (stratospheric) nitrates for the Antarctic atmosphere, thus affecting also nitrate content in the snowpack through different processes (e.g., aerosol dry deposition, uptake of gaseous nitric acid by the uppermost snow layers). Hence, STT could indirectly affect near-surface O 3 variability by representing an input of nitrates that can be recycled by the snowpack and released as NO X under sunlight conditions (Honrath et al., 2000), thus favoring the occurrence of OEEs.

Final remarks and conclusions
Measurements of O 3 mixing ratios carried out at the WMO/GAW Concordia station from 2006 to 2013 allowed to characterize near- Fig. 8. Conditional probability field for the occurrence of OEEs at DMC (A) and average altitude (a.g.l.) of back-trajectories during OEEs (B). Please note that for the conditional probability field only cells visited by at least 45 back-trajectories were retained. Fig. 9. Scatterplot of monthly total time spent by air masses over the plateau before the arrival at DMC versus monthly OEEs frequency for November (triangles), December (stars) and January (circles). surface O 3 variability over the eastern Antarctic Plateau, strengthening the availability of atmospheric composition observations in this remote region.
The systematic comparison of typical monthly diurnal cycles of O 3 with wind speed confirms the presence of a diurnal cycle that can be associated with photochemical production related to emissions of NO X from nitrates in the snowpack and diurnal variability of the convective boundary layer. However, as deduced by the analysis of variability occurring at hourly and daily scales, the latter contribution dominated, suggesting that even during the photochemical "active" summer period, the synoptic transport somehow controls the overall O 3 variability (Neff et al., 2008). Since the occurrence of summer O 3 enhancement events (OEEs) showed significant interannual variability both in the monthly frequency and in the related O 3 mixing ratios, we analyzed a subset of specific processes able to influence them: (i) variability in TCO over the region, (ii) variability of synoptic-scale air-mass transport and (iii) occurrence of "deep" STT. Jones and Wolff (2003) proposed that trends in downwelling UV irradiance due to TCO depletion at the South Pole could drive the summer net production of near-surface O 3 via changes in − JNO 3 and NO X fluxes from the snowpack. In our data we did not find clear evidence for an impact of TCO and the related increase of UV surface fluxes with summer OEEs. This can be partially related also to the different location of DMC with respect to the South Pole. Indeed, the South Pole is characterized by modest but significant katabatic flows, implying a larger "catchment" area for surface air masses before arriving at the measurement site. Conversely, we found a positive tendency for increasing monthly TCO values and increasing occurrence of OEEs. Basing on 6 years of continuous O 3 and TCO observations, these results would support the fact (as already reported in Frey et al., 2015) that TCO is not the main driver for photochemical O 3 production in summer. At the moment, the positive relationship between TCO and OEEs occurrence for November is still not fully understood. We argue that this can be related to dynamical and synoptic-scale air mass circulations, which favor the advection of air masses rich in O 3 to the measurement site, under conditions of enhanced TCO. Indeed, the permanence of air masses over the continental plateau appears to be an important driver for the occurrence of OEEs at DMC. In agreement with Legrand et al. (2016), we found a rather clear relationship between the total time spent by air masses over the plateau and the interannual variability of the monthly occurrence of OEEs.
As deduced by the application of the STEFLUX tool, "deep" STT events only play a marginal role in steering the occurrence of OEEs at DMC via "direct" transport of O 3 from the stratosphere, or from the free troposphere, to the surface. It should be taken into account that, basing on the global STE climatology produced by Škerlak et al. (2014), STEFLUX is only able to consider relatively "young" (i.e., 4-day old) events. This can lead to an underestimation of the real influence of STT at DMC. However, as already suggested by Mihalikova and Kirkwood (2013), this Lagrangian tool based on ERA-Interim data may not have high enough spatial and vertical resolution to resolve the complex transport of stratospheric air masses down to the Antarctic plateau during STT. Moreover, Roscoe (2004) stated that stratospheric "material" transport can also occur via: (i) downward transport across the central Antarctic tropopause, (ii) subsidence in the stratosphere, and (iii) surface level suction to resupply katabatic winds draining the central Antarctic boundary layer. These processes cannot be captured by STEFLUX, and they cannot be reflected by "peak" events in the O 3 observations. Since it was proposed that STT can represent a source of nitrates for the Antarctic snowpack (thus possibly affecting summer NO X re-emission and photochemical O 3 production), it is important to carry out further studies to better assess this process.