Several aspects have to be considered in discussing causes of the surprising decrease of RMSA in December/January compared to values in October and March at Concordia compared to what is observed at the coast. They mainly include segregation between MSA and sulfate during transport towards DC via either formation or deposition of the two species. In addition, as proposed to explain the loss of MSA from the snowpack, we cannot exclude the possibility of an evaporative loss from aerosol.

Here we may first invoke a different size distribution of nssSO4 and MSA leading to a change of their respective abundance during transport between the ocean and central Antarctica. Indeed, several studies (Rankin and Wolff, 2003; Kerminen et al., 2000) pointed out an enrichment of MSA with respect to nssSO4 in supermicron compared to submicron particles at coastal Antarctic sites in summer. In this way, the decrease of RMSA at Concordia in December and January could be explained by (1) a larger abundance of MSA with respect to nssSO4 in large than in small particles at the coast and (2) a transport of marine air mass between the coast and Concordia becoming less efficient in December and January compared to October and March. An alternative possibility involves a selective formation of sulfate with respect to MSA under mid-summer conditions in inland Antarctica. As discussed by Davis et al. (1998), the RMSA ratio in Antarctica is strongly influenced by the respective importance of liquid- and gas-phase sulfur chemistry. Briefly, the OH oxidation of DMS produces SO2 (abstraction pathway), DMSO, and SO2 (addition pathway). DMSO is further oxidized by OH either in the gas phase or in the aqueous or aerosol phase. Legrand et al. (2001) reported that the heterogeneous DMSO oxidation efficiently produces MSA in summer at the coast. The efficiency of this heterogeneous process has been confirmed by kinetic studies (Bardouki et al., 2002). Therefore, due to a larger presence of aerosol, it is expected that the oxidation of DMSO would produce much more MSA in the atmospheric boundary layer compared to the buffer layer above. That was supported by field observations made by Davis et al. (1998) at the Palmer site showing a rapid increase of DMSO when vertical downward transport brought buffer layer air mass within the boundary layer. These observations imply a longer lifetime of DMSO in the buffer layer than in the boundary layer, following a strong weakening of the heterogeneous reaction of DMSO caused by a far lower aerosol surface (and liquid water) available there. From that, Preunkert et al. (2008) proposed that the weakening of marine advection in December–January compared to March associated with an ongoing oxidation of SO2 into sulfate in the buffer layer, where the heterogeneous chemistry of DMSO is very limited, would account for the drop of RMSA observed in mid-summer at Concordia. Note, however, that this hypothesis was based on observations that were limited to 1 year.

The previously mentioned enrichment of MSA with respect to nssSO4 seen in supermicron compared to submicron particles at the coast in summer is also observed at DDU, a coastal site located in margin regions facing the Indian and Pacific oceanic sectors from which most of the marine air mass reaching Concordia comes. As shown in Fig. 4, the second supermicron mode of MSA in aerosol at DDU remains centered around 1–2 µm and differs from the coarse sea-salt aerosol mode observed at 6–8 µm by Jourdain and Legrand (2001, 2002). The contribution of the second mode to the total MSA mass remains rather weak (less than 18 %). At Concordia in summer, the size distribution is very similar to the one observed at DDU with no significant decreasing contribution of the supermicron particles to total MSA mass between the two sites (Fig. 4). Furthermore, as seen in Fig. 5, the size distributions of the two sulfur species typically observed at Concordia indicate no significant decreasing contribution of the supermicron mode to the total MSA mass in summer compared to other seasons (16 % in winter, 18 % in spring and summer, and 10 % in fall). That suggests an absence of selective deposition of MSA during transport between the coast and the inland Antarctic plateau in mid-summer. The weak contribution of supermicron particles to the total MSA mass, as observed at the coast, likely limits the segregation between MSA and nssSO4 during the transport between ocean and Concordia. It is also interesting to notice that the dominant presence of MSA in submicron sulfuric acid particles does not confirm the statement generally presented in previous studies that, after its formation in the gas phase, in contrast to sulfuric acid, MSA is more easily incorporated in larger, less acidic particles (Jefferson et al., 1998). Since DMSO is soluble in acidic solution, we may in fact expect that its solubilization in hydrated acidic aerosol followed by a rapid oxidation into MSA explains the presence of this latter in sulfuric acid particles.

Figure 6 shows that the large drop of RMSA observed on bulk aerosol in summer (Fig. 2, see also Fig. 8 in Sect. 3.2.3.) is consistently revealed by impactor data (open triangles in Fig. 6). It also shows a far more pronounced decrease of RMSA in submicron than micron particles. An example of this strong depletion of MSA relative to nssSO4 in submicron compared to micron particles in summer is seen in Fig. 5c. As a consequence, since the contribution of submicron particles dominates the total mass of MSA, the drop RMSA values in mid-summer at Concordia is mainly due to a drop of RMSA in submicron particles. A drop of RMSA may result from an increase of sulfate and/or a drop of MSA. Impactor data corresponding to the March–November time period (Fig. 7a and b) show that RMSA is very poorly related to the nssSO4 content (R2 of 0.01 and 0.06 for submicron and micron particles, respectively). Conversely, the higher is the MSA content the higher is RMSA (R2 of 0.55 and 0.47 for submicron and micron particles, respectively). In fact, as seen in Fig. 7c and d, when distinguishing samples with a high and low nssSO4 content (higher and lower than 100 ng m-3), a strong relationship is found between MSA and RMSA. Note that most of samples containing less than 100 ng m-3 of nssSO4 correspond to the November–December period, whereas those containing more than 100 ng m-3 of nssSO4 to the January–March period. In this way, the drop of RMSA from November to December and from February–March to January is examined separately. In both cases, the RMSA drop is mainly due to a decrease of MSA. For instance, in submicron particles R2 equal to 0.86 is calculated for samples containing less than 100 ng m-3 of nssSO4 (0.84 for those having more than 100 ng m-3 of nssSO4). The same is seen for supermicron particles (R2 equal to 0.89 and 0.83 for samples containing less and more than 100 ng m-3 of nssSO4, respectively).

The year-round record of RMSA (HV filter data) was examined at the light of different parameters related to the history of air mass present during sampling at Concordia (Fig. 8). This was done by using 10-day backward trajectory as well as chemical characteristics (i.e., sodium, MSA, nssSO4, and ozone) and air temperature at the site. We restrict the discussion to the last 5 years, for which the chemical records are the most continuous (Fig. 8). Sodium is here used to evaluate the importance of marine advection from spring to fall and its interannual variability. As discussed by Legrand et al. (2009), ozone at Concordia exhibits a seasonal cycle characterized by a maximum in July followed by a decrease until October and the occurrence of a secondary maximum in November–January (Legrand et al., 2009), similar to what is observed at the South Pole (Crawford et al., 2001). Whereas it is expected that such very remote regions experience winter accumulation of O3 transported from other regions followed by photochemical destruction in spring and summer, the occurrence of a secondary maximum in November–January is surprising. That was attributed to a photochemical ozone production induced by the high NOx levels generated by the photo-denitrification of the Antarctic snowpack (Davis et al., 2001).

The sodium record reported in Fig. 8f does not support the assumption that the drop of RMSA values coincides with a weakening of marine air advection. Indeed, whereas there are in general much more sodium (late October–early November) just before the drop of RMSA (see the grey areas in Fig. 8c), the recovery of high RMSA values in fall (February–March) is never accompanied by a recovery of sodium levels. For example, RMSA jumped from < 0.10 mid-February 2014 to 0.40 at the end of February 2014, whereas the sodium levels remained over the whole period between 1 and 3 ng m-3. Consistently with what is observed on impactor (Fig. 7), HV data indicate that the drop of RMSA in mid-summer at Concordia is related to a decrease of MSA rather than an increase of nssSO4 levels (Fig. 8d). The lack of a recovery of marine air advection in fall and the fact that low RMSA values are more related to low MSA rather than high nssSO4 levels do not support the preceding assumption of a chemistry favoring sulfate during transport in mid-summer as resulting from a weakening of marine advection and a chemistry of DMSO promoting formation of sulfate in the absence of heterogeneous chemistry. Note also that the temperature record (Fig. 8e) does not support the possibility of a drop of MSA caused by evaporative loss from aerosol.

The fact that the drop of RMSA values in mid-summer at Concordia is mainly due to the disappearance of MSA in the fine aerosol and not to an increase of sulfate permits us to reject the assumptions of (1) a selective deposition of MSA with respect to sulfate or (2) a preferential production of sulfate with respect to MSA during transport between the coast and the inland Antarctic plateau. As seen in Fig. 8, the only significant change that coincides fairly well with the drop of RMSA in mid-summer is the occurrence of the secondary maximum of ozone mixing ratio that is attributed to a local photochemical activity driven by NOx emissions from the snowpack of the Antarctic plateau. To illustrate the timing and the amplitude of the photochemical ozone production we have reported in Fig. 8a the de-seasonalized ozone record. It is seen that the sudden appearance of low RMSA values that generally occurred at the beginning of November and ended in February (see the grey areas in Fig. 8) coincides with the periods over which the excess of ozone related to the strong photochemical activity took place. The link between low RMSA values and high photochemical activity is also seen in the interannual variability, with particularly low RMSA values in November–December 2011 and 2012 (∼ 0.03) compared to November–December 2013 and 2014 (0.08) corresponding to larger excess ozone (10.5 ppbv in November–December 2011 and 2012 versus 8.0 ppbv in November–December 2013 and 2014). As discussed by Legrand et al. (2016), the more time spent by the air mass above 3200 m elevation prior to its arrival at Concordia, the higher the ozone mixing ratio was. That can be seen in Fig. 8b when comparing the fraction of time spent by the air mass above 3200 m a.s.l. in November–December 2011 and 2012 with November–December 2013 and 2014 (6 days instead of 4–5 days).

The preceding observations of a drop of RMSA driven by a decrease of MSA level in submicron particles around the beginning of November and its recovery in February, simultaneous with the high photochemical activity at mid-summer at Concordia, suggest the occurrence of a (photo)chemical destruction of MSA taking place in submicron particles at that time. Under conditions encountered in the marine atmosphere, in contrast to its fast heterogeneous production, a significant (but slow) in cloud destruction of MSA is suspected to take place (Von Glasow and Crutzen, 2004; Barnes et al., 2006; Hoffmann et al., 2016). After its production, MSA present in air masses traveling over inland Antarctica will encounter more oxidative conditions, especially when air masses remained for a few days over the high plateau, and is thus characterized by high ozone mixing ratio (see discussions in Legrand et al., 2016). Assuming both the fact that aerosol particles spend about 3 h per day as cloud droplets and an aqueous-phase OH (OHaq) concentration of 6×10-13 M, Zhu et al. (2005) calculated a mean lifetime of MSA of 14 days in the marine boundary layer. Clearly, conditions encountered over the Antarctic plateau are very different and it is out of the scope of this paper to identify the involved chemical mechanisms leading to a destruction of MSA in submicron sulfuric acid particles over central Antarctica. We can note, however, that over the Antarctic plateau, while the chance of aerosol experiencing aqueous-phase chemistry in cloud droplets is far lower than in the marine boundary layer, the production of OHaq from the reaction of ozone with O2- would be favored compared to conditions encountered in the marine boundary layer due to far more acidic conditions (Ervens et al., 2003). Note also that the dissolution of H2O2 that also contributes to the budget of OHaq would be 2 orders of magnitude higher at Antarctic temperatures than at temperatures encountered in the marine boundary layer.

Apart from marine biogenic emissions, sulfate present over Antarctica can also originate from southern hemispheric continents or the stratospheric reservoir. 210Pb data permit one to derive an estimate of the contribution of sulfate long-range transported from continents by comparing the 210Pb concentrations at Concordia (27 µBq m-3) after having corrected them from marine 222Rn exhalation (∼ 15 %; Weller et al., 2014) (i.e., 23 µBq m-3) with those observed at Chacaltaya (407 µBq m-3; Feely et al., 1988), a remote site located at 5220 m a.s.l. in Bolivia. At this site, a typical sulfate level of 250 ng m-3 can be assumed (see Minikin et al., 1998, and references therein). Note that this value is similar to the one reported by Huebert and Lazrus (1980) for the free troposphere (240 ng m-3, at 5–6 km elevation over the Pacific ocean). Assuming a sulfate concentration of 250 ng m-3 for the continental free troposphere of the southern hemisphere, and applying a dilution factor of 18 based on 210Pb data (407 µBq m-3 at Chacaltaya compared to 23 µBq m-3 at Concordia), we calculate a mean sulfate concentration of 14 ng m-3. The 210Pb activities at Concordia (Fig. 2) show an enhancement by a factor of 2.5 from June–September to November–February, leading to an estimated concentration of continental sulfate of 8 and 20 ng m-3, respectively. These values may be overestimated since an influence of the city of La Paz (3600 m a.s.l., located only at 25 km away from the Chacaltaya site) on the sulfate concentration remains here possible. Note also that other continents, such as Australia, certainly contribute to the long-range transport of continental 210Pb and sulfate towards Antarctica, in particular in the case of East Antarctica (Heimann et al., 1990).

For the first time, 10Be concentrations are documented in the atmosphere of the high East Antarctic plateau (Fig. 9). The seasonal cycle characterized by a winter minimum and a January–February maximum is similar to what was observed at the coastal site of Neumayer by Elsässer et al. (2011) but with a far stronger winter–summer amplitude (a factor of 10 at Concordia instead of 2–3 at NM). This difference is likely reflecting (1) the particularly strong inversion layer in winter at inland Antarctica and (2) higher summer concentrations at 3200 m a.s.l. than at the sea level. From the observation of a mean winter concentration of 10Be of 0.6×104 atoms m-3 at Concordia (Fig. 9) and considering the 10Be concentration of 107 and 0.5×107 atoms m-3 observed between 11 and 19 km elevation at 65∘ N by Raisbeck et al. (1981) and Field et al. (2006), respectively, we derive a dilution factor in the range of 800–1700 between the lower stratosphere and the atmosphere at Concordia in winter. Lazrus et al. (1979) measured worldwide background (non-volcanic) sulfate mixing ratios from 0.1 ppbm at 11 km elevation to 0.5 ppbm at 19 km elevation. Considering a mean sulfate mixing ratio of 0.3 ppbm for the lower stratosphere, and the typical dilution factor observed for 10Be between the lower stratosphere and the atmosphere at Concordia, we estimate that stratospheric–tropospheric exchange may account for 0.4 ng m-3 of sulfate in winter at Concordia. As shown in Table 3, similarly to what was previously shown for NM (Wagenbach, 1996), an increase of 7Be / 210Pb and 10Be / 7Be ratios from winter to summer is seen at Concordia, suggesting a 2 times stronger downward transport from the stratosphere in summer than in winter there.

Beryllium data do not account for sulfate transported downward from the lower stratosphere during sedimentation of polar stratospheric clouds that occurs without stratospheric–tropospheric air mass exchange. Measurements of 35S that offer the possibility to estimate stratospheric input of sulfur were done on aerosol collected at Concordia, showing that as upper estimates 3.3 ng m-3 of sulfate in winter and 16.5 ng m-3 in summer/fall come from the lower stratosphere (Hill-Falkenthal et al., 2013).

The preceding discussions have shown how uncertain estimates of the contribution of non-biogenic source of sulfate over the Antarctic plateau remain based on radionuclide data, except for the downward transport from the stratosphere as traced back using beryllium (10Be and 7Be).

The initial motivation to conduct simultaneous measurements of MSA and sulfate in Antarctic air or snow was to separate the marine biogenic source of non-sea-salt sulfate from others like volcanic emissions, anthropogenic sources, and terrestrial sources. It was expected that the examination of the relationship between nssSO4 and MSA would help. However, several previous studies pointed out that the quantification of the non-biogenic sulfate sources by examination of the y intercept of the relationship between nssSO4 and MSA is complicated by the fact that RMSA is seasonally dependent and varies with the MSA concentration range; see for instance Legrand and Pasteur (1998) and Piel et al. (2006). Figure 10 illustrates how poor the correlation between the two sulfur species is at Concordia ([nssSO4] = 6.0 [MSA] + 14 with R2=0.33; Table 4). As recommended by Ayers (2001), we here used a bivariate regression (so called reduced major axis regression). As clearly shown by Fig. 10a, the scattering of the correlation is largely due to summer samples (R2=0.37 for summer samples as shown with red points). Since, as previously discussed, a destruction of MSA takes place in summer at Concordia, leading to unusually low RMSA values (even lower than in winter), we have scrutinized the correlation only considering data corresponding to the rest of the year (i.e., considering only black points shown in Fig. 10a). For these samples, which cover the spring to fall time period, a far better correlation is observed ([nssSO4] = 3.5 [MSA] + 7 with R2=0.80, Table 4). However, the significant change of RMSA from spring to fall still leads to an overestimation of non-biogenic sulfate when considering the y intercept of the linear regression line. In the following, we therefore scrutinize separately spring, fall, and winter data (red triangles, black triangles, and blue circles in Fig. 11, respectively).

In winter, both nssSO4 and MSA exhibit low concentrations (Fig. 2). Nevertheless, as suggested by increases of 210Pb, the isolation of the high Antarctic plateau from the free Antarctic troposphere sometimes breaks down, often resulting from arrival of warm air associated with a marine intrusion. As seen in Fig. 12, under these conditions, the increase of 210Pb is accompanied by an increase of sulfate, suggesting that the lower troposphere at Concordia was temporary filled with winter free-tropospheric air. Interestingly, Fig. 2 shows that the highest mean winter levels of nssSO4 occurred in 2011 (8.0 ± 3.9 ng m-3) and 2013 (8.1 ± 5.6 ng m-3), when the 210Pb levels were also the highest of the record (22–24 µBq m-3). Conversely, the lowest nssSO4 mean winter level is observed in 2008 when the level of 210Pb was particularly low (11 µBq m-3). In the following we examine the origin of sulfate present in the free-tropospheric winter atmosphere. As shown in Fig. 11c, the slope of the linear relationship between nssSO4 and MSA ([nssSO4] = 15.2 [MSA] + 0 with R2=0.65, Table 4) corresponds to a RMSA value of 0.066 ± 0.004. Such a RMSA value below 0.10 in winter is typically observed in the remote marine boundary layer at low- to mid-southern latitudes in winter (from April to September): 0.077 at 40∘ S (Cap Grim; Ayers et al., 1991), 0.037 at 29∘ S (Norfolk; Saltzman et al., 1986), and 0.026 at 22∘ S (New Caledonia; Saltzman et al., 1986). The y intercept of the linear relationship (0 ± 1 ng m-3; Table 4) suggests that in winter, when marine biogenic emissions are located far away from the Antarctic continent, the contribution of non-biogenic source to nssSO4 remains at the best limited to around 1 ng m-3 (for a total nssSO4 concentration of 6 ± 4 ng m-3).

In spring, the linear relationship between nssSO4 and MSA ([nssSO4] = 5.1 [MSA] + 9 with R2=0.89; Table 4) suggests an increase of RMSA compared to winter (0.20 instead of 0.07 in winter). This increase of RMSA from winter to spring likely reflects the enhanced contribution of DMS emissions from marine area located south of 50∘ S as shown by the increase of oceanic chlorophyll (Fig. 2). In contrast to winter, the linear relationship between nssSO4 and MSA observed in spring suggests the existence of a non-biogenic sulfate source accounting for a few ng m-3. The non-linearity of the relation between nssSO4 and MSA tends, however, to overestimate the y intercept (Fig. 11a); a y intercept of around 3 ng m-3 is obtained when samples with less than 5 ng m-3 of MSA are considered.

In fall, the relationship between nssSO4 and MSA becomes even less linear than in spring, rendering rather inaccurate the use of the y intercept in evaluating the non-biogenic sulfate contribution (Fig. 11b). However, Fig. 11b suggests that, if significant, the non-biogenic sulfate concentration remains well below 5 ng m-3 at that time. The slope of the relationship between nssSO4 and MSA in fall (2.9 ± 0.2, Table 4) indicates a further increase of RMSA compared to spring (0.34 instead of 0.20 in spring). Comparing November and March, Fig. 2 indicates that a decrease of oceanic chlorophyll concentrations have started between 50 and 60∘ S whereas those at latitudes higher than 60∘ S are maintained or slightly higher in March than in November. That can explain the higher RMSA values in March compared to November. Note that such a higher RMSA in March compared to November is also observed at coastal sites (Fig. 3).

To minimize the uncertainties related to the variability of RMSA between the different samples, we also examined the relationship between non-sea-salt sulfate and MSA concentrations on the 12 stages of impactor run. This approach reduces the uncertainty linked to variability of RMSA value over time. A few impactor runs show a y intercept that was significantly different from zero, reaching 1 ng m-3 in winter and 1 to 4 ng m-3 at other seasons.

It can therefore be concluded that, whatever the season, marine biogenic emissions of DMS dominate the atmospheric budget of sulfate over inland Antarctica. If it exists, the contribution of non-biogenic sulfate source remains limited to 1 ng m-3 in winter and possibly reaches a few ng m-3 particular from spring to fall. Previous discussions on radionuclide data gained at Concordia suggest the long-range transported sulfate from continent in spring and/or summer and downward transport from the lower stratosphere particularly in summer and/or fall as non-biogenic sources of sulfate.