Characterization of long-term and seasonal variations of black carbon ( BC ) concentrations at Neumayer , Antarctica

Introduction Conclusions References


Introduction
Among sulfate, mineral dust, and water soluble inorganic compounds (e.g.ionic compounds originating mainly from sea salt aerosol production as well as ammonia and nitrate), carbonaceous aerosols represent one of the major fractions of tropospheric aerosols (Ramanathan et al., 2001;Forster et al., 2007).Aerosol particles influence the global radiation balance directly and can additionally act as condensation nuclei for cloud droplets, which in turn affect climate (Haywood and Boucher, 2000;Hatzianastassiou et al., 2004).Extinction of solar radiation in the visible and near UV range by carbonaceous aerosols is dominated by absorption and not by scattering (Horvath, 1993;Bond and Bergstrom, 2006).Actually, apart from carbonaceous aerosol, only mineral dust shows significant absorption of visible light, caused by variable amounts of haematite (Fe 2 O 3 ) which result in a considerable imaginary part of the complex refractive index in the blue and green spectral range (Clarke and Charlson, 1985;Petzold et al., 2009).
There exist various terminologies for the light-absorbing fraction of the carbonaceous aerosol (Bond and Bergstrom, 2006;Andreae and Gelencsér, 2006).Despite the ambiguous definition, we refer to the term black carbon which is defined as combustion-produced black particulate carbon having a graphitic-like microstructure (Novakov, 1984) because it is widely used in climate research.However, it has to be noted that precisely speaking we refer to equivalent BC because for all applied optical methods the principal measure is light absorption which then is converted into mass of lightabsorbing BC by applying a mass-specific absorption coefficient (MAC); see e.g.Bond and Bergstrom (2006).

R. Weller et al.: Characterization of long-term and seasonal variations of BC
Black carbon is exclusively produced by combustion of fossil fuels or biomass (Cachier, 1995).The remaining and larger fraction of carbonaceous aerosol (about 70-90 %) consists of highly polymerized organic material, so-called organic carbon (OC), which mainly scatters solar radiation comparable to sulfate aerosols (Cachier, 1995).Recent model calculations assessing the influence of aerosols on the global radiation budget have shown that the radiative forcing of carbonaceous aerosols is contrary to all other aerosol species: While the latter, including the weak light absorbing mineral dust, provoke a net cooling and thus mitigate the impact of greenhouse gases, the radiative forcing of carbonaceous aerosols is mainly positive and consequently leads to an amplification of the greenhouse warming (Andreae, 2001;Hatzianastassiou et al., 2004;Hansen et al., 2005;Seinfeld, 2008;Mahowald et al., 2011).Hansen et al. (2005), however, showed that for carbonaceous aerosol the situation is somewhat complicated by the fact, that in global models fossil fuel BC + OC clearly provoke positive forcing, while for biomass burning BC + OC a negative forcing dominates.The reliability of the amount of forcing on a global scale and especially its regional dependence is largely determined by the accuracy of the modelled aerosol distribution which finally relies on the availability of global BC emission and concentration data.While the northern hemisphere is clearly dominated by industrial BC emissions, within the southern hemisphere BC emissions by biomass burning (natural, fire clearing and household) are decisive (Koch et al., 2005).Especially in Polar Regions the positive forcing of absorbing aerosol should be most striking due to inherently high surface albedo caused by snow, sea ice and glacial ice coverage (Warren and Clarke, 1990;Forster et al., 2007).Indeed, there are some indications that BC and its deposition on snow may have substantially contributed to the rapid warming entailed by sea ice loss in the Arctic (Hansen et al., 2005).
Concerning the global distribution of carbonaceous aerosols, there is a major lack of observations at high southern latitudes, most notably Antarctica.This is especially disturbing, considering the recent enormous boost in the mainly ship borne Antarctic tourism activities during austral summer (Shirsat and Graf, 2009;Graf et al., 2010) Bodhaine, 1995), slightly increasing to coastal Antarctic sites (around 1 ng m −3 at Halley, Wolff and Cachier, 1998), but significantly higher at the Antarctic Peninsula (8.3 ng m −3 , Pereira et al., 2006).Neither of these investigations provided a trend estimate for BC concentrations, mainly due to short observation periods.Very recently, a study on global decadal trends of aerosol optical properties revealed no significant long term trend for BC concentrations observed at Neumayer (Collaud Coen et al., 2012).Individual ice core studies showed highly decadal scale BC deposition variabil-ity during the last 150 yr, but an anthropogenic impact could not be assessed (Bisiaux et al., 2012a and b).
In this paper we present an analysis of our BC concentration record from the German Antarctic Station Neumayer (NM), starting in 1999 and comprising now 13 yr of continuous measurements.First of all, this analysis will comprise a reanalysis of the BC long term trend, now including our observations made by a Multi-Angle Absorption Photometer (MAAP) started in March 2006, which was not included in the trend analysis mentioned above (Collaud Coen et al., 2012).In addition we will primarily focus on seasonal aspects and finally on the impact of BC on the bulk optical properties of the aerosol at NM.

Site description
BC measurements were conducted at the Air Chemistry Observatory, Neumayer Station (70 • 39 S, 8 • 15 W, http://www.awi.de/en/go/air chemistry observatory) which participates in the Global Atmosphere Watch (GAW) programme.A detailed description of the measuring site, meteorological conditions, contamination free sampling was already presented and we refer here to Wagenbach et al. (1988), König-Langlo et al. (1998) and Weller et al. (2008).Because in particular BC measurements are highly prone to the impact of local contamination, we will respond to this crucial aspect here again.First of all, the Air Chemistry Observatory is situated in the clean air district approximately 1.5 km south of NM and power supply is provided by cable from the main station.
Due to the fact that northerly wind directions are very rare, contamination from the base can be excluded for most of the time.Furthermore, contamination-free sampling is additionally controlled by the permanently recorded wind velocity, wind direction and by the condensation particle (CP) concentration.Contamination was indicated for each of the following criteria: Wind direction within a 330 • -30 • sector, wind velocity < 2.0 m s −1 and/or CP concentrations (measured by a TSI CPC 3022A particle counter) > 2500 cm −3 during summer, > 800 cm −3 during spring/autumn and > 400 cm −3 during winter.The CP threshold values were chosen based on our more than 25-yr long CP record from Neumayer, demonstrating that CP concentrations above the corresponding levels can usually be traced back to local pollution.In fact, only during such conditions, BC concentrations frequently showed suspicious spikes exceeding 100 ng m −3 by far.In the end, about 2 % of data loss was actually caused by potential contamination according to the criteria mentioned previously.

Aerosol absorption photometer
Black carbon measurements with aerosol absorption photometers are based on the optical attenuation method described in general by Hansen et al. (1982Hansen et al. ( , 1984)).In short, BC is collected by drawing ambient air through a filter.The transmission of light through the loaded filter is related to the transmission through a clean part of the filter material.Knowledge of the specific BC attenuation cross section on the filter in use (Q BC ) finally allows the calculation of BC concentrations.This approach requires that the main aerosol species absorbing visible light at the considered wavelength is actually BC.In fact only haematite (Fe 2 O 3 ), naturally occurring as a minor compound in mineral dust, has a significant, but two orders of magnitude lower absorption cross section in the visible compared to BC (Clarke and Charlson, 1985).Recent studies on the wavelength-dependence of the light absorption by mineral dust demonstrated that crosssensitivities of light absorption methods for BC measurement are relevant for the wavelength of 550 nm and shorter while they can be almost neglected for the red spectral region (Petzold et al., 2011).Considering the generally low mineral dust concentrations at NM, typically in the lower ng m −3 range (Weller et al., 2008) and the operational wavelengths of the applied instruments (Aethalometer: around 800 nm; MAAP: 637 nm), this possible interference was neglected throughout this work.At Neumayer BC concentrations were continuously monitored by a white light Aethalometer (Magee Scientific, type AE10) from 1999 through 2009.Apart from several problems concerning the relation of the measured absorption signal to actual BC concentrations (e.g.Weingartner et al., 2003;Collaud Coen et al., 2010), this instrument is wellestablished and was used worldwide for many years, including pristine Antarctic sites like South Pole (Bodhaine, 1995) and Halley (Wolff and Cachier, 1998).The AE10 accumulated the aerosol on a circular spot with an area determined to be 107 ± 2 mm 2 .We employed high volume filters (Pallflex, type T60A20; material: Pure borosilicate glass fibers with moisture-resistant tetrafluoroethylene coating) with a flow rate of 2.37 m 3 h −1 .The detection limit (DL), i.e. three times the standard deviation (std) of the noise level was derived from the output signals when the Aethalometer was operated with zero air (ambient air passed through two quartz filters in series or one HEPA filter).Referring to the original temporal resolution (sampling period t) of 4 h the DL was around 5 ng m −3 which corresponds to a DL of around 10 ng m −3 for 1 h resolution (assuming a 1/ √ t dependence, supported by the fact that the noise signal was normally distributed).This raw data set was finally combined into daily means and medians.The manufacturer notes that under normal performance the instrument should be able to detect approximately 1 ng BC, but recommend a more conservative value of 5 ng BC.This would result in a nominal DL of 0.6 ng m −3 BC for a 4 h sampling interval, which appeared highly optimistic as already stated before by Nyeki et al. (1998).In addition it was found that the sensitivity of the method decreased when the BC loading on the filter becomes high.We found that the noise level slightly increased when the sample filter was preloaded with black carbon, e.g. a test loading of 1850 ng to 3350 ng BC deteriorated the detection limit by a factor of 1.1 to 1.4.To prevent an appreciable loss in sensitivity, we changed the filters if the optical attenuation -ln(I/I 0 ) (where I and I 0 are the transmitted light intensities for an exposed and a blank portion of the filter) exceeded a value 0.22, corresponding to a BC quantity of 1570 ng.More problematic is the impact of increasing light scattering with increasing aerosol load of the filter, biasing the results to the higher end.Up to now there exists no generally accepted procedure to correct this interference (Collaud Coen et al., 2010;Müller et al., 2011).The reduction of optical path length in the filter with increasing load was considered by the so-called Weingartner correction (Weingartner et al., 2003;Müller et al., 2011) and a more recent correction by Collaud Coen et al. (2010;equation 13 therein).Both corrections depend on the single scattering albedo of the aerosol ω 0 , which was derived from nephelometer data combined with MAAP results (see below) in order to avoid circular reasoning.It turns out that these corrections were well below 1 % throughout and thus negligible in case of the NM measurements.Due to the fact that such a procedure was impossible before March 2006 (start of the MAAP measurements) we finally refrain from correcting the AE10 data.
Since March 2006 ongoing a more sophisticated MAAP (Model 5012, Thermo Electron Corp.) is in operation, especially circumventing successfully the stray light problem (Petzold and Schönlinner, 2004;Petzold et al., 2005).Aerosol was sampled on a glass filter tape (Schleicher and Schüll, type GF 10) at an inlet air flow of 1.0 m 3 h −1 .The filter tape was automatically moved forward when the transmission exceeded 20 %.The measured absorption refers to a wavelength of 637 nm (Müller et al., 2011).Raw data were originally sampled in one minute resolution.Finally hourly, as well as daily means and medians were extracted.Hourly averaged MAAP data went rarely to negative values (12.8 % of the data), whereas for the AE10, even the 4-h samples fluctuated around zero (26.9 % of the data).A similar behavior was observed during side-by-side operations of an Aethalometer and a MAAP at the Jungfraujoch observatory (Petzold and Schönlinner, 2004).These fluctuations indicate BC values close to or below the DL.Consequently, we generally did not remove negative values because such a procedure would bias the data to higher values.Daily median MAAP data never showed negative values, while still about 8 % of the AE10 values were below zero (range between −1.0 and 0 ng m −3 ).Thus all further data evaluation was based on daily medians.We found a MAAP detection limit of around 6.3 ng m −3 BC for hourly resolution, which appeared somewhat better than the DL of < 20 ng m −3 for 30 min means reported in the original paper (Petzold and Schönlinner, 2004).Extending the averaging period to 24 h with a flow of 1.0 m 3 h −1 would result in a reduced DL of 0.25 ng m −3 .
The accuracy of BC measurements by aerosol absorption photometers depends on the validity of the specific BC attenuation cross section (Q BC ) which is dependent on the used filter material.Q BC is 14 m 2 g −1 for the AE10 operated with Pallfex filters (provided by the manufacturer; note that a factor of 19 m 2 g −1 was recommended for often used quartz filters.Both values include filter artefacts).According to Müller et al. (2011) a Q BC of 6.6 m 2 g −1 was used for the MAAP.However, it should be mentioned that values in the range from 5 m 2 g −1 up to 20 m 2 g −1 have been found in ambient samples (Liousse et al., 1993;Petzold et al., 1997).Apart from this, it has to be stated very clearly that Q BC is different from MAC for which a recent review recommend a value of 7.5 ± 1.2 m 2 g −1 at a wavelength of 550 nm (Bond and Bergstrom, 2006).According to Bodhaine et al. (1995) we suppose that the method dependent Q BC vary proportional to physically defined MAC so the ratio might be expected to be fairly constant.Both instruments operated at NM were not absolutely calibrated on-site, hence a stringent accuracy designation could not be given.As a reasonable assumption we refer to the DL as a crude measure for accuracy.For daily averaged BC concentration, the accuracy for the AE10 and MAAP would thus correspond to ± 2.1 ng m −3 and ± 1.3 ng m −3 , respectively.We emphasize that this estimate includes uncertainties concerning light transmission, flow rate, and BC deposition area on the filter but not uncertainties arising from the actual specific absorption coefficients Q BC and MAC.Due to the inferior quality of the AE10 data, we considered the MAAP results as more reliable and focus for the main part of our evaluation and reasoning on this data set.Nevertheless we are confident that also the AE10 data represent a consistent and valuable BC time series.
The AE10 and the MAAP were both connected to the central inlet duct of the observatory, a ventilated electro-polished stainless steel inlet stack (total height about 8 m above the snow surface) with a 50 % aerodynamic cut-off diameter around 7-10 µm at wind velocities between 4-10 m s −1 (determined with an aerodynamic particle sizer TSI, type APS 3321).From the stack to the MAAP and AE10 we used silicone conductive tubes each with a length of around 50 cm, hence the inlet characteristic of both instruments were virtually identical.
Apart from the contamination screening procedure an additional data selection was necessary for the AE10: During stormy weather conditions (typically at wind velocities above 20 m s −1 ), sometimes extraordinarily noisy signal output occurred, possibly caused by sea salt deposition on the filter.In such cases spikes in the original data set typically exceeding some 100 ng m −3 were immediately followed (i.e. in the sub-sequent measuring interval) by comparably large spikes in the opposite direction.Such cases could be easily identified and were treated as fake signals and discarded throughout, because an evaluation was not meaningful.

Auxiliary measurements
In order to assess the impact of BC on the optical properties of the aerosol, BC concentrations have to be converted to particle absorption coefficients σ ap (λ) (unit: 1 Mm −1 = 10 6 m −1 ) by using the specific BC absorption on the filter material (Q BC ).Note that although light attenuation caused by aerosol deposition on the filter was primarily measured, the recorded output signals finally correspond to BC concentrations internally calculated by using the respecting Q BC .This part of the evaluation was restricted to the MAAP data, because the operating wavelength and spectral sensitivity of the AE10 detection optic is ill defined (Weingartner et al., 2003).For the MAAP, we relied on the operative wavelength of 637 nm.In addition to BC data, we employed particle scattering coefficients measured with an integrating three wavelength nephelometer (TSI, model 3563), in operation since 2001, and snow surface albedo measurements R s , started in 1982.A thorough discussion of these time series is a matter of its own and beyond the scope of this paper.Hence we merely present the time series of these parameters in the Supplement (Figs.S.1 to S.3).
Operation of the nephelometer and data evaluation have previously been described in detail by Weller and Lampert (2008).In short, particle scattering σ sp (λ) and hemispheric back-scattering σ bsp (λ) coefficients were originally determined in 10 min averages at the wavelengths 450 nm, 550 nm, and 700 nm.After correction for non-Lambertian illumination and truncation errors (according to Anderson and Ogren, 1998) hourly averages were extracted and used for further evaluation.The hemispheric back-scattering fraction b λ was calculated according to:  sky conditions (Pirazzini, 2004;Wuttke et al., 2006).Such impacts are obviated by using daily means of radiation data and additionally excluding all global radiation values below a daily mean of 10 W m −2 .

Data presentation
Figure 1a shows the BC time series measured by the AE10, while in Fig. 1b the results from the MAAP are presented.Individual data points refer to daily medians, which could be described by a log-normal distribution (see Supplement S.4). Figure 2  (2.60 ng m −3 ) and 2.34 ng m −3 (2.77 ng m −3 ) for the overlapping period.
A linear reduced major axis (RMA) regression of BC aeth vs. BC MAAP (daily medians) resulted in a slope of 0.82 ± 0.03 with an intercept of 0.86 ± 0.1 (r 2 = 0.19, n = 928; Fig. 3).Note that a RMA regression based on weekly medians resulted in a comparable poor correlation (slope = 0.96 ± 0.07, intercept = −0.53± 0.36, r 2 = 0.26), potentially related to the fact that both instruments operate close to their lower limit of detection which caused very noisy data.Due to the poor correlation between both data sets, a reliable homogenization emerged unsuccessful and we preferred a separate trend analysis on each time series.A nonparametric rank-order Mann Kendall test with Sen's slope trend analysis (Hirsch et al., 1982) revealed that for the individual segments measured by the MAAP or AE10 no significant overall trend could be detected, in accordance with the trend analysis already presented for the AE10 time series (Collaud Coen et al., 2012; note that therein a statistically significant but marginal negative trend of −0.01 ng m −3 yr −1 was found and finally discarded due to high percentage of negative values).A monthly resolved analysis showed significant (p< 0.05) but opposite trends for the months June and August (−0.51ng m −3 yr −1 and + 0.14 ng m −3 yr −1 , respectively) for the MAAP data.As for the AE10 time series, slight negative trends for the months November and December (−0.09 ng m −3 yr −1 and −0.11 ng m −3 yr −1 , respectively) could be detected.Apart from the obvious discrepancy of both data sets and the weak correlation between them, the mean seasonality was in agreement (Fig. 4), indicating that the AE10 data reliably described the annual BC cycle at NM.With both instruments a primary annual maximum between October and November and a minimum in April was visible with a maximum/minimum ratio of 4.5/1.6 = 3.8 and 2.7/0.64 = 4.2 for the MAAP and AE10 data, respectively.closer inspection revealed that BC concentrations measured by the MAAP appeared also relatively high for the months January and February a feature which was not reflected by the AE10 data (Fig. 4).It seems that this double maximum was not an artifact provoked by the averaging process but was also present during the individual austral summers 2007/2008, 2008/2009and 2009/2010 (Fig. 2) (Fig. 2).

Long term trend
A discussion of trend and seasonality requires that the specific BC attenuation cross section (Q BC ) was all about the same throughout the measuring period.This assumption ap-peared reasonable considering the fact that BC from biomass burning is well aged (long range transport time to coastal Antarctica is typically > 5 days according to Stohl and Sodemann, 2010) and the physico-chemical state of BC aerosol should be consistent.Indeed, Liousse et al. (1993) and Petzold et al. (1997) found Q BC to be constant in a given atmospheric environment.Furthermore we argue that the impact of local BC was generally negligible, except from some few sporadic events (see below).This conclusion was supported by the annual cycle of the observed BC concentrations showing a broad maximum around October/November (Fig. 4).In contrast, local BC pollution at NM should have a striking maximum starting in the middle of December through the end of February.Supply and removal activities of the station happened within this time frame each year, associated with heavy freight traffic.Nevertheless there are some indications pointing to a minor but not negligible impact of local contamination.These are occasional secondary BC maxima in the MAAP data around January and apparently high BC concentrations observed from November 2007 through February 2009 (Figs. 1 and 2), the period when the new construction of Neumayer III Station was afoot.
Although ice core records indicate a growing BC deposition in Antarctica (Bisiaux et al., 2012a and b), our results revealed a virtually constant long-term atmospheric BC level around 2 ng m −3 for the MAAP and 1.55 ng m −3 for the AE10 time series.Up to now no atmospheric BC trend analyses from other Antarctic sites are available, but BC concentrations at NM were comparable to previous atmospheric measurements from Halley (AE10 record, 1992-1995) with a mean around 1 ng m −3 , but definitely lower than the annual mean BC concentrations of 8.3 ng m −3 observed with an AE9 in 1993, 1997, and 1998 on King George Island (Antarctic Peninsula) (Table 1).Chaubey et al. (2010) reported outstanding high BC concentrations during austral summer 2008/2009 from the coastal sites Maitri (75 ± 33 ng m −3 ) and Larsemann Hills (13 ± 4 ng m −3 ) measured with a Magee Scientific type AE41 aethalometer.At least for Maitri, a considerable impact of local pollution by human activities has to be considered (Chaubey et al., 2010).We conclude that BC concentrations between 1 ng m −3 and 2 ng m −3 can be regarded representative for coastal Antarctica except the northern parts of the Antarctic Peninsula (King George Island).Lowest BC concentrations (0.65 ng m −3 measured with a Magee Scientific aethalometer, type not specified but most probably an AE10; Bodhaine 1995) were found at South Pole, a site which may be representative for the Antarctic Plateau. Figure 5 may serve as simplified survey of this geographical BC concentration gradient.Based on our data, there is up to now no conclusive evidence of significantly enhanced atmospheric BC pollution caused by increasing shipping activities (mostly tourism) in the South Atlantic peaking between December and March (Shirsat and Graf, 2009).Note, however, that Antarctic tourism is up to now mainly focused on the marine   1.
realm of the Antarctic Peninsula and long range transport from this region to NM seems to be less efficient.

Seasonality and dependence on general weather situation
Black carbon concentrations at NM showed a broad annual maximum around October/November, consistent in relative amplitude and phase with measurements conducted at Halley (Wolff and Cachier, 1998), South Pole (Bodhaine 1995), and Antarctic Peninsula (Pereira et al., 2006).Biomass burning in Africa, South America and Australia is the main BC source in the southern hemisphere (van der Werf et al., 2006 and2008).Given the fact that large scale meridional transport was mainly provoked by cyclonic activity (Pereira et al., 2006;Fiebig et al., 2009;Hara et al., 2010), we expect higher BC concentrations at NM during passing of low pressure systems which was generally characterized by strong easterly winds at NM.In fact BC concentrations were significantly higher (2.85 ng m −3 ) during storms (wind velocity > 15 m s −1 and wind direction between 0 • and 120 • ) compared to the annual median (2.14 ng m −3 ) or to 1.64 ng m −3 ).Another potential transport mechanism, in particular during polar night, could be subsidence of middle and upper tropospheric air above central Antarctica associated with horizontal advection by katabatic wind to coastal Antarctica (James, 1989;Van de Berg et al., 2007).Actually, we could not find any indications that southerly winds (indicative for downslope or katabatic flow, König-Langlo et al., 1998) were associated with enhanced BC concentrations at Neumayer (the same is true for mineral dust tracers) and the seasonality of the mentioned tracers is in conflict with air mass transport from the Antarctic Plateau by catabatic outflow.Results from Pereira et al. (2006) and Fiebig et al. (2009) indicate that South America is the main BC source region for the Weddell Sea and Dronning Maud Land region.There, biomass burning typically peaks in late winter/early spring (September-October), i.e. around 1-2 months earlier compared to BC concentration maxima typically observed in Antarctica (van der Werf et al., 2006 and2008).The dominant biomass burning region in South America is Amazonia, extending to northern Argentina around 45 • S (van der Werf et al., 2006; see also http://rapidfire.sci.gsfc.nasa.gov/firemaps/).Apart from BC, the dominant source region for R. Weller et al.: Characterization of long-term and seasonal variations of BC mineral dust aerosol entry in Antarctica is also South America (Smith et al., 2003).Figure 6 presents the mean annual cycle of BC in combination with the rare earth element La (a tracer for mineral dust; data from Weller et al. 2008) and a further terrigenous tracer, 210 Pb (data from Elsässer et al., 2011) all observed at NM.Given that for these three tracers South America was the main source region, a comparison of their annual cycle observed at NM may give some insight on the interplay of source variability (including vertical exchange processes with the free troposphere), large scale meridional transport efficiency (including depositional loss en route), and boundary layer inversion strength at NM (affecting downward mixing) in determining the concentrations finally measured at NM.For 210 Pb and the mineral dust tracer La a broad summer maximum was evident, slightly shifted to January/February for La (Fig. 6).The main seasonal maximum of BC was about 3 months early.For biomass burning a strong seasonality and patchy, sporadic occurrence is evident (van der Werf et al., 2006 and2008).While also mineral dust sources are generally patchy and sporadic, their seasonality in South America is somewhat ambiguous, though a broad maximum between October and March is discernible (Gaiero et al., 2003).Atmospheric mixing height should also be at maximum during summer supporting vertical exchange of air masses.Deposition efficiency should be larger for the supramicrometer mineral dust (Sanak et al., 1981) compared to BC, a typical sub-micrometer aerosol species (Seinfeld and Pandis, 2006a;Adachi and Buseck, 2008).As for 210 Pb, a decay product of the noble gas 222 Rn which is continuously emitted from soil, the annual cycle of the continental source strength is relatively weak.Note that at NM, oceanic emissions peaking after the retreat of the sea ice in late summer might add to the local atmospheric 222 Rn inventory during that time (Elsässer et al., 2011).The effect on the 210 Pb seasonality is, however, not clear yet.Long range transport of 210 Pb and maybe also of BC should be more efficient compared to mineral dust due to the smaller atmospheric residence time of the latter.Large scale meridional transport towards Antarctica, most probably via the free troposphere should be most favorable after decay of the polar vortex in summer (Genthon, 1992;Krinner and Genthon, 2003;Stohl and Sodemann, 2010).This is also the time when surface inversion strength at NM was at minimum (Elsässer et al., 2011), facilitating down mixing from the free troposphere into the planetary boundary layer.Therefore the seasonality of virtually all processes relevant for mineral dust generation and transport as well as 210 Pb transport to NM were in phase and consistent with the observed annual maximum (Weller et al., 2008;Elsässer et al., 2011).On the other hand, the annual BC cycle observed at NM was presumably shaped by a convolution of the more pronounced and earlier biomass burning maximum in September/October and the seasonal maximum of large scale meridional transport efficiency during austral summer.The impact of biomass burning seasonality appeared considerable if not dominant, hence observed BC concentrations apparently reflect essentially the variability of biomass burning in South America.

Impact on aerosol optical properties
In order to assess the impact of BC on optical aerosol properties at NM, we calculated the single scattering albedo (ω 550 ) as well as the critical single scattering albedo (ω crit ) for the wavelength λ = 550 nm (Seinfeld and Pandis, 2006b): (2) The critical single scattering albedo ω crit provides a crude estimate whether aerosol forcing was negative (ω crit < ω 550 ) or positive (ω crit > ω 550 ) (Seinfeld and Pandis, 2006b).To this end we made use of aerosol light scattering coefficient, hemispheric backscattering fraction b 550 at 550 nm, and surface albedo (R s ) values from our continuous nephelometer and radiation measurements, respectively.BC concentrations reported by the MAAP were re-converted to aerosol absorption coefficients at 550 nm (σ ap (550)) using the firmware coefficient of 6.6 m 2 g −1 at λ = 637 nm and extrapolated to λ = 550 nm accepting a λ −1 dependence of σ ap (λ) (Horvath, 1993;Schnaiter et al., 2003).Note that for R s only broadband values (305 nm to 2800 nm) were available.Spectral snow surface albedo measurements at NM indicate a maximum close to 1.0 at 500 nm, decreasing to around 0.95 at 300 nm and between 0.45 and 0.75 at 1000 nm (Wuttke et al., 2006).As already mentioned in Section 2.2.1, the time series of the respective parameters σ sp (550), σ bsp (550), and R s are presented in the Supplement (Figs.S.1 to S.3).
In short, the long-term mean values and standard deviations were: σ sp (550) = 4.37 ± 5.8 Mm −1 (range: 0.1 Mm −1 to 59 Mm −1 ), σ bsp (550) = 0.62 ± 0.71 Mm −1 (range: 0.1 Mm −1 to 15.4 Mm −1 ), and R s = 0.857 ± 0.053 (range: 0.51 to 1.0).In addition, a trend analysis of aerosol optical properties by nephelometer measurements at NM can be found in Collaud Coen et al. (2012).The calculated daily mean values for ω 550 and ω crit are presented in Fig. 7.In addition Fig. 8 may serve as an overview for the dependence of ω crit on R s and hemispheric backscattering fraction b 550 values typical for NM as well as the range of the derived ω 550 data.Based on this simple estimate, in most of the cases a net heating effect (positive forcing) of boundary layer aerosol at NM (ω 550 < ω crit in 69.5 % of the cases) was evident.Nevertheless, even under the highest measured σ ap (550) (around 0.15 Mm −1 , for daily medians) merely 0.015 % of the incident solar radiation at 550 nm would be absorbed in an atmospheric layer with an optical path length of 1 km.We conclude that any heating effect within the PBL was eventually negligible.

Conclusions
In light of the dramatically increasing human activities in Antarctica during the last few decades, the nearly constant annual median BC concentrations observed at NM from 1999 through 2011 appeared somewhat peculiar.Apart from the sporadic impact of local pollution, growing combustion derived emissions from ship borne activities in the Southern Ocean are, up to now, only of exiguous relevance for BC concentrations at NM. Large scale meridional transport of biomass-burning derived black carbon, preferentially from South America, seems to determine the BC burden in Antarctica and caused a distinct and consistent spring / early summer concentration maximum.Nevertheless, without much doubt anthropogenic combustion derived emissions will perpetually grow around and within continental Antarctica.Hence there is a demand to continue observations with a view to document the impact of such activities on the Antarctic environment.In addition to atmospheric BC measurements such observations should especially include the critical role of BC deposition and its impact on surface albedo in Antarctica.Apart from the contribution of absorbing aerosol on the radiation balance of the Antarctic atmosphere at present, BC profiles measured in Antarctic ice cores could elucidate the potential increase since pre-industrial era and thus clearly constrain the respective human effect.Moreover, this effort might reveal the biomass burning history in ancient times including its relevance on climate forcing.

Fig. 1 .
Fig. 1.BC time series, based on daily medians, measured by the Magee Scientific AE10 Aethalometer (a) and by the MAAP (b).Vertical grey lines mark the turn of the year.
Figure1ashows the BC time series measured by the AE10, while in Fig.1bthe results from the MAAP are presented.Individual data points refer to daily medians, which could be described by a log-normal distribution (see Supplement S.4).Figure2shows the monthly means of both time series together.Obviously the AE10 data with an overall median (mean) of 1.60 ng m −3 (2.20 ng m −3 ) for the whole period (1999 through 2009) and 1.55 ng m −3 (2.23 ng m −3 ) regarding the overlapping period with the MAAP (March 2006 through 2009) appeared somewhat lower compared to the MAAP data with a long-term median of 2.14 ng m −3 (2.60 ng m −3 ) and 2.34 ng m −3 (2.77 ng m −3 ) for the overlapping period.A linear reduced major axis (RMA) regression of BC aeth vs. BC MAAP (daily medians) resulted in a slope of 0.82 ± 0.03 with an intercept of 0.86 ± 0.1 (r 2 = 0.19, n = 928; Fig.3).Note that a RMA regression based on weekly medians resulted in a comparable poor correlation (slope = 0.96 ± 0.07, intercept = −0.53± 0.36, r 2 = 0.26), potentially related to the fact that both instruments operate close to their lower limit of detection which caused very noisy data.

Fig. 2 .Fig. 3 .
Fig. 2. Complete BC time series based on monthly means.The black and red line represents the data measured by the AE10 and the MAAP, respectively.

Fig. 4 .
Fig. 4. Box plots of the mean seasonality of BC concentrations based on daily medians: Results measured by the AE10 for the years 1999-2009 (a) and by the MAAP for the period 2006-2011.Lines in the middle of the boxes represent sample medians (mean: red line), lower and upper lines of the boxes are the 25th and 75th percentiles, whiskers indicate the 10th and 90th percentiles, dots 5th and 95th percentiles.The graph comprises 18 months starting from January to show the summer maximum without cease.

Fig. 5 .
Fig. 5. Spatial BC concentration gradient in Antarctica.Only sites from which annual mean values were available are shown; references can be found in Table1.

Fig. 6 .
Fig. 6.Seasonality (monthly means) of typical terrestrial tracers measured at NM: BC (grey line, measured by the MAAP), the rare earth element La (red line), a tracer for mineral dust, and 210 Pb (blue line) a progeny of the terrigenous noble gas 222 Rn.The graph comprises 18 months starting from January to show the seasonal maxima without cease.

Fig. 7 .Fig. 8 .
Fig. 7. Time series of the measured single scattering albedo at 550 nm, ω 550 (black circles) along with the calculated critical single scattering albedo ω crit (red circles).Note that ω crit values were not available between end of April and early August of each year, because then solar radiation was below 10 W m −2 and R s could not be reliably determined.
. Previous measurements indicate a geographical gradient of atmospheric BC concentrations from South Pole to Southern Ocean with minimum BC concentrations at South Pole (around 0.65 ng m −3 ,

Table 1 .
BC concentration measured at different Antarctic sites.