Interactive comment on “ Reactive bromine in the low troposphere of Antarctica . Estimations at two research sites

This manuscript gives an overview of the retrieval of BrO vertical profiles and column densities from ground-based MAX-DOAS measurements at 2 locations on the Antarctic coast. They examine the impacts of meteorology, aerosols, and ozone on the observed BrO, as well as doing some simple modeling to determine ozone lifetimes due to BrOx at each site. They find differences in BrO between the two sites and suggest these differences could be linked to differing sea ice condions between the two sites. They also find that BrO enhancements occur under low wind speed conditions and that blowing snow is not needed for substantial enhancements, but rather that surface emissions and vertical mixing can be responsible for the observed enhancements. This work is

Abstract. For decades, reactive halogen species (RHSs) have been the subject of detailed scientific research due to their influence on the oxidizing capacity of the atmosphere and on the climate. From the RHSs, those containing bromine are of particular interest in the polar troposphere as a result of their link to ozone-depletion events (ODEs) and to the perturbation of the cycle of toxic mercury, for example. Given its remoteness and related limited accessibility compared to the Arctic region, the RHSs in the Antarctic troposphere are still poorly characterized. This work presents ground-based observations of tropospheric BrO from two different Antarctic locations: Marambio Base (64 • 13 S, 56 • 37 W) and Belgrano II Base (77 • 52 S, 34 • 7 W) during the sunlit period of 2015. By means of MAX-DOAS (Multi-axis Differential Optical Absorption Spectroscopy) measurements of BrO performed from the two research sites, the seasonal variation in this reactive trace gas is described along with its vertical and geographical distribution in the Antarctic environment. Results show an overall vertical profile of BrO mixing ratio decreasing with altitude, with a median value of 1.6 pmol mol −1 in the lowest layers of the troposphere. Additionally, observations show that the polar sunrise triggers a geographical heterogeneous increase in bromine content in the Antarctic troposphere yielding a maximum BrO at Marambio (26 pmol mol −1 ), amounting to 3-fold the values observed at Belgrano at dawn. Data presented herein are combined with previous studies and ancillary data to update and expand our knowledge of the geographical and vertical distribution of BrO in the Antarctic troposphere, revealing Marambio as one of the locations with the highest BrO reported so far in Antarctica. Furthermore, the observations gathered during 2015 serve as a proxy to investigate the budget of reactive bromine (BrO x = Br + BrO) and the brominemediated ozone loss rate in the Antarctic troposphere.

Introduction
The importance of the halogens (X = Cl, Br, I) in atmospheric chemistry and climate became clear decades ago after observations of these substances were made in the stratosphere and also in the troposphere (e.g., Molina and Rowland, 1974;Farman et al., 1985;Barrie et al., 1988;Oltmans et al., 1989;Fan and Jacob, 1992;Hausmann and Platt, 1994;Solomon, 1999). Indeed, reactive halogen species (RHSs) are of special interest in the troposphere for limiting the lifetime of species such ozone (O 3 ), mercury (Hg), dimethyl sulfide (DMS) and organic compounds; for affecting the partitioning of NO x (NO + NO 2 ) and HO x (OH + HO 2 ); and, in the case of iodine, for participating in new aerosol formation. As such, the presence and the impact of the tropospheric halogen chemistry have been the subject of numerous studies with focus on remote regions and on environments under anthropogenic influence (e.g., Simpson et al., 2015 and references therein). Particular attention has been paid by the scientific community to the role of halogens in the polar regions. Although not unique to these regions, it is in the polar areas where bromine becomes particularly relevant, directing the oxidizing capacity of the atmosphere during springtime and causing ozone-and mercury-depletion events (ODEs and AMDEs, respectively). For details on sources, sinks and historical background, the reader is kindly referred to the compendium works of Simpson et al. (2007Simpson et al. ( , 2015, Steffen et al. (2008), and Ariya et al. (2015) and references therein, for example.
Briefly, while the presence of reactive bromine in the global pristine troposphere is primarily due to the photolysis and oxidation of very-short-lived bromocarbons emitted from the oceans (e.g., Carpenter et al., 2014), in the polar regions its dominant source is of inorganic origin and is linked to heterogeneous chemistry. Through experimental and modeling studies, it is known that a set of heterogeneous reactions based on acidic substrates comprising hypobromous acid (HOBr) and bromide (Br − ) take place in sea ice, open leads, brine, frost flowers, snowpacks or sea-salt aerosols, for example (e.g., Fan and Jacob, 1992;Vogt et al., 1996;Platt and Lehrer, 1997;Abbatt et al., 2012;Pratt et al., 2013, Toyota et al., 2014Simpson et al., 2015, Thompson et al., 2015Custard et al., 2017;Wang and Pratt, 2017). These reactions are summarized as Br 2 Cl − aq ↔ (Br 2 ) aq + (Cl − )aq, yielding the possibility that molecular bromine (Br 2 ) transforms from the aqueous (aq) to the gas phase. When this is followed by the photolysis of Br 2 into two bromine atoms, an autocatalytic release of bromine is triggered, resulting in an exponential buildup of reactive bromine BrO x (Br + BrO) in the troposphere and the so-called "bromine explosion" events (e.g., Fan and Jacob, 1992;Platt and Lehrer, 1997;Wennberg, 1999;Simpson et al., 2015). These events were first observed in the Arctic region by correlating the detection of filterable bromine with ODEs (e.g., Barrie et al., 1988) and, later on, by observing a very high amount of BrO (tens of picomoles per mole) in the boundary layer just after the polar sunrise (e.g., Hausmann and Platt, 1994;Tuckermann et al., 1997). Since then, several studies have tried to determine the chemical sources, sinks and pathways of these compounds (e.g., Simpson et al., 2007Simpson et al., , 2015. In particular, the main BrO source reactions involve In pristine environments (i.e., very low nitrogen oxide), along with photodissociation (in polar spring J BrO ∼ 3 × 10 −2 s −1 ; e.g., Thompson et al., 2015), the BrO sink reactions associated with the catalytic ODEs are → Br + Cl, BrO + HO 2 → HOBr, where Reaction 10 (R10) represents the main channel for the abovementioned bromine explosions causing ODEs (e.g., Bottenheim et al., 1986;Barrie et al., 1988;Oltmans et al., 1989;Platt and Hönninger, 2003;Simpson al., 2007), where the ozone loss rate is limited by the BrO itself and cross reactions (R8) and (R9) and estimated as (e.g, Hausmann and Platt, 1994;Le Bras and Platt, 1995;Platt and Jenssen, 1995;Platt and Lehrer, 1997 Overall, the attempts from the scientific community to estimate the presence of BrO x in the Antarctic troposphere were initiated 20 years ago with ground-based DOAS measurements of BrO from Arrival Heights (77.8 • S, 166.7 • E), observations compatible with the presence of 30 pmol mol −1 in the first 2 km of the troposphere (Kreher et al., 1997). Due to the complexity of performing measurements in such a hostile and remote environment, very few ground-based scientific works have followed that study (summarized in Table 1). In addition to the sparse ground-based measurements, the presence of tropospheric BrO in the Antarctic region has also C. Prados-Roman et al.: Reactive bromine in the low troposphere of Antarctica 8551 been addressed through satellite observations (e.g., Wagner and Platt, 1998;Wagner et al., 2001;Richter et al., 2002;Theys et al., 2011), shipborne measurements (e.g., Wagner et al., 2007) and, more recently, by airborne DOAS measurements (e.g., Hüneke et al., 2017). In spite of the elapsed years and the efforts of the scientific community, compared to its northern counterpart, the current characterization of BrO x in the Antarctic troposphere is very poor given the very scarce geographical coverage available with vertical information. Moreover, most of the observations are campaign based in random years and hence the time coverage is also quite limited (see Table 1). The present work aims at improving this geographical, vertical and time coverage by adding two Antarctic sites to those few observing BrO in the Antarctic troposphere. These observations were made by endurable, stable and sensitive DOAS instrumentation built specifically for long-term measurements in hostile environments. Particularly, herein we present the observations performed during 2015 from two stations. The measurement sites and methodologies are introduced in Sect. 2. Section 3 puts forward the results obtained in terms of time series of BrO along with time series of the aerosol optical thickness, near-surface O 3 and meteorology parameters. Then it deepens into the details of the vertical information gained after these observations and assesses the budget and distribution of inorganic reactive bromine in the troposphere of Antarctica. Section 4 summarizes the work.

Observations from two Antarctic stations
During 2015, ground-based spectroscopic measurements were performed from two Antarctic research stations: Marambio and Belgrano II. Details of the measurement sites and methods are provided below, along with ancillary observations.

Measurement sites
In 2010, in collaboration with the National Antarctic Direction of Argentina/Argentinian Antarctic Institute (DNA/IAA), the National Institute for Aerospace Technology (INTA) deployed a MAX-DOAS (Multi-axis Differential Optical Absorption Spectroscopy; e.g., Platt and Stutz, 2008) , 2018). Note that, given their location around the Weddell Sea, long-term trace gas observations from these Antarctic sites provide a great opportunity for investigating the troposphere-sea ice interactions in two different scenarios: a station surrounded by seasonal sea ice (Marambio) and another where the perennial (edged) sea ice dominates (Belgrano). Figure 1 shows the locations at which INTA has instrumentation deployed in Antarctica and it shows the sea ice concentration (Spreen et al., 2008) surrounding these stations by the end of the austral summer and by mid-winter of the year 2015, which was the first year that both instruments operated simultaneously.

Measuring method
The spectral measurement technique used for the observations presented in this work was MAX-DOAS, gathering UV/VIS scattered skylight in the sunlit atmosphere. Through this technique, tropospheric vertical profiles of aerosol extinction coefficients (AECs) and BrO volume mixing ratios (vmr's) can be inferred. Specific details of the instruments' deployment and the spectral analysis and inversion scheme are provided in the following.

MAX-DOAS instruments
Although on a few occasions tropospheric BrO has been measured in remote regions with in situ techniques (e.g., chemical ionization mass spectrometry; Liao et al., 2011), the operational activities in remote and hostile environments render the DOAS (differential optical absorption spectroscopy) technique a very suitable approach given its sensitivity, versatility and instrumental endurance (e.g., Platt and Stutz, 2008). Either with active setups (e.g., long-path DOAS) or with passive ones (e.g., zenith DOAS, MAX-DOAS, satellite observations), the DOAS technique has been used broadly to research the troposphere in remote environments (e.g., Wagner and Platt, 1998;Bobrowsky et al., 2003;Wagner et al., 2007;Saiz-Lopez et al., 2007a, b;Puentedura et al., 2012;Prados-Roman et al., 2015;Peterson et al., 2017). In particular, the MAX-DOAS instrumental setup referred to in this work consists of a telescope scanning the atmosphere at different elevation angles, inferring with it vertically resolved information of the status of the atmosphere regarding aerosols and trace gases (e.g., Hönninger et al., 2004;Wagner et al., 2004). This is indeed an advantage that the MAX-DOAS configuration offers over the standard setup of the long-path DOAS, for example, and also over in situ instruments (e.g., chemical ionization mass spectrometry, CIMS), whose information is commonly limited to the instrument's altitude. Also, the MAX-DOAS specific ability to characterize the low troposphere overcomes the often limited sensitivity of the satellite observations to the planetary boundary layer.    3, 5, 10, 15, 30, 60, 90 1, 2, 3, 5, 10, 20, 30, 90 The hardware and software of the MAX-DOAS instruments referred to in this work were developed by INTA. For decades, the group has been investigating the atmosphere from different sites of the world using the DOAS technique, particularly from polar regions (e.g., Gil et al., 1996Gil et al., , 2008Yela et al., 2017). The two MAX-DOAS instruments referred to in this study consist of an outdoor unit with a temperate pointing system developed and built at INTA (Fig. 2), comprising a stepper motor and a telescope with an 8 cm focal length fused silica lens yielding a field of view of 1 • . The sunlight is focused in a quartz fibre bundle, which is directed into the indoor unit comprising a temperaturestabilized Czerny-Turner monochromator and a CCD camera fully developed by INTA based on a Hamamatsu S7031-1008 sensor, kept at −40 • C ± 0.05 • C with a temperature control developed and built at INTA. Both instruments operate in off-axis mode scanning the atmosphere from the horizon to the zenith every 15 min while the solar zenith angle (SZA) is lower than 85 • . For a SZA higher than 85 • , the telescopes are fixed at zenith position and during the polar night no measurements are performed (i.e., April-August at Marambio and March-September at Belgrano). Further details of the MAX-DOAS instruments installed at Antarctica are provided in Table 2.

Spectral analysis and vertical profile inversion
The spectral analysis of the DOAS observations shown in this work was performed with INTA's software LANA (e.g., Gil et al., 2008;Peters et al., 2017). The retrieval of BrO was centered in the 335-358 nm spectral range, including the absorption cross sections of BrO (Fleischmann et al., 2004), O 4 (Thalman and Volkamer, 2013), CH 2 O (Meller and Moortgat, 2000), OClO (Kromminga et al., 2003), NO 2 (Vandaele et al., 1998), O 3 (Bogumil et al., 2003) and of a pseudo-Ring spectra (Chance and Spurr, 1997), along with a fifth-degree closure term and constant intensity offset. In order to decrease possible instrumental instabilities and to minimize the influence of stratospheric trace gases in the retrieval, the zenith spectrum from each scan was used as a reference. Moreover, only data gathered in off-axis mode with a SZA < 75 • were used in this work.
Similarly, the O 4 differential slant column densities (dSCDs) were also retrieved (337-370 nm spectral window) in order to invert the vertical profile of the AEC and therefore to characterize the scattering properties of the atmosphere and the light path of the photons reaching the detector. The reliability of the aerosol vertical information retrieved by MAX-DOAS observations has already been demonstrated under different visibility conditions (e.g., Frieß et al., 2016). This retrieval is based on the concept that the concentration O 4 is known and stable in the atmosphere. Hence, a variation in the O 4 dSCDs is usually related to a change of the optical path, generally due to the presence of aerosols (e.g., Hönninger et al., 2004;Wagner et al., 2004). The inverted AEC vertical profile was then used as input for the linear inversion of the vertical profile of the BrO vmr. In this work, the inversion of the vertical profiles of the AEC and the BrO concentration was based on the optimal estimation method (e.g., Rodgers, 2000). The radiative transfer model (RTM) used was LIDORT (Spurr, 2008) and the inversion scheme was BePRO (BIRA, Clémer et al., 2010). The procedure consisted of a two-step approach. First, the AECs were retrieved from the observed O 4 dSCDs through an iterative nonlinear process (e.g., Hendrick et al., 2014;Córdoba-Jabonero et al., 2016). The inferred AEC was then used to invert the targeted BrO profiles.
The RTM input parameters characterizing the profile retrievals were carefully chosen for polar conditions and always bearing in mind that the aim is to gain long-term observations in the hostile conditions of Antarctica. For the measurements performed from the Marambio station, the pressure (P ), temperature (T ), O 3 and NO 2 vertical profiles were obtained from the standard atmosphere for subarctic latitudes (Anderson et al., 1986). For the observations made from Belgrano, the considered P , T and O 3 profiles were obtained from monthly averaged available ozonesonde records (from 1999 to 2006, e.g., Parrondo et al., 2014), while the NO 2 profile was taken from the same standard atmosphere. The modeled atmosphere was stratified into layers of 100 m from 0 to 4 km altitude, layers of 1 km from 4 to 6 km (wider grid related to less sensibility at those altitudes) and layers of the same width of those of the standard atmosphere above this altitude. The retrieved profiles were obtained up to an altitude of 6 km. In the inversion scheme, the diagonal elements of the measurement uncertainty covariance matrix were the square of the dSCD error after the DOAS fit (1σ ). The statistics of the DOAS fit are included in the Supplement. The diagonal elements of the a priori covariance matrix in the inversion were calculated as 100 % of the a priori profile for BrO and based on Clémer et al. (2010) for the aerosol extinction retrieval. Note that, given the large variability in the visibility conditions at Antarctica, the true aerosol profile can strongly differ from the a priori profile. In order to allow the extinction profile retrieval to capture these variations, we followed the method described in Clémer et al. (2010). In this method, the diagonal element of S a closest to the surface (i.e., S a (1,1)) is set equal to the square of a scaling factor (β) times the maximum partial aerosol optical depth (AOD) of the extinction profile obtained in the precedent iteration. In this study, β has been set to 1. The other diagonal elements decrease linearly with altitude down to 20 % of S a (1,1). Note that, despite not being a statistically Bayesian method, it allows the profiles corresponding to large AOD to differ significantly from the a priori AOD profile, while the profiles with smaller AOD present lower variations from the a priori AOD profile. The non-diagonal elements were calculated following a Gaussian distribution with a correlation length of 100 m for aerosols and 300 m for trace gases (e.g., Hendrick et al., 2004). Therefore, the error of the retrieved profiles provided in this work contains the measurement error (experimental dSCD error) and the smoothing error of the retrieval. The last takes into account that the retrieval is an estimate of the true profile smoothed by the averaging kernel functions.
Given that the measurements in Antarctic stations are frequently affected by blowing snow, the aerosol optical properties were obtained using Henyey-Greenstein phase functions for the single-scattering albedo SSA = 0.999982 and asymmetry parameter g = 0.89, corresponding to typical values of clean ice crystal (e.g., Frieß et al., 2011). After several tests considering typical snow albedos (between 0.8 and 0.9), the surface albedo was set to 0.8. Note that this value is consis-tent with observations of the sea ice albedo performed in the UVA spectral range in the Arctic region (e.g., Ehrlich, 2009). In order to avoid unrealistic values, an upper limit for AOD was set to 0.5, therefore neglecting all the observations made in such complicated conditions from the light scattering point of view. Also, only retrievals with degrees of freedom higher than 1 were taken into consideration. The degrees of freedom of the data set presented in this work are summarized in the Supplement (Table S1).
After the AEC vertical profile was estimated at 360 nm by means of the measured O 4 dSCDs, the aimed BrO vmr vertical profiles were retrieved at 338 nm using the calculated AEC as input of the RTM. In order to properly include the inferred AEC in the retrieval of the BrO vmr profiles, they were calculated at the corresponding wavelength using an Ångström parameter of 2.2 (e.g., Hegg et al., 2010). An exponential decreasing profile corresponding to AOD = 0.02 was chosen as the a priori AEC vertical profile for the AEC vertical profile. The scale height of the a priori AEC was set to 0.5 km for Belgrano and to 2 km for Marambio since these values provided the lower differences between observed and modeled O 4 dSCD for each station. The a priori BrO vertical density corresponded to an exponentially decreasing profile with a scale height of 1 km and a surface value of ∼ 1.5 pmol mol −1 .

Ancillary data
As a consequence of its logistically complicated location and very harsh weather conditions, there are very scarce observational data describing the atmosphere at Antarctica. In order to interpret the bromine results, in addition to the spectra gathered by the MAX-DOAS measurements, near-surface O 3 vmr's measured at both stations were also compiled. At both sites, the surface O 3 was measured with ozone analyzers (Thermo Environmental Instrument, Thermo Fisher Scientific, model 49; i.e., TEI49). The operation principle of this in situ instrumentation consists of the attenuation of an ultraviolet light beam (254 nm) by an air sample containing ozone and has a manufactured sensitivity and limit of detection of 1 nmol mol −1 . In the case of the Marambio station, which contributes to the GAW network (Global Atmosphere Watch, WMO World Data Centre for Greenhouse Gases, WDCGG), the measurements of surface O 3 were carried out by the National Meteorological Service of Argentina (SMN) and can be retrieved from the WMO WDCGG (https: //ds.data.jma.go.jp/gmd/wdcgg/, last access: 18 September 2017). These data from the TEI49 from Marambio are compared every year against the regional standard (WMO, RCC-BsAs, TEI49PS). At the research site of Belgrano, the yearround surface O 3 has been measured by INTA since February 2007 (e.g., Jones et al., 2013). At this site, the inlet of the analyzer, protected from rain, snow and dust, is placed 0.85 m above the roof of the base in the cleanest area of the station, free of pollution from the research site.
Additionally, the weather information was obtained from the observations performed by the SMN at Marambio (WMO station 89055) and by INTA at Belgrano. In the case of Marambio, the weather station is installed in the so-called scientific pavilion of the Marambio Antarctic station together with an automated met station (AMS), which measures the temperature, humidity, precipitation, wind speed and direction in addition to the atmospheric pressure. The data acquisition system is carried out through a Campbell Scientific CR100 datalogger. In the case of Belgrano, the weather parameters are gathered by a Vaisala weather station installed at the site in 2009. In this case, the weather station is installed on the roof of the base, on a 210 cm mast and it provides wind speed and direction, atmospheric pressure, temperature, and relative humidity.

Results and discussions
This section is divided into three main parts. First, it presents the time series of the DOAS measurements and the ancillary observations performed during 2015, offering an overview of the information gathered within the frame of this study. Then, the details and discussion of the retrieved BrO and AEC vertical distributions at the two sites are provided. Finally, the activation of bromine in the Antarctic troposphere during 2015 is investigated along with the reactivity of the Antarctic troposphere with regard to inorganic reactive bromine.

Time series
This section presents the observations gathered during 2015 at each station. It first shows the DOAS measurements in terms of AOD and BrO vertical column densities (VCDs). Later on, it shows the results of the ancillary observations (weather parameters and surface ozone).

DOAS observations: BrO VCD 2 km and AOD 2 km
Herein, we present the AOD and the BrO VCD measured with the DOAS technique during the sunlit period of 2015. This period lasted for about 8.5 months at Marambio and 7.5 months at Belgrano (Figs. 3 and 4). Due to instrumental issues, there were missing data at the beginning of the year at both stations. Note, however, that the polar sunrise and therefore the peak season of the bromine activation was well covered at both sites (e.g., Simpson et al., 2007).
As mentioned in Sect. 2.2.2, the sensitivity of the MAX-DOAS observations decreases with altitude. Hence, aiming also at comparing both stations, here we refer to the AOD and BrO VCD inferred in the first 2 km of the troposphere at each site (i.e., AOD 2 km and VCD 2 km , respectively). Figure 3 shows the AOD 2 km retrieved at both stations (mean relative error of 40 %). Observations indicate that the aerosol optical thickness of the low troposphere at Belgrano was generally higher than at Marambio. At Belgrano, 62 % of the AOD 2 km was lower than 0.05 (12 % between 0.05 and 0.1) compared to 90 % of the AOD 2 km at Marambio that was below 0.05. In addition to this geographical dependence of the AOD, Fig. 3 also suggests that the period with higher aerosol thickness lasted longer in the southernmost station of Belgrano. As seen in the figure, while high AODs were observed at Belgrano during most of the sunlit period, the AOD at Marambio intensified from September until December. The geographical variability in the aerosol load within Antarctica has already been reported (e.g., Savoie et al., 1993;Minikin et al., 1998), although further studies on the distribution of aerosols are needed in order to understand the interannual variability in the different sources (e.g., Giordano et al., 2017). Further insights into the aerosol properties within the Antarctic troposphere during 2015 will be provided in a following work (Gómez-Martín et al., 2018).
Regarding the BrO, the pseudo-vertical column densities (VCD 2 km ) are calculated by integrating the BrO concentration obtained within the first 2 km of the troposphere. Results, shown in Fig. 4, indicate that BrO was present in the sunlit Antarctic troposphere at both stations. The median BrO VCD 2 km values at both sites were quite similar (∼ 0.5 × 10 13 molec cm −2 , with a mean relative error of 10 %) and 75 % of the observations at both stations fell below a similar value (0.8 × 10 13 molec cm −2 ). Also, at both sites, the maximum BrO VCD 2 km values were observed after the polar sunrise (i.e., SZA < 75 • herein) and the BrO levels were undetectable just before the polar sunset (i.e., SZA > 75 • herein) and immediately after the polar sunrise. However, the magnitude of and variability in the BrO VCD 2 km maximums direct the difference between both stations, with an absolute maximum BrO VCD 2 km observed at Marambio that is 3.2× higher than at Belgrano. As can be observed in Fig. 4, it is also worth noticing the clear photolytic activation of BrO at Marambio during austral spring, with levels an order of magnitude higher than the median BrO VCD 2 km values at the station. Insights into the vertical distribution of BrO and aerosol extinction in the Antarctic troposphere are provided later on in Sect. 3.2.

Ancillary observations: meteorological parameters and surface O 3
Aiming at contextualizing both research stations, this section briefly presents some weather parameters and the nearsurface ozone characterizing the sites of Marambio and Belgrano during 2015.
Regarding the weather information at each station, the mean observed values of meteorological parameters such as temperature (T ) and precipitation are provided in Table 3. Overall, observations indicate that Belgrano station sits at a dryer and colder location where, in 2015, temperatures dropped below −40 • C. Regarding the wind measurements, the wind rose of the 2015 measurements at each station is shown in Fig. 5 for low, medium and high wind speeds. Note , the median wind speed at Marambio (7.2 m s −1 ) was in general 50 % higher than at Belgrano. Concerning the wind direction observed at each station during 2015 (Fig. 5), while the air masses arriving at Marambio had no clear dominant direction, those arriving at Belgrano usually came from the south-southwest. Hence, based on the wind rose (Fig. 5), during 2015 Marambio was mostly influenced by air masses coming along the west edge of the Weddell Sea but also from the surrounding Scotia and Amundsen seas, while the air masses from con-tinental Antarctica dominated the observations performed at Belgrano station. As for the near-surface O 3 , the annual variation at both stations showed the seasonal pattern expected at high-latitude stations (Antarctic and sub-Antarctic regions, Fig. 6) and is typical of remote, low NO x environments with O 3 being accumulated during winter (maximum) and destroyed (minimum) during summer. Also, the observed amplitude of the surface O 3 annual cycle at both stations was also characteristic of an Antarctic station (e.g., Helmig et al., 2007;Legrand et al., 2016). While in 2015 the median values of surface O 3 were quite similar at both locations (23 nmol mol −1 at Marambio and 24 nmol mol −1 at Belgrano), the maximum values reported at Marambio (36.8 nmol mol −1 ) were about Table 3. Meteorological parameters (temperature T and snowfall) observed during 2015 at both stations. Data are provided by the World Meteorological Organization (WMO), the Argentinian Meteorological Centre and INTA's meteorological station (Belgrano). The snowfall is based on surface synoptic observations.

Station
Mean a 10 % higher than those observed at the Belgrano station. Regarding the absolute minimum surface O 3 detected, 2015 observations at Marambio indicate ozone depletion events (ODEs) with measurements very close to or below instrumental detection limit (1 nmol mol −1 ), while the minimum surface O 3 detected at Belgrano was not lower than 6 nmol mol −1 . This suggests that, compared to Belgrano, Marambio is either a more photochemically active region or a region more exposed to ozone-depleted air masses. Noteworthy is also the high variability in surface ozone observed at Marambio during the polar sunrise compared to observations at the Belgrano station. This behavior is characteristic of a coastal Antarctic station as reported by Helmig et al. (2007), for example.

Vertical profiles of BrO in the Antarctic troposphere
The time series of BrO vmr retrieved during 2015 in the lowest kilometers of the troposphere of Marambio and Belgrano are provided in Fig. 7, along with the AEC. For clarity, Fig. 8 shows, in more detail, some examples of the time evolution of BrO as measured at Belgrano and at Marambio and in Fig. 9 the sea ice conditions during those days. The wind speed on those selected days was below 10 m s −1 . As can be seen in Fig. 8, the maximum of BrO was located close to the surface, although its specific altitude depended on the day, always located below 1 km of altitude. Some days the peak of BrO was located just above the surface (e.g., 11 November at Belgrano or 25 September at Marambio in Fig. 8; both days with winds ≤ 6 m s −1 ), while in others that BrO maximum was slightly elevated, suggesting heterogeneous reactions aloft (e.g., 29 October at Belgrano with wind speed ≤ 2 m s −1 or 28 November at Marambio with wind speed 6.5-9 m s −1 ; in Fig. 8). Worth noticing is also the time and seasonal variability in the occurrence of the maximum of BrO vmr. As shown in Fig. 8, on 29 October at Belgrano, some days the BrO vmr followed the diurnal evolution with a noon maximum predicted by model studies (e.g., Saiz-Lopez et al., 2008) and observations (e.g., Buys et al., 2013) in which the BrO formation is linked to the solar irradiance and the photolysis of bromine sources (e.g., Br 2 , BrCl). Based on the calm wind conditions observed during that particular day (29 October), those sources are most probably not far from the station of Belgrano (Fig. 9). Conversely, on 28 November at Marambio, for example (Fig. 8), on other days BrO was present in the low troposphere with a double maximum (morning and evening), characteristic of a late spring behavior and more related to bromine being recycled through HOBr, for example (e.g., von Glasow et al., 2002;Pöhler et al., 2010;Liao et al., 2012;Buys et al., 2013). The slightly higher southeasterly winds during those observations on 28 November at Marambio point towards bromine sources in the Weddell Sea (Fig. 9). This shown time and altitude dependence of the BrO distribution in the troposphere reinforces the benefits of the sort of instrumentation employed in this work, which offers vertically resolved information and is able to perform long-term observations. Figure 10 shows a summary of the BrO vmr vertical profiles observed at each station during the sunlit period of 2015. The median BrO vmr in the lowest layers of the troposphere (< 0.5 km) was similar at both stations (∼ 1.6 pmol mol −1 above the surface) with 75 % of the BrO data below 2.5 pmol mol −1 . However, as shown in Fig. 7, the maximum BrO values observed after the polar sunrise (i.e., SZA < 75 • ) at Marambio (26.0 ± 0.4 pmol mol −1 ) were over 3-fold of those observed after the sun rose at Belgrano (see also Fig. 4). This maximum BrO was detected during austral spring at ∼ 200 m of altitude, with a magnitude dependent on the station. This slightly elevated peak of BrO (e.g., Fig. 10), also mentioned above, has already been foreseen by studies accounting for the vertical gradient of the acidity of the aerosols and/or the effect of convection (e.g., von Glasow et al., 2002;Wagner et al., 2007). Also note that, as shown in Fig. 7, while at Belgrano the maximum BrO observed during October-November (8.1 ± 0.6 pmol mol −1 ) quadrupled its mean value measured during the rest of the sunlit period, the BrO values observed at Marambio just after the sunrise were over 15× higher than the BrO mean values at that station. All this suggests that the halogen reactivity at Marambio is considerably stronger than at Belgrano (see also Sect. 3.3). The BrO vmr ranges reported herein (Fig. 10) are comparable to previous tropospheric Arctic studies (e.g., Tuckermann et al., 1997;Hönninger and Platt, 2002;Prados-Roman et al., 2011;Liao et al., 2012;Peterson et al., 2017;Simpson et al., 2017) and consistent with the few existing Antarctic measurements (e.g., Table 1). By adding the BrO measurements provided in the frame of this work to the few previous ground-based observations performed at Antarctica from different sites, Fig. 11 depicts an updated map of the maximum values of BrO observed in the lower troposphere of Antarctica, pointing once more to its heterogeneity with regard to reactive bromine load. Section 3.3 offers a closer look at this heterogeneity.
Overall, the observations presented in this study indicate that the vertical profile of BrO in the Antarctic troposphere descended with altitude (Fig. 10). Note that, in this work, the detection limit of BrO is regarded as the threshold value above which the inferred BrO is significantly higher than the noise of the inversion. In this case, it is defined as double the inversion error (Sect. 2.2.2) and corresponds to a mean value of 1 pmol mol −1 . In previous studies of the Antarctic troposphere (e.g., Frieß et al., 2004;Roscoe et al., 2014), the presence of uplifted reactive bromine was suggested. While the work of Frieß et al. (2004) remarked on the presence of uplifted reactive bromine (> 4 km altitude) due to advection processes, in the work of Roscoe et al. (2014), the authors were able to differentiate between two types of BrO vertical profiles: those with only near-surface BrO (centered at 200 m) and those with a double peak (centered at 2 km and near surface). In the work presented here, no uplifted layers (> 2 km) of BrO were detected, although, given the limited vertical information content of the MAX-DOAS observations (Supplement), double-peak types of BrO profiles cannot be ruled out. Although the definition of the height of the boundary layer over ice and snow surfaces (e.g., Anderson and Neff, 2008) is out of the scope of this work, previous studies place the top of the boundary layer in Antarctica show the altitude and are forced to be the same for the sake of comparison. The horizontal scales indicate the periods of the measurements, which depended on the station. The color code of the upper figures corresponds to the BrO vmr and is forced to be the same for the sake of comparison. The same applies to the color code of the lower figures, which indicate the AEC at each station. The BrO vmr higher than 5 pmol mol −1 and AEC higher than 0.5 km −1 are shown in dark red, while values below the detection limit are shown in black. Time periods with no observations (SZA > 75 • ) are indicated with white areas and those with data below quality filters are shown with grey areas. The vertical grid of the retrieval is indicated with the small ticks in the vertical axis.
between 100 m and 2 km, depending on the boundary layer parameterization, station and time of year (e.g., King et al., 2006;Nygård et al., 2013). Nygård et al. (2013) marked the height of the boundary layer at Marambio between 1 and 1.4 km for the non-winter periods of our interest, and the ozonesondes performed at Belgrano set that height between 1 and 2.5 km at that site (e.g., Parrondo et al., 2014). However, the altitude threshold between the boundary layer and the free troposphere cannot be assessed at the time resolution of our observations. Additional investigations at the two sites would be needed to confirm whether BrO reaches the free troposphere and, given the case, to assess the budget of BrO in the Antarctic free troposphere. Note that previous work on polar environments set BrO below 1.5 and 2 pmol mol −1 in the Arctic and Antarctic free troposphere, respectively (e.g., Frieß et al., 2011;Prados-Roman et al., 2011;Peterson et al., 2017;Hüneke et al., 2017).
In addition to the BrO knowledge gained after this work, the information related to the vertical AEC is also noteworthy (Fig. 7, lower panels), sustaining the particularity of the surroundings at each station. In addition to the aforementioned different aerosol optical thickness at both stations (Sect. 3.1.1), there is also a noticeable difference regarding the seasonality and altitude of the maximum AEC at the two sites. While at Marambio the peak of the AEC appeared close to the surface with a clear maximum extinction observed in November, the observations performed from Belgrano suggest that, at this site, the height of the aerosol layer  Table 4. Rates (k) of reactions provided in Sect. 1 and employed in Sect. 3.3. The temperature used for the calculations was T = 262 K, similar to the mean temperature observed during 2015 at each station (Table 3).

Reaction
Rate constant Reference (cm 3 molec −1 s −1 )  Atkinson et al. (2007) was much more variable than at Marambio, manifesting once more the relevance of vertically resolved observations within the Antarctic troposphere. As mentioned before, the work of Gómez-Martín et al. (2018) will address these issues.

BrO x in Antarctica
In order to investigate the hinted heterogeneity of the Antarctic lower troposphere regarding reactive bromine (BrO x = Br + BrO), in this section the budget of bromine [Br] is estimated considering the steady state of BrO in a pristine atmosphere with virtually no NO (e.g., Zeng at al., 2006), a concentration of ClO of 1.7 × 10 8 molec cm −3 (typical of Arctic conditions; e.g., Halfacre et al., 2014), and concentrations of HO 2 and OH of 2.2 × 10 7 and 3.9 × 10 5 molec cm −3 , respectively (mean values observed at the Antarctic station of Halley; e.g., Bloss et al., 2007). Hence, [Br] can be estimated from the observed BrO and O 3 concentrations as (e.g., Hausmann and Platt, 1994;Le Bras and Platt, 1995;Zeng et al., 2006;Stephens et al., 2012) [ where J represents the rate of photolysis (J BrO = 3.10 −2 s −1 for noontime in the polar spring; e.g., Thompson et al., 2015)  During the timeframe of that sea ice transformation, BrO VCD2km peaked at Marambio (Fig. 4). Also, note how the edge of the sea ice near Belgrano transforms towards summer (e.g., lower right). and k the different reaction rates (Table 4). Since the measurements of O 3 were performed near the surface, accordingly only the BrO retrieved in the lowest atmospheric grid (i.e., 100 m) is considered for the calculation of the Br and BrO x budgets at each site. The possible influence of horizontal advection and blowing snow is limited in the data set by applying an upper limit for the wind speed of 6 m s −1 .
Note that previous studies pointed to 8 m s −1 as the wind threshold for blowing snow (e.g., Jones et al., 2009) and others indicated that the steady-state approximation is valid for wind speeds lower than the 6 m s −1 threshold considered here (e.g., Liao et al., 2012). Based on Eq.
(2) and observed near-surface BrO and O 3 under low wind conditions (time stamp of the DOAS measurements), Fig. 12 shows the 2015 seasonal evolution of the BrO x budget at each station and Fig. 13 presents the BrO-Br-BrO x statistical analysis in the form of box charts at each site (only data for SZA < 75 • are considered). Since the kinetic calculations used herein are based on observations performed under low wind conditions, these budgets may be considered to be representative of the surroundings of each station. Figure 12 indicates that, in agreement with Peterson et al. (2015), the presence of reactive bromine at both stations does not only correspond to advected bromine-enriched air masses or blowing snow. As expected from previous polar studies (e.g., Simpson et al., 2007, and references therein) and shown in Fig. 12, the maximum bromine-related reactivity of the troposphere at both stations takes place just after the photolysis is triggered with the polar sunrise. As shown in Fig. 12, this maximum reactivity does occur at a medium O 3 regime at both stations (10-25 nmol mol −1 ). The study of the BrO-Br-BrO x data (Fig. 13) indicates that, during the sunlit period of 2015, the mean budget of BrO x at Belgrano (2.0 pmol mol −1 ) was ∼ 17 % higher than at Marambio. However, just after sunrise (i.e., SZA < 75 • ), the BrO x budget (and hence the oxidizing capacity) at Marambio triplicated the one at Belgrano (e.g., Fig. 12). Estimated values for atomic bromine radicals present in the lowermost troposphere during the sunlit period of 2015 were up to 1.4 pmol mol −1 at Belgrano and up to 3.4 pmol mol −1 at Marambio (Fig. 13). These ranges are in line with previous model studies for Antarctic latitudes (e.g., von Glasow et al., 2004;Saiz-Lopez et al., 2008) and in the lower limit of Arctic model studies (e.g., Thompson et al., 2017). Overall, these estimations indicate that the BrO x partitioning was clearly driven by BrO at both sites, indicating that ozone in general was not fully depleted as confirmed by the observations (Sect. 3.1.2). The evolution of the ratio Br to BrO after the polar sunrise (SZA < 75 • ) is shown in Fig. 14 for each site. The initial Br : BrO after dawn was ∼ 0.05 at both stations. Throughout the polar spring, during ODEs, that ratio rose over 4-fold at both sites. The baseline of the Br-to-BrO ratio during the sunlit period could be approximated by an exponential growth with a time constant of about 10 days at Belgrano and 17 days at Marambio (blue line in Fig. 14). Towards summer, that baseline increased up to 0.17 at Belgrano and to 0.10 at Marambio. In the simpli- Figure 11. Maximum values of BrO vmr reported in the low troposphere of Antarctica as measured with ground-based observations. The different sites where BrO has been reported in the low troposphere are indicated with a colored dot. The color code of each dot (station) refers to the maximum BrO vmr reported in literature (Arrival Heights: Kreher et al., 1997;Neumayer: Frieß et al., 2004;Dumont d'Urville: Grilli et al., 2013;Halley: Roscoe et al., 2014;Dome-C: Frey et al., 2015;Marambio and Belgrano: this work). Note that only the present study provides contemporary observations from different sites. Further details are provided on Table 1. fied scheme suggested by Eq. (2) and discussed in this section, this Br : BrO increase could be explained by the overall summer decrease in surface O 3 compared to springtime (Fig. 6). Additional investigations on the variability and geographical distribution of the bromine source gases throughout the year are suggested to address the bromine pathways in the Antarctic troposphere and their consequences. Bearing in mind this simplified scheme, based on Eq. (1) and the same fixed concentration of ClO as for Eq. (2) (i.e., to typical values in the Arctic environment), the brominemediated ozone loss rate can be assessed at each research site for the different BrO and O 3 regimes observed at low wind speed. Similar median BrO values measured during 2015 at both stations (1.6 pmol mol −1 ) yield a similar ozone loss rate of 0.4 nmol mol −1 day −1 at both sites. During the more active bromine season of October-November at Belgrano (e.g., Fig. 12), this rate speeds up to 2.9 nmol mol −1 day −1 . During September at Marambio (peak bromine season at that station), the bromine-mediated ozone loss occurs at a much faster rate of between 0.7 and 17.4 nmol mol −1 day −1 (i.e., up to 6× faster than at Belgrano). Former works have estimated that the bromine-driven ozone loss in the polar atmosphere represents 44 % of the total O 3 chemical loss (e.g., Liao et al., 2012;Thompson et al., 2017). Therefore, in the sites referred to in this work the shortest (i.e., at highest BrO x and low wind speed) ozone chemical lifetime τ O 3 expected is 2.6 days at Belgrano and 0.7 days at Marambio. Note that the estimations provided herein are limited by the information content inherent in the in situ technique measuring near-surface ozone (i.e., information at the exact instrument's location) compared to the DOAS data, which integrate information several kilometers away from the instrument (depending on scattering conditions) in the horizontal field of view and also in the vertical field of view (see retrieved averaging kernels in the Supplement). Further studies would be needed to confirm these numbers, including investigations on different sources and sinks of bromine radicals in the Antarctic environment, which herein are based on ozone depletion through (only) the BrO-BrO and BrO-ClO channels, dominant however in the polar spring (e.g., Simpson et al., 2007). All these kinetic approximations are historically based on conclusions from numerical models and laboratory and campaigned-based observations obtained in the polar regions (mainly the Arctic; e.g., Simpson et al., 2007). Nevertheless, the year-round erratic behavior of the wind speed in Antarctica at each station makes the verification of these estimated (low wind) τ O 3 with observations complicated. However, the exemplary days provided in Fig. 8 with higher BrO at each station (upper figures) may serve the purpose (low wind speeds). For instance, based on the ozone observations (Fig. 6), the rate of O 3 depletion measured at Marambio (25 September) was 4.1 nmol mol −1 h −1 and at Belgrano (29 Oc- tober) it was 0.58 nmol mol −1 h −1 . Therefore, as suggested by the above related theoretical calculations, the destruction of surface O 3 during the bromine peak season was indeed much faster at Marambio (7× faster than at Belgrano). Considering the mean O 3 vmr observed at each station on those days, the observed τ O 3 at Belgrano was 1.3 days while at Marambio it was 10-fold shorter. Note that, as shown in Fig. 8 (upper figures), on those specific days the BrO load at Marambio was also over an order of magnitude higher than at Belgrano. Comparing these observed τ O 3 values with the τ O 3 estimated above from kinetics, the measurements show shorter τ O 3 at both stations (50 % shorter at Belgrano and 18 % shorter at Marambio). Despite the low statistical meaning of this sort of "case study" exercise, the resemblance of the observed and calculated τ O 3 at Marambio suggests that the assumptions made at Marambio's surroundings (e.g., the Br-Cl channel dominates the ozone depletion) is close to reality, which seems not to be the case for Belgrano's surroundings. This reinforces, once more, the need for further long-term investigations for a better understanding of all the processes and key parameters involved in the halogens' pathways in the Antarctic troposphere.

Summary and outlook
As a result of its remoteness and a more complex logistics compared to the Arctic region, the characterization of the Antarctic troposphere with regard to halogen compounds is still very scarce. Based on contemporary ground-based observations performed at two Antarctic sites during 2015 (new sites as far as tropospheric BrO observations are concerned), this study reports on the presence and vertical distribution of reactive inorganic bromine in the low troposphere at the two sites and discusses the geographical distribution of BrO x . Prior to this study, throughout Antarc-tica only five sites had reported ground-based observations of BrO in the low troposphere. With the appropriate instrumental setup at the research stations of Belgrano and Marambio, INTA has expanded this net considerably. Moreover, to the authors' knowledge, this is the first study in which these bromine observations are reported simultaneously from two Antarctic stations, making it possible to gain an insight into the geographical distribution of reactive bromine in the Antarctic troposphere. Additionally, through the 2015 MAX-DOAS measurements performed at the two sites, this work presents vertically resolved observations of BrO at two different Antarctic stations with a dedicated inversion scheme for inferring the vertical distribution of BrO throughout the Antarctic troposphere. Furthermore, the aerosol extinction and the surface ozone at the two sites are also provided. Overall, results show the expected seasonal and daytime variation in BrO related to the photolytic activation of reactive bromine triggered by the polar sunrise at the two sites. However, as referred to above and unlike some former studies, during the sunlit period of 2015 no elevated plumes of BrO were detected above 2 km. Also, this study reports on the positive detection of BrO in the low troposphere (< 2 km) of Antarctica even under low wind conditions, suggesting that the presence of this trace gas is not only related to horizontal advection but also to surface emissions and/or vertical mixing. As for the vertical and geographical distribution of BrO in the lower layers of the troposphere, observations indicate a slightly elevated BrO peak at 200 m at both stations, with a maximum value measured at Marambio considerably higher than the observed value at Belgrano (26 pmol mol −1 vs. 8 pmol mol −1 , respectively).
In general, the observations and assessments presented in this work reveal a remarkable geographical heterogeneity of the Antarctic low troposphere with regard to the budget of reactive bromine. Beyond blowing snow, the inferred 3-fold enhancement of BrO x at Marambio compared to Belgrano after the polar sunrise also denotes a geographical heterogeneity on the bromine sources. Marambio sits on a region surrounded by open waters and seasonal sea ice while the dominant sea ice near Belgrano is perennial (Fig. 9). Since bromine explosions are linked to heterogeneous reactions related to sea ice, open leads and snow surfaces, for example, the type of sea ice and its seasonal evolution around each station may be a good starting point to tackle the bromine sources riddle and to investigate how climate change may affect the budget of BrO x in the troposphere of Antarctica. Moreover, the geographical distribution of BrO x and its partitioning addressed in this work also suggest that the reactivity of the troposphere at Marambio is particularly enhanced compared to other Antarctic sites ("hot spot"). Since the presence of BrO x in the polar atmosphere represents a sink for elemental mercury, this study also reveals the tip of the Antarctic Peninsula (Marambio) as a region for potentially enhanced mercury deposition (bioaccumulation) worth looking into. Also, dedicated investigations combining models and collocated observations of halogenated substances (not only BrO), organic compounds, DMS, NO x , HO x , particles and sea ice properties at different stations, for example, could assist a thorough study of the bromine sources and pathways in Antarctica, their geographical distribution and their projections under a changing environment.
In addition to the bromine-related information gained from this work, this study also emphasizes the benefits of deploying quality instrumentation in pristine and remote locations able to provide not only surface but also vertically resolved information. It also shows the scientific benefits of maintaining long-term observations despite the efforts related to sustaining research activities in such a hostile environment. The data provided by the two ground-based instruments presented herein may, for instance, assist the satellite retrievals to distinguish between tropospheric and stratospheric BrO signal and hence facilitate a more accurate assessment of stratospheric BrO and ozone trends, for example. Additionally, they could also serve chemistry-climate models for constraining the chemistry behind processes specifically related to polar regions, areas where global models are often weak (particularly in Antarctica).
Data availability. Data are available from the corresponding author upon request.
Competing interests. The authors declare that they have no conflict of interest.