Chemistry of the Antarctic Boundary Layer and the Interface with Snow: An Overview of the CHABLIS campaign

. CHABLIS (Chemistry of the Antarctic Boundary Layer and the Interface with Snow) was a collaborative UK research project aimed at probing the detailed chemistry of the Antarctic boundary layer and the exchange of trace gases at the snow surface. The centre-piece to CHABLIS was the measurement campaign, conducted at the British Antarctic Survey station, Halley, in coastal Antarctica, from January 2004 through to February 2005. The campaign measurements covered an extremely wide range of species allowing investigations to be carried out within the broad context of boundary layer chemistry. Here we present an overview of the CHABLIS campaign. We provide details of the measurement location and introduce the Clean Air Sector Laboratory (CASLab) where the majority of the instruments were housed. We describe the meteorological conditions experienced during the campaign and present supporting chemical data, both of which provide a context within which to view the campaign results. Finally we provide a brief sum-Correspondence high halogen species a snowpack.


Introduction
The Antarctic boundary layer is one of the regions of the world's atmosphere least affected by human activity. The continent is surrounded by the Southern Ocean with the nearest population centres being at much lower latitudes in the South American, African and Australian continents. The Antarctic boundary layer is a place of extremes; during the winter, the sun does not rise for months on end, contrasting with the summer which is light for 24 hours each day. Temperatures remain below zero for the majority of the year, Published by Copernicus Publications on behalf of the European Geosciences Union.  With this background, a study of Antarctic boundary layer chemistry can provide many things -a view of a natural atmosphere to be contrasted with more polluted regions; a challenging test bed for numerical models which simulate the atmosphere and climate of the future; and in addition, it provides a chance to deconstruct the chemistry, deposition and post-depositional processes which determine the record of impurities retrieved from deep ice cores.
The chemistry of the Antarctic troposphere has been relatively little studied to date. The earlier studies focused mainly on aerosols (e.g. Wolff et al., 1998) and long-lived radiatively and stratospherically important gases (e.g. Elkins et al., 1993;CMDL, 1996). Of the more reactive trace gases, only surface ozone was measured with any real vigour across the continent, with continuous sampling inland at South Pole (Oltmans and Komhyr, 1976) and on the coast, for example at Neumayer (Wyputta, 1994) and Syowa (Murayama et al., 1992). Data from the coastal stations are now recognised to also show springtime destruction of tropospheric ozone similar to that seen in the Arctic (Wessel et al., 1998). Recent years, however, have seen the first attempts to study the broader active background tropospheric chemistry and the interaction between air and snow. The SCATE (Sulfur Chemistry in the Antarctic Troposphere Experiment) campaign at Palmer Station on the Antarctic Peninsula (austral summer1993/94)  focused on sulphur chemistry but included measurements for O 3 , CO, NO, OH. Further year-round work on sulphur chemistry has been carried out at Dumont d'Urville station (Jourdain and Atmos. Chem. Phys., 8, 3789-3803, 2008 www.atmos-chem-phys.net/8/3789/2008/ Legrand, 2001). SCATE was followed up at South Pole by further summer programmes with a progressive shift towards focus on oxidants: ISCAT (Investigation of Sulfur Chemistry in the Antarctic Troposphere) in 1998 and 2000 (e.g. Davis et al., 2001;Davis et al., 2004). These studies have since been complemented by ANTCI (Antarctic Troposheric Chemistry Investigation) 2003 and 2005. Two campaigns were conducted at Neumayer, in austral summers 1997 and 1999 aimed at studying the budget and chemistry of NO y (e.g. Jones et al., 1999;Weller et al., 1999;Jacobi et al., 2000). In addition, measurements of tropospheric H 2 O 2 and HCHO were run at Neumayer throughout 1997 (Riedel et al., 1999;Riedel et al., 2000), and of NO and NO y throughout 1999 (Weller et al., 2002), recognising the need for yearround measurement programmes. From the limited number of previous studies it was already clear that the Antarctic boundary layer behaved in very unexpected ways, driven by the extreme cold, long periods of darkness alternating with continuous sunlight, and a strong chemical coupling between the snowpack and the overlying atmosphere. But in spite of these studies, the detailed chemistry of the Antarctic boundary layer remained little explored, in particular beyond the summer season. Progress was limited by the difficulty of assembling suitable facilities, instruments and researchers in the Antarctic, and in doing so for a time period long enough to unambiguously observe the underlying interactions at work. In response to this, CHABLIS (Chemistry of the Antarctic Boundary Layer and the Interface with Snow) was designed to explore the atmospheric chemistry of the Antarctic boundary layer in greater detail and for a longer period of time than had been previously achieved. It was a collaborative project between the British Antarctic Survey and research groups from five UK universities, and involved measurements made by an extensive range of chemical analysers as well as numerical modelling (see Table 1). The centre-piece to CHABLIS was a year-long field campaign to carry out measurements of Antarctic boundary layer chemistry and experiments to explore exchange processes between the snowpack and the overlying atmosphere.

CHABLIS scientific aims
The broad aims for CHABLIS outlined above, were distilled into three specific science foci: i) seasonal studies of oxidant chemistry which considered the potential role for NO 3 during polar twilight/night, halogen chemistry during spring and summer, and HOx chemistry during polar summer; ii) year-round studies of the NO y budget aimed at determining the dominant NO y components, how they varied throughout the year, and what this meant both to sources of NO x and to ice core nitrate; iii) air/snow transfer studies, to assess the influence of the snowpack on coastal Antarctic boundary layer chemistry. Details of the measurement techniques are given in Table 1. The measurement strategy was for a co-ordinated  to 13 August. Prior to sun-up and after sun-down, scattered light can be observed at Halley for several hours around solar noon.

The Clean Air Sector Laboratory
The CHABLIS measurements were made at the new Clean Air Sector Laboratory (CASLab) (Fig. 3) which is designed specifically for studies of background atmospheric chemistry and air/snow exchange. The laboratory is constructed from three shipping containers, and mounted on a steel, legged platform that is raised approximately every two years in order to keep the platform roughly 3 m above the snow surface. The laboratory is situated 1 km to the south east of the main station, in a sector which receives minimal air flow from the base (see Fig. 2). Routine access is on ski or by foot, so interference from vehicle emissions is kept to an absolute minimum. For all trace gas analysers housed in the CASLab (i.e. all but FAGE which was in its own container), ambient air was sampled off the main trace gas inlet stack. The stack comprises a 100 mm internal diameter (i.d.) electropolished stainless steel tube ventilated by a fan at the far end (see Fig. 4). The air flow through this central stack was ∼314 m 3 /h, i.e. sufficiently high that the residence time within the stack was less than 1 s. Additionally, the volume of air drawn through the stack was such that the samples removed less than 10% of the total air flow. Access to the air was achieved via individual ports, spaced 10 cm apart along the length of the stack, with 1/4 stainless steel tubing that penetrated to the centre of the air flow. Inlets to individual samplers were connected to these ports using Swagelock connectors. A Pitot tube monitored air flow through the stack during the CHABLIS campaign. The inlet of the central stack itself was initially ∼8 m above the snow surface. As the campaign progressed and snow accumulated during the year, this height reduced such that by the end of the campaign the inlet was ∼7 m above the snow surface.
Aerosol measurements were made from a separate stack specifically designed for particulate sampling. It comprises of a 200 mm i.d. stainless steel chimney with a cowl at the air intake to prevent snow ingress (see Fig. 5). Ventilation was achieved using a fan with a variable flow rate that can be altered according to the volume of air being drawn from the stack for sampling in order to allow isokinetic sampling. During CHABLIS, a stack flow rate of 240 l/min was appropriate for the aerosol instrument suite: the low-volume filter sampler (flow rate 20 l/min), the aethalometer (flow rate 20 l/min) and the condensation particle counter (CPC, flow rate 1 l/min). The samplers accessed the stack via stainless steel cones that were installed at the base of the chimney (see Fig. 5a and b). Each cone was specific to a particular instrument and was designed to maintain isokinetic flow up to the point of sampling. To achieve this, the ratio of the area of the  cone opening to the area of the stack tube needs to be equal to the ratio of the sampler flow rate to stack flow rate. For the low volume air sampler at 20 l/min, this led to a diameter of the sampling cone of 57.6 mm. The samplers then connected to the cones via swagelock connectors situated below the stack. Particularly challenging for the CHABLIS set-up phase was the installation of the boundary layer differential optical absorption spectrometer (BL DOAS) system. This instrument records differential spectra between a light beam, emitted by a 400 W Xe arc lamp, and the return beam. In order to allow sufficient path length to achieve a satisfactory detection limit, the retro-reflector was located 4 km away from the CASLab. In order to keep the surfaces of the retro free from snow/ice, it was necessary to warm it. This was done by supplying 90 W of power from the CASLab to heaters at the retro-reflector along 4 mm 2 cables over 5 km long.
In addition to the main CASLab facility, the Clean Air Sector includes a suite of external micro-meteorological instruments for probing the physics of the boundary layer. The suite of boundary layer measurements made during CHABLIS are summarised in Table 2. The instruments were mounted on a 30 m mast 50 m south of the laboratory, and located just beyond the air-flow distortion associated with the facility itself. In addition, the acoustic sounder and cospatial microbarograph array were situated 1 km north of the CASLab and operated for periods of the CHABLIS campaign. The combination of chemistry and physics instrumentation at the CASLab provided a powerful tool for interpreting the chemical measurements within the context of the physical behaviour of the boundary layer.

Meteorological conditions during CHABLIS
The general meteorological conditions at Halley during the CHABLIS campaign are shown in Figs. 6 and 7, and are similar to those described from earlier years (König-Langlo et al., 1998). Figure 6 shows that surface temperatures ranged from around zero during the summer months to almost −50 • C during the late spring. Surface winds were generally between 0 and 10 m/s, with a number of storms throughout the year. The windiest period was during the spring, but a number of storms occurred also at other times of the year, including a particularly strong and sustained storm during May 2004. These storms tend to be associated with low pressure systems in the Southern Ocean that track the Antarctic coast and bring maritime air to Halley. Of interest also in this plot, and shown clearly for the May 2004 storm, is that periods of high winds are often associated with warmer temperatures; conversely, quiescent periods with low wind speeds tend to be colder.
The wind rose shown in Fig. 7 provides information not only on wind direction, but also on the strength of the winds (shown by the width of the bars). It demonstrates very clearly that the majority of winds during the campaign came from the east, and also that these were the strongest winds. It is also clear that for a significant proportion of the time during CHABLIS, winds came from the south west. Although local wind direction is not generally a good indicator of air mass origin, air arriving at Halley from the south west is likely to have had contact with open water in the Precious Bay region within the previous few hours. An important aspect is that    very few winds were from the northwesterly direction of the main station buildings. The broader air mass origins can be derived from back trajectory calculations. 8-day back trajectories were calculated for the entire CHABLIS campaign using the web-based trajectory service of the British Atmospheric Data Centre (BADC) using input data from the European Centre for

Supporting chemical measurements
Accompanying papers in this special issue explore the details of chemical composition and processes during the CHABLIS campaign. Here we present the chemical climatology as demonstrated by various measurements made during the campaign, and explore what some of these mean for boundary layer chemistry.

Surface ozone
Surface ozone was measured throughout the CHABLIS campaign using a 2B Technologies model 202 ozone monitor. This instrument was selected in part because of its extremely low power consumption (∼4 W) which is a great advantage when operating in remote regions. The analyser was connected directly to the trace gas inlet manifold and captured data every 10 seconds. The hourly-averaged time series is shown in Fig. 8. Surface ozone reaches its maximum concentration during the winter months and is at its minimum during the summer. This is the classic seasonal cycle for a trace gas whose concentration is balanced by increases arising from air mass transport and destruction by the direct action of the sun or by sunlight-initiated chemistry. Such a cycle is apparent at all other coastal Antarctic sites where surface ozone is measured (e.g. Neumayer, Syowa, McMurdo and Sanae (Helmig et al., 2007)). The cycle differs from that of South Pole, where surface ozone values decrease during spring from a winter peak before escalating rapidly towards a sharp annual maximum in summer (Crawford et al., 2001;Helmig et al., 2007). This unusual behaviour is associated with high mixing ratios of NO x which are themselves driven by emissions from the snowpack Oncley et al., 2004); at sufficiently high NO 2 mixing ratios, in situ ozone production occurs which dominates over loss processes (Crawford et al., 2001). Early results suggested that elevated NO x mixing ratios occurred under conditions of shallow boundary layer depths (20 m to 50 m) which were themselves associated with strong surface stability and light winds Oncley et al., 2004). The picture was that trace gases produced from the snowpack were concentrated to a relatively high degree within this shallow layer. However, a recent study has found that the picture is somewhat more complicated, and that elevated NO mixing ratios can also occur under conditions of higher winds, weaker surface stability and deeper mixed layers (Neff et al., 2007). At Halley, the summertime boundary layer is normally deeper than that observed at South Pole; this factor will in part explain the lower surface O 3 mixing ratios observed during summertime at Halley as NO x mixing ratios do not achieve the elevated levels observed at South Pole. However, an additional factor influencing surface ozone at Halley (and most likely all other coastal sites) is the presence of halogens. Measurements of IO and BrO made at Halley suggested that they were present throughout the summer at mixing ratios of the order 2 pptv, with typical peaks of 5 pptv . This inevitably suggests an associated presence of I and Br atoms which would be a significant sink for surface ozone . The influence of halogens on surface ozone is also clearly apparent during springtime. Figure 8 shows the springtime ozone depletion events (ODEs), previously seen at Halley (Jones et al., 2006) and indeed a commonly-recognised feature at coastal sites in both polar regions (e.g. Simpson et al., 2007). Interestingly, a deep and rapid ODE lasting for just under a day was measured on 19 October which is surprisingly late in the season. Two sampling protocols were adopted. Up until 6 August 2004, the measurement duty cycle included 5 min long, hourly calibration/zeroing sequences, the data were logged at 10 s intervals with no smoothing; this yielded a standard deviation for a one hour data sample (at low atmospheric variability) of 0.7 ppbv. After 6 August 2004, the duty cycle comprised 10 min long, 12 hourly calibration/zeroing sequences, the data were logged at 30 s intervals, with a 60 s running average; this yielded a standard deviation for a one hour data sample (at low atmospheric variability) of 0.2 ppbv. The timeseries measured during the CHABLIS campaign is shown in Fig. 9, with the anticipated mid-winter maximum and summer minimum. The very short-lived spikes are associated with exhaust emissions from the main station, and were used to produce a pollution inventory against which other data could be filtered. These data and issues of CO calibration and standardisation are discussed by Bauguitte et al. (2008)  Nitric acid is notoriously difficult to collect and measure because of its tendency to stick to any available surface (see e.g. Ryerson et al., 1999). Apart from the more expensive routes (such as Chemical Ionisation Mass Spectrometry), three main methods have been used to sample HNO 3 in the past: nylon filters, annular denuders, and mist chambers. For CHABLIS, annular denuders (URG corporation) were used to sample HNO 3 . For the routine measurements two denuders were used in series, each denuder being 150 mm long, 30 mm outside diameter and with 3 concentric channels. The second denuder acted as a back-up from which collection efficiency could be derived.
Before use, each denuder was coated with a solution made up from 1 g Na 2 CO 3 (sodium carbonate), 1 g glycerol in 50 mL Milli-Q water plus 50 mL methanol; the denuders were then dried in a clean air stream and capped at either end. For sampling, the denuder train was installed on the roof of the CASLab to allow open-face sampling and ensure minimal losses of HNO 3 to surfaces. Air was pumped through the denuder train at a flow rate of 20 l/min. Denuders were exposed for periods of 1 week throughout the campaign, and in each season for a week, daily sampling was conducted as well as one day at 6-hourly resolution. After sampling, the denuder coating was rinsed off using a known amount of MilliQ water (18 M ) and the liquid was frozen and returned to the UK for analysis. This was carried out using an ion chromatograph (IC).
The collection efficiency was assessed throughout the campaign by comparing HNO 3 sampled on the first denuder with that sampled on the second (both initially blankcorrected): collection efficiency = 1 − (denuder 2/denuder 1) During CHABLIS, HNO 3 was sampled with a collection efficiency of 91%. Blanks were assessed every 3 months by connecting an additional denuder to the sampling train. The third denuder acted as the blank, the assumption being that no significant amount of ambient HNO 3 would break through into this denuder.
When working up the data, the amount of HNO 3 in the boundary layer was calculated according to: total sample = (denuder 1 − blank)/collection efficiency The detection limit for HNO 3 sampling (derived from 2 × standard deviation of the blank) varied from 0.14 pptv for weekly sampling to 1 pptv for daily sampling. For the 6-hourly sampling, unfortunately all the data were below the detection limit of 4 pptv. The HNO 3 measurements made during CHABLIS are shown at their highest resolution in Fig. 10a and are further discussed in Jones et al. (2007).

Particulate nitrate
Particulate nitrate was sampled through the main aerosol inlet within the CASLab, on 1 µm pore size, 37 mm diameter Zefluor filters (Pall-Gelman Corp.), at a flow rate of ∼20 l/min. Again, filters were exposed for periods of 1 week throughout the majority of the campaign, but with higher resolution (daily and 6-hourly) in each season. Handling Atmos. Chem. Phys., 8, 3789-3803, 2008 www.atmos-chem-phys.net/8/3789/2008/ 14-Feb-04 14-Mar-04 14-Apr-04 14-May-04 14-Jun-04 14-Jul-04 14-Aug-04 14-Sep-04 14-Oct-04 14-Nov-04 14-Dec-04 14-Jan-05 gaseous nitric acid (pptv) Figure 10a. Year-round HNO3 from denuders at weekly and daily sampling frequency, plotted on same y-axis scale as Figure 10b to ease comparison. Only limited numbers of daily samples were above the detection limit, (6,7,10 April; 6, 8 October). 14-Mar-04 14-Apr-04 14-May-04 14-Jun-04 14-Jul-04 14-Aug-04 14-Sep-04 14-Oct-04 14-Nov-04 14-Dec-04 14-Jan-05 particulate nitrate (pptv) Figure 10b. Year-round particulate nitrate from filter sampling at weekly, daily and 6hourly sampling resolution. Higher resolution sampling was carried out around mid April, mid July, end September/early October, and mid January. 30 Fig. 10a. Year-round HNO 3 from denuders at weekly and daily sampling frequency, plotted on same y-axis scale as Fig. 10b to ease comparison. Only limited numbers of daily samples were above the detection limit, (6, 7, 10 April; 6, 8 October). 14-Mar-04 14-Apr-04 14-May-04 14-Jun-04 14-Jul-04 14-Aug-04 14-Sep-04 14-Oct-04 14-Nov-04 14-Dec-04 14-Jan-05 gaseous nitric acid (pptv) Figure 10a. Year-round HNO3 from denuders at weekly and daily sampling frequency, plotted on same y-axis scale as Figure 10b to ease comparison. Only limited numbers of daily samples were above the detection limit, (6,7,10 April; 6, 8 October). 14-Mar-04 14-Apr-04 14-May-04 14-Jun-04 14-Jul-04 14-Aug-04 14-Sep-04 14-Oct-04 14-Nov-04 14-Dec-04 14-Jan-05 particulate nitrate (pptv) Figure 10b. Year-round particulate nitrate from filter sampling at weekly, daily and 6hourly sampling resolution. Higher resolution sampling was carried out around mid April, mid July, end September/early October, and mid January. 30 Fig. 10b. Year-round particulate nitrate from filter sampling at weekly, daily and 6-hourly sampling resolution. Higher resolution sampling was carried out around mid April, mid July, end September/early October, and mid January. blanks were run approximately every 3 months, during which the filter was installed as normal but the pump switched on only for a few seconds before changing the filter. Once exposed, all filters were stored frozen and returned to the UK for analysis. Filters were extracted into ∼8 ml of MilliQ water (18 M ) by agitating in an ultrasound bath. The extract was then analysed for major cations (e.g. Na + , Ca + ) and anions (e.g. NO − 3 , methane sulphonic acid) using an IC. Lowvolume aerosol filters operated in this way are assumed to sample aerosol with 100% efficiency. Throughout this paper particulate nitrate measurements are expressed in pptv, where 1 pptv is equivalent to 2.78 ng m −3 . Detection limits were derived from 2 × standard deviation of the blanks, and varied from 0.01 pptv for weekly sampling to 2.4 pptv for 6-hourly sampling (see Table 1). The low-volume aerosol p-NO − 3 data, blank corrected, are shown at their highest resolution in Fig. 10b. These and other low-vol data are more fully analysed in other papers in this special issue Wolff et al., 2008).  Figure 11. The sum of nitric acid and particulate nitrate, often referred to as total inorganic nitrate, as measured on denuders and low-volume filters during CHABLIS. Fig. 11. The sum of nitric acid and particulate nitrate, often referred to as total inorganic nitrate, as measured on denuders and low-volume filters during CHABLIS.

Total inorganic nitrate
Here we briefly consider the monthly averaged "total inorganic nitrate" (TIN = HNO 3 + p-NO − 3 ) data for CHABLIS, which is shown in Fig. 11. The peak in summer (Nov-Dec-Jan) and shoulder suggesting a possible secondary peak in August are consistent with observations at both Neumayer  and Dumont d'Urville, a coastal station in the Pacific sector of Antarctica. High volume aerosol filters (assumed to capture both particulate and gaseous nitrate) from Dumont d'Urville were analysed both for their nitrate concentration and isotopic signature (Savarino et al., 2007). The authors concluded that the two nitrate peaks were driven by quite separate mechanisms. The summertime peak was interpreted as arising from snowpack emissions on the polar plateau which were recycled and transported to the coastal region. Polar stratospheric cloud (PSC) sedimentation was concluded to be responsible for the late winter/early spring TIN peak. Certainly measurements of tritium, a marker of stratospheric water and hence PSCs, reported for Halley from 1983 to 1992 ) reveal a maximum in August/September. 6.4 Snow sampling 6.4.1 Surface snow Surface snow was collected on a daily basis during CHABLIS from March 2004 until the end of the measurement campaign. An area of clean snow was selected, roughly 50 m to the south west of the CASLab. This area was remote from the station, undisturbed by vehicle traffic of any sort, and therefore provided pristine snow for sampling.
Using clean procedures (i.e. wearing clean-room gloves and not breathing into the sample), snow was collected directly into clean accuvettes to a depth of roughly 1 cm. The samples were stored frozen and returned to the UK at the end of the measurement campaign where they were analysed on an IC for major anions and cations (Na + , K + , Ca 2+ , Mg 2+ , F − , MSA, Cl − , NO − 3 , SO 2− 4 ). Surface snow data are discussed in various papers in this special issue, e.g.  and Wolff et al. (2008).

Snow pit measurements (nitrate and nitrite)
Snow pits were dug each month during the CHABLIS campaign, in a similar location to the surface snow sampling and again using clean procedures to limit contamination. The pits were sampled at a series of depths between the surface and 80 cm, by pushing cleaned Perspex bottles directly into the wall of the pit. As the diameter of the bottles is 3 cm, the resolution of each snow pit sample was the desired depth ±1.5 cm. Snow samples for nitrate analysis were stored frozen and returned to the UK for analysis on the BAS IC. This approach relies on the fact that nitrate is stable within snow when stored frozen and in the dark. Snow samples for nitrite, however, were analysed at Halley using the HONO analyser which, given it uses a liquid method for sample analysis, was ideal for determining nitrite in melted snow. The samples were stored frozen and in the dark and were analysed within a few days. Figure 12a shows the depth profile of nitrite in the top 80 cm of the Halley snow pack for June, July (winter) and January (summer). The June and July profiles effectively overlap, displaying essentially no variation within this depth range. January, however, is markedly different, with a very clear maximum in the top 0 to 15 cms, which gradually decreases with depth until it coincides with the wintertime pro-files at around 60 cm. This behaviour is characteristic of a snow impurity that has a maximum during the summer months. At Halley, with roughly a meter of snowfall each year, sampling to only 80 cm depth would not reach the previous summer's maximum assuming that it was preserved. The trend in the nitrite profile is very similar to that for nitrate, shown in Fig. 12b; concentrations of nitrate in surface snow have a maximum during the summer and minima in the winter .
Of particular interest in Fig. 12 are the comparative concentrations of nitrate and nitrite, with summertime nitrate exceeding that of nitrite by 2 orders of magnitude. Abundant data exist for concentrations of nitrate in polar snow and ice, however, there are extremely few reported measurements of nitrite. A likely reason for this are the very low concentrations of nitrite, that would take it below the detection limit of many analysis systems. One previous set of measurements was made during the spring in Barrow, Alaska (Li, 1993) where surface snow nitrate concentrations were typically around the instrument detection limit of 40 nM. Two recent papers have calculated nitrite concentrations in snow. Jacobi and Hilker (2007) used a model derived from laboratory experiments and applied it to field data to calculate nitrite concentrations in summertime snow at Summit, Greenland, of less than 0.0002 nM. Chu and Anastasio (2007) calculated concentrations of snow grain nitrite based on steady state analysis of sources and sinks for four polar locations. Their estimates ranged from 0.7 nM (for Alert) to 13 nM (for Summit). For Neumayer (70 • S, 8 • W), a coastal site in the same sector of Antarctica as Halley, nitrite concentrations of 5 nM were calculated for noon solstice (21 December) conditions.
The interest in comparing nitrite and nitrate snow pack concentrations stems from the fact that both are potential sources of OH (although their direct contribution has been shown to be considerably less than that from H 2 O 2 for example (Chu and Anastasio, 2007), and also sources of NO x . Nitrate is photolysed to directly produce NO 2 via its major channel: and nitrite photolysis generates NO: Nitrite itself can be formed from the minor photolysis channel of nitrate (NO − 3 +hν→NO − 2 +O( 3 P)) as well as from hydrolysis of NO 2 generated by Reaction 1.
The fact that nitrite is photolysed at somewhat longer wavelengths than nitrate (320-400 nm as opposed to 280-320 nm) results in significantly higher photolysis rates for nitrite than for nitrate under natural sunlit conditions. Chu and Anastasio (2007) calculated photolysis rates for R1 and R2 appropriate for Neumayer (for noon 21 December) and found that those for nitrite were two orders of magnitude Atmos. Chem. Phys., 8, 3789-3803, 2008 www.atmos-chem-phys.net/8/3789/2008/ higher than for nitrate. We use their calculated photolysis rates and our measurements of nitrite and nitrate in the top of the summertime snow pits to derive relative production rates for NO x (see Table 3). The ratio of NO x produced by photolysis of nitrite to that from nitrate under these conditions is 1: 0.69, i.e. an equivalent order of magnitude from both sources. This approach considers only surface snow and does not take into account photolysis occurring deeper within the snow pack. However, the production with depth is expected to scale linearly for both species, so this additional factor is not likely to significantly alter the ratio. However, it must be remembered that surface snow nitrate concentrations display considerable heterogeneity; for example, at Halley the average surface snow nitrate concentration for the months of 4 December and 5 January is 2630±982 nM, with a range from 801 nM to 6015 nM. Unfortunately, we have no additional nitrite data to explore whether equivalent heterogeneity exists for this species such that the NO x production ratio is preserved, or whether the ratio varies. However, the results do illustrate that, at least at Halley, nitrite, albeit at low concentrations in the snow, can be a very significant source of NO x . This conclusion, arrived at from measurements, agrees with that of Chu and Anastasio (2007) who made this suggestion based on calculations. What it means is that numerical models that aim to calculate fluxes of NO x from snow (and indeed those probing the chemistry of snow-pack and its interstitial air) should include a comprehensive suite of reactions to account for nitrite formation and subsequent photolysis.
6.5 Measurements of the ratio of up-welling to downwelling actinic flux During CHABLIS, a spectroradiometer was used to derive photolysis frequencies for a number of trace gases, and to measure the ratio of up-welling to down-welling actinic flux. The purpose of the flux-ratio measurements was to provide parameterisations suitable for use by modellers. For this purpose, an accuracy of a few percent is quite adequate, unlike in climate and snow-character studies where an accuracy of better than 1% would be important (for some modelling studies, even an estimate of up-welling flux within 10% might be considered sufficient).
The albedo A is defined by the ratio of upward to downward irradiance, i.e. measurements by horizontal cosineresponse (flat plate) sensors. The ratio of up-welling to down-welling actinic flux measurements, by isotropicresponse (hemispheric) sensors, only equals A in the idealised case of an isotropic diffuse sky with no reflectivity (Madronich 1987). The ratio is 2A cos(SZA) under clear skies if Rayleigh scattering by air molecules is ignored. If the surface is snow, the ratio is close to 1.0 in the more realistic diffuse sky case of thick cloud, even if A is less than 0.9, because multiple reflections between snow and cloud ensure near-equal up-welling and down-welling fluxes. Table 3. Comparing photolysis rates, surface snow concentrations (measured at the surface of the snowpits), and the consequent ratio in NO x production rate from nitrite and nitrate in snow. The photolysis rates are for j (NO − 2 →NO) and j (NO − 3 →NO 2 ) respectively for noon 21 December conditions (Chu and Anastasio, 2007 Some models calculate actinic flux from first principles, and so include a comprehensive treatment of up-welling actinic flux, including the calculation of Rayleigh scattering and 2A cos(SZA) together with multiple reflections between cloud and snow. These models need the albedo A as an input. Irradiance measurements by cosine-response sensors have been carried out routinely at Halley for many years, and were evaluated from 1963 to 1982 by Gardiner and Shanklin (1989) in their Table 9b. Monthly-mean albedos varied from 0.81 to 0.83, 0.03 to 0.05, with no clear seasonal cycle. We therefore recommend a value of albedo of 0.82 0.06.
However, some users of CHABLIS data need as input an estimate of total actinic flux from measurements. This can be derived from the near-continuous CHABLIS measurements of down-welling flux by multiplying by (1 + upwelling/down-welling flux), hence they need an empirical evaluation of up-welling/down-welling flux. Below we show measurements of the flux ratio and derive a simple parameterisation.
The actinic flux instrument used was a single monochromator with a diode array detector (meteorologieconsult gmbh, (Hofzumahaus et al., 2004)). It uses a 512 pixels diode array detector, mounted in a fixed-grating monolithic monochromator (Zeiss), covering the spectral range from 270 to 700 nm. The input optics were heated to prevent riming. Spectra were digitised to 16 bits (MOE C161, TEC 5 AG, Germany), and the values transmitted to the CASLab computer, 150 m away, via RS422 serial communication. The instrument was calibrated at 6 month intervals using a spectral irradiance standard consisting of a 1000 Watt quartz halogen tungsten filament lamp model FEL A (Optronics Laboratories).
Normally, the actinic flux spectrometer measured the down-welling hemispheric flux by observing vertically upward via a quartz hemisphere ground on the inside (conventionally known as a 2π dome). A black horizontal plate and ring defined the lowest angle seen. From time to time during the summer, measurements of the up-welling flux were made by pointing the spectrometer vertically downward towards the snow 2.5 m below. The process of unclamping the spectrometer from its mounting frame, rotating it and re-clamping took 2 to 4 min. Measurements of up-welling flux were made for 5 to 15 min before restoring the spectrometer to its upwards pointing position. Sometimes three or four successive up-welling measurements were made at intervals of 10 to 15 min.
Rather than derive up-welling/down-welling flux ratios at each of the 512 wavelengths measured by the spectrometer (which would have been unwieldy for the user), we derived flux ratios from j-values for specific molecules calculated from the spectra. These quantities are much more useful for most atmospheric chemistry models. Ideally the ratio would be derived from simultaneous down-welling and up-welling measurements, but as our measurements were not simultaneous we had to use a modified approach. Instead, measurements of up-welling flux were compared with those of the down-welling flux at the start and end of each set of upwelling measurements. The up-welling data were then ad-justed by various trial factors, by eye, until they best matched the down-welling flux. The factor that best achieved this match was taken as the flux ratio. To assess the accuracy of this approach, the adjusted up-welling data were compared with an interpolation line achieved by linearly interpolating across missing down-welling data. Disagreement from this interpolation line was sometimes as large as 1%, which suggests this value as an upper limit to the error on flux ratio measurement. A set of successive up-welling measurements was treated as one flux ratio measurement.
Results are given in Fig. 13, divided into measurements during more cloud and less cloud. The dividing value for cloudiness was between 6 and 7 octas (1 octa is 1/8 cloud cover), chosen by examination. The cloudy measurements gave a flux ratio of about 0.98, more or less independent of solar zenith angle (SZA).
The less-cloudy measurements (shown in Fig. 13a) did not give a constant flux ratio, but had minimum flux ratios between 72 • and 78 • SZA. The reduction in flux ratio with increasing SZA is expected from the cos(SZA) dependency (Madronich 1987) discussed above. The increase at larger SZA can be explained by a greater proportion of the illumination coming from light scattered isotropically by air molecules (Rayleigh scattering) as SZA increases. The decrease in flux ratio is smaller at UV wavelengths (e.g. j O 1 D) because the Rayleigh scattering is much larger for UV light.
Some users of CHABLIS data will require parameterised equations for the flux ratio, and fits over the range of SZA in Fig. 13 allow interpolation outside the summer period. We experimented with fitting polynomials, but they were generally less representative than the pairs of straight lines shown in the figure. These were fitted by eye, as there are too few points at SZA>75 • for a formal least-squares fit, and the observations are within a few percent of the lines chosen. Coordinates of the line pairs are listed in Table 4. The average flux ratio derived from these data is 0.89.
Variability around the lines was several percent, which is probably due to variability in the age of the snow since it fell, as aging reduces albedo. For example, in the snow model of Flanner and Zender (2006), the albedo of fresh snow decreased by 2 to 9% over 5 to 10 days, depending on temperature and grain size. If chemical model comparisons of CHABLIS data demanded flux ratios accurate to better than a few percent, the time since snowfall could be used to improve the parameterisation in Table 4.

Key findings from CHABLIS
As described above, CHABLIS had three specific science foci, i) seasonal studies of oxidant chemistry, ii) year-round studies of the NO y budget, and iii) air/snow transfer studies. Advances were made in each of these three areas of study and the broad conclusions are presented here. The year-round measurements of BrO and IO made during CHABLIS revealed the key role of halogens in determining boundary layer chemistry at Halley. A springtime maximum in both radicals was measured, the observed IO peak at 20 pptv  having only recently been exceeded (Whalley et al., 2007). The NMHC data gathered during CHABLIS allowed a kinetic analysis that revealed significant concentrations of both Cl and Br atoms during ozone depletion events (ODEs), with Cl in the range 1.7×10 3 to 3.4×10 4 atom cm −3 and Br in the range 1.4×10 6 to 2.9×10 7 atom cm −3 (Read et al., 2007). A further analysis using NMHC data revealed a persistent background Cl atom concentration in August of 2.3×10 3 atom cm −3 (Read et al., 2007).
Measurements of OH and HO 2 by FAGE during the austral summer produced the longest duration dataset yet achieved by that instrument . Concentrations were lower than measured at South Pole , and were more in line with data captured at Palmer Station, another coastal Antarctic site (Jefferson et al., 1998). Elevated levels of NO x released from the snowpack (Bauguitte et al., 2008 2 ) contributed to enhancements in OH concentrations through HOx cycling. However, CH 3 O 2 −HO 2 −OH conversion was dominated by the halogen oxides, IO and BrO , which were still present during the summer at mixing ratios of around 2 pptv, with peaks to 5 pptv . The halogens were also responsible for controlling the NO/NO 2 ratio at this site (Bauguitte et al., 2007 2 ). This overarching key role for halogens in boundary layer oxidant chemistry was completely unanticipated, and the results have led to a step-change in our thinking and understanding.
The partitioning of boundary layer NO y is explored in greater detail than previously reported for an Antarctic site . The nitrate radical remained below the 2 pptv instrumental detection limit throughout the duration of the campaign. Unfortunately this threshold is too high to explore upper limits to NO 3 chemistry including any role for this oxidant within the boundary layer. Organic components (peroxyacetyl nitrate and methyl nitrate) overwhelmingly dominate the budget during the winter months, but are in no way associated with concentrations of nitrate measured in newly fallen snow. There thus appears to be no direct link between organic NO y and the ice core nitrate record. The measurements of nitrate in snow (and hence the nitrate record in ice) appear to be closely linked to inorganic species, with summertime peaks apparent both in surface snow nitrate and the inorganic species HNO 3 and HONO. The control over these peaks, and whether the atmosphere is controlling snow or vice versa requires further study. Nitrite in snow is present in very low concentrations, but with its relatively fast photolysis rate, can also be a significant source of NO x to the boundary layer. The range of species measured for the NO y budget study allowed an assessment of NO x sources to the boundary layer. Kinetic analyses were carried out for two case studies, one in summer and one in spring. In both cases, the production rate for boundary layer NO x was dominated by emissions from the snowpack; they were more important than any other gas phase reservoir and indeed greater than the integrated gasphase source. This finding suggests that for certain periods in the Earth's past history, it will be possible to calculate concentrations of Antarctic boundary layer NO x by using the ice core nitrate record.
Taken as a whole, the CHABLIS campaign clearly reveals that the chemistry of the coastal Antarctic boundary layer can only be understood if we consider both the active snow photochemistry that dominates the continental interior region, and the halogen chemistry that dominates the sea ice zone. To understand what is eventually preserved in snowpack and hence in ice cores, it remains essential to follow all these processes, and the resulting air-snow fluxes, throughout the year.

Summary
CHABLIS was an ambitious programme of atmospheric and snowpack sampling that built strongly on existing research. The results have provided exciting new insights into the way the polar atmosphere functions in coastal regions and provided step changes in our understanding about some of the chemical processes occurring there. Antarctica is a vast continent, and a real understanding of its boundary layer chemistry is essential if we are, for example, to access the rich ice core record of past atmospheric, climatic and environmental changes. The CHABLIS campaign has demonstrated what can be achieved with a very limited number of researchers in the field. It provides a model for others to follow in order to extend atmospheric chemistry research in this inhospitable environment.