Notice: Wiley Online Library will be unavailable on Saturday 30th July 2016 from 08:00-11:00 BST / 03:00-06:00 EST / 15:00-18:00 SGT for essential maintenance. Apologies for the inconvenience.
 Although it has recently been established that iodine plays an important role in the atmospheric chemistry of coastal Antarctica, where it occurs at levels which cause significant ozone (O3) depletion and changes in the atmospheric oxidising capacity, iodine oxides have not previously been observed conclusively in the Arctic boundary layer (BL). This paper describes differential optical absorption spectroscopy (DOAS) observations of iodine monoxide (IO), along with gas chromatographic measurements of iodocarbons, in the sub-Arctic environment at Kuujjuarapik, Hudson Bay, Canada. Episodes of elevated levels of IO (up to 3.4 ± 1.2 ppt) accompanied by a variety of iodocarbons were observed. Air mass back trajectories show that the observed iodine compounds originate from open water polynyas that form in the sea ice on Hudson Bay. A combination of long-path DOAS and multiaxis DOAS observations suggested that the IO is limited to about 100 m in height. The observations are interpreted using a one-dimensional model, which indicates that the iodocarbon sources from these exposed waters can account for the observed concentrations of IO. These levels of IO deplete O3 at rates comparable to bromine oxide (BrO) and, more importantly, strongly enhance the effect of bromine-catalyzed O3 depletion in the Arctic BL, an effect which has not been quantitatively considered hitherto. However, the measurements and modeling results indicate that the effects of iodine chemistry are on a much more localized scale than bromine chemistry in the Arctic environment.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
 Iodine chemistry in the lower troposphere has attracted increasing attention over the last decade because of its role in destroying ozone and changing the oxidising capacity of the atmosphere by affecting HOx (= OH + HO2) and NOx (= NO + NO2) chemistry [e.g., Davis et al., 1996; Saiz-Lopez et al., 2008]. Iodine compounds can also lead to the activation of other halogen species like bromine and chlorine through heterogeneous reactions [Vogt et al., 1999]. Furthermore, iodine oxides are implicated in the formation of ultrafine particles which can grow to act as cloud condensation nuclei (CCN) and thus affect climate [McFiggans, 2005; O'Dowd et al., 2002]. High levels of iodine species have been observed at coastal locations in midlatitude areas, where ultra-fine particle bursts have also been seen [Alicke et al., 1999; Mahajan et al., 2009; McFiggans et al., 2004; Peters et al., 2005; Saiz-Lopez and Plane, 2004]. Recently, open ocean measurements have revealed significant levels of iodine monoxide (IO), indicating its possible global importance in the marine environment [Read et al., 2008].
 Ozone depletion and mercury depletion events have been reported in the Arctic BL, which have been attributed to catalytic reactions involving bromine [e.g., Barrie et al., 1988; Bottenheim and Chan, 2006; Schroeder et al., 1998]. The presence of iodine compounds in the Arctic would be important, not only because of its individual effect on O3, but also because it couples with bromine through the reaction IO + BrO [Gilles et al., 1997], leading to enhanced O3 and mercury depletion, as has been suggested for Antarctica [Saiz-Lopez et al., 2008].
 Sources of gas-phase iodine in the BL are thought to be mainly biogenic, in the form of I2 and photolabile iodocarbons emitted by exposed macroalgae or by phytoplankton as a response to oxidation stress [Carpenter, 2003; Karlsson et al., 2008; Moore et al., 1996; Saiz-Lopez and Plane, 2004]. Recently, O3 uptake on the ocean surface has also been recognized to cause emission of iodine-containing compounds [Martino et al., 2009]. In polar regions, the release of iodine-containing organic compounds from ice algae and seawater is postulated to be the main source of gaseous iodine [Carpenter et al., 2007; Moore and Zafiriou, 1994; Reifenhäuser and Heumann, 1992]. Molecular iodine, biogenically produced by algal colonies on the underside of sea ice and transported to the surface through brine channels, is speculated to be the source of high levels of IO in Antarctica. The greater average thickness of Arctic sea ice compared to Antarctic sea ice is a potential reason for the lack of observable IO in the Arctic, although this hypothesis has yet to be confirmed (A. Saiz-Lopez, Laboratory for Atmospheric and Climate Science, personal communication, 2010). Furthermore, algal colonies on the underside of sea ice are less predominant in the Arctic than around Antarctica [Thomas and Dieckmann, 2003].
2. Site Description and Measurements
 A field study was performed as part of the “impact of combined iodine and bromine release on the Arctic atmosphere” (COBRA) project, a UK contribution to the International Polar Year (IPY). The study in February–March 2008 was based near Centre d'Études Nordiques, Kuujjuarapik (55°18′N, 77°44′W), which is located on the eastern coast of Hudson Bay, Canada (Figure 1). The site was located at the sea ice edge, which was frozen throughout the campaign. Almost the entire surface of Hudson Bay is covered with sea ice in winter, although in early spring large open water areas, termed polynyas, and smaller cracks in the sea ice, termed leads, start to appear as the ice melts. These open water areas, which can be more than 100 km long in the case of polynyas, are temporary in nature and generally lasted for 1–3 days during the COBRA campaign.
2.1. IO Measurements: LP-DOAS
 Measurements of IO were made for 11 days using the long path-DOAS technique. The LP-DOAS optical path ran from the measurement site to a retroreflector array, which was placed on an island about 5.5 km to the north, giving a total optical path of 11 km (Figure 1). The height of the light beam ranged from about 5 m above the sea ice near the instrument to about 7 m at the retroreflectors. Details of the LP-DOAS and the measurement procedure are given elsewhere [Plane and Saiz-Lopez, 2006]. Briefly, the instrument employs a 0.5 m Czerny-Turner spectrometer with a 1200 grooves mm−1 grating, resulting in a spectral resolution of 0.25 nm covering a spectral region spanning ∼40 nm. Recorded spectra were converted into optical densities and the contributions of the individual absorbing species were then determined by using singular value decomposition to fit simultaneously a library of reference absorption cross sections. The reference spectra fitted in the 425–440 nm window were IO [Gómez Martin et al., 2005], NO2 [Vandaele et al., 1998], and H2O [Rothman et al., 2003]. Examples of the fit quality are shown in Figure 2. The spectra were averaged for ∼30 min to improve the signal-to-noise ratio. The typical detection limit of the instrument was around 1 ppt depending on the visibility and spectra quality, corresponding to a root mean square (RMS) of the residual of 1.1 × 10−4.
2.2. IO Measurements: MAX-DOAS
 The scattered sunlight DOAS technique is a passive absorption spectroscopy technique using the Sun as light source. This method was originally deployed in zenith geometry only to study the stratospheric composition [e.g., Solomon et al., 1987]. Pointing a telescope a few degrees above the horizon significantly enhances the tropospheric light path and hence provides information about absorbers in the boundary layer. By means of using the multiaxis (MAX) observation geometry, i.e., collecting photons from several discrete viewing directions, some information about the vertical distribution of a tropospheric species can be collected by simulating the photon path from the Sun into the telescope with the help of a radiative transfer model.
 Here the deployed MAX-DOAS instrument is a scanning system with an optical fiber bundle mounted to a stepper motor. This provides the opportunity to collect scattered sunlight from discrete elevation angles between −5° and 40°, as well as the zenith.
 A 2.5 cm diameter quartz lens with a 7.5 cm focal length was placed in front of the fiber optic to limit the field of view to ∼1°. This minitelescope with the stepper motor was placed in a heated housing that also held a Ne-Hg lamp for wavelength calibrations. Spectra of this lamp were recorded on a daily basis after sunset. The combination of the temperature-stabilized spectrometer with the cooled CCD detector was identical to the LP-DOAS. The MAX-DOAS pointed toward the north, parallel to the LP-DOAS path length (Figure 1). For the IO analysis, the NO2 at 220 K [Vandaele et al., 1998], O3 at 221 K [Burrows et al., 1999], the Ring [Vountas et al., 1998], and IO [Gómez Martin et al., 2005] cross sections were simultaneously fitted to the measured optical densities together with a polynomial of the order of 5. The spectral deconvolution was performed in a wavelength window between 416 and 439 nm with the daily zenith noon spectrum as reference. A straylight correction was also applied. In this setup, the selected angles were 1°, 3°, 5°, 7°, 9°, 15°, 30°, and 90° and spectra were recorded with a 30 s exposure time per angle. Typically, the RMS residual varied between 2.5 and 5 × 10−4. IO could not be detected in the spectra. The upper limit or detection limit for the IO SCDs was estimated to be 4 × 1013 molecule cm−2.
2.3. Halocarbon Measurements
 Approximately hourly in situ measurements of reactive halocarbons were made at the experimental container site using a Perkin Elmer (USA) Turbomass gas chomatograph/mass spectrometer (GC/MS) system connected to a Perkin Elmer thermal desorption (TD) unit. Air was sampled at 100 mL min−1 through a clean metal bellows pump, connected via a T-piece to a 10 m PFA Teflon sample line (0.5″ id) located ∼3 m above ground. This sample line was separately pumped at a flow rate of several L min−1. Sample air was dried by passage through a Nafion™ membrane (Brunswick Instruments, USA) molecular sieve dryer before preconcentration (trapping) of 3 L onto a three-stage carbon-based adsorbent microtrap (Air monitoring trap, Perkin Elmer UK) held at −30°C. Volatiles were thermally desorbed by flash heating to 350°C and transferred to a 25 m, 0.32 mm (id) Porabond Q column (Chrompack). The MS was operated in selective ion recording mode (SIR) mode with a dwell time of 0.15 s for each ion monitored for identification and quantification. A gas standard containing low-ppt mixing ratios of the target halocarbons was prepared prior to the campaign by purging zero grade nitrogen (BOC UK) through a 20 mL mixture of pure halocarbons diluted in HPLC (Fisher UK) water into an evacuated, cleaned Aculife cylinder (10 L, CK Gases), which was then pressurized to 70 bar with zero grade nitrogen (BOC UK). This gas standard was used in the field to calibrate the GC/MS and was quantified in the University of York laboratory against an in-house permeation system [Wevill and Carpenter, 2004] before and after the campaign.
2.4. OH, HO2, NO, NO2, and O3 Measurements
 In situ measurements of OH, HO2, NO, NO2, and O3 were also made at the experimental container site. The technique used by the Leeds team for OH detection was Fluorescence Assay by Gas Expansion (FAGE), which employs on-resonance pulsed laser-induced fluorescence at low pressure [Dusanter et al., 2008; Faloona et al., 2004; Hard et al., 1984; Smith, 2007]. Measurements of NO were made using a single channel chemiluminescence NO analyser (Ecophysics CLD 780TR) [Dias-Lalcaca et al., 1998; Jones et al., 2001]. Periodically the sample flow was switched through a photolytic converter (Ecophysics PLC 760SR), which uses filtered UV light from a Xe lamp (300–400 nm) to convert NO2 to NO, hence allowing measurement of ambient NO2 [Ryerson et al., 2000]. Measurements of O3 were made using a UV absorption instrument (Thermo Environmental Instruments 49C). The instrument was calibrated using a primary O3 standard (Thermo Environmental Instruments 49C Primary Standard), which itself was certified at the National Physical Laboratory, UK.
 Besides the instruments listed above, standard meteorological data including wind direction, wind speed, solar radiation, relative humidity, and temperature were also available.
Figure 3 (second frame) shows that IO was observed above the detection limit of the LP-DOAS instrument (∼1–1.5 ppt) on 6 days, although on some days it was detected only for a short period of 1–2 h. The maximum concentration seen during the campaign was 3.4 ± 1.2 ppt. IO was never observed above the detection limit during the night. Using the MAX-DOAS, only an upper limit for the IO differential slant column densities (dSCD) of 4 × 1013 molecules cm−2 for an individual measurement of 30 s could be estimated. Additionally, elevated levels of iodocarbons were observed (Figure 3, third frame), with maximum concentrations of CH3I = 3.83 ppt (DL = 0.15 ppt), CH2IBr = 0.11 ppt (DL = 0.01 ppt), and CH2ICl = 0.165 ppt (DL = 0.008 ppt). The most photolabile iodocarbon, CH2I2, could not be detected above an estimated upper limit of 0.05 ppt during the entire campaign. The elevated levels of iodocarbons were observed mostly during the night, with an exception on 19 February 2009, when relatively high levels were also present during the day.
 IO and the iodocarbons were not observed ubiquitously during the campaign. Elevated levels were seen only in air masses which, according to air mass back trajectories, had passed over polynyas or open leads in relatively close proximity to Kuujjuarapik. The larger polynyas and smaller open leads were identified from the sea ice concentration maps generated from NASA's AMSR-E sensor on a wider scale (5 km resolution) and from high resolution synthetic aperture radar images of ESA's Envisat ASAR system on a finer scale (150 m resolution). The differences in the type of air masses arriving at the measurement site were studied using their 5 day back trajectory histories as predicted by the National Oceanic and Atmospheric Administration (NOAA) HYSPLIT model [Draxler and Rolph, 2003] using the Global Data Assimilation System (GDAS) meteorological data field (Figure 4). The back trajectories showed that the height of the air parcels were mostly less than 50 m during transit. Measurable concentrations of IO were seen only in air masses that had passed over a polynya at a distance of 200–250 km away, while elevated levels of iodocarbons were present in air masses that had been over polynyas or open leads less than 250 km away. Elevated concentrations of IO and iodocarbons were not present in air masses that had passed over ice, land, or polynyas much further from the site (∼1000 km; Figures 3 and 4). A point to note is that on 19 February, elevated levels of iodocarbons were seen during the daytime, whereas IO was not seen above the detection limit (1.5 ppt). During this period, the air mass passed over an open lead ∼40 km away from the measurement site.
Figure 3 (top) shows the record of O3 measurements over the same period as the IO and iodocarbon measurements. The O3 mixing ratios were generally lower and variable (5–40 ppb) in the presence of elevated levels of IO e.g., on 20 and 24 February (Figure 3). However, significant levels of iodocarbons and IO were not seen during the largest ozone depletion event (26–27 February), with IO measurements made only during the morning of the event. The air mass back trajectories during this period indicate the influence of a large polynya far to the north, about 1000 km away from the measurement site.
4. Modeling and Discussion
 To interpret the observations and quantify the impact of iodocarbons on formation of reactive iodine species and their impact on the oxidizing capacity of the BL, we use the one-dimensional Tropospheric Halogen Chemistry Model (THAMO) [Saiz-Lopez et al., 2008]. The iodine chemistry scheme in THAMO is updated according to Mahajan et al. .
 The combined MAX-DOAS and LP-DOAS measurements constrain the vertical distribution of IO, in the following way. Simultaneous MAX-DOAS measurements of the O4 dSCD can be used to determine the radiative transfer in the atmosphere [Wagner et al., 2007]. On average, the aerosol extinction was very small and an O4 dSCD as large as 1.5 × 1044 molecule2 cm−5 at 446 nm (where the O4 column density is expressed in terms of molecules of O2, by multiplying by the equilibrium constant for O2 dimerization [Greenblatt et al., 1990]) was retrieved for an elevation angle of 1° around noon. Simulations of the O4 dSCDs with the radiative transfer model SCIATRAN [Rozanov et al., 2002] could best reproduce the measurements if scattering on aerosols was excluded from the calculations. This allows the dSCDs of other trace gases to be interpreted with a simple geometric approach. If the average IO mixing ratio of 2.8 ppt (≈7 × 107 molecule cm−3) measured with the LP-DOAS on 24 February is assumed to be well mixed along the line of sight then, to be consistent with the MAX-DOAS IO column density of ≤4 × 1013 molecule cm−2, the slant path length would be only ∼5.8 km. For an elevation angle of 1.0 ± 0.5°, this slant path can be converted to a block profile with a layer thickness of 101 ± 51 m, indicating that the IO is confined to layers close to the surface.
 Hence, for modeling purposes, we assume that the BL was capped at 100 m during the times when IO was observed by the LP-DOAS. The vertical eddy diffusion coefficient (Kz) is calculated using the measured wind speed data at 3 m during COBRA, according to the procedure described by Stull  and ranged from 7.5 × 102 cm2 s−1 at the ground to a maximum of 1.5 × 104 cm2 s−1 at 30 m after which it is assumed to decrease to a small value at 100 m (2.0 cm2 s−1), so that there is insignificant exchange with the free troposphere. The sensitivity of the model to the Kz profile is discussed later. THAMO is employed for these model runs with a vertical resolution of 0.5 m. The concentrations of all iodine species, O3 and NOx were allowed to vary. The model was initialized with [O3] = 30 ppb, [NO2] = 50 ppt, [HO2] = 5 ppt (midday), [OH] = 0.03 ppt (midday), aerosol surface area = 1 × 10−6 cm2 cm−3, [CO] = 160 ppb, [CH4] = 1750 ppb, [HCHO] = 300 ppt (upper limit according to the DOAS measurements), [ethane] = 3 ppb, [propane] = 1.8 ppb, [propene] = 200 ppt, temperature = −20°C, and relative humidity = 70%. These are all typical of the measurements during the COBRA campaign.
 The levels of IO and iodocarbons measured during this study depend strongly on the history of the air mass, indicating that the emissions from polynyas are their major source, since neither IO nor reactive iodocarbons were detected in air masses passing over ice or land. It should be noted that although some mechanisms for the abiotic release of iodocarbons in this region have been suggested in the past [Carpenter et al., 2005; Martino et al., 2009], the precise mechanisms are still unknown and discussion of them is beyond the scope of this paper. The polynyas were typically ∼15–20 km wide (Figure 4). Hence, at the average observed wind speed of 4 m s−1, an air parcel would have taken about 1 h to travel over this area of exposed water. In order to simulate this in the model, we prescribe an iodocarbon pulse lasting for 1 h. This air parcel would then take 14–17 h to reach the measurement site in cases where the polynyas are 200–250 km away (this travel time is confirmed by the air mass back trajectories).
 We consider two scenarios for testing whether the model can reproduce both the IO and the iodocarbon measurements. In scenario 1, when the measurements of IO peaked in the afternoon (e.g., 24 February, Figure 3), we inject the iodocarbons an hour after model initialization at midnight; they then arrive at the measurement site at 1500 h. In scenario 2, where the iodocarbons were seen to peak during nighttime (Figure 3), we inject the iodocarbons an hour after model initialization at midday, so that they arrive at the measurement site at 0300 h. The iodocarbon fluxes were not measured during the campaign and hence the model initialization flux was tuned to reproduce the iodocarbon observations downwind at the site. Fluxes of CH2I2 = 1 × 106 molecule cm−2 s−1, CH2IBr = 2 × 109 molecule cm−2 s−1, CH2ICl = 5 × 107 molecule cm−2 s−1, and CH3I = 2 × 108 molecule cm−2 s−1 are needed to reproduce both the daytime and nighttime levels of iodocarbons on the 24 February (Figure 3).
 Scenario 1 model predictions for IO, O3 and the two iodocarbons, CH2IBr (lifetime ∼ 1 h) and CH2ICl (lifetime ∼ 2.7 h), which contribute most to the production of IO due to their relatively short lifetimes, are shown in Figure 5. Immediately after the injection of iodocarbons, the concentrations of CH2IBr and CH2ICl reach as high as ∼30 and ∼0.8 ppt at 10 m. They are then diluted vertically through the 100 m BL with time (Figure 5). At 14 h after injection, the model predicts only 4 × 10−4 ppt of CH2IBr and 0.011 ppt of CH2ICl at the height of measurements, which is in agreement with the daytime levels of <0.001 and 0.01 ppt for CH2IBr and CH2ICl, respectively (Figure 3, third frame). CH2I2 is predicted to be under the upper limit of 0.05 ppt, as it undergoes rapid photolysis. The model also predicts the presence of 2.8 ppt of IO, which is comparable to the daytime observations (Figure 3, second frame). The model indicates that the IO column would be well mixed in the first 100 m. When the IO is confined to the BL, this would correspond to a column density of ∼6 × 1011 molecules cm−2, which is close to the upper limit of the MAX-DOAS instrument (7 × 1011 molecules cm−2). Note that this modeled column density is also below the satellite detection limit of ∼1.7 × 1012 molecules cm−2 for a single measurement (allowing an air mass factor of 4) and is thus consistent with the lack of satellite observations of IO in the Arctic [Schönhardt et al., 2008].
 The required iodocarbon flux is sensitive to the Kz profile used. For the base case, the Kz is considered to reduce with height after 30 m to restrict mixing into the free troposphere. If we assume that Kz does not decrease with height but stays constant between 30 and 100 m, the fluxes necessary to reproduce the IO and iodocarbons measurements are CH2I2 = 1 × 107 molecule cm−2 s−1, CH2IBr = 6 × 109 molecule cm−2 s−1, CH2ICl = 3 × 108 molecule cm−2 s−1, and CH3I = 4.2 × 109 molecule cm−2 s−1. That is, approximately one order of magnitude larger.
 In THAMO, the iodine species (e.g., HOI, INO3) are considered to deposit on the surface at a flux calculated using a deposition velocity of 0.5 cm s−1 in the lowest box. These species are assumed to then recycle back into the gas phase, similar to recycling on aerosol surfaces, as interhalogens (IBr or ICl). We do not consider production of extra iodine from the surface as a result of this recycling, as has been reported for bromine compounds over saline surfaces [Wennberg, 1999]. If extra production of iodine occurs, then the flux necessary to explain the [IO] observed is greatly reduced, but the model also predicts that IO would be seen ubiquitously during the daytime in all types of air masses, which was not the case.
 In scenario 2, after 14 h (i.e., at 0300 h) the model predicts 0.06 ppt of CH2IBr and 0.09 ppt of CH2ICl at the height of measurements, in agreement with the nighttime levels of ∼0.06 and ∼0.08, respectively, on 24 February (Figure 6). Again, CH2I2 is predicted to be under the upper limit of 0.05 ppt. In scenario 2, no IO is produced during the night, as confirmed by measurements. Interestingly, the model predicts that higher levels of IO would be seen near the large polynyas, with the predicted IO concentrations peaking at ∼11 ppt at 10 m, 1 h after the iodocarbon injection. Nevertheless, the vertical column would still be under the satellite detection limit.
 On 19 February, elevated levels of iodocarbons were seen during the day, but IO was not observed above the LP-DOAS detection limit. During this period, the air mass back trajectories show the influence of an open lead only 40 km from the site (Figure 4). This lead is much smaller than the polynyas seen on the western or northern edge of Hudson Bay (e.g., 24 February). These open leads were typically only 1–4 km wide and hence would likely contribute lower levels of iodocarbons to air masses passing over them. Indeed, using similar fluxes to scenarios 1 and 2, an exposure time of about 10 min is necessary to reproduce the observed daytime iodocarbons on 19 February. The resulting IO concentration at the measurement site is predicted to be only ∼1 ppt, which was not above the detection limit of the BL-DOAS.
 After a period of more than 24 h, the model predicts that all the IO generated through the photolysis of iodocarbons is lost to particle formation. This loss to particles is considered to be irreversible in the model and hence the model predicts that after 24 h the IO concentrations would be under the detection limit. This was the case on 26 February, when the air masses that had passed over the polynyas ∼1000 km away (Figure 4) would have taken 70 h to reach the site. The iodocarbons are also permanently lost through photolysis and hence were not observed at elevated levels (Figure 3).
 In THAMO, the formation and growth of iodine oxide particles (IOPs) occurs through a mechanism recently proposed by Mahajan et al. (manuscript in preparation, 2010). I2O4 is treated as the primary polymerizing species, on which I2O3 can condense. Both species can condense on larger polymers, which can all also coagulate with each other. These reactions are assumed to occur at their respective collision frequencies (or kernels) (R. Saunders, University of Leeds, personal communication, 2009). The IOPs are assumed to grow through the uptake of acids (principally sulfuric acid) and water, once they have reached a critical size (diameter = 2 nm). Thus the survival of the particles depends on competition between growth through the uptake of acids and water against loss to background aerosols. Assuming an average [H2SO4] of 1 × 106 molecules cm−3 based on past measurements in the Arctic [Weber et al., 2003], the model predicts that ∼120 particles cm−3 of diameter >20 nm are generated 2 h downwind of the polynyas in scenario 2 (at a height of 10 m). The production of these relatively large IOPs would then have ceased, because it is a highly nonlinear function of [IO]. Subsequent mixing and dilution during the remaining 12 h of transit to Kuujjuarapik probably explains why no significant increase in Aitken size range particles was observed during the periods when IO was above the LP-DOAS detection limit (J. R. Dorsey, University of Manchester, unpublished data, 2009).
Figure 5 also shows the effect of iodine species on O3. Over 1 day, the O3 concentration reduces by ∼1.6 ppb. This assumes that there is no entrainment of O3 into the BL or diffusion of halogens into the free troposphere. As mentioned above, the MAX-DOAS data indicates that the IO was limited to a surface layer below about 100 m, implying a capped boundary layer. Note that if the BL were uncapped, IO would diffuse into the free troposphere and would not be observed above the detection limit of the LP-DOAS unless the iodocarbon fluxes were substantially larger. However, this would lead to a large IO vertical column density, which was not observed by the MAX-DOAS. Thus, the effect of iodine will be strongly dependent on the meteorological conditions.
Figure 3 shows that elevated IO coincides with some of the observed ozone depletion events. On 24 February, this phenomenon is clearly noticeable with a strong anticorrelation between IO and O3 (Figure 3). However, during the largest ozone depletion event (26 February to 27 February), IO was not observed above the detection limit when measurements were made during the morning of the 26 (the polynyas were ∼1000 km away). During this period, the DOAS instruments measured high levels of BrO (25–30 ppt) (H. Oetjen, manuscript in preparation, 2010), which most likely caused the large ozone depletion event. Unfortunately, simultaneous measurements of IO and BrO were not made as the species are retrieved in different DOAS spectral windows, and both the DOAS instruments were set to measure the same species at any given time.
 The decrease of ∼10 ppb of O3 on 24 February indicates that O3 depletion is observed during elevated IO episodes. However, iodine chemistry on its own cannot explain the loss of 10 ppb over the day: the model shows that only ∼1.6 ppb would be destroyed. The larger O3 depletion is almost certainly due to the presence of bromine species in the same air mass. In the presence of IO, the effect of BrO on O3 is enhanced due to the coupling reaction between IO and BrO [Gilles et al., 1997]. The presence of iodine species (e.g., HOI and INO3) in the BL can also lead to the activation of bromine and chlorine compounds [Vogt et al., 1999].
 Several measurements in the past have shown elevated levels of BrO in the Arctic BL, and bromine chemistry is widely regarded as the main O3 destruction process [Barrie et al., 1988; Bottenheim and Chan, 2006; Tuckermann et al., 1997]. However, our model calculations show that under conditions seen in COBRA, IO greatly increases the O3 depletion potential of BrO. At midday, 3 ppt of IO would deplete O3 at a rate of 0.28 ppb h−1, while 25 ppt of BrO, as was measured during COBRA and has also been reported previously [Hönninger et al., 2004], would lead to O3 destruction at a rate of 0.62 ppb h−1. Importantly, if IO is present in the same air mass as BrO, then the total effect of both halogens leads to O3 destruction at a rate of ∼2 ppb h−1. This is more than 3 times the effect of BrO alone, demonstrating that the presence of a relatively small amount of IO causes a large increase in the O3 destruction potential of BrO. The increased effect of IO on the O3 depleting potential of BrO at different concentrations is presented in Figure 7, which shows that the measured levels of IO should cause a substantial change to the oxidising capacity of the BL.
 Finally, the observations indicate that the source of IO is the emission of iodine-containing compounds from polynyas formed in the sea ice. Past observations have shown that there has been an increase in the number of polynyas and open leads, along with a decline in the sea ice thickness, in the Arctic over the last few decades [Gloersen and Cambell, 1991; Lindsay and Zhang, 2005; Parkinson and Cavalieri, 2008]. If iodine compounds are indeed emitted from polynyas and open leads, our measurements and model results indicate that iodine chemistry could play a much larger role in the Arctic environment in the future.
 Elevated levels of iodocarbons and IO were observed in the sub-Arctic BL, indicating that iodine chemistry can play an important role in this region. Air mass back trajectories indicate that high IO events were associated with passage of air over open water polynas in the Hudson Bay area. A combination of LP-DOAS and MAX-DOAS observations suggest that the IO was limited to a layer near the surface, less than 100 m in height. A model assuming that polynas were the main sources of IO precursors was used to study the evolution of the iodine compounds and to quantify their effect on the oxidizing capacity of the BL. The model indicates that iodocarbons were the major source of the observed levels of IO and that the observed levels of IO can greatly accelerate the bromine-catalyzed depletion of O3. However, iodine chemistry is predicted to occur only on a local scale.
 We thank Claude Tremblay and Centre d'Études Nordiques, Whapmagoostui-Kuujjuarapik for logistical support and acknowledge the UK NERC for financial support (COBRA project NE/D005914/1). We would like to thank James Dorsey for the provision of aerosol measurements and Alfonso Saiz-Lopez, who worked on the initial versions of THAMO. A.S.M. thanks the School of Chemistry, University of Leeds, for a Ph.D. studentship. P.E. thanks NERC for a Ph.D. studentship.