Interactions of bromine , chlorine , and iodine photochemistry during ozone depletions in Barrow , Alaska

The springtime depletion of tropospheric ozone in the Arctic is known to be caused by active halogen photochemistry resulting from halogen atom precursors emitted from snow, ice, or aerosol surfaces. The role of bromine in driving ozone depletion events (ODEs) has been generally accepted, but much less is known about the role of chlorine radicals in ozone depletion chemistry. While the potential impact of iodine in the High Arctic is more uncertain, there have been indications of active iodine chemistry through observed enhancements in filterable iodide, probable detection of tropospheric IO, and recently, observation of snowpack photochemical production of I2. Despite decades of research, significant uncertainty remains regarding the chemical mechanisms associated with the bromine-catalyzed depletion of ozone, as well as the complex interactions that occur in the polar boundary layer due to halogen chemistry. To investigate this, we developed a zero-dimensional photochemical model, constrained with measurements from the 2009 OASIS field campaign in Barrow, Alaska. We simulated a 7-day period during late March that included a full ozone depletion event lasting 3 days and subsequent ozone recovery to study the interactions of halogen radicals under these different conditions. In addition, the effects of iodine added to our Base Model were investigated. While bromine atoms were primarily responsible for ODEs, chlorine and iodine were found to enhance the depletion rates and iodine was found to be more efficient per atom at depleting ozone than Br. The interaction between chlorine and bromine is complex, as the presence of chlorine can increase the recycling and production of Br atoms, while also increasing reactive bromine sinks under certain conditions. Chlorine chemistry was also found to have significant impacts on both HO2 and RO2, with organic compounds serving as the primary reaction partner for Cl atoms. The results of this work highlight the need for future studies on the production mechanisms of Br2 and Cl2, as well as on the potential impact of iodine in the High Arctic. Published by Copernicus Publications on behalf of the European Geosciences Union. 9652 C. R. Thompson et al.: Interactions of bromine, chlorine, and iodine photochemistry


Introduction
The importance of halogen chemistry in polar regions has been well established over the past few decades since the observation of near-surface boundary layer ozone depletion in the 1980s (Oltmans and Komhyr, 1986;Barrie et al., 1988;Bottenheim et al., 1990).Since that time, ozone depletion events (ODEs) have been observed at numerous Arctic and Antarctic locations (e.g., Bottenheim et al., 2002;Saiz-Lopez et al., 2007b;Simpson et al., 2007;Oltmans et al., 2012).ODEs are characterized by episodic depletions of tropospheric ozone from background mole ratios of approximately 35 parts per billion by volume (ppbv) to less than 2 ppbv over periods of hours to days.These events are known to occur following the onset of polar sunrise, and continue through polar spring, when temperatures are low and snow and sea ice are still present, with a stable atmospheric boundary layer (Simpson et al., 2007).
Photochemical reactions involving halogen radicals, notably bromine, are thought to be the primary cause of ODEs (see Simpson et al., 2007, for a review).The chemical destruction of O 3 by Br can be described by Reactions (R1)-(R3) (Platt and Hönninger, 2003).
The efficiency of Br atoms in destroying ozone is due primarily to its relative lack of atmospheric sinks, and thus its relatively high gas-phase concentration, as well as its ability to recycle and regenerate from temporary sink species.Bromine atoms do not react appreciably with methane or other saturated hydrocarbons; thus, its primary sinks (other than O 3 ) consist of a few oxygenated volatile organic compounds (OVOCs) (e.g., aldehydes) or unsaturated hydrocarbons, which result in production of HBr, and HO x species.BrO can react to regenerate a Br atom (via Reaction R3, followed by Reaction R1), or react with HO 2 to produce HOBr, which can in turn lead to the production of two Br atoms through a heterogeneous reaction mechanism termed the bromine explosion (Tang and McConnell, 1996;Vogt et al., 1996), as shown below.
The presence of chlorine chemistry in the Arctic has been well recognized through indirect measurements of hydrocarbons (Jobson et al., 1994;Ariya et al., 1998;Keil and Shepson, 2006;Tackett et al., 2007) and through detection of photolyzable chlorine species (defined as [Cl 2 + HOCl]) (Impey et al., 1997); however, few direct measurements of chlorine species have been reported.The only currently reported measurements of ClO were by Tuckermann et al. (1997), who detected ClO at Spitsbergen.Unlike bromine, chlorine radicals efficiently oxidize a wide range of pollutants and volatile organic compounds (VOCs), often with faster rate coefficients than analogous reactions by the hydroxyl radical (OH); thus, chlorine has an abundance of atmospheric sinks.Estimates of polar region Cl atom concentrations using hydrocarbon decay methods are typically in the range of 10 4 -10 5 molecules cm −3 (Jobson et al., 1994;Ariya et al., 1998;Boudries and Bottenheim, 2000), approximately 2-3 orders of magnitude lower than analogous estimates of Br (Cavender et al., 2008).
Like bromine, chlorine can react directly with O 3 , generating a ClO radical via Reaction (R8).The presence of ClO may also promote bromine-induced depletion of O 3 through the fast cross-reaction of BrO and ClO that serves to regenerate Br atoms (Reaction R9) (Le Bras and Platt, 1995;Platt and Hönninger, 2003).
Due to analytical challenges, few tropospheric observations of Cl 2 and ClO exist; therefore, the role of Cl in ozone depletion events remains uncertain and has been much debated.Typical estimated Cl concentrations are likely too low for chlorine to be a significant direct contributor to ozone depletion.However, elevated levels of Cl 2 (exceeding 100 pptv) were recently observed during the Ocean-Atmosphere-Sea Ice-Snowpack (OASIS) 2009 campaign in Barrow, Alaska, using chemical ionization mass spectrometry (CIMS) (Liao et al., 2014).The impact that such high levels of Cl 2 could have on ODEs or on the oxidation chemistry of the Arctic troposphere has not been fully investigated.Although Reaction (R9) can act to enhance the rate of ozone depletion by propagating the Br cycle, formaldehyde, propanal, acetaldehyde and HO 2 are all produced as by-products of VOC oxidation by Cl and are efficient sinks for Br or BrO radicals (Shepson et al., 1996;Sumner et al., 2002).Therefore, it is likely that the interaction between chlorine and bromine is complex.Modeling studies simulating ozone depletion events often use BrCl as the primary source of Cl atoms (Calvert and Lindberg, 2003;Piot and Von Glasow, 2008), and thus [Cl] is quite low and often insignificant.Few have added significant Cl 2 sources (Sander et al., 1997;Piot and von Glasow, 2009).As a result, our understanding of chlorine chemistry in the Arctic is limited.
Iodine may play a significant role in the photochemical reactions leading to ODEs (Sander et al., 1997;Saiz-Lopez et al., 2007b, 2008;Mahajan et al., 2010) based on recent ground-based and satellite observations of IO of up to 20-50 pptv over snow-and ice-covered surfaces in coastal Antarctica (Saiz-Lopez et al., 2007a, b;Schönhardt et al., 2008;Frieß et al., 2010).Although IO has been routinely detected at high levels in Antarctica, there has been no successful set of IO measurements in the High Arctic to date.However, this is possibly due to limitations of current analytical methods rather than an absolute absence of iodine chemistry.Several studies in the Arctic have indicated the presence of iodine species through measurements of springtime peaks in filterable iodine (Sturges and Barrie, 1988;Barrie et al., 1994;Sirois and Barrie, 1999) and total gaseous iodine (Martinez et al., 1999).Hönninger (2002) was able to detect IO during only one instance above the detection limit at Alert in 2000 using long path DOAS, corresponding to 0.7 pptv of IO if averaged over the 10 km light path.Mahajan et al. (2010) also observed up to 3.4 pptv IO at the sub-Arctic location of Kujjuarapik, Canada.Most recently, I 2 at levels ≤ 0.5 pptv has been observed by our research group at Barrow, Alaska, by CIMS, lending direct evidence supporting the presence of at least low levels of iodine chemistry in the Arctic (Raso et al., 2015).
Due to the lack of iodine observations in the Arctic, models often omit iodine chemistry when simulating ODEs.However, this may significantly underestimate the rate of ozone depletion because the BrO + IO cross-reaction (Reaction R10), which propagates the Br chain, has a rate constant that is approximately 2 orders of magnitude faster than the BrO self-reaction.
Besides causing tropospheric ODEs, halogens can also impact HO x and NO x chemical cycles.The presence of reactive halogen species has the general effect of shifting the [HO 2 ] / [OH] ratio towards OH (Platt and Hönninger, 2003;Thomas et al., 2012), primarily through Reactions (R11) and (R12), where X can be Br, Cl, or I.
Bromine and chlorine atoms can react with formaldehyde (and higher molecular-weight aldehydes), leading to production of hydrogen halides and HO x .Alternatively, halogen atom reaction with HO 2 can act as a HO x sink.The inclusion of chlorine chemistry further impacts the HO x chemistry because RO 2 , HO 2 , and carbonyl compounds are produced as side-products of Cl oxidation of VOCs (e.g., Reactions 13-15). Cl Halogen chemistry also generally increases the Leighton ratio ([NO 2 ] / [NO]) through conversion of NO to NO 2 via Reactions (R14) and (R16) (Platt and Hönninger, 2003), where X is Br, Cl, or I. X or XO can also react with NO 2 to form XNO 2 or XONO 2 , respectively (Reactions R17-R18); however, the net effect of halogen chemistry is an increase in the [NO 2 ] / [NO] ratio.
XO + NO → X + NO 2 (R16) In this work, we utilize a zero-dimensional model to further investigate the chemical interactions occurring in the Arctic related to the chemistry of halogen radicals and the interactions between bromine, chlorine, and iodine.The OA-SIS 2009 campaign, conducted in March and April in Barrow, Alaska, provides a valuable opportunity to perform a unique study of halogen chemistry using direct observations of a variety of atmospheric species, including HO x and XO x radicals, and molecular halogens, with high time resolution measured concurrently and from a single location.During OASIS, measurements of a large suite of saturated, unsaturated, and oxygenated VOCs, numerous halogenated species (e.g., Br 2 , Cl 2 , BrCl, HOBr, BrO, ClO), OH, HO 2 , NO, NO 2 , HONO, O 3 , and actinic radiation were conducted.We constrain our model with time-resolved data from OASIS, in order to investigate the following questions.

Field campaign and measurements description
The OASIS (Ocean-Atmosphere-Sea Ice-Snowpack) 2009 campaign occurred during the months of February through April as part of the most recent International Polar Year (2007)(2008)(2009).The motivation for OASIS was to investigate the chemical and physical processes leading to ozone and mercury depletion events in polar spring.The OASIS campaign represents the largest suite of simultaneous and colocated atmospheric trace gas measurements conducted in the Arctic to date, and as such, has provided a rich data set for understanding chemical processes in the Arctic troposphere.
The atmospheric measurements employed in this work were conducted from mobile instrument trailers located near the Barrow Arctic Research Consortium (BARC) facility on the Naval Arctic Research Laboratory (NARL) campus.Winds arriving at the measurement site are primarily northeasterly (i.e., from the sea ice), and, thus, should represent background conditions with influence from only natural processes and snow-ice interactions.Occasional winds from the west (i.e., from the town of Barrow) are easily identifiable by coincident enhancements in both NO x and CO.
Table 5 lists the measurements conducted at OASIS that were used in this modeling study.This is not an exhaustive list of all measurements performed during the campaign.As the methods have all been described previously, they will only be described very briefly here.Ozone, NO, and NO 2 were measured using the well-established chemiluminescence technique as described in Ridley et al. (1992) and Ryerson et al. (2000).O 3 and NO x were measured at three heights (2, 6, and 18 ft) from a fixed tower.The data from the 6 ft height were used in this analysis.Carbon monoxide (CO) concentrations were measured using a standard commercial CO analyzer (Thermo Scientific) with infrared absorption detection.The inorganic halogen species Br 2 , Cl 2 , BrO, and HOBr were measured in situ at a height of ∼ 1.5 m above the snowpack with a chemical ionization mass spectrometer (CIMS) employing I − ion chemistry.This method, as used in Barrow, and these observations have been published in Liao et al. (2011Liao et al. ( , 2012Liao et al. ( , 2014)).ClO was measured at a height of ∼ 3 m above the snowpack via chemical conversion to a stable, halogenated product followed by online separation and detection by gas chromatography with electron capture detection (Stephens et al., 2012).Formalde- . Ambient measurements of O 3 , Cl 2 , Br 2 , NO 2 , NO, CO, VOCs, and OVOCs that are incorporated into the model.Mixing ratios for the NMHCs that were collected as canister samples were interpolated between samples.Available canister sample data points are indicated on their respective plot.The photolysis rate coefficient (J ) is shown for NO 2 as an example in Panel (a).Time is expressed in Alaska Standard Time.hyde (HCHO) concentrations were measured at 1 Hz frequency from the fixed tower using a tunable diode laser absorption spectrometer as described in Fried et al. (1997) and Lancaster et al. (2000).As with the O 3 and NO x measurements, only data at 6 ft were used in this analysis.HONO was measured using the Long-Path Absorption Photometer (LOPAP) technique described in Villena et al. (2011) with a reported 10 % measurement uncertainty.Finally, a large suite of organic compounds was measured both in situ by fast GC-MS (Apel et al., 2010) and via whole air canister samples with offline GC-MS (Russo et al., 2010).Those species that were directly used in this analysis are listed in Table 5.

Model description
The model used for this study is a zero-dimensional, photochemical model solved using the FACSIMILE commercial software.It incorporates much of the known gas-phase chemistry that occurs in the Arctic in 220 gas-phase reactions and  1 and 2).The model also includes deposition/uptake to snow/aerosol surfaces for certain species, aqueous-phase halide reactions (Tables 3 and 4), and a simple dilution rate for long-lived species (e.g., PAN).

Atmos
The model was developed to investigate, as a particularly useful example case study, the period 25 through 31 March 2009, which included a 3-day ozone depletion event during which O 3 remained below 1 ppbv.This time period was chosen to investigate the changes in halogen interactions under different ozone regimes, i.e., background ozone (∼ 30-35 ppbv), ozone depleting periods, full ozone depletion (O 3 < 5 ppbv), and ozone recovery, during which winds were consistently (with the exception of the first half of 25 March) arriving at the measurement site directly from the frozen Arctic Ocean.To enable focused pursuit of specific questions about gas-phase free radical chemistry, the model was constrained to observed values of O 3 , C 2 H 2 , C 2 H 4 , C 2 H 6 , C 3 H 8 , C 3 H 6 , n-C 4 H 10 , i-C 4 H 10 , HCHO, CH 3 CHO, CH 3 COCH 3 , methyl ethyl ketone, Cl 2 , Br 2 , NO, NO 2 , and CO, as well as calculated time-varying photolysis rate coefficients (J ) for O 3 and NO 2 .The in situ measurements that were used to constrain the model are presented in Fig. 1.NO 2 mole ratios in Barrow are often quite variable and relatively high compared to other polar measurement locations (Beine et al., 2002).At times, winds travel from the southwesterly direction, bringing air influenced by Barrow emissions over the measurement site.This is the case for the high NO 2 observed at the beginning of 25 March, which correlates with enhanced CO.Observed values were incorporated at 10 min time steps, calculated as 10 min averages of the reported data, over the entire 7-day period.Constraining the model with these time-varying observations precludes the need for parameterization of atmospheric transport due to advection, and thus justifies the use of a zero-dimensional model for this study.It is important to note that, because our objective was to investigate the halogen chemistry occurring during this time period rather than to attempt to simulate an ozone depletion event, we have constrained our model with the observed ozone mole ratios in order to fully study the fast chemical interactions occurring under these observed conditions, which are strong functions of O 3 .
Mole ratios of CH 4 were held constant at an average value for this time of 1.89 ppmv as reported by the NOAA-ESRL Barrow Observatory.[H 2 O] was calculated for 25 March from observed meteorological conditions of 78 % relative humidity and an ambient temperature of −19.5 • C (NOAA; Barrow airport data), corresponding to a water vapor concentration of 2.23 × 10 16 molecules cm −3 .This was held constant throughout the simulation.Temperature was not varied in the model.
The gas-phase chemical reactions and corresponding rate constants used in the model are shown in Table 1.Unless otherwise noted, all rate constants were calculated for a temperature of 248 K, consistent with average ambient conditions in Barrow for this time (NOAA Barrow Observatory).This mechanism includes halogen, HO x , NO x , and VOC chemistry associated with ozone depletions in the Arctic spring.The inorganic iodine reaction scheme used here is adapted from McFiggans et al. (2000McFiggans et al. ( , 2002)), Calvert andLindberg (2004a, b), andSaiz-Lopez et al. (2008).Organic iodine compounds are not included.Although some organic iodine compounds have been observed in coastal and marine locations (Carpenter et al., 1999;Jones et al., 2010), I 2 is likely the major source of atmospheric iodine (Saiz-Lopez and Plane, 2004) and, thus, was assumed to be the reactive I source in this model.It should also be noted that while I does not react appreciably with VOCs, it is likely that I and IO do react with RO 2 radicals (Sommariva et al., 2012).However, these reactions are not included here.Thus, though the iodine reaction scheme agrees with previous studies, it is likely incomplete; therefore, modeled iodine chemistry should be taken only as potential impacts to help direct future research efforts.
Photodissociation reactions included in this model are listed in Table 2.For many of the species, time-varying J coefficients were calculated using a modified version of the Tropospheric Ultraviolet and Visible (TUV) radiation model (Madronich and Flocke, 1999) based on in situ 0.1 Hz measurements of downwelling actinic flux conducted throughout the duration of the OASIS campaign and a surface albedo of 0.8.Upwelling flux was estimated as a function of solar zenith angle assuming clear sky conditions.The sum of upwelling and downwelling radiation was used by the TUV model to calculate the total photolysis frequencies.J coefficients for solar noon on 25 March (listed as J max ) are provided in Table 2 as an example.J coefficients were input into the model at 10 min time steps for O 3 and NO 2 .All other J values were scaled to J NO2 in the modeling code.
J coefficients from OASIS were not available for OClO, HOCl, or iodine compounds.An estimate for J max for OClO was taken from Pöhler et al. (2010) and that for HOCl was taken from an Arctic model study by Lehrer et al. (2004), which were then scaled to J NO2 .J coefficients for iodine compounds were calculated according to the work of Calvert and Lindberg (2004a, b), who also simulated conditions for late March in Barrow, although we note that there is a larger uncertainty for the photolysis coefficients for the iodine species, with the exception of I 2 .Time-varying J coefficients for the iodine species were calculated using a fourthorder polynomial and varying the solar zenith angle from 98.4 to 72.3 • .
Deposition of species to the snowpack is estimated based on measured dry deposition velocities and applied using Eq.(1) to calculate transfer coefficients (k t ) from the gas to aqueous phase.position velocities to snow in the Arctic range from 0 to 0.2 cm s −1 (Gong et al., 1997;Helmig et al., 2007).Here, we use an average value of 0.05 cm s −1 for ozone, similar to the modeling studies of Cavender et al. (2008) and Michalowski et al. (2000).However, there is large uncertainty in this parameter and often-contradictory observations from field measurements (Helmig et al., 2007(Helmig et al., , 2012)).For 25 March, the boundary layer height in Barrow was estimated at 300 m based on radiosonde data and model simulations (R. Stae-bler, Environment Canada, personal communication, 2010).This corresponds to a k t for ozone of 1.67 × 10 −6 s −1 .Because dry deposition velocities to snow have not been determined for the halogen acids, we use the estimation method of Michalowski et al. (2000) and assume a transfer coefficient 10 times greater than that for ozone.Thus, for HBr, HCl, HOBr, HOCl, and HOI, k t = 1.67 × 10 −5 s −1 (Table 3).Similarly, we assume an equivalent deposition velocity for the oxidized acidic nitrogen compounds (i.e., HNO 3 , HO 2 NO 2 , 5.50 1.67 × 10 −5 HOBr (g) → HOBr (s) 1.67 × 10 −5 HOCl (g) → HOCl (s) 1.67 × 10 −5 HOI (g) → HOI (s) 1.67 HONO, N 2 O 5 ), though a full mechanism of aqueous-phase nitrate chemistry is not included in this model (see Custard et al., 2015, for a more detailed investigation of nitrate chemistry).
As mentioned above, in the final version of our model we used the observed values of Br 2 and Cl 2 to constrain the halo-gen precursors; however, our model was initially developed to utilize only multiphase chemistry to produce Br 2 (and Cl 2 ) with a mechanism adapted from Michalowski et al. (2000).Transfer coefficients into and out of the particle phase were calculated as described in Jacob (2000), and the snowpack volume available for heterogeneous reaction was limited to a column 10 cm deep, based on the effective UV extinction depth for Arctic snow (King and Simpson, 2001).Constant concentrations of chloride and bromide ions in the aerosol and snowpack were used throughout the duration of the simulation, as it is likely that these represent an inexhaustible source of chloride and bromide (Lehrer et al., 2004).Concentrations of Br − and Cl − in the snowpack were calculated by assuming all bromide and chloride from bulk snow measurements over "thick first-year ice" near Barrow (Krnavek et al., 2011) were contained within the quasi-liquid layer (QLL) (Cho et al., 2002).We used a pH of 4 for the QLL as an estimate, which is in line with previous modeling studies (Thomas et al., 2011(Thomas et al., , 2012)).Aerosol halide concentrations were used as reported for the bulk aerosol chloride and bromide measurements from Barrow, Alaska (Li and Winchester, 1989), and we used an estimate of H + for Arctic aerosols from the ARCPAC flight campaign (Fisher et al., 2011).Additionally, upward mixing of chemical species emitted from the snowpack heterogeneous reactions was simplified such that it was assumed to be limited only by vertical mixing from the surface and did not include diffusion through the snowpack interstitial air, which is dependent mostly on wind pumping (a parameter not included in this zero-dimensional model).The aqueous-phase reactions included in the model are listed in Table 4.
Using this multiphase mechanism, we were able to produce sufficient gas-phase Br 2 in the model solely from the snowpack-aerosol heterogeneous reactions with aqueous bromide; however, we could not reproduce the day-to-day variability in [Br 2 ] from only the heterogeneous chemistry; thus, since our main objective was to examine interactions between halogen radical species, we chose to constrain the model to observed Br 2 .This was necessary to ensure that [Br] and [BrO] were accurately represented, so that we could properly examine interactions with other radicals, e.g., HO 2 , ClO, and IO.There was a period of missing Br 2 data during 29 and 30 March.For these days, we used an average Br 2 diurnal cycle from the previous days, which was then scaled until the modeled BrO was in agreement with observations.It should be noted, also, that the uncertainty in the BrO measurements is high during ODEs as the observed values are very near the detection limit (LOD of ∼ 2 pptv with an uncertainty of −3/+1 pptv near the LOD), and thus the uncertainty in modeled [Br] would be greatest during these periods.Only daytime BrCl was used as produced in the model multiphase mechanism.BrCl measurements from OASIS are sparse; however, the daytime simulated BrCl mole ratios of 0-10 pptv are in agreement with the available observations for the campaign.a Third-order rate constant, expressed in units of cm 6 molecule −2 s −1 .b Second-order rate constant, expressed in units of cm 3 molecule −1 s −1 .c First-order rate constant, expressed in units of s −1 .Volumetric fluxes were also introduced into the model for HONO and I 2 .The fluxes of HONO and I 2 were scaled to J NO2 since HONO, and likely I 2 , are photochemically produced (Honrath et al., 1999;Zhou et al., 2001;Saiz-Lopez et al., 2011).The HONO flux was adjusted to agree with observed gas-phase concentrations (Villena et al., 2011), with daytime averages near 10 pptv and spikes in excess of 20 pptv, as shown in Fig. 3a.Although the flux is included, and is necessary for measured HONO to agree with observations across all 7 days of the simulation, the contribution of the flux is small in comparison to gas-phase production through NO x chemistry, with the exception of 26-28 March, when gas-phase NO x is especially low. Figure 3b shows a time series of the NO and NO 2 observations that has been re-scaled to a smaller range on the y axis so that the relatively higher NO x on 29 and 30 March can be discerned.A recent paper by Li et al. (2014) speculates a possible additional gas-phase source of HONO from the reaction of NO 2 with an HO 2 × H 2 O complex.However, as laboratory studies are still required to quantify the efficiency of this mecha- nism, the rate constant of this reaction, and the product yield of HONO, this mechanism was not included in our model.

Atmos
To investigate the impacts that different levels of iodine chemistry could have on ozone depletions, two iodine scenarios were used, which will be referred to as "Low Iodine" and "High Iodine."For the Low Iodine case, the I 2 flux was chosen such that average daytime gas-phase mole ratios of IO remain near 1 pptv (i.e., near or below the detection limit of DOAS and in line with previous indications of [IO] in the Arctic) for the majority of days (as shown in Fig. 4).Only during the ozone depletion period did simulated IO reach close to 2 pptv, for the chosen I 2 flux.The higher [IO] during ozone-depleted periods is a result of the constant I 2 flux that is utilized for iodine, in contrast to bromine and chlorine species, which use actual observations.Because the I 2 flux is constant during the ODE, while Br 2 and especially Cl 2 are decreased, iodine becomes more dominant for reaction with the remaining O 3 (∼ 1-2 ppbv) and [IO] increases.I 2 averages ∼ 0.025 pptv throughout the Low Iodine simulation.The High Iodine scenario uses a flux that is adjusted so that I 2 averages 0.5 pptv across the simulated period.This results in IO mole ratios ranging from 5 to 10 pptv (see Fig. 3).

Comparison of modeled and observed mole ratios for select species
Those species that are not specifically prescribed in the model (e.g., all radical species and many inorganic halogen compounds) by inputting time-varying observations or by introducing a flux are allowed to freely evolve.Here we compare modeled vs. observed mole ratios for only those species that are most important for this analysis in halogen interactions.
Since the contribution of each halogen atom to ozone destruction is a function of its concentration, it is important that the model simulates halogen radicals at levels that are consistent with observations.Comparisons of the multi-day model output with observed BrO and ClO are shown in Fig. 2b  and e. ClO observations are limited, but the model captures the occurrence and general shape of the ClO peak observed on 29 March.However, the model generally underpredicts ClO where there are available observations.A question to address is the extent to which there are other sources of Cl atoms during this time, such as HOCl or chlorinated organic compounds that are not included here.In Fig. 2b, the simulated BrO output generated through bromine chemistry constrained by observed Br 2 is compared with BrO observations.The model represents the overall temporal profile and magnitude of [BrO] throughout this period.This is important for the analysis of the interactions between BrO, ClO, and IO, which is the focus of this work.It has been suggested that daytime Br 2 observed by CIMS contains a contribution from HOBr conversion to Br 2 on the inlet, and that daytime Br should be below detection limits on average due its fast photolysis rate (Liao et al., 2012).However, this modeling study suggests that Br 2 should indeed be present in the daytime (given the agreement with observed [BrO]), though it is acknowledged that there is some degree of interference from HOBr, as is apparent from the model overprediction of BrO on 31 March.Considering an e-folding photolytic lifetime of Br 2 at solar noon of 23 s (J max = 0.044 s −1 ), and using the method of Guimbaud et al. (2002), the effective daytime mixing height (Z*) of Br 2 in the stable air typical of the Arctic (K c = 95 cm 2 s −1 ) is ∼ 0.5 m.Assuming simple first-order kinetics, the [Br 2 ] remaining after mixing up from the surface to the intake of the CIMS (∼ 1 m or 2 lifetimes) is 10 % that at the surface.A recent study examining Br 2 production from surface snow in Barrow demonstrates that enhanced Br 2 production is observed in the presence of solar radiation (Pratt et al., 2013).Given that the Br 2 concentrations in the snowpack interstitial air should be elevated due to heterogeneous production mechanisms (e.g., the bromine explosion), and that production should be greater during sunlit periods, it seems reasonable to conclude that Br 2 should sometimes be observable during the day.Indeed, Br 2 mole ratios needed in the model to reproduce BrO observations agree with the "uncorrected" Br 2 observations (shown as the red data in Fig. 2a), as reported in Liao et al. (2012).
Figure 2a and d compare the model output for Br and Cl atom concentrations (black trace), with steady-state approximations of [Br] and [Cl] (red trace) following the method of Stephens et al. (2012), as direct observations of Br and Cl atoms have not yet been achieved.In both cases, the modelsimulated output is greater than the steady-state approximations on a few days, sometimes by as much as 90 %, although both methods capture the diurnal trends and fluctuations.A notable feature of Fig. 2a are the rather high Br atom concentrations during the 3-day ODE, with concentrations up to 3 × 10 9 molecules cm −3 (∼ 100 pptv), though as mentioned previously, this should be considered an upper limit due to the uncertainty of the Br 2 and BrO measurements and the potentially positive model bias during this time.Nonetheless, higher [Br] (and [I]) is expected during periods of complete ozone depletion due to the loss of this dominant Br atom chemical sink.However, these concentrations are on the high end of previous estimates using hydrocarbon measurements, which range from 1 × 10 6 to 1 × 10 9 for ODEs (Jobson et al., 1994;Ariya et al., 1998;Rudolph et al., 1999).While [Br] peaks during periods of low O 3 , [Cl] is enhanced when O 3 is present due to the elevated Cl 2 mole ratios that are observed only when O 3 is above ∼ 10 ppbv (Fig. 1a; Liao et al., 2014).The model predicts Cl atom concentrations of 2 × 10 5 to 6 × 10 5 molecules cm −3 , which is also higher than previous estimates of 1 × 10 3 to 1 × 10 5 determined from hydrocarbon measurements (Jobson et al., 1994;Ariya et al., 1998;Rudolph et al., 1999;Boudries and Bottenheim, 2000;Keil and Shepson, 2006).As discussed in Stephens et al. (2012),  (solid line) and "High Iodine" (dashed line) scenarios.The "Low Iodine" scenario and the "High Iodine" scenario correspond to a daytime average of 0.025 and 0.5 pptv I 2 , respectively.Time is expressed in Alaska Standard Time.
the hydrocarbon-based methods average over the transport path, which can be aloft and consist of several days, and thus should be lower than that observed near the surface, if the surface is the Cl 2 and Br 2 source.
Model simulations of [OH] and [HO 2 ] (Fig. 2f and g) lie within the bounds of measurement uncertainty for the ambient data, with the exception of 29 March, when the model underpredicts HO 2 by a factor of ∼ 3. The accurate simulation of OH and HO 2 is important since halogen chemistry, especially that involving chlorine, can have a significant impact on the HO x budget.The discrepancy between the modeled and observed HO 2 for 29 March may lead to a less than realistic impact of HO 2 on oxidation chemistry for this day.The model captures the temporal cycle of HOBr well; however, it generally overpredicts the daytime peak mole ratios (Fig. 2c).It is unclear why this should be the case, given that both BrO and HO 2 are well represented, with the exception of the BrO overprediction early on 31 March.This suggests that either our simple parameterization of deposition to snow and particle surfaces is slower than that occurring in nature, or that we are missing some other important HOBr sink.

Contribution of Br, Cl, and I to ozone depletion and the rate and timescale of ozone loss
An important question regarding ozone depletion events is the contribution of the halogen radicals, Br, Cl, and I, to the total destruction of ozone and the rate of ozone depletion.Based on field measurements of reactive bromine compounds, primarily BrO, it is generally accepted that bromine is the dominant reactant leading to ozone destruction (Simpson et al., 2007).However, this has not been quantified over highly variable chemical conditions, since prior to OASIS, simultaneous high-time resolution measurements of such a large suite of chemical species had not been performed.Few studies have examined the impact of Cl on ozone depletion, and these have not had the benefit of in situ observations of Cl 2 and ClO with which to compare their models.This ob-servational data set and model thus represent a unique opportunity.
In this study, "Base Model" runs are those that include both bromine and chlorine."Br Only" simulations are performed by removing Cl 2 and chloride; likewise, "Cl Only" simulations are performed by removing Br 2 , bromide, and organobromine compounds.Simulations with iodine are conducted by introducing fluxes of I 2 of a magnitude necessary to keep the average daytime IO near 1 pptv (Low Iodine) or at 0.5 pptv I 2 (High Iodine), as described previously (reference Fig. 4).It is important to note that the model is adjusted to fit observations using the Base Model only.No re-adjustments are made to the model when either bromine or chlorine is turned off, or when iodine is turned on, to observe the effect of a one-variable change.
The time-varying fraction of ozone depleted by bromine, chlorine, and iodine was calculated by considering those reactions that destroy ozone (i.e., Reaction R2 for Br and the analogous reactions for Cl and I) while correcting for those reactions that result in the regeneration of O 3 .Here, photolysis of BrO, ClO, or IO, as well as reaction of XO with NO, ultimately lead to O 3 production.The rate of O 3 loss by halogen X is determined by Eq. ( 2), where X = Br, Cl, or I, and the total chemical O 3 loss rate is calculated using Eq. ( 3).The fractional contribution for each halogen is then determined by dividing the two quantities.

Rate of O 3 loss by
(2) While we recognize that considering only the sum of the rates of XO + XO reactions (where X = Br, Cl, or I) has generally be used in previous studies to estimate the rate of ozone destruction (Le Bras and Platt, 1995;Platt and Janssen, 1995), we have chosen to use Eqs.( 2) and (3) as it is likely that these complex interactions are oversimplified by the XO + XO method.Indeed, we show in a forthcoming manuscript (Thompson et al., 2015) that the XO + XO method underestimates chemical ozone loss by Br and Cl atoms when compared to Eq. (2).
Figure 5 shows the fraction of the rate of O 3 depletion by Br and Cl for the Base, the Base + Low Iodine, and the Base + High Iodine model scenarios as a diurnal average for the daytime hours of only 25 and 29-30 March, for conditions in which O 3 > 5 ppbv, since this quantity is only meaningful when O 3 is present.Br accounts for the vast majority of O 3 depletion in the Base scenario, at about 95 %, whereas Cl only accounts for 5-10 % of ozone loss on average.In the Low Iodine plot, it can be seen that the levels of iodine considered (1 ppt IO) contribute more to O 3 destruction than does chlorine on average, with a contribution of 20-40 %.This also decreases the Br contribution to 60-80 %.When High Iodine is included, the relative importance of Br and I switches, such that I now contributes ∼ 60 % to ozone loss, whereas Br is reduced to 30-40 %.Thus, it is clear that iodine can have a significant contribution to ozone depletion even at low levels.
To investigate this in further detail, we used a focused version of our model that simulated a 1.5 h period fixed at mid-day conditions to facilitate instantaneous calculations of ozone depletion.We used chemical mole ratios similar to those of 29 March.This day has Br 2 at 6.7 pptv and Cl 2 at 80 pptv, and ozone at background concentrations near 30 ppbv.We chose not to use 25 March, as this day was impacted by local pollution during the first half of the day (as evidenced by NO x and CO in Fig. 1).The mid-day version of our model was run with constant molecular halogen and VOC concentrations; thus, although the conditions are similar to 29 March, the results are not directly comparable.
The fractional contribution of each halogen to ozone destruction was calculated using Eq. ( 2) for ten different model scenarios with different combinations of halogens present and with no halogens.In Eq. ( 2), P , Q, R, S, T , U , and V represent the total value for model "counters" for each process in parentheses, and the denominator is the sum of chemical ozone loss terms.Because photolysis of O 3 can result in fast reformation of O 3 through the reaction of O( 3 P) + O 2 , the O( 1 D) + H 2 O reaction is used to represent the ozone that is destroyed via photolysis.
Results of this calculation are shown in  6 that Br is the primary driver of O 3 loss, with an at least 64 % contribution in all cases in which it is present, reaching up to 87 % in the "Br and Cl" (Base) scenario and 97 % in the "Br Only" simulation.In the "Br and Cl" run, Cl accounts for 11.5 % of the loss (at Cl 2 levels somewhat enhanced at 80 pptv), which is a minor, though significant, contribution.On an absolute basis, the addition of chlorine chemistry also enhances the O 3 depletion caused by bromine, with the total ozone depleted by Br increasing from 2.21 ppbv in the Br Only scenario to 3.61 ppbv in the Br and Cl scenario (data not shown).This is primarily a result of the efficient cross-reaction between BrO and ClO that regenerates Br and thereby increases the [Br] / [BrO] ratio.
Comparing "Br Only" with "Br and Low I", it is apparent that iodine chemistry directly contributes a significant amount to ozone depletion, at a fractional contribution of 11.9 % with just 1 ppt IO present, due to both the fast reaction between I and O 3 , and IO and BrO, as well as the lack of known competing sinks for I.When iodine is increased, as in the "Br and High I" scenario, the iodine contribution jumps to 35.0 %.From examination of the fractional ozone depletion, it appears that iodine chemistry decreases the Br destruction of ozone; however, on an absolute basis, there is an increase in the total amount of ozone destroyed by Br in the "Br and Low I" simulation (2.55 ppbv O 3 destroyed) compared to the "Br Only" simulation (2.21 ppbv O 3 ); this is even greater in the "Br and High I" scenario, at 3.92 ppbv O 3 .The O 3 also increases by a factor of 1.3 for the "Br and Low I" case and 2.8 for the "Br and High I" case over the integrated 1.5 h simulation period.When Low Iodine is added to the Base Model, there is a factor of 1.2 increase in O 3 , with iodine contributing ∼ 9 % to the ozone depletion; this increases to a factor of 3.3 increase in O 3 and 28.8 % contribution from I when High Iodine is included.In the "Br, Cl, and Low I" case, Cl atoms actually contribute more to the ozone depletion; however, this is a function of the much higher Cl 2 (80 pptv) than I 2 (0.025 pptv) in this scenario.On a per atom basis, iodine is more efficient than chlorine at depleting ozone.That is partly due to the fact that ∼ 50-100 % of I atoms are lost via reaction with O 3 , while only ∼ 20 % of Cl atoms are lost via O 3 reaction.
The presence of multiple halogen species leads to synergistic effects, where the total amount of O 3 destroyed is greater than the sum of the species run in isolation.This synergism is primarily due to the efficient cross-reactions between BrO, ClO, and IO that reform O 3 -depleting halogen atoms.The absolute ozone destruction by Br is enhanced more by iodine than by chlorine as a result of the very fast reaction between BrO and IO.Indeed, the increase of 1.41 ppbv of O 3 destroyed by Br when chlorine is included ("Br and Cl" case vs. "Br Only" case) corresponds to an increase of only 0.02 ppbv of O 3 destroyed per pptv of Cl 2 , whereas the enhancement in ozone destruction by Br provided by low iodine is nearly 7 ppbv of O 3 per pptv of I 2 .Therefore, while Cl acts primarily to enhance Br chemistry, I both increases the efficiency of bromine catalyzed ozone destruction and directly depletes ozone.
Chlorine is an especially interesting case because it contributes to both O 3 depletion (though relatively minor) and O 3 production.Indeed, in the "Cl Only" run, there is a net increase in O 3 that is greater than in the "No Halogens" scenario by a factor of nearly 8.This increase is due to the efficient reaction of Cl with nearly all VOCs, and the resultant production of RO 2 and HO 2 .Br, in contrast, does not react appreciably with most VOCs, although its reaction with HCHO will also lead to HO 2 production, as evidenced by the increase in O 3 production in the "Br Only" scenario compared to the "No Halogens" case.I does not react to produce HO 2 ; thus, there is no additional O 3 production in the "Low I Only" case, and only a slight increase in the "High I Only" case, likely as a result of IO reaction with NO to form NO 2 .It is clear that fully understanding ODE chemistry will re-quire a complete understanding of all three halogen radical families.Platt and Janssen (1995) indicate that ∼ 99 % of Br atoms react with O 3 when present at background levels, while only ∼ 50 % of Cl atoms react with O 3 due to efficient reactions with many VOCs.However, this quantity is highly dependent on the mole ratios of O 3 , HCHO, CH 3 CHO, and the VOCs, which can fluctuate independently of each other.To further investigate how the three halogens contribute to ozone depletion, we considered the fraction of available Br, Cl, and I atoms that react with O 3 over all other competing pathways across the 7-day simulation with iodine included (Low Iodine) as shown in Fig. 6.The inclusion of iodine does not significantly change the result for Br and Cl; thus, only the simulation including I is shown.Br atoms were simulated to react with O 3 55-95 % of the time when O 3 is not depleted, but this fraction fluctuates significantly and is at times below 25 % when O 3 is depleted (i.e., < 5 ppbv).Cl atom reaction with O 3 varies between 10 and 20 % on average, and is always less than 30 % for this time period.These numbers are lower than those estimated previously, likely because we are using actual measurements of all of the known Br and Cl atom sinks that contribute to this quantity, e.g., HCHO, CH 3 CHO, numerous VOCs, and NO x .Based on our chemical mechanism, I atoms can react with O 3 up to 100 % of the time, and usually up to 90 % (though with high variability during O 3 -depleted periods), consistent with the estimates of Platt and Janssen (1995), who determined 99 % for I.However, as stated previously, it is likely that our iodine reaction scheme is incomplete, e.g., for IO reaction with peroxy radicals.
Not all halogen atom reactions with O 3 result in a net loss of ozone.Though up to 90 % of Br atoms react with O 3 , most of these reactions do not ultimately destroy ozone, as it can be reformed via, e.g., BrO photolysis or BrO reaction with NO.When considering only Br and Cl in our mid-day model, 70 % of the BrO formed regenerates O 3 through photolysis or reaction with NO, whereas this quantity is only 12 % for ClO, which photolyzes much more slowly.While this suggests that Cl atoms are actually more efficient at destroying ozone on a per atom basis, the lower concentration of Cl due to the numerous competing Cl-atom sinks and the low fraction of Cl atoms that react with O 3 make it a minor player in ozone depletion.When iodine is included at low levels, thus opening the BrO + IO cross-reaction, the percentage of BrO that reforms O 3 drops to 64 % and that for IO is comparable at 65 %.
A highly significant finding from the OASIS 2009 campaign was the observation of unexpectedly high levels of Cl 2 (Liao et al., 2014).Given the observed Cl 2 maxima of 100-400 pptv, a sensitivity study was performed to investigate the impact that such high concentrations could have on ozone depletion chemistry.Table 7 shows the results of modeling runs for the mid-day of 29 March performed by varying Cl 2 between 25 and 400 pptv (recall that the Base scenario here for 29 March has 80 pptv Cl 2 ).These particular runs included only bromine and chlorine chemistry.Here it can be seen that the high Cl 2 levels observed in Barrow can have significant impacts on O 3 depletion, with Cl atoms accounting for over 30 % of O 3 loss when present at 400 pptv.At Cl 2 levels less than 100 pptv, however, Cl destruction of O 3 is ≤ 11 %; thus, chlorine must be relatively elevated to make an appreciable impact on O 3 depletion.While the amount of O 3 production also increases at higher Cl 2 (by an additional 0.58 ppbv O 3 at 400 pptv Cl 2 compared to 25 pptv Cl 2 ), the increase in the O 3 destruction by both the additional Cl atoms (2.8 ppbv O 3 ) and by the enhanced efficiency of Br atoms through the synergy of the halogen cross-reactions (4.37 ppbv O 3 ) far outweighs the O 3 production.Thus, when both bromine and chlorine are present, our modeling results do not show indications of a "chlorine counter-cycle" hindering ozone depletion as suggested by Piot and von Glasow (2009).
It is useful to also consider the rate of ozone depletion.Often, fast apparent ozone depletions, in which O 3 is observed to decrease over timescales of hours, have been attributed to air mass transport of ozone-depleted air, whereas local chemistry is believed to result in a more gradual depletion (Bottenheim and Chan, 2006;Simpson et al., 2007;Halfacre et al., 2014).The ozone depletion rate and the resulting timescale for depletion induced by Br, Cl, and I, both in isolation (for Br and I) and when allowed to interact, were investigated with model runs conducted for "Br Only" and "I Only", as well as with different combinations of the above, including both Low Iodine and High Iodine conditions.Cl was not run in isolation here, since as was shown in Table 6, the "Cl Only" simulation did not result in a net loss of O 3 .Here, we considered only the ozone depletion rate at midday of 29 March, determined using the delta O 3 for the 1.  simulation time (as shown in Table 6), to calculate the depletion timescale.Table 8 shows the resulting ozone depletion rates in ppbv h −1 for the different permutations of halogen radicals studied, along with the resultant timescale for total ozone depletion from a background of 35 to 5 ppbv, our definition of a major ODE (assuming a constant ozone depletion rate, which represents an upper limit as the rate of ozone depletion decreases as [O 3 ] decreases).
For the chemical conditions observed on this day, bromine chemistry alone is capable of depleting ozone on a timescale of 16.2 h at a rate of 1.85 ppbv h −1 , whereas low iodine levels would require 13.9 days at a rate of 0.09 ppbv h −1 .High iodine alone depletes ozone at a rate of 1.72 ppbv h −1 in 17.4 h.Again, the synergy that exists between the halogen species is apparent, such that the combination of any halogen species depletes ozone at a faster rate than the sum of the components run in isolation.This is a result that was previously found for the interaction of bromine and iodine in modeling studies by Calvert and Lindberg (2004a, b), Saiz-Lopez et al. (2008) and Mahajan et al. (2010).The greatest ozone depletion rate of 6.07 ppbv h −1 , with a timescale for complete depletion of only 4.9 h, results when all three halogens are present, with I 2 set at the High Iodine level of 0.5 pptv.Based on these results, it can be concluded that iodine has the potential to have a much greater impact on ozone depletion than chlorine.The combination of Br, Cl and low iodine, resulting in an 8.5 h depletion timescale, is consistent with the fast-end timescales observed over the Beaufort Sea by Halfacre et al. (2014).
It should be noted, however, that the ozone lifetimes discussed above simply represent the calculated ozone lifetime, as determined from measurements conducted within the surface layer.As discussed in detail in Tackett et al. (2007), the current view of the boundary layer is one that is very chemically stratified, with the most important source of Br 2 and Cl 2 likely from the snowpack surface.Thus, the concept of a "boundary layer" is one that has been defined in terms of the height over which ozone is observed to be depleted, i.e., typically ∼ 400 m (Bottenheim et al., 2002;Helmig et al., 2012).It is likely, however, that ozone is depleted largely in the very near-surface layer, and the timescale for that depletion is thus determined in significant part by the timescale for downward diffusion of ozone in the 0-400 m range to the near-surface layer in which BrO x concentrations are large.In short, when averaged over the entire boundary layer, the actual ozone depletion timescale may potentially be slower than that calculated here using only surface measurements, because of diffusion limitation.
In Table 9, we show the average ozone depletion rate (calculated as discussed for Table 8) for the five different scenarios with varying [Cl 2 ].When Cl 2 is present at only 25 pptv, the timescale for depletion is 13.2 h, which is longer than for the Base Model (with 80 pptv Cl 2 ), but still less than 1 day, primarily due to the activity of bromine.100 pptv Cl 2 decreases the timescale for depletion to 9.55 h, while at 400 pptv, the timescale is only 5.24 h (again assuming constant Cl 2 mole ratios).For comparison, the Base Model with low iodine as shown in Table 8 has an ozone depletion timescale of 8.47 h, whereas the Base Model with High Iodine has a depletion timescale of 4.94 h.Thus, the presence of only 0.025 pptv of I 2 has a greater effect on the ozone depletion rate than does 100 pptv of Cl 2 , and sustained levels of Cl 2 greater than 400 pptv would be required to have a comparable impact on ozone depletion, as does 0.5 pptv I 2 .Although the depletion timescales reported here, both for Base Model conditions and for elevated [Cl 2 ], represent relatively fast ozone depletion and, as stated, represent an upper limit assuming a constant ozone depletion rate, total ozone depletions are often observed to occur on timescales of a day or less (Tang and McConnell, 1996;Simpson et al., 2007;Halfacre et al., 2014), and a fast ozone depletion of ∼ 7 h attributed to local chemistry was reported over the Arctic Ocean by Jacobi et al. (2006).Thus, the ozone depletion rates calculated here (including for elevated [Cl 2 ]) show that rapid photochemical ozone depletion events are possible, and thus one should not assume that all fast ODEs represent transport, without appropriate supporting information.Backtrajectories for periods of very high [Cl 2 ] during OASIS indicate air mass transport over the surface of the Arctic Ocean (Stephens et al., 2012).Ozone instruments onboard the O-Buoy network of sea ice-tethered buoys (Knepp et al., 2010) in the Arctic Ocean have also indicated very fast ozone depletions, with a median timescale of 10.4 h and numerous individual events much faster than that (Halfacre et al., 2014).Unfortunately, we are limited by a lack of Cl 2 measurements from across the frozen Arctic Ocean, and, thus, it is not possible to speculate how widespread this elevated Cl 2 may be.

The impact of chlorine chemistry on oxidative capacity
One of the primary objectives of this work was to assess the impact of chlorine on bromine chemical cycles relating to ODEs in light of the recent discovery of unexpectedly high [Cl 2 ] in Barrow during OASIS.We discussed the contribution of observed levels of chlorine to ozone depletion in Sect.3.2 and the impact of varying concentrations of Cl 2 on the rate and timescale of ozone depletion.These results show that chlorine itself plays only a minor role in the direct destruction of ozone until Cl 2 reaches over 100 pptv.That Cl atoms do not directly contribute significantly to ozone depletions is not surprising, given their lower ambient concentration compared to Br (see Fig. 2), due to the multitude of Cl atom sinks.In the model, the rate of Cl atom production from Cl 2 and BrCl (assumed to be surface emitted species) is also on average only 20 % of the rate of Br production from Br 2 and BrCl.The Br / Cl ratio for the Arctic has been estimated to range from 80 to 1200 when ozone is not fully depleted (Jobson et al., 1994;Keil and Shepson, 2006;Cavender et al., 2008), based on observations of halocarbons and hydrocarbon decay.Our model predicts daytime Br / Cl ratios for non-O 3 -depleted days ranging from a low of 18 up to 1300, consistent with previous estimates.When ozone is fully depleted (i.e., < 5 ppbv), [Br] becomes much greater than [BrO], due to the loss of its primary atmospheric sink and the resultant lack of production of BrO, and the Br / Cl ratio increases dramatically to 8 × 10 3 -2.5 × 10 5 .During these periods when the ozone concentration is low, CH 3 CHO becomes an important Br atom sink, with a rate of reaction with Br comparable to that of Br + O 3 (Fig. 7).This supports the hypothesis by Shepson et al. (1996)  A detailed study into the impact of these increased NO x levels in Barrow is the subject of Custard et al. (2015), and thus is not discussed extensively here.Although Cl chemistry can generate HCHO and CH 3 CHO in the gas phase through oxidation of methane and ethane, respectively (e.g., Reactions R13-R15), it is likely not the case that these important Br sinks are the result of such reactions during ODEs.From the time series of ambient Cl 2 observations (Fig. 1a), and as discussed in Liao et al. (2014), it is apparent that substantial mole ratios of Cl 2 are only observed when O 3 and radiation are present.When O 3 is fully depleted during 26-28 March, Cl 2 is nearly absent.Moreover, the efficiency of production of HCHO and CH 3 CHO is NO xdependent, with hydroperoxide production more important at low NO x levels typical of more remote Arctic environments.Thus, while some significant gas-phase production of HCHO and CH 3 CHO can occur if [Cl] is high (Sumner et al., 2002), it is far more likely that in the Arctic, surface concentrations of these compounds are primarily derived from snowpack emissions (Grannas et al., 2007;Barret et al., 2011).The production of HCHO and CH 3 CHO from the snowpack has been documented in previous studies (Sumner and Shepson, 1999;Grannas et al., 2002) and strong vertical fluxes of both compounds were observed during OASIS (Barret et al., 2011;Gao et al., 2012).
As discussed previously, and shown in Fig. 6, only a small fraction of available Cl atoms react with O 3 .This is due to the very efficient reactions of Cl with numerous organics, with rate constants much faster than analogous oxidation by OH.In Fig. 8, we investigate the most important Cl atom sinks.Indeed, reaction with organic compounds dominates the Cl atom reactivity for both non-ODE and ODE days.Reactions with reduced hydrocarbons (red trace) are the primary sink under both circumstances, at 63 and 78 %, respectively.When O 3 is present, the O 3 reaction sink accounts for ∼ 27 % of Cl atom reaction (calculated as the median across 29 and 30 March), with oxygenated organic compounds (OVOCs) accounting for 9 %.When ozone is depleted, the OVOCs make up 22 % of the Cl atom reactivity.The remaining sinks (i.e., HO x , NO y , and other inorganic halogen compounds) are insignificant in comparison.Thus, one of the most important impacts of Cl chemistry in this environment is its direct contribution to the oxidative capacity of the near-surface boundary layer via oxidation of organic compounds.Based on the reaction scheme shown in Reactions (R13)-(R15), and similar VOC oxidation pathways involving Cl atoms, Cl chemistry generates HO 2 as a by-product, a very significant BrO sink (though relatively unimportant for Br atoms) and itself an important atmospheric oxidant; however, this too is dependent on NO x .It has been suggested that the presence of chlorine will significantly increase HO 2 (Rudolph et al., 1999;Piot and von Glasow, 2009).Figure 9a presents results of a sensitivity study in which simulations with different combinations of halogens present were performed to investigate the impact on HO 2 .For days when O 3 is not fully depleted (i.e., 25, 29, and 30 March), and when Cl 2 is present, the "Cl Only" simulation results in HO 2 concentrations up to 3 × 10 8 molecules cm −3 (∼ 10 pptv) for Cl 2 nearing 100 pptv (29 and 30 March).Especially for 29 and 30 March, it is apparent that the "Br Only" simulations have much lower daytime HO 2 , generally remaining below 1 × 10 8 molecules cm −3 .When Cl 2 is present (i.e., when O 3 is not depleted), the "Cl Only" simulation is nearly indistinguishable from the Base Model, indicating the chlorine is dominating the halogen contribution to HO 2 .However, in some cases, specifically when O 3 is depleted (and thus Cl 2 is The corresponding pie charts represent the fractional importance of each sink, calculated as median values across the days shown.HO x is defined as the sum of OH and HO 2 , NO y is the sum of all nitrogen oxide compounds, VOCs include all non-functionalized alkanes, alkenes, and acetylene, OVOCs include all non-radical organic compounds with an oxygen (carbonyls, ketones, carboxylic acids), and halogens include all inorganic halogenated species.absent as discussed above), HO 2 is indistinguishable from a "Br Only" case.During the 26-28 March period, there would be little HO x production from O 3 photolysis, and as Cl 2 is also nearly absent, bromine chemistry and photolysis of HOx precursors from the snowpack (e.g., HCHO or HONO) become the primary sources of HO x production.
The production of HO 2 as a result of chlorine chemistry is also apparent when examining the [HO 2 ] / [OH] ratio in simulations performed with and without chlorine.Table 10  ratio was 115; this nearly doubled to 230 when only chlorine chemistry was added.This effect is in contrast to the "Br Only", "Low Iodine Only", and "High Iodine Only" simulations, where the addition of bromine or iodine served to decrease the [HO 2 ] / [OH] ratio.Futhermore, the addition of chlorine to any scenario shifted this ratio towards HO 2 .For example, "Br Only" decreased this ratio from 115 to 63; upon the addition of chlorine for the "Br and Cl (Base)" scenario, the ratio was increased to 104.It is important to note that NO x chemistry has a very significant impact on both HO x and halogen chemistry.For these calculations, NO and NO 2 were held constant (at 25 and 50 pptv, respectively) so that we could investigate the impacts of just adding or removing halogens without concomitant changes in NO x .If NO x is not constrained, the removal of halogens has the effect of leading to increased NO x levels (through removal of this sink), which then decreases HO x .This leads to a confounding influence on the HO x ratio that is due to increased NO x chemistry rather than removal of halogens.
To investigate further, time-varying rates of production of OH and HO 2 were determined from their primary source reactions (Fig. 10).Under our modeling conditions for the 7 days studied, when O 3 is not fully depleted, HO x is produced primarily through photolysis of HONO (as postulated by Zhou et al., 2001, andVillena et al., 2011) and as a product of Br reaction with HCHO (orange and blue traces in Fig. 8, 29 and 41 %, respectively, for 30 March), with significant production from photolysis of HCHO (15 % for 30 March) as well, confirming the importance of snowpack-emitted carbonyl compounds for the oxidation capacity of the Arctic boundary layer (Sumner and Shepson, 1999).When O 3 is fully depleted, HCHO is the primary direct HO x source via reaction with Br (84 % for 27 March).This result explains the observation in Fig. 8a that on O 3 -depleted days, the "Br Only" and Base simulations give essentially identical results for [HO 2 ].Although these modeling results clearly indicate that Br oxidation is the dominant HO 2 source during ODEs, the absolute value of the HO 2 production from this pathway during the ozone-depleted days should be regarded as an upper limit, since, as explained previously, the uncertainty in the BrO measurements is high during this time and it is possible that [Br] is somewhat overestimated.Indeed, this is likely the reason behind the high bias of the model HO 2 compared to measurements for 26-28 March (Fig. 2g).Finally, an additional significant impact of chlorine chemistry on the oxidation capacity of the polar boundary layer (PBL) is the production of RO 2 through Reaction (R13) and analogous oxidation reactions with higher order hydrocarbons.RO 2 reaction with BrO can also lead to production of HOBr through Reaction (R19), which could enhance the heterogeneous production of Br through HOBr deposition and the bromine explosion mechanism.
However, the presence of chlorine chemistry also decreases the [BrO] / [Br] ratio through the BrO + ClO reaction, so Reaction (R19) may affect [HOBr] most under certain chemical conditions (e.g., lower [ClO]).[RO 2 ] (calculated as the sum of methyl through butyl forms) for simulations conducted with and without chlorine is shown in Fig. 9b.This figure illustrates that when Cl 2 is high, i.e., on 25, 29, and 30 March, there is sometimes a large impact on RO 2 , as can by seen by comparing the "Br Only" case (blue trace) with the Base Model (black trace).Again, [Cl 2 ] is very low when ozone is depleted, so simulated [RO 2 ] during ozone-depleted days becomes nearly indistinguishable for the Base and "Br Only" scenarios, as was shown already for HO 2 .These results support the hypothesis that Cl can substantially increase the oxidation capacity of the Arctic troposphere when present at elevated mixing ratios.The levels of RO 2 that are present when Cl is not included in the model are primarily a result of both Br-and OH-oxidation chemistry; the RO 2 spike on 31 March that is seen in the "Br Only' simulation, for example, is a result of the observed spike in CH 3 CHO and reaction with Br.The impact of the increased RO 2 due to Cl chemistry appears to be less significant as a BrO sink, however, as CH 3 OO (the dominant RO 2 species) represents only a minor sink for BrO (Thompson et al., 2015), and BrO is likewise a minor sink for CH 3 OO in comparison to HO 2 and NO (the median sink contribution of BrO to CH 3 OO is 0.05 %).
Based on our analysis, it appears possible that the presence of chlorine can promote the production of reactive bromine species through two distinct mechanisms: (1) oxidation chemistry due to chlorine, which under certain chemical conditions can (e.g., high Cl 2 , elevated NO x ) increase [HO 2 ] and/or [RO 2 ], thereby increasing the production of HOBr, and thus, the heterogeneous recycling of bromine, and (2) gas-phase HOCl can react with Br − in the aqueous phase to produce BrCl, thus producing reactive bromine (though this pathway is not as significant for Br atom production as is Br 2 ).Conversely, the presence of both chloride and bromide ions in the aqueous phase can lead to a competition between the production of Br 2 , BrCl, or Cl 2 .When we initially ran our model with only heterogeneous production of molecular halogens, large increases in gas-phase [Br] were produced when aerosol chloride was decreased by 3 orders of magnitude, but no difference in [Br] occurred when snowpackphase chloride was decreased by an equivalent amount.The snowpack is by far the primary source of both Br and Cl atom precursors in our model, as shown in Michalowski et al. (2000); however, the greater sensitivity to the aerosol chloride loading suggests that the aerosol could be a potentially important source of Cl atom precursors to the atmosphere.This may likely be related to the aerosol pH.Laboratory and field studies would be required to test this hypothesis.It is clear that the interactions between bromine and chlorine are quite complex in that chlorine chemistry can produce reactive bromine sinks (when at elevated concentrations), but the presence of chlorine can also increase Br atom production under certain circumstances, thereby increasing the rate of ozone depletion in an indirect fashion.

The impact of iodine chemistry and bromine-iodine interactions
Several modeling studies have also investigated the potential impact of iodine on ozone depletion in the troposphere (e.g., Chameides and Davis, 1980;Davis et al., 1996;Sander et al., 1997;Calvert and Lindberg, 2004b;Saiz-Lopez et al., 2008), in each case concluding that it could be very important.Iodine is potentially one of the most important species in Arctic ozone chemistry, and yet there is very little observational information.To date, no conclusive measurements of IO in the High Arctic have been achieved, though recently our group has detected substantial concentrations of I 2 in the snowpack air in Barrow with CIMS (Raso et al., 2015); thus, there is evidence that iodine chemistry is present to some extent.
As discussed in Sect.3.2, iodine chemistry can greatly enhance the rate of ozone depletion, both through the direct reaction of I with ozone (given that ozone is the primary I atom sink, Fig. 4) and through the indirect effect of increasing available Br atoms through the cross-reaction of IO and BrO.This can be seen in Table 10, where it is shown that the [BrO] / [Br] ratio is decreased from 7.65 in the "Br Only" simulation to 7.34 or 6.12 upon addition of low iodine or high iodine levels, respectively.In a similar fashion, iodine chemistry also shifts the ClO  (Vogt, 1999).Under the conditions employed here on 29 March, the [IO] \ [I] ratio ranges from 1.57 to 2.60.
Previous modeling studies have concluded that iodine chemistry can impact the partitioning of important atmospheric oxidants by decreasing the [HO 2 ] \ [OH] ratio through the formation of HOI and subsequent photolysis.The rate constant for the IO + HO 2 reaction is a factor of ∼ 2.5 and 10 faster than the analogous BrO and ClO reactions, respectively; thus, iodine has the potential to have a significant impact on the [HO 2 ] \ [OH] ratio, and HO 2 can represent a primary IO sink during periods of low [IO] and [BrO] (Vogt, 1999).As mentioned above and shown in Table 10, the [HO 2 ] \ [OH] ratio at mid-day on 29 March with no halogens present is 115.The presence of bromine alone decreases this ratio to 63, and inclusion of iodine further decreases this to 60.8 or 53.1 when at low or high levels, respectively.There of course can be wide variability in the absolute ratios depending on specific chemical conditions, such as NO x or VOC levels.Saiz-Lopez et al. (2007b) determined a [HO 2 ] \ [OH] ratio of 33 due to bromine-iodine chemistry in Antarctica.
The decrease in the ClO x ratio effected by iodine could also impact the oxidizing capacity of the PBL.In fact, we see a slight increase in [RO 2 ] when iodine is present at high levels in the multiday model, with a greater increase during O 3depleted days when iodine is included (green trace, Fig. 9b).Because iodine atoms do not oxidize VOCs, the increased RO 2 is likely a result of the combined effects of a shift in the ClO x ratio toward Cl, and a shift in the HO x ratio toward OH, resulting in a decrease in RO 2 sinks and an increase in production.These impacts suggest an important indirect role for iodine in mediating the oxidation potential of the PBL, in the presence of the other halogens.
We note, however, that the observed [Cl 2 ] and [Br 2 ] are lower during ozone-depleted days than during non-ozonedepleted days.This suggests a surface activation mechanism controlled by O 3 , presumably in part through a snowpack halogen explosion mechanism, such as has been observed in laboratory studies by, e.g., Oum et al. (1998a, b) and the recent Pratt et al. (2013) study using natural Barrow snow.In our model, because the production mechanism of I 2 is undefined and we have no observations with which to compare, the I 2 flux is of the same magnitude each day, independent of [O 3 ].Therefore, during the O 3 -depleted days when [Br 2 ] and [Cl 2 ] are much reduced, iodine chemistry becomes dramatically more important in relative terms.However, laboratory studies have indicated that I 2 can be produced via O 3 oxidation of aqueous iodide (Garland and Curtis, 1981;Martino et al., 2009;Carpenter et al., 2013).If it were indeed the case that I 2 is also produced through an O 3 -mediated activation mechanism, this would likely eliminate the large difference between depleted and non-depleted days that is seen here.More studies are required to determine the dominant mechanism and kinetics for the heterogeneous surface production of I 2 .
As discussed previously, the calculations in Table 10 were performed by fixing NO and NO 2 concentrations so that we could investigate halogen impacts on HO x .However, we did examine the impact of halogen chemistry on the [NO 2 ] \ [NO] ratio.This was performed simply by initiating the model with NO and NO 2 mole ratios of 25 and 50 pptv, respectively, and then allowing them to freely evolve.For these sensitivity studies, the presence of halogen chemistry in our model increased the [NO 2 ] \ [NO] ratio, primarily through Reaction (R16), though Cl chemistry can further increase this ratio through the production of RO 2 (e.g., Reactions R13-R14).However, though bromine increases the mid-day NO x ratio from 0.68 to 1.81 for 29 March compared to a model run conducted with no halogens, the further addition of low iodine only increases it to 1.85.The absolute concentrations of [NO 2 ] and [NO] both decrease with the addition of halogens, but the [NO 2 ] \ [NO] ratio increases because XO reaction with NO is generally faster than reaction with NO 2 .However, in the case of iodine, IO reaction with NO 2 is faster than its reaction with NO, and I atoms also react very efficiently with NO 2 .Thus, chlorine should have a greater effect on increasing the NO x ratio than iodine.Indeed, when the model was run with bromine and chlorine (the Base scenario), the [NO 2 ] \ [NO] ratio significantly increased to 8.4.

Conclusions
The goal of this work was to investigate the interactions and impacts of halogen chemistry on ozone depletion using a model that is constrained to a unique set of observed, timevarying chemical conditions at the time of the event.With this approach, we have been able to dissect some of the important chemical pathways pertaining to ODEs, focusing on the interactions between the halogen radicals.It is clear that the interactions between bromine, chlorine, and iodine are very complex and highly dependent on the concurrent conditions of relevant species, such as O 3 , HO x , NO x , and the VOCs.As these species fluctuate, the partitioning of halogen species will also change, and so too will their impact on chemistry of the PBL.Thus, a full understanding of halogen chemistry requires the careful measurement of all these species (including Cl 2 , Br 2 , HOBr, and HOCl).
This work has demonstrated that bromine chemistry is clearly the dominant destruction pathway for ozone depletion episodes, but that chlorine and, especially, iodine, can contribute significantly to both the rate and timescale of ozone depletion.The new observations of high chlorine levels at Barrow potentially change the way we view ODEs.Often, observations of rapid decline in ozone at various locations have been attributed to transport because bromine chemistry alone was not enough to account for the ozone depletion rate (Hausmann and Platt, 1994;Tuckermann et al., 1997;Bottenheim et al., 2002) depletions can occur on timescales of much less than a day, and at elevated chlorine levels, even as short as several hours.This result suggests that more of the ODEs that are observed could be the result of local-scale chemistry; however, as noted previously, our calculated timescales for ozone depletion are based upon measurements conducted within the near-surface layer, and may not necessarily reflect the overall depletion timescale through the boundary layer.While chlorine is clearly not necessary to cause ozone depletion, it can impact the rate of ozone depletion through synergistic effects of cross-reactions that enhance Br atom recycling.Moreover, the presence of elevated chlorine levels can impact important Arctic chemical budgets, including HO x , NO x , and VOCs, with implications for the oxidative capacity of the PBL.More field measurements, including over the frozen Arctic Ocean, are necessary to evaluate the ubiquity of these elevated chlorine levels.In light of these new data, it is crucial that future Arctic modeling studies take into account the activity of chlorine.
In our model, we prescribed a Cl 2 mixing ratio that was varied from 25 to 400 pptv for 29 March in sensitivity studies to investigate the impact of elevated Cl 2 .However, we also determined the Cl 2 fluxes that would be necessary to produce the desired gas-phase [Cl 2 ].In our model, volumetric fluxes ranging from 5.5 × 10 5 molecules cm −3 s −1 to 1.45 × 10 6 molecules cm −3 s −1 were required.If we assume a boundary layer mixing height of 300 m, the corresponding surface fluxes would be 1.65 × 10 10 -4.35 × 10 10 molecules cm −2 s −1 .
Similarly, the fluxes necessary for Br 2 range from 1.0 × 10 3 to 9.0 × 10 6 molecules cm −3 s −1 .These numbers are consistent with those used by Piot and von Glasow (2009).However, there are no Br 2 or Cl 2 measurements aloft of the surface in the Arctic.Utilizing the method of Guimbaud et al. (2002), the effective mixing height, Z*, for Cl 2 at mid-day is only 2.15 m, making the calculated Cl 2 fluxes 1.2 × 10 8 -3.1 × 10 8 molecules cm −2 s −1 ; Z* for Br 2 is 0.5 m, leading to a calculated Br 2 flux of 5.0 × 10 4 -4.5 × 10 8 molecules cm −2 s −1 .Therefore, field studies aimed at determining the magnitude of the surface flux of Cl 2 and Br 2 are warranted.This disparity also points out the importance of measurements of the vertical profiles of molecular halogens above the snowpack/sea ice surface.
Finally, we find that iodine could be the most efficient halogen for depleting ozone on a per atom basis, given our assumed fluxes.We assume very low [IO] in our model (≤ 1 pptv), below the detection limit of the DOAS instrumentation, and yet the resulting enhancements in ozone depletion are quite significant.Higher levels of iodine would certainly have a dramatic effect on ozone chemistry, as illustrated by our simulations incorporating 0.5 pptv I 2 (∼ 10 pptv IO).A possible mechanism for the production of I 2 in the polar regions is the activity of ice algae and phytoplankton that produce iodine compounds, which are then wicked to the surface through brine channels (Mahajan et al., 2010).It has also been suggested that the primary productivity in Antarctic ice and waters may be higher than in the Arctic (Arrigo et al., 1997;Gosselin et al., 1997;Lizotte, 2001;Mahajan et al., 2010), possibly accounting for the difference in apparent iodine activity between the two poles.However, as the ice in the Arctic continues to thin, and as more multiyear ice has been replaced by seasonal sea ice (Nghiem et al., 2012), the algae and phytoplankton productivity in the Arctic has increased (Arrigo et al., 2008(Arrigo et al., , 2012)).This could lead to an increase in iodine emissions in the future, and thus to a greater occurrence of springtime ODEs.The further development of analytical methods capable of measuring the very low [IO] and [I 2 ] potentially present in the Arctic should be a high priority for the further advancement of this research.

Figure 2 .
Figure 2. Modeled (black) compared to observed (red) time series for BrO, HOBr, ClO, OH, and HO 2 .Modeled Br and Cl (black) are compared to steady-state approximations as calculated in Stephens et al. (2012) (red).All concentrations are shown in molecules cm −3 on the left axis and as mole ratios in pptv on the right axis.Radiation is shown as the cyan trace as a reference.Time is expressed in Alaska Standard Time.
Figure 3. (a) Time series of in situ HONO observations (red data) and modeled HONO including an additional flux (black trace) and without an added flux (blue data).(b) Time series of observed NO (black trace) and NO 2 (red trace) that has been re-scaled to a smaller range on the y axis to show the relatively higher NO x mole ratios on 29 and 30 March.

Figure 4 .
Figure 4. Simulated time series of IO shown for the "Low Iodine" (solid line) and "High Iodine" (dashed line) scenarios.The "Low Iodine" scenario and the "High Iodine" scenario correspond to a daytime average of 0.025 and 0.5 pptv I 2 , respectively.Time is expressed in Alaska Standard Time.

Figure 5 .
Figure 5. Average diurnal cycles of the fraction of O 3 depleted by Br, Cl and I atoms for the Base Model scenario, the Base + Low Iodine scenario and the Base + High Iodine scenario.Br is shown in the blue trace, Cl in the red trace, and I in the green trace; radiation is shown in the cyan trace as a reference.Data shown are only for periods when O 3 > 5 ppbv.Time is expressed in Alaska Standard Time.

Figure 6 .
Figure 6.(a) Time-varying fraction of Br (blue), Cl (red), and I (green) atoms that react with O 3 across the 7 days of the simulation.(b) Time series of observed ozone mole ratios with radiation is shown as the cyan trace as a reference.Time is expressed in Alaska Standard Time.

Figure 7 .
Figure 7. Time-varying rates (in molecules cm −3 s −1 ) of the important Br atom sinks during ODEs (O 3 < 5 ppbv).The inset pie chart represents the fractional importance of each sink calculated as the median value across the 3 days shown.Time is expressed in Alaska Standard Time.

Figure 8 .
Figure 8.Rates of reaction for the most important Cl atom reaction partners during ODE days (a; O 3 > 5 ppbv) and non-ODE days (b).The corresponding pie charts represent the fractional importance of each sink, calculated as median values across the days shown.HO x is defined as the sum of OH and HO 2 , NO y is the sum of all nitrogen oxide compounds, VOCs include all non-functionalized alkanes, alkenes, and acetylene, OVOCs include all non-radical organic compounds with an oxygen (carbonyls, ketones, carboxylic acids), and halogens include all inorganic halogenated species.

Figure 9 .
Figure 9.Comparison of modeled HO 2 (a) and RO 2 (b) from four different modeling scenarios: Base (Br and Cl present; black trace), Br Only (blue trace), Cl Only (red trace), and Base + High Iodine (green trace).RO 2 is defined here as the sum of the methyl through butyl forms of alkyl peroxy radicals.Time is expressed in Alaska Standard Time.

Figure 10 .
Figure 10.Comparison of HO x production rates (in molecules cm −3 s −1 ) for five primary pathways: OH production from HONO photolysis (orange trace), OH production from O 3 photolysis and subsequent reaction with H 2 O (black trace), HO 2 production from HCHO photolysis (green trace), HO 2 production from Br + HCHO reaction (blue trace), and HO 2 production from Cl + HCHO reaction (red trace).The corresponding pie charts represent the fractional importance of each source for a select ODE day (27 March; O 3 > 5 ppbv) and a select non-ODE day (30 March) calculated as the median value across each day.Time is expressed in Alaska Standard Time.
x partitioning, decreasing the midday [ClO] \ [Cl] for 29 March from 247 in the Base Model (Br and Cl) to 244 or 230 when low or high levels are iodine are present, respectively, due to the IO + ClO reaction and decreased [O 3 ].The decreased [ClO] \ [Cl] ratio could then impact the radical distribution through RO 2 production.The [IO]\[I] ratio is much lower than either the BrO x or ClO x ratio due to fast photolysis of IO and far fewer known I atom sinks . As we have shown, complete ozone www.atmos-chem-phys.net/15

Table 1 .
Gas-phase chemical reactions used in the model.All rate constants are calculated for a temperature of 248 K unless otherwise noted and are expressed in units of cm 3 molecule −1 s −1 .

Table 2 .
Photochemical reactions.J max values for 25 March are shown as an example.J coefficients are expressed in units of s −1 .

Table 3 .
Mass transfer reactions.All rate constants are expressed in units of s −1 .

Table 5 .
Summary of the ambient measurements from OASIS that were used to constrain the model and the instrumental method used.

9679, 2015 9666 C. R. Thompson et al.: Interactions of bromine, chlorine, and iodine photochemistryTable 6 .
Fraction of ozone depleted by each halogen calculated for mid-day of 29 March via Eq.(4)."Br and Cl" is the Base Model.

Table 7 .
Fraction of ozone depleted by bromine and chlorine calculated for mid-day of 29 March via Eq.(4) for modeling scenarios performed by varying Cl 2 mole ratios.

Table 8 .
Ozone depletion rates in ppbv h −1 calculated by Eq. (3) and the corresponding timescale for ozone depletion from 35 to 5 ppbv for mid-day of 29 March for eight different modeling scenarios with different combinations of halogen radicals present."Br and Cl" is the Base Model.

Table 9 .
Ozone depletion rates in ppbv h −1 calculated by Eq. (3) and the corresponding timescale for ozone depletion from 35 to 5 ppbv for mid-day of 29 March for five different modeling scenarios performed by varying Cl 2 mole ratios.
that aldehydic compounds, including CH 3 CHO, represent important Br sinks during ODEs.HCHO is at times a major Br atom sink, but only when ambient concentrations are high enough to compete with CH 3 CHO (e.g., towards the second half of 28 March where [HCHO] / [CH 3 CHO] = ∼ 0.8).Interestingly, NO 2 also represents a major Br sink during ODEs.In comparison to other Arctic locations, Barrow can have relatively high NO x , and thus NO 2 would likely not be as important a Br sink in more pristine Arctic environments.

Table 10 .
Partitioning of the HO x , BrO x , ClO x , and IO x radical partners for mid-day of 29 March for a variety of modeling scenarios performed by including or excluding different halogen species.