Pollution‐enhanced reactive chlorine chemistry in the eastern tropical Atlantic boundary layer

This study examines atmospheric reactive chlorine chemistry at the Cape Verde Atmospheric Observatory in the eastern tropical Atlantic. During May–June, 2007, Cl2 levels ranged from below detection (∼2 ppt) to 30 ppt. Elevated Cl2 was associated with high HNO3 (40 to 120 ppt) in polluted continental outflow transported in the marine boundary layer (MBL) to the site. Lower Cl2 was observed in recently subsided air masses with multiday free tropospheric oceanic trajectories and in air containing Saharan dust. Model simulations show that the observations of elevated Cl2 in polluted marine air are consistent with initiation of Cl chemistry by OH + HCl and subsequent heterogeneous, autocatalytic Cl cycling involving marine aerosols. Model estimates suggest that Cl atom reactions significantly impact the fates of methane and dimethylsulfide at Cape Verde and are moderately important for ozone cycling.


Introduction
[2] Reactive chlorine (Cl x : Cl, Cl 2 , HOCl, ClO, etc.) is potentially important to tropospheric photochemistry, but its abundance is poorly quantified. Cl atoms have >10x higher reaction rate constants than OH with climate-active gases such as dimethylsulfide (DMS), methane, and non-methane hydrocarbons (NMHCs) [Sander et al., 2006]. Atomic chlorine can cause ozone loss via catalytic pathways such as: ClO þ HO 2 ! HOCl þ O 2 ð2Þ The net impact of Cl on tropospheric ozone depends primarily on NO x levels. Under high-NO x conditions, the oxidation of hydrocarbons by Cl results in ozone production due to enhanced formation of peroxy radicals [Knipping and Dabdub, 2003].
[4] Few direct observations of reactive chlorine gases in marine air have been reported. Cl* (the sum of Cl 2 and HOCl) has been measured at mixing ratios of up to 420 ppt (pmol mol À1 ) in polluted coastal New England air [Keene et al., 2007]. Spicer et al. [1998] reported up to 150 ppt Cl 2 in onshore flow in Long Island, and Finley and Saltzman [2006] reported mixing ratios up to 15 ppt Cl 2 in polluted coastal California air. Recently, Osthoff et al. [2008] reported up to 1200 ppt ClNO 2 in highly polluted air off Houston, Texas. These coastal measurements provide evidence of active chlorine cycling in polluted air. However, reactive chlorine levels are highly variable, and the controlling factors are not well understood.
[5] Multiphase photochemical models suggest that reactive halogen compounds can be generated in marine air via autocatalytic mechanisms [Vogt et al., 1996]: HOCl aq Atomic Cl (and Br) subsequently regenerate HOCl (and HOBr) via reactions (1) and (2). Model calculations suggest that Br should dominate the halogen chemistry in clean marine air but that Cl chemistry is enhanced via aerosol acidification under more polluted conditions [von Glasow et al., 2002;Pechtl and von Glasow, 2007].
[6] Here we report measurements of gas phase Cl 2 , Cl*, acids and particulate species at Cape Verde Atmospheric Observatory (CVAO) in the eastern tropical Atlantic. The goals of the study were to assess the importance of reactive chlorine in marine boundary layer air and to elucidate halogen activation mechanisms. The data encompass a variety of air mass types and suggest a relationship between pollutant levels and the intensity of reactive chlorine cycling in MBL air.

Field Site and Conditions
[7] Air was sampled about 50m from the ocean on the windward northeast coast of São Vicente Island, Republic of Cape Verde (16.848°N, 24.871°W), from May 20 to June 9, 2007 (day of year 140-160). Northeasterly trade winds maintained onshore air flow throughout the campaign. Boundary layer heights were estimated at 500 -1200 m based on daily soundings at Sal Island (16.73°N, 22.95°W) and local aircraft soundings [Read et al., 2008]. The boundary layer was well mixed and capped by a strong inversion. Skies were mostly clear with occasional periods of scattered cumulus clouds. Moderate levels of African dust were observed on days 147-151, but no major dust outbreaks occurred during the campaign. There were no local sources of pollution upwind of the site other than occasional ship plumes, and NO x levels were extremely low (<30 ppt 95% of the time), with rare, brief ($1 hr) excursions of <900 ppt.

Analytical Methods
[8] Cl 2 was measured using atmospheric pressure chemical ionization tandem mass spectrometry (APCI/MS/MS) [Finley and Saltzman, 2006]. The inlet was located at about 3m above ground (13m above sea level). Ambient air was drawn through a 3m long, multi-stage laminar flow inlet and ionized using a 63 Ni beta-emitting foil to form Cl 2 À ions from atmospheric Cl 2 . Cl 2 À was mass filtered, then collisionally dissociated to Cl À , which was mass filtered and detected. The instrument blank was assessed by sampling ambient air through carbonate-coated glass wool. The instrument was field-calibrated using gas standards (14 -57 ppt) generated from a permeation tube calibrated both gravimetrically and by reacting its output with a neutral KI solution and measuring the production of I 3 À by absorbance. The mean detection limit (DL) was 1.9 ± 0.7 ppt (1s) for reported 15-minute means. BrCl was also monitored but was not detected above the estimated <2 ppt DL.
[9] Gas-phase Cl*, water-soluble forms of volatile inorganic Cl and oxidized N (dominated by and hereafter referred to as HCl and HNO 3 , respectively), and HCOOH were sampled atop a 30m tower (40m above sea level) over 2-hour intervals with tandem mist-chamber systems [Keene et al., 2007]. Inertial size-fractioning inlets removed super-mmdiameter aerosols and in-line Teflon filters removed sub-mm aerosols from sample air. Exposed mist solutions were analyzed on site by ion chromatography and data were corrected for dynamic handling blanks. DLs for Cl*, HCl, HNO 3 , and HCOOH were 14, 26, 12, and 29 ppt, respectively.
[11] NO x was sampled 3m above ground (13m above sea level) and quantified with a chemiluminescence NO detector with a photolytic NO 2 converter with a DL of <14 ppt for reported NO x (10-minute means), and O 3 was measured at 3m via UV absorption and is reported as 15-minute means. Air mass back trajectories were calculated using the British Atmospheric Data Centre (BADC) trajectory service.

Observations
[12] Cl 2 levels ranged from <2 to 30 ppt (Figure 2), with a mean value of 4 ppt. Cl 2 exhibited a marked diel cycle, with nighttime maxima and daytime minima. During the periods with highest nighttime Cl 2 levels (e.g., days 156-157), Cl 2 was occasionally detectable during daylight hours. Cl 2 and HNO 3 were positively correlated at night, but Cl 2 and NO x were not correlated. Cl* is a measure of photolyzable inorganic chlorine, which in this environment is equivalent to P (2Cl 2 + HOCl). Cl* ranged from <14 to 222 ppt Cl and was almost always greater than 2Cl 2 . Cl* and Cl 2 exhibited similar temporal trends.
[13] Cl 2 mixing ratios varied on time scales of a few days in association with changes in atmospheric circulation. Eight-day air mass back trajectories, acidic gases, and aerosol chemistry were used to classify the study period in terms of four air mass types: Iberian Influenced, Open Ocean-low seasalt (-lss), Dusty, and Open Ocean-high seasalt (-hss) (Figure 1). In general, air mass types remained coherent for a few days at a time.

Open Ocean-low seasalt (-lss)
[14] These air masses typically originated in the subtropical or midlatitude free troposphere over the western North Atlantic ocean or North America, circulated around a mid-Atlantic high pressure system, and subsided into the boundary layer 1.5-2.5 days north of Cape Verde (days 142-144, Figure 1). These air masses reached Cape Verde without passing over Europe and exhibited low (mostly <12 ppt)  Table 1). Nocturnal Cl 2 mixing ratios were usually below 5 ppt (Table 1 and Figure 2).

Open Ocean-high seasalt (-hss)
[15] Elevated levels of seasalt (up to 256 nmol m À3 Na) were observed on days 152 -154 (Table 1). These air masses originated over the central or eastern North Atlantic, skirted the West African coast northeast of Cape Verde, and entered the MBL 2 -2.5 days upwind of the site (Figure 1). Acid gas levels were low ($15 ppt HNO 3 , <29 ppt HCOOH), and O 3 mixing ratios were $35 ppb. Cl 2 had nighttime maxima of 5 -15 ppt (Table 1 and Figure 2).

Iberian Influenced
[16] These conditions occurred when free tropospheric air from the central North Atlantic subsided into the boundary layer over Iberia 3 -4.5 days upwind of Cape Verde (days 141 and 156 -158, Figure 1). This air was characterized by elevated levels of HNO 3 (up to 124 ppt) and HCOOH (up to 796 ppt), $35 ppb O 3 , and moderate aerosol Na (119-186 nmol m À3 ; Table 1). These air masses exhibited the highest nocturnal Cl 2 levels (up to 30 ppt) and, occasionally, detectable midday Cl 2 levels (1 -3 ppt; Table 1 and Figure 2).

Dusty
[17] Dusty air was sampled on days 147-151 and was characterized by elevated levels of Al and Mn (Table 1). Most air mass trajectories during this period originated in the free troposphere over the Northwestern Sahara, then passed over Iberia before subsiding into the boundary layer $2 days north of Cape Verde (Figure 1). HNO 3 mixing ratios were near the 12 ppt DL, HCOOH levels were usually <29 ppt, O 3 levels were $35 ppb, and Na increased from low to high levels (82 -247 nmol m À3 ). Cl 2 levels were near the DL, except for day 150 when nocturnal mixing ratios reached 13 ppt (Table 1 and Figure 2). Days 140 and 159 were transitional between Polluted, Open Ocean-lss, and Open Ocean-hss conditions. For these days, Cl 2 nighttime maxima ranged from 10 -15 ppt. Day 146 was transitional between Open Ocean-lss and Dusty conditions. Day 145 uniquely had an eight-day back trajectory entirely in the MBL, with low Cl 2 , low acids, and low seasalt.

Model Simulations
[18] Numerical simulations were carried out to assess the extent to which multiphase halogen cycling mechanisms can explain the observed association between Cl 2 and pollutants. A one-dimensional Lagrangian model with detailed gas phase and aerosol halogen (Cl x and Br x ) reactions was used (MISTRA [von Glasow et al., 2002]). The complete chemical mechanism is reported in the supplement to Pechtl et al. [2006]. For the present study, iodine chemistry was turned off. The model includes dynamics, thermodynamics, and a detailed microphysical module that handles seasalt generation at the ocean surface, calculates particle growth explicitly, and accounts for interactions between radiation and particles. Boundary conditions were chosen such that no clouds formed. Each model run simulated the advection of an air column in the MBL over the ocean for five days, with initialization at 13:00.
[19] Two simulations were carried out, dubbed ''Clean'' and ''Polluted'' cases, to represent the conditions in the observed Open Ocean-hss/lss and Iberian Influenced air   -22.6 a Full ranges given except for NO x , for which 5% -95% range is given due to rare large spikes; one Cl* outlier noted. mass types, respectively. The two cases were initialized using the following mixing ratios: Clean -20 ppt NO x , 35 ppt HNO 3 , 0.5 ppb NMHC, 45 ppb O 3 , and 80 ppt SO 2 , and Polluted -0.8 ppb NO x , 8 ppb HNO 3 , 2 ppb NMHC, 75 ppb O 3 , and 0.5 ppb SO 2 . The initial sulfate aerosol concentration was higher in the polluted case. Back trajectories suggest that model days 3 -4.5 and 1.5-2.5 should be compared to observations for the Polluted and Clean cases, respectively.
[20] The Polluted case exhibited very active Cl chemistry, with Cl 2 levels reaching 125 ppt over the second night ( Figure 3). On the 4th and 5th nights (3.5 and 4.5 days into the run) Cl 2 levels reached 30 and 17 ppt (Figure 3), respectively, in good agreement with the observations at Cape Verde. Simulated HOCl was roughly 50 ppt during the daytime 4 days into the run (Figure 3), and the sum of HOCl and 2Cl 2 in the model ($30-70 ppt) was in reasonable agreement with Cl* (Table 1). HCl in this simulation was $550 ppt, in good agreement with the observed range of 400 -600 ppt. The modeled Cl chemistry is initiated by acid displacement of HCl from seasalt aerosol and the subsequent reaction HCl + OH ! Cl. This process enables the heterogeneous autocatalytic cycling described in reactions (1) -(6) .
[21] In the Clean case, a maximum of 6 ppt of Cl 2 was generated after 1.5 days, and 7.5 ppt after 2.5 days ( Figure  3). The lower Cl 2 levels in the Clean case are a consequence of less HCl being available to react with OH to release Cl atoms and of the slower Cl cycling in less-acidified aero-sols. This result is in rough agreement with the observations at Cape Verde under the Open Ocean-lss conditions and slightly lower than observations under Open Ocean-hss conditions (Table 1). Simulated midday HOCl after 2 days was about 20 ppt, and HOCl + 2Cl 2 was $15-20 ppt, somewhat lower than observed Cl* (<14 -84) under Open Ocean conditions. Model predictions of HCl were $250 ppt, within the range of the observed levels.
[22] BrCl levels were $10 ppt under Polluted conditions and $4 ppt under Clean conditions, both higher than observed levels (<2 ppt). This discrepancy may indicate that the model overestimates the extent of bromide depletion in sea salt aerosols, causing a shift from Br 2 production to BrCl production.

Impacts on Methane, DMS, and Ozone
[23] For the Polluted case four days into the run, the simulated 24-hr mean Cl concentration was 3.5 Â 10 4 cm À3 . At this concentration and a temperature of 25°C, the lifetime of CH 4 in the MBL would be about 9 years with respect to Cl oxidation [Sander et al., 2006]. If the mean Cl levels in the Polluted case were present year-round in a 1 km deep MBL from 30°N to 30°S, the global atmospheric lifetime of CH 4 with respect to oxidation by Cl atoms (t CH4-Cl ) would be $150 years, as compared with t CH4-OH of $10 years [Lawrence et al., 2001]. In that scenario, low latitude MBL Cl would represent $7% of the global CH 4 sink. For the Clean case two days into the run, Cl levels were roughly 4 times lower, and the same calculation gives t CH4-Cl of $600 years ($2% of the global CH 4 sink).
[24] The Cl levels estimated in these model runs would also impact the fate of DMS at Cape Verde. In the Clean case on day 2, DMS oxidation occurs primarily via OH (54%), with BrO and Cl accounting for 36% and 9%, respectively. In the Polluted case on day 4, Cl is more important, accounting for 24% of DMS oxidation, with BrO and OH oxidizing 26% and 48% respectively. The total DMS oxidation rate is similar in the two cases. The oxidation of DMS by Cl may affect the distribution of oxidation products (DMSO, DMSO 2 , MSA, SO 2 , MSA), with implications for aerosol growth and CCN formation. The Cl + DMS and OH + DMS reactions both proceed via addition and abstraction, but the lifetime and fate of the Cl-DMS adduct are not known [Stickel et al., 1992].
[25] Cl x chemistry was responsible for 5% of total O 3 destruction under day 2 Clean conditions and 18% under day 4 Polluted conditions. HO x chemistry was the dominant O 3 sink in both cases, with Br x contributing 11% in the Clean case and 12% in the Polluted case. These model results are similar to those based on long-term measurements of BrO and IO at Cape Verde [Read et al., 2008], which indicate that the combined influences of transformations involving Br x and I x account for $30% of the total O 3 loss with 11% of the total attributed to Br x alone.

Conclusions
[26] This study suggests a link between chlorine cycling and the advection of polluted air in the MBL over the eastern Tropical Atlantic. In particular, higher Cl 2 levels were observed in aged polluted air than in clean marine air, most likely as a result of aerosol acidification. Therefore, regions of very active chlorine chemistry over the oceans are likely to be spatially heterogeneous and time-varying. Simultaneous gas-phase, speciated Cl x measurements and observationally-constrained aerosol pH calculations are needed to quantitatively assess the mechanisms involved in this process.