Atomic chlorine concentrations derived from ethane and hydroxyl measurements over the equatorial Pacific Ocean: Implication for dimethyl sulfide and bromine monoxide

[ 1 ] Atomic chlorine and bromine monoxide concentrations ([Cl] and [BrO]) and dimethyl sulfide (DMS) sea-air fluxes are estimated from data collected during a Lagrangian flight made near Christmas Island (2 (cid:1) N, 157 (cid:1) W) during August 1996 aboard the NASA P3-B aircraft. Intensive hydrocarbon sampling began in the surface layer (SL) one-half hour after sunrise and continued until (cid:1) 1300 local solar time. Our empirical model includes in situ observations of hydroxyl [HO] and precise measurements of ethane (C 2 H 6 ) mixing ratios. Ethane was (cid:1) 40 pptv higher in the buffer layer (BuL) than in the SL, thus vertical exchange tended to replace any C 2 H 6 that was photochemically removed in the SL. In spite of this, SL C 2 H 6 mixing ratios decreased significantly during the flight. Using only the measured [HO] and estimated vertical mixing, our mass balance equation cannot explain all of the observed SL C 2 H 6 loss. However, when an initial [Cl] of 8.4 (±2.0) (cid:2) 10 4 Cl cm (cid:3) 3 , decreasing to 5.7 (±2.0) (cid:2) 10 4 Cl cm (cid:3) 3 at midday, is included, the observed and estimated C 2 H 6 values are in excellent agreement. Using our [Cl], we estimate a DMS flux a factor of 2 higher than when HO


Introduction
[2] Atomic chlorine (Cl), a radical species, is a powerful oxidant reacting with many volatile organic compounds at rates approaching the collisional limit. Because much of the Cl in the atmosphere originates from reactions involving sea salt aerosol [e.g., Rossi, 2003;Finlayson-Pitts, 2003], Cl is expected to be at its highest concentration in the marine boundary layer (MBL) [Singh and Kasting, 1988;Platt et al., 2004]. Therefore Cl should be an important oxidant in the MBL for those trace gases that react quickly with Cl. For example, DMS reacts about 60 times faster with Cl than with HO [Stickel et al., 1992;Sander et al., 2002].
[3] Because of its low abundance and high reactivity, [Cl] cannot be measured directly with present instrumentation and has only been characterized in a few field experiments in the MBL using observation of nonmethane hydrocarbons and halocarbons, while simultaneously solving for air mass mixing and [HO] [Singh et al., 1996;Wingenter et al., 1996Wingenter et al., , 1999. During polar sunrise, a different approach employing nonmethane hydrocarbons and their reactivities to Cl and Br yielded average [Cl] and [Br] estimated over a few days [Jobson et al., 1994;Ariya et al., 1998;Ramacher et al., 1999]. Bromine monoxide (BrO) has been characterized by atomic resonance fluorescence during Arctic sunrise [Avallone et al., 2003]. Hausmann and Platt [1994] also quantified BrO in the Arctic region using differential optical absorption spectroscopy (DOAS). This technique has been used to measure chlorine monoxide (ClO) and BrO near large dry salt lake beds [Hebestreit et al., 1999;Stutz et al., 2002]. Balloon-mounted DOAS instruments have measured a tropospheric average background BrO mixing ratio of 1 -2 pptv [Fitzenberger et al., 2000]. A wealth of knowledge of the distribution of BrO worldwide, especially in the polar regions, has been gained by DOAS instruments deployed on satellites [Hegels et al., 1998;Wagner et al., 2001]. Mist chamber measurements of Cl* (the sum of Cl, Cl 2 , ClO, HOCl) and HCl* (mostly HCl; hydrogen chloride) have been reported recently from an experiment in Hawaii [Pszenny et al., 2004]. Measurement of unique products of isoprene oxidation by Cl in a coastal urban setting is another method to infer [Cl] [Ragains and Finlayson-Pitts, 1997;Tanaka et al., 2003]. In situ measurements in the MBL, which utilized an atmospheric-pressure chemical ionization mass spectrometer [Spicer et al., 1998], revealed high levels of Cl 2 in a coastal region.
[4] Measurements of [Cl] and [Br] in aerosols indicate that these halogen species were being depleted from sea salt particles [Duce et al., 1963]. Several mechanisms have been shown to produce photolabile atomic chlorine precursors. Laboratory studies of reactions of NO y (NO, NO 2 , N 2 O 5 , HNO 3 , etc.) with sea salt aerosol reveal the production of photochemically active chlorine species, and in some studies analogous active bromine species were produced. The NO y and salt reactions generally proceed as NO y gas ð Þ þ NaCl aq ð Þ ! ZCl gas ð Þ þ products ðR1Þ where ZCl is an easily photolyzed Cl species, such as Cl 2 , NOCl, ClNO 2 , or HOCl, which can accumulate overnight and dissociate within about 1 hour after sunrise (for reviews see Rossi [2003], Finlayson-Pitts [2003, and von Glasow and Crutzen [2003]).
[5] Laboratory and modeling results [Abbatt and Waschewsky, 1998;Fickert et al., 1999;Vogt et al., 1996] indicate that the reactions of HOCl and HOBr are important in releasing chloride and bromide from sea-salt aerosol, Here X is either a Cl or Br atom.
[7] Another process that recycles Cl and mimics the HO diurnal cycle is Hydrogen chloride is the largest reservoir of inorganic Cl and is generated when Cl abstracts a hydrogen atom from an organic molecule, usually methane (CH 4 ). Because the recycling of Cl via (R5) is relatively slow, formation of HCl is considered a sink of Cl. Acidified aerosols also generate gas phase HCl. In the clean marine environment, the acid source is derived from the oxidation of DMS, which can form sulfuric acid (H 2 SO 4 ). Subsequent accumulation of H 2 SO 4 on sea-salt aerosol decreases aerosol pH. These processes lead to chloride and bromide deficits observed in marine aerosols [Duce et al., 1963;Keene et al., 1990].
[8] Methane (CH 4 ) is an import greenhouse gas, and measurements of its carbon isotopes provide constraints on sources and loss processes. The oxidation of CH 4 leads to an enrichment of atmospheric 13 CH 4 . Moreover, oxidation by Cl has an order of magnitude greater kinetic isotope effect than loss by HO [Cantrell et al., 1990;Saueressig et al., 1995Saueressig et al., , 2001Tyler et al., 2000] Ratios of 13 CH 4 / 12 CH 4 that cannot be explained without including removal of CH 4 by Cl have been observed in the extratropical Southern Hemisphere [Platt et al., 2004].
[9] Chlorine chemistry can initiate increased radical chemistry in coastal urban areas, such as southern California and Houston, Texas [Knipping and Dabdub, 2003;Tanaka et al., 2003], and causes greater oxidation of volatile organic carbon species (VOCs). This can result in an increased rate of formation of two chief urban pollutants, ozone and organic aerosol. Reaction of Cl with hydrocarbons in the Arctic also leads to enhanced radical production [Sander et al., 1997;Wingenter et al., 2003], thereby fueling local photochemistry.
[10] The air-sea flux of dimethyl sulfide provides much of the sulfur to the natural atmosphere [Bates et al., 1987b;Charlson et al., 1987;Chin et al., 2000] and is produced by phytoplankton [Keller et al., 1989] and during the grazing of phytoplankton by microzooplankton [Archer et al., 2001]. DMS can be oxidized to form new particles or to contribute to the growth of existing particles von Glasow and Crutzen, 2004]. In the unpolluted marine environment, sulfur from the sea-air flux of DMS is a major source of cloud condensation nuclei (CCN) [Clarke et al., 1998]. Production and growth of CCN resulting from DMS oxidation is important in the remote MBL because it modifies cloud cover and planetary albedo, thus altering the radiative properties of the Earth [Bates et al., 1987a;Charlson et al., 1987;Wingenter et al., 2004]. Therefore, understanding its photochemical loss processes and accurately determining its sea-air flux are essential to understanding current, past, and future climates.
[11] In this paper we employ time series observations of C 2 H 6 and in situ measurements of HO concentrations made during a Lagrangian flight to estimate the [Cl] time series and DMS air-sea flux using empirical mass balance approaches. (In a Lagrangian reference frame, there is no relative motion between that which is being observed and the observer. Oftentimes mixing must be parameterized in a Lagrangian reference frame to obtain meaningful results [Wingenter et al., 1996;Davis et al., 1999].) The difference between this DMS sea-air flux estimate and that made using a micro-meteorological approach [Lenschow et al., 1999] can be resolved if loss of DMS by BrO oxidation is included in a mass balance equation. The estimated values of [Cl] are employed in the New Mexico Tech/Georgia Tech (NMT/ GT) photochemical point, or zero-dimensional, model in order to approximate the partitioning of inorganic Cl species in the gas phase, which may be helpful in planning future in situ measurements of these compounds.

Experimental
[12] In situ measurements of HO, DMS, and other chemical and physical parameters [Mauldin et al., 1999;Davis et al., 1999;Hoell et al., 1999] were made aboard the P-3B aircraft during the NASA Global Tropospheric Experiment (GTE) Pacific Exploratory Mission (PEM) Tropics A field campaign near Christmas Island (2°N, 157°W) on 24 August 1996. One hundred and twenty whole air samples were also collected and analyzed for concentrations of nonmethane hydrocarbons, halocarbons, and alkyl nitrates . Limits of detection were 1-2 Â 10 5 HO cm À3 for HO, 2 pptv for DMS, and 3 pptv for C 2 H 6 . The measurement precision for DMS and C 2 H 6 was 2%. The precision of the individual DMS and C 2 H 6 measurements was estimated by dividing the standard deviations of the measurements taken during a flux circle by the average mixing ratio. Thus our precision estimates include the natural variability of the air mass. Therefore this is an upper limit of the actual measurement precision. The flight data, including the merged data sets, are archived at www-gte.larc.nasa.gov/pem/pemta_dat.htm.
[13] The P-3B departed from Christmas Island (2°N, 158°W) just before dawn heading ESE to an air mass with a forecasted trajectory near Christmas Island at the time the flight would end (see Figure 1 of Davis et al. [1999]). The P-3B flew a series of six 0.5-hour surface layer (SL) flux circles below 200 m, advecting WNW in a Lagrangian manner ( Figure 1). During each half hour flux circle, four to six whole air samples were collected, while approximately the same number of DMS samples were analyzed in situ by the research group from Drexel University ]. An intensively sampled ascent and descent to nearly 5 km were also made during the flight to obtain detailed vertical information regarding the chemical composition of this region (Figures 2 and 3). Details of this flight have been reported previously by Davis et al. [1999] and Lenschow et al. [1999].

Atomic Chlorine
[14] Atomic chlorine concentrations as a function of time were determined using a mass balance equation for C 2 H 6 . Briefly, the change of concentration of a species, C in the SL, with respect to time is equal to its production (P), minus loss (L), plus any change as a result of mixing with an air mass of different concentration (m), Equation (1) can be written explicitly for C 2 H 6 as Equation (2) assumes that the marine source of ethane is negligible. The bimolecular rate constants for the reactions of Cl or HO with C 2 H 6 [Sander et al., 2002] are referred to  as either k Cl+C2H6 or k HO+C2H6 and are 5.7 Â 10 À11 cm 3 molecule À1 s À1 for k Cl+C2H6 and 2.4 Â 10 À13 cm 3 molecule À1 s À1 for k HO+C2H6 at the average SL temperature, $298 K, during this flight. The mixing parameter (M), of 1.2 ± 0.6 cm s À1 , was determined previously (G. Chen, personal communication, 2000) . The height of the SL (h) and the BuL were estimated from vertical profiles of Theta (potential temperature; q) and temperature (T) measured during the ascent to nearly 5 km ( Figure 2). Theta remained constant with height up to about 550 m, marking the top of the SL. The height of the BuL extended to approximately 1800 m where T remained fairly constant (until 2400 m) inhibiting mixing from the free troposphere. The difference in concentration of C 2 H 6 in the BuL and SL is estimated from the mixing ratios observed during the ascent and descent occurring near 1000 local solar time (LST) (Figure 3). Concentrations of HO and C 2 H 6 were taken from the best fit of their measured time series (Figures 4 and 5).
The calculated [Cl] time profile appears in Figure 6. The maximum calculated [Cl] of 8.4 ± 2.0 Â 10 4 Cl cm À3 occurs at the start of the Lagrangian experiment, which began about 0.5 hours after sunrise, decreasing about 30% to 5.7 ± 2.0 Â 10 4 Cl cm À3 at midday. The uncertainties are estimated by performing sensitivity studies and propagation of error estimates with nearly all of the uncertainty derived from M. The natural variation in C 2 H 6 concentrations and analytical precision contributed negligibly to the error.
[16] An additional sensitivity analysis was performed by setting M equal to zero. This scenario suggests that even if no vertical replenishment of C 2 H 6 had occurred, which seems highly unlikely given the observed vertical distribution of C 2 H 6 and Theta, the levels of chlorine estimated in this scenario are still 31% of the estimate when mixing is considered. If no mixing had occurred, the empirical model suggests the average [Cl] over the 6-hour time period of 2 Â 10 4 Cl cm À3 .
[17] The sea-air flux of C 2 H 6 into the atmospheric surface layer in this region and during this flight was considered negligible and was not included in the calculations. The source of C 2 H 6 to the MBL was from the free troposphere as evidenced by the C 2 H 6 vertical profile ( Figure 3). A significant marine flux of C 2 H 6 would result in an underestimation of [Cl]. The concentrations of other measured hydrocarbons and halocarbons (i.e., propane, ethyne, C 2 Cl 4 , CH 4 ) were either too close to their detection limits or react too slowly with Cl to show significant change over the course of the observations, and therefore cannot be used in our mass balance analyses.  Each data point was weighted based on the stdeom. The curve is the best-weighted linear least squares fit (R 2 = 0.97). The prediction intervals associated with the curve are at the 2-sigma level, meaning there is a 95% probability that the actual best linear fit should fall between the intervals.

DMS Flux Estimate Including [Cl]
[18] The DMS flux was estimated using the calculated [Cl] time series (see above) and a mass balance equation, The production term representing the sea-air flux of DMS is F divided by h. The Cl + DMS rate constant (3.3 Â 10 À10 cm 3 molecule À1 s À1 ) used was determined by Stickel et al. [1992]. The reaction of chlorine with DMS can proceed by two channels. At low pressure, the hydrogen abstraction is important and results in the formation of HCl. At a pressure of $203 torr the addition of Cl to DMS becomes important, resulting in a decreased HCl yield of $0.5. At a pressure of 700 torr, the rate increases to 3.3 Â 10 À10 cm 3 molecule À1 s À1 . The atmospheric fate of the DMS-Cl adduct is uncertain [Urbanski and Wine, 1999], but in our analysis we assume that the addition channel is irreversible (I. Barnes, personal communication, 2004). The HO + DMS rate constant used was 5.0 Â 10 À12 cm 3 molecule À1 s À1 [Sander et al., 2002]. The minimum in chisquare occurs when a constant DMS flux of 4.5 (±1.5) Â 10 9 DMS m À2 s À1 is used in the empirical model. The measured and empirically derived [DMS] time series are quite comparable, as shown in Figure 7.

DMS Flux Estimate Including Cl and BrO
[19] The difference between the DMS flux estimate of Lenschow et al. [1999], 6.1 (±1.9) Â 10 9 DMS m À2 s À1 , and our flux estimate (4.5 (±1.5) Â 10 9 DMS m À2 s À1 ), which does not include loss of DMS by reacting with BrO, is used to estimate the concentration of BrO throughout the flight using a modified mass balance equation. It should be noted that the difference between the two flux estimates may not be significant. The mass balance equation now includes loss of DMS by BrO oxidation, The BrO + DMS rate constant used is from Ingham et al. [1999]. They report that the reaction proceeds with a rate constant of 4.4 Â 10 À13 cm 3 molecule À1 s À1 at our temperature of interest, 298 K, and the chief products are dimethyl sulfoxide (DMSO) and Br.

Photostationary State Point Model
[20] In order to investigate the gas phase partitioning of chlorine species, the New Mexico Tech/Georgia Tech (NMT/GT) photostationary state (PSS) point model was used. A version of the model has been described previously by Davis et al. [1996]. The updated version includes full HO x /NO x /CH 4 /NMHC/Halogen chemistry. Chlorine chemistry is simulated using 46 mechanisms (Table 1). The NO instrument was not operational during flight 7. Model input for NO was treated similarly as by Davis et al. [1999] with NO mixing ratios of $3 pptv and NO x of 10 pptv measured by Bradshaw et al. [1999] during a DC-8 flight near Christmas Island 5 days later. Although, the PSS model does not explicitly incorporate aerosol reactions, it does include loss terms for washout/rainout and deposition.
[21] Output of our PSS model was informally compared to the Model Of Chemistry Considering Aerosols (MOCCA: www.mpch-mainz.mpg.de/$sander/mocca/) results using input data described in case M1 of Table 1b of Pszenny et al. [2004]. In the NMT/GT model, Cl is not generated from sea salt aerosol as in the case of the MOCCA model ; rather, the [Cl] simulated above are used as input to examine the partitioning of Cl species calculated at photostationary state. The amount of Cl flux necessary for the PSS model to maintain photostationary state is added to each model run. Because we do not know the nature of this source (or sources of Cl), we refer to it as the unknown source. It is equal to the washout/rainout and deposition terms of HOCl and HCl. Because there was no precipitate during this flight, the washout/rainout terms were set to 0 s À1 . Our estimates of ClO, and HOCl, concentrations are within $20% of Pszenny et al.'s and are 50% higher for HCl. Our PSS model results are also in good agreement with other clean marine tropical model results (R. von Glasow, personal communication, 2005). Some of the differences are a result of different rate constants and J-values used in the models. The NMT/GT model uses rate constants and photolysis cross-sections recommend by the JPL panel [Sander et al., 2002] and the rate constant for the thermal decomposition of ClONO 2 is from Anderson and Fahey [1990]. The deposition term for HCl in the NMT/GT model may be too slow in comparison to surface losses used in the M1 case of the MOCCA model.

Discussion
[22] The [Cl] time series estimated from our mass balance calculations ( Figure 6) has a higher morning concentration that slowly declines to a midday minimum, which is consistent with the buildup of Cl precursors at night [Keene et al., 1990] and a strong midday Cl source [Knipping et al., 2000] or rapid Cl recycling on aerosol particles [Vogt et al., 1996]. The increasing tail after midday in our [Cl] time series (Figure 6) is likely an artifact attributed to the greater variability of the [HO] near the end of the flight (Figure 4) and its impact on the mass balance calculation.
[23] As noted earlier, our PSS model does not simulate Cl production from aerosol. Rather it calculates the amount of Cl that must be generated from an unknown source in order to keep the model in steady state based on the [Cl] input into the model (from the mass balance of C 2 H 6 ) and the partitioning of Cl into the inorganic Cl species (i.e., ClO, HOCl, ClONO 2 , HCl). Thus this source is equal to the sum of the loss process R Cl 43 through R Cl 46 in Table 1. On average, the unknown source term supplied $4 Â 10 5 Cl cm À3 s À1 . This unknown source term can be interpreted as the production of Cl from the photolysis of Cl 2 , BrCl, and other photolabile Cl species, produced from (R3) and (R4).
Recall that (R4) increases from morning to midday when hydroxyl concentrations peak.
[24] Plots of NMT/GT model output indicate that all Cl containing species have their highest mixing ratios during the early morning (Figure 8). Chlorine mixing ratios (as determined by mass balance) decreased from about 3.4 parts per quadrillion by volume (ppqv) to about 2.3 ppqv during the course of the flight. Early morning mixing ratios of ClONO 2 and HOCl are estimated from the PSS model to be $2.6 and 41 pptv, respectively, decaying to levels of about 2.3 pptv and 26 pptv by 0718 LST. Near noon (1223 LST), ClONO 2 and HOCl concentrations appear to have reached minimum mixing ratios of 1.2 and 8 pptv, respectively. Chlorine monoxide concentrations decay in the model simulations from high mixing ratios of $3 pptv in the morning to levels about half that by midday. Chlorine monoxide concentrations decrease primarily because it partitions quickly with Cl, which decreases by midday. Early morning production of HCl (R Cl 4, 6 -14) declines primarily in response to the decrease in [Cl] throughout the day. Hydrogen chloride is a relatively long-lived molecule (about 1 day) and is not treated accurately with a photostationary state model. Even so, our noontime [HCl] estimate was only $50% higher than the MOCCA model test case The J values were calculated at a solar zenith angle of 30°. Because of the lack of precipitation during this flight, the washout/rainout terms (R Cl 43 and R Cl 45) were set to 0 s À1 . All gas phase constants are from Sander et al. [2002]. The deposition rainout/washout rates are adapted from Davis et al. [1996]. b Read 1.3E-12 as 1.3 Â 10 À12 .
for noon. Its abundance is expected to accumulate during the day, while it is being produced, and we estimate that its concentration was about 1 ppbv at noon. Comparison of our model-generated [HO] with measured [HO] is an indication of the accuracy of the model in estimating short-lived species. The modeled [HO] was in excellent agreement with the measured [HO], with the modeled [HO] between 4 and 13% lower than the measured [HO] (Figure 9).
[26] By 1222 LST, [Cl] dropped merely 30%, indicating that Cl is being produced and/or recycled quickly during the day. A budgetary analysis using the PSS model attributes 41% of the Cl source at midday as coming from the unknown source. Photolysis of HOCl (R Cl 39), and the reactions ClO + CH 3 O 2 (R Cl 41) each contributed about 10% to the [Cl] at midday as estimated during this flight ( Table 2). The recycling of Cl via of HCl + HO (R Cl 40) near noon contributed $20% to the [Cl]. The importance of HOCl photolysis increases somewhat during the day as its J-value increases and also as the proportion of Cl in this reservoir increases (when HCl is not considered). The contribution of R Cl 39 as a source increases in importance from about 1% to 20%, as [HO] increase diurnally and the HCl fraction of the total inorganic chlorine concentration grows. The fraction of Cl contributed by R Cl 41 remains steady throughout the day.
Much of the morning CH 3 O 2 radical production comes from the abstraction of H from CH 4 by Cl (R Cl 6) in the presence of O 2 (Table 1).
[  [HOCl]) is 4.6 Â 10 8 Cl* cm À3 , which is mostly HOCl, is in the middle of the range of mist chamber measurements made in Hawaii (21°22.0 0 N, 157°42.8 0 W) during September of 1999 by Pszenny et al. [2004] of less than 1.5 Â 10 8 to 9.5 Â 10 8 Cl* cm À3 . The present  noontime [Cl] estimate of 5.6 (±2.0) Â 10 4 Cl cm À3 is remarkably similar to our previous noontime estimate [Wingenter et al., 1996] of [Cl] over the North Atlantic near the Azores ($36°N, 21°W) of 6.5 (±1.5) Â 10 4 Cl cm À3 . This estimate employed hydrocarbon and halocarbon data from flights during the Atlantic Stratocumulus Transition Experiment/Marine Aerosol and Gas Exchange (ASTEX-MAGE) field experiment. However, the Atlantic Lagrangian flights observed an aged continental air mass (which likely influenced the aerosol acidity) and was flown below a stratocumulus layer, which tended to decrease photochemistry in general. Our PEM Tropics A morning estimate [Cl] is also quite comparable to our ASTEX-MAGE estimate. During ASTEX we estimated for one air mass a morning [Cl] of 3.3 (±1.1) Â 10 4 Cl cm À3 over an 11-hour time period from midnight to 11 a.m. Normalizing this estimate over only the five daylight hours yields a morning average [Cl] of 7.3 (±1.1) Â 10 4 Cl cm À3 (within errors), nearly identical to our PEM Tropics A morning average of 7.4 (±2.0) Â 10 4 Cl cm À3 .
[28] A model analysis by Platt et al. [2004] needed to invoke a [Cl] of 1.2 to 1.8 Â 10 4 Cl cm À3 to explain the methane 13 C/ 12 C ratio in the troposphere of the extra tropical Southern Hemisphere. By contrast, our present estimate of average diurnal [Cl] is $3.4 Â 10 4 Cl cm À3 . Because of the assumed lower aerosol pH attributed to higher [DMS] and the increased photochemistry in the tropics, it is reasonable that our average [Cl] should be higher than the Platt et al. [2004] estimate. In contrast, our present diurnal average [Cl] is considerably higher than our [Cl] estimate made during the First Aerosol Characterization Experiment (ACE 1) field campaign of 7.2 (±1.0) Â 10 2 Cl cm À3 [Wingenter et al., 1999]. The ACE 1 estimate was made using data collected on flights over the Southern Ocean when DMS emissions were at a relative low for late spring and aerosol pH was higher [Hainsworth et al., 1998]. It should be noted that our [HO] estimate for ACE 1 [Wingenter et al., 1999] was in excellent agreement with the average photochemically modeled [HO], which was normalized to HO measurements made on the aircraft [Mauldin et al., 1998]. The low average [HO] concentrations (6.1 (±0.3) Â 10 5 HO cm À3 ) estimated during ACE 1 may have also contributed to lower Cl emissions from sea salt aerosols. Knipping and Dabdub [2002] derived a linear dependence between the rate of atmospheric Cl 2 formation and gas phase [HO] via (R4).
[29] The difference between our DMS flux estimate (neglecting BrO) of 4.5 (±1.5) Â 10 9 DMS m À2 s À1 and a micro-meteorological approach by Lenschow et al. [1999] of 6.1 (±1.9) Â 10 9 DMS m À2 s À1 can be used to infer the average [BrO]. The difference between the two flux esti-mates, which may not be significant, suggests that an important loss process may have been overlooked in our mass balance analysis, particularly the oxidation of DMS by BrO [von Glasow and Crutzen, 2004]. Using equation (5) and increasing the flux term to the estimate of Lenschow et al., the minimum in chi-square occurs when the average BrO mixing ratio is 1.3 (±1.3) pptv. The photolysis of bromoform (CHBr 3 ) at noon, with a mixing ratio of 4.5 pptv, would only have provided about 0.01% of the Br source.
[30] In the work of Pszenny et al. [2004], a BrO mixing ratio of 0.3 pptv was estimated from relative changes in the BrO signal of the DOAS instrument, although the measured BrO was never above the detection limit of 2 pptv during the Hawaii experiment. The tropospheric ''background'' mixing ratio is estimated to be 1 -2 pptv [Fitzenberger et al., 2000]. Thus our BrO estimate of 1.3 (±1.3) pptv, with its large uncertainty, is within the range of these observations.
[31] It is estimated that over the course of the flight, approximately 44% of the DMS was oxidized by HO and 56% by Cl, when only these two oxidants are taken into account (Table 3). When 1.3 pptv of BrO is also considered, it is estimated that HO oxidized 33% of the DMS, while chlorine oxidized 41%, and BrO 26% (Table 3). At these temperatures (approximately 298 K) much of the HO oxidation proceeds by H abstraction  and about half of the oxidation of DMS by Cl is by hydrogen abstraction. However, addition of BrO to DMS leads to the formation of DMSO [Ingham et al., 1999;von Glasow and Crutzen, 2004]. The DMSO channel leads mainly to particle growth, whereas H abstraction has a greater probability of forming new particles. Therefore knowledge of the type and abundance of halogen chemistry that is present is important in the understanding of key processes that determine the radiative budget of the Earth.

Conclusion
[32] Atomic chlorine concentrations were estimated during a Lagrangian flight near Christmas Island using measurements of [HO] and [C 2 H 6 ] and were found to be 8.4 (±2.0) Â 10 4 Cl cm À3 in the early morning, decreasing about 30% to 5.7 (±2.0) Â 10 4 Cl cm À3 at midday. This diurnal change can be attributed to the buildup of Cl precursors overnight and efficient heterogeneous generation or regeneration of photolabile chlorine species from aerosol. Because of the associated uncertainty, mostly from the mixing term, the actual Cl diurnal profile may deviate from that determined using mass balance. The [Cl] time series derived above served as an input in our PSS model used to demonstrate the potential partitioning of inorganic chlorine species within the limitations of this type of model. The empirically derived [Cl] time series was also used in another mass balance analysis yielding a DMS flux estimate of 4.5 (±1.5) Â 10 9 DMS m À2 s À1 . This flux estimate was found  to be a factor of 2 higher than when HO is the only oxidant considered in such calculations . Comparison of this DMS flux with that of another DMS flux estimate, employing a micro-meteorological technique [Lenschow et al., 1999], indicated another loss term may be needed in our mass balance approach, although the differences between the two estimates may not be significant. In order to minimize differences between our DMS flux estimates and that of Lenschow et al.'s, we postulate that an average BrO mixing ratio of 1.3 (±1.3) pptv must be present. When considering all three oxidants, 33% of the DMS was oxidized by HO, 41% by Cl, and 26% by BrO. Oxidation of DMS and subsequent formation of new particles or growth of existing particles is not only temperature dependent but also oxidant dependent. Accurate knowledge of the flux of DMS and type and quantity of oxidants present is needed to better understand the impact of the DMS on particle formation, growth, and the radiative properties of the Earth.