Shipboard measurements of dimethyl sulfide and SOz southwest of Tasmania during the First Aerosol Characterization Experiment (ACE 1)

. Measurements of seawater dimethylsulfide (DMS), atmospheric dimetry!sulfide, and sulfur dioxide (SO2) were made on board the R/V Discoverer in the Southern Ocean, southeast of Australia, as part of the First Aerosol Characterization Experiment (ACE 1). The measurements covered a latitude range of 40øS-55øS during November-December 1995. Seawater DMS concentrations ranged from 0.4 to 6.8 nM, with a mean of 1.7 _+ 1.1 nM (lo). The highest DMS concentrations were found in subtropical convergence zone waters north of 44øS, and the lowest were found in polar waters south of 49øS. In general, seawater DMS concentrations increased during the course of the study, presumably due to the onset of austral spring warming. Atmospheric DMS concentrations ranged from 24 to 350 parts per trillion by volume (pptv), with a mean of 112 _+ 61 pptv (1 o). Atmospheric SO2 was predominantly of marine origin with occasional anthropogenic input, as evidenced by correlation with elevated and trajectories. Concentrations ranged from 3 to 1000 pptv with a mean of 48.8 _+ 149 pptv (lo) and a median 15.8 pptv. The mean S(): concentration observed in tindisturbed marine air was 11.9 _+ 7.6 pptv (lo), and the mean DMS to SO2 ratio in these conditions was 13 +_ 9 (10). Diurnal variations in SO2 were observed, with a daytime maximum and early morning minimum in agreement with model simulations of DMS oxidation in the marine boundary layer. Steady state calculations and photochemical box mtxlel simulations suggest that the DMS to SO2 conversion efficiency in this region is 30-50%. Comparison of these results with results from warmer regions The dashed curve was obtained with an SO2 production efficiency of 40% and an aerosol sink of 5.6 x 103 molecules c(cid:127)n '3 s 'l and the solid curve was obtained with a production efficiency of 30(cid:127) and an aerosol sink a production efficiency of 50% and an aerosol sink of 1.3 x l() 4 molecules cm '3 s -(cid:127) and the solid curve was obtained with a production efficiency of 40% and an aerosol sink of I x 104 molecules cm -3 s 'l.

the analytical system to eliminate oxidant interferences [Saltzman and Cooper, 1988]. The air sample volume ranged from 0.5 to 1.5 I_, depending on the DMS concentration. Seawater samples were collected from the ship's seawater pumping system at a depth of 5 m. The seawater line ran to the analytical system where 5.1 mL of sample were valved into a teflon gas stripp•er. The samples were purged with hydrogen at 80 mL min-for 5 min. Water vapor in both the air and water samples was removed by passing the flow through a-25ø(7 Teflon tube filled with silanized glass wool.

DMS was then trapped in a -25øC Teflon tube filled with
Tenax. During the sample trapping period, 6.2 pmol of methylethyl sulfide were valved into the hydrogen stream as an internal standard. At the end of the sampling/purge period, the coolant was pushed away from the trap, and the trap was electrically heated. [)MS was desorbed onto a DB-1 megabore fused silica column where the sulthr compounds were separated isothermally at 50øC and quantified with a sullhr chemiluminesence detector. The detection limit during ACE 1 was approximately 0.8 pmol. The system was calibrated using gravimetrically calibrated permeation tubes. The precision of the analysis, based on both replicate analyses of a single water sample and replicate analyses of a standard introduced at the inlet of the air sample line, was typically +_8%. The performance of the system was monitored regularly by running blanks and standards through the entire system. Values reported here have been corrected for recovery losses. System blanks were below the detection limit. One atmospheric I)MS measurement was made every 30 min.

Sulfur Dioxide
Sullhr dioxide was measured using an HPI,C/fluorescence technique with pre-column derivatization. Details of the instrument design and perthfinance are discussed by Saltzman et al. 119931 and Gallagt•er et al. [19971. Here we briefly describe the principles of operation of the instrument and its performance during this experiment.
Ambient air was drawn into the instrument through a coarse Teflon filter to remove particulates, a nation membrane dryer to remove water vapor and a gas-liquid exchange coil. All gas lines up to and including the entrance to the nation dryer were heated to 50øC to prevent the condensation of water vapor and resultant loss of SO2. In the gas-liquid exchange coil, S()• was absorbed t¾om the ambient air stream into an aqueous lbrmaldehyde scrubbing solution (10g/V/, pli=4.6). The scrubbing solution was mixed serially with l mM ethanolamine in a pit 9 borate buftEr, 1.1 mM orthopthalaldehyde (OPA), and a sodium acetate solution (2.5 M) buffbred at ptI 5.7. The pH 9 buffer converts the liquidphase SO2 to sulfite, which reacts with the ethanolamine and ()PA to tbrm a stable, highly fluorescent iso-indole derivative. The resultant mixture flows through a delay coil (2 rain) to optimize the derivative yield. The sodium acetate solution is added to lower the ptl of the sample, slowing thrther reaction of the derivative with OPA and reducing degradation of the HPIA-' column material. The reaction stream flows through a 600 gL injection loop, which was periodically injected into an isocratic reverse-phase (Spherisorb C-18) column. The isoindole derivative was detected by fluorescence at 380 nm (X•.,, = 330nm: Hitachi FI000 or Hitachi L7480).
The instrument used a calibration system based on the double dilution of an SO2 permeation device, in which the analyte never passes through a mass flow controller. Using a 12 ng rain 4 device, the calibration system can generate gasphase standards ranging from 3 to 600 pptv. During the cruise, ambient air was anal•ed continuously, except during calibration periods and periods of instrument maintenance and obvious adverse wind direction. Calibrations were run daily, and SO2 loss in the inlet was monitored by adding standards to the inlet before the particle filter. The inlet was cleaned, and the particle filter was changed whenever SO2 loss became significant (>15%). Blanks were generated by scrubbing ambient air with carbonate filters and were run 4-5 times a day.
The system is thlly automated, has a linear response over the normal range of atmospheric concentrations, a 4 min sample integration time, and a 7 rain measurement period. The instrument has a 3-5 pptv detection limit (3c• of the daily mean blank), and the precision is estimated to be less than 10% at 20 pptv. The detection limit achieved during this project represents an improve•nent over that recently reported by Gallagher et al. [1997]. This improvement was achieved by using lower reagent concentrations, which resulted in a lower system blank. In Figure 1, we show chromatograms tbr a blank, a 5 pptv, a 17 pptv, and a 30 pptv air sample.
During this project, the instrument was positioned in an air-conditioned van on the starboard side of the flying bridge well tbrward of the stack, and ambient air was sampled directly through a removable port on the Ibrward wall of the van. The van was approximately 17 •n above sea level. Elevated SO2 levels associated with stack gases were observed during episodes of low or highly variable wind speed and direction, relative to the motion of the ship. Contaminated samples have been filtered from the data set using shipboard wind speed and direction. Stack gas The study region is located in the westerlies, and meteorological conditions during this study were aftk•cted by numerous low-pressure systems and cold t¾onts. According to Hainsworth el al. [this issuel, t¾ontal activity was somewhat more active than normal during the period of the ACE I study, due to the presence of a long wave trough west of the study area. This trough moved to the east through the region during the course of this cruise. Episodically, upt;&e by rain droplets also removes SO2 t¾om the atmosphere. A steady state expression Ibr the relationship between atmospheric 1)MS and SO: can be written as tbllows, assuming that vertical entrainment of SO: is sinall relative to SO: produced by DMS oxidation:

IDMS1
(f, + f• + v a / H + kso21 OH 1) where t] is the first-order loss of SO2 to sea-salt aerosols, f• is the first-order loss of SO2 to in-cloud oxidation, Va is the dry deposition velocity for SO2 to the sea surface, H is the boundary layer height, 13 is the efficiency of SO2 production, kr,•s is the DMS+OH bimolecular rate constant, and kso2 is the SO2+OH bimolecular rate constant. Equation (1) is valid as long as the average SO2 level is close to the average steady state SO2 concentration, which is true as long as the time required to reach steady state is short relative to the timescale of the experiment and perturbations to the system. The relatively long periods of time in which the SO2 levels did not change significantly suggest that this is the case. DMS/SO2 ratios measured in this study are higher than those observed in recent studies in the tropics and subtropics Values tbr the paratneters in equation (1), calculated from regional averages of measurements or obtained from the literature, are given in Table 1 Table 1, we estimate that it would require an SO2 production efficiency or' 33-42% to support the observed average DMS/SO2 ratio.  Hynes et al. [1986] The uncertainty in this estimation of • is dominated by the uncertainty in the calculation of the S02 sea-salt scavenging rate. As mentioned above, the estimation given in Table 1 Table 1 to 24-33%. An alternative approach to assessing the SO2 production efficiency and total SO2 sink is to look at time-dependent variations in the data. This approach allows tbr an independent assessment of the total SO2 sink from the data.

Diurnal Variations in SO2
In general, SO2 should exhibit a consistent diurnal variation in the marine boundary layer, with a maximum in the late afternoon reflecting the photochemical oxidation of DMS, and a minimum at dawn resulting from nighttime losses due to heterogeneous processes. Assuming that vertical entrainment is negligible, the nighttime loss is a direct measure of the total SO2 sink. This diurnal variability in SO2 was recently observed at a tropical site on Christmas Island IBandv et al., 1996] and during a transect prior to the ACE-1 intensive (W.J. De Bruyn ½t al., manuscript in preparation, 1997). Analysis of that data suggested SO2 production efficiencies of 60-80% and total deposition velocities tbr SO2 of 0.7-1 cm s '•.
The Southern Ocean is an environment in which one might not expect to find highly reproducible diurnal cycles because of the rapidly changing meteorological conditions, and the high degree of spatial variability in the sea surface DMS source. In addition, the cruise track required tbr nonphotochemical ACE 1 goals precluded long sampling periods in steady conditions. However, in spite of these factors the expected diurnal variation in the SO2 data was often observed during this study. In Figure 5a, we show SO2 data with a clear atlernoon maximum and early morning minimum for a 3.5 day period (days 337.5-341) during the experiment. The solid lines in the figure are a plot of solar irradiance. The dift•rence in absolute SO2 levels and amplitude of the diurnal cycle between days 338 and days 339-340 is due to varying oceanographic and metex)rological conditions. On day 338, the Discoverer steamed northeast through Subantarctic waters under cloudy skies, while for most of days 339 and 340 the Discoverer was stationed southwest of Cape Grim in the subtropical convergence zone in a well defined cloud-free high-pressure system (Figures 1  and 2). Day 338 is more representative of average conditions encountered for the ACE 1 intensive. Interestingly, atmospheric DMS did not exhibit consistent diurnal cycles during this period. This may reflect the heterogeneity of nearby sources and/or the longer lifetime of DMS. In Figures 5b and 5c, the SO2 data for days 338 and 339-340 have been averaged into 2 hour bins over a 24 hour period and compared to the output from a time-dependent photochemical box model. Error bars are standard errors of the mean. The photochemical box model simulates the photochemistry of the marine-boundary layer using a multistream radiation code, 11 photolysis reactions, and 140 thermal reactions. The model includes detailed OH chemistry and has previously been described in detail by Yvon and 50% or less at approximately 10øC suggests that the SO2 production efficiency has a positive temperature dependence. This is consistent with the simplified mechanism given in Figure 6 and

Summary
Measurements of seawater dimethylsulfide (DMS), atmospheric dimethylsulfide and sulfur dioxide (SO2) were made on board the R/V Discoverer in the Southern Ocean, southeast of Australia as part of the First Aerosol Characterization Experiment (ACE 1). Seawater DMS concentrations ranged from 0.4 to 6.8 nM and increased during the course of the study, presumably due to the onset of austral spring warming. Atmospheric DMS concentrations ranged from 24 to 350 pptv, with a mean of 112 + 61 pptv (1 o). Atmospheric SO2 concentrations ranged from 3 to 1000 pptv. The mean "background" SO2 concentration observed was 11.9 +_ 7.6 pptv (l o), and the mean DMS/SO2 ratio was 13.2 + 9.0 (lc•). Diumal variations in SO2 were observed, with a daytime maximum and early morning minimum, in good agreement with model simulations of DMS oxidation in the marine boundary layer. Steady state calculations and photochemical box model simulations suggest that the DMS to SO2 conversion efficiency in this region is 30-50%, which is lower than that estimated for field studies in tropical regions

Implications for the Mechanism of DMS Oxidation
In Figure 6, we show a simplified DMS oxidation mechanism highlighting some key features. I2boratory kinetic studies have shown that the initial step in the mechanism proceeds via hydrogen abstraction and OH addition [Hynes et al., 1986;Barone et al., 1996]. It is generally assumed. that SO2 is produced only through the abstraction channel and via the thermal decomposition of the intermediate CH3SO2.
}towever, reaction rates of the intermediate C}13SO2 and final product yields are poorly known [Yin et al., 1990a]. In this simplified mechanism, the efficiency of SO2 production is the product of the branching ratio tbr hydrogen atom abstraction and CH3SO2 thermal decomposition.
The SO2 production efficiencies estimated from this study (30-50%) agree masonably well with the branching ratio tbr hydrogen atom abstraction (40%) at the temperatures encountered during this study. Assuming that no SO2 is produced through the addition channel, a production efficiency of 30% requires that 75% of the abstraction channel produces