DMS and SO2 Measurements in the Tropical Marine Boundary Layer

Dimethyl sulfide (DMS) and sulfur dioxide (SO2) mixing ratios were measured in the boundary layer on Oahu, Hawaii in April and May 2000. Average DMS and SO2 levels were 22 ± 7 (n = 488) pmol/mol and 23 ± 7 (n = 471) pmol/mol respectively. Anti-correlated DMS and SO2 diurnal cycles, consistent with DMS + OH oxidation were observed on most days. Photochemical box model simulations suggest that the yield of SO2 and total SO2 sink are ∼85% and ∼2 × 104 molec cm− 3 s− 1 respectively. On several days the rate of decrease in DMS and increase in SO2 levels in the early morning were larger that predicted by the model. Dynamical and chemical causes for the anomalous early morning data are explored.


Introduction
The atmospheric oxidation of dimethylsulfide emitted from the sea surface is an important process in the formation of sulfate aerosols in marine air. Both laboratory and field research efforts have been directed at elucidating the reaction mechanism and product distributions, but our understanding is far from complete. In marine air, DMS oxidation is initiated primarily by the OH radical. In polluted air, the nitrate radical is also believed to contribute significantly to DMS oxidation at night. Several workers have suggested that halogen radicals (Cl, ClO, Br, BrO) also contribute significantly to DMS oxidation (Keene et al., 1996(Keene et al., , 1990Pszenny et al., 1993;Chin et al., 1996Chin et al., , 1998. The evidence for oxidation via halogen radicals is largely indirect, consisting of model calculations and estimates derived from hydrocarbon measurements (Vogt et al., 1996;Sander and Crutzen, 1996;Jobson et al., 1994, Wingenter et al., 1996. Direct observation of Cl 2 in coastal air has been reported, at levels which suggest that Cl radical chemistry should occur (Spicer et al., 1998).
The diurnal cycle of DMS in marine air can also provide indirect insight into the oxidation process. Several previous studies have demonstrated that the diurnal cycle of DMS in clean marine air exhibits late afternoon minimum and early morning maximum expected for a well-mixed boundary layer with a sea surface DMS source and daytime oxidation by OH Bandy et al., 1996, Yvon et al., 1996. Although previous modeling studies have shown that the amplitude of the diel cycle is larger than predicted using the currently accepted rate constant for DMS + OH (Hynes et al., 1986), the phase and shape of the observed diel profiles in earlier studies is consistent with OH as the principle oxidant of DMS in clean marine air.
In this study we present measurements of DMS and SO 2 mixing ratios made in the remote marine boundary layer. Anti-correlated diel cycles, generally consistent with a DMS + OH oxidation process were observed on most days. The phase and shape of the cycles suggest a second chemical or dynamical process occurring in the early morning. The cycles are compared to the output from a photochemical box model to estimate the DMS flux, SO 2 yields, the total SO 2 sink and to explore whether the second process is chemical or dynamical.

FIELD SITE LOCATION AND METEOROLOGY
Measurements were carried out in the North Pacific trade winds at the University of Hawaii's atmospheric research station at Bellows Air Force Base on Oahu during the period of April 20-26, 2000. The site is located on the eastern side of the island on a sandy stretch of beach in Waimanalo Bay. The instrument was housed in a van immediately below the site-sampling tower, which was located approximately 7 m from the high tide mark. The air inlet was positioned on the tower at a height of 13 m. The site experienced clean trade winds off the ocean for the entire sampling period. Back trajectory calculations suggest that the sampled air originated over the North Pacific Ocean and did not pass over any significant land mass.

ANALYSIS OF DMS AND SO 2
Atmospheric DMS and SO 2 were measured using a gas chromatograph with mass spectrometric detection (GC/MS). An isotope dilution technique was used, in which a stable isotopically labeled isotopomer of each analyte ( 12 C 2 H 3 32 S 12 C 2 H 3 and 34 SO 2 ) is added to the air stream prior to sampling. This provides an internal standard and compensates for losses in the process of sample preconcentration and analysis. This technique has been used extensively in ground-based and aircraft studies (Bandy et al., 1993;Thornton et al., 1996;Bandy et al., 1996, De Bruyn et al., 2001, and has been evaluated in two formal instrument intercomparisons (Hoell et al., 1993, Stecher et al., 1996. Details of the system and its performance are given in De Bruyn et al. (2001). In brief, air is drawn through two traps, one for DMS and one for SO 2 , held at liquid argon temperatures to pre-concentrate the samples. The air streams are spiked with the isotopically labeled standards and then dried and scrubbed of oxidants prior to reaching the traps. During this experiment two-liter samples were trapped every 8 minutes. The cryo-traps are resistance heated and injected onto the GC/MS columns for analysis. The estimated lower limit of detection is 3-5pmol/mol and the accuracy and precision of measurements is better than ±10% and ±5% respectively.

MODELING
A time-dependent photochemical box model has been used to simulate the observed DMS and SO 2 diel cycles. The model simulates the photochemistry of the boundary layer using a multi-stream radiation code, 11 photolysis reactions and 140 thermal reactions (Yvon et al., 1996). Input parameters were taken from measurements made at the site (temperature, relative humidity), the literature (ozone, methane) or long-term observations made at Cape Kumukalu (CMDL) on the island of Hawaii (column ozone, carbon monoxide). In all simulations the column ozone and carbon monoxide levels were set at 280 Dobson Units and 130 ppb respectively. Boundary layer ozone levels were estimated from Logan (1996) to be 30 ppb and methane levels were estimated from Koga and Tanaka (1993) and Spivakovsky et al. (1990) to be 1.7 ppb. In these simulations, NOx was conserved, and held at 20 pmol/mol. The average temperature and relative humidity measured at the site were 24 • C and 70% respectively. DMS and SO 2 levels above the boundary layer were assumed to be 0 and 60 pmol/mol respectively and the boundary layer height and entrainment rate were assumed to be 1 km and 0.5 cm s −1 respectively. The model only considers loss of DMS via OH, NO 3 oxidation and entrainment into the layer above. SO 2 is produced in the model from NO 3 oxidation, OH oxidation of DMS, entrainment from above, and is lost via heterogeneous processes and OH oxidation. The DMS flux, yield of SO 2 from DMS oxidation and total SO 2 sink are adjustable parameters.

Observations and Analysis
Three days of DMS and SO 2 measurements (days 111, 112 and 115) are shown in Figure 1. DMS levels reached maxima of 30-35 pmol/mol in the early morning and minima of 10-15 pmol/mol in the late afternoon. SO 2 reached maxima of 30-35 pmol/mol in the late afternoon and minima of 10-15 pmol/mol in the early morning. The observed diel patterns and anticorrelation of the two compounds is expected if DMS is emitted from the sea surface continuously, and is oxidized photochemically during daylight hours. The diel amplitude of DMS may be damped by nighttime oxidation by NO 3 radicals , but in low NO x environments like Hawaii, this is a minor effect. By contrast, SO 2 is produced via DMS oxidation during the day, and is lost principally by heterogeneous deposition (to clouds, aerosols, and the sea surface). Heterogeneous loss of SO 2 is not believed to vary on a diel basis.
Also shown in the figure are model-produced DMS, SO 2 and OH levels. The DMS flux in the model was adjusted until the simulated DMS profile best matched the observed DMS levels, then the SO 2 production efficiency and total SO 2 sink were adjusted until the model best reproduced the observed SO 2 levels. The DMS fluxes, SO 2 sinks and SO 2 yield used to simulate the observed DMS and SO 2 levels are given in Table I. In order to reproduce the amplitude of the observed DMS diurnal cycle the oxidation rate in the model had to be increased by a factor of 1.75. Similar observations have been made in several previous attempts to reproduce dimethyl sulfide diurnal cycles with photochemical models (Suhre et al., 1995;Yvon et al., 1996;Chin et al., 1996Chin et al., ,1998. Explanations put forward for this inability to model  (1987) for this region and total SO 2 sinks and SO 2 yields are consistent with those reported by Bandy et al. (1996) and Davis et al. (1999) for the tropical Pacific Ocean. The uncertainties given in Table I reflect only the uncertainty in the fit of the model to the data. Simulations were carried out to test the sensitivity of the DMS flux, SO 2 sink and SO 2 yield to changes in the assumed parameters. With a boundary layer height of 1.5 km the DMS flux had to be increased by 38% and the SO 2 sink had to be decreased by 10% to balance sources and sinks. The SO 2 yield did not change. Decreasing the entrainment rate to 0.3 cm s −1 increased the DMS flux by 31%, decreased the SO 2 sink by 10% and also did not change the yield significantly. Decreasing the NO x level to 10pmol/mol decreased the flux by 11%, and decreased the SO 2 sink by 9%. Simulations were also carried out without increasing the oxidation rate in the model. With a normal oxidation rate the model would require a 20% lower DMS flux and a 50% higher SO 2 yield in order to support the observations. However, the overall quality of the DMS fit to the data deteriorates. The uncertainty in the DMS flux is controlled by the uncertainties in entrainment rate, boundary layer height and DMS oxidation rate. Uncertainties in the SO 2 sink are relatively small and are controlled by the uncertainty in boundary layer height, entrainment rate, NO x level and goodness of fit to the data. The overall uncertainty in the SO 2 yield is fairly large and is dominated by the uncertainty in the DMS oxidation rate.
On most days during this study, there was a significant discrepancy between measurements and model simulations at sunrise. The observed DMS mixing ratio decreases earlier in the morning than the model produced DMS levels. Conversely, the observed SO 2 mixing ratios increase earlier in the morning than predicted by the model. The assumed or adjustable parameters in the model have little impact on these differences. There are two possible types of explanations for the anomalous early morning observations, dynamical and photochemical. A dynamical explanation would require the dilution of the boundary layer with DMS-free air containing SO 2 . Such air presumably exists aloft, above the boundary layer, where DMS has been largely or completely oxidized to SO 2 and other products. A photochemical explanation requires the presence of an oxidant other than OH, having a different diel profile favoring morning oxidation. The most likely candidates are halogen radicals.
To determine the magnitude of the anomalous early morning DMS loss, the DMS and SO 2 measurements from 3 days (Figure 1) were averaged into hourly bins over a single 24-hour period (Figure 2). The solid lines are best-fit model simulations obtained with a DMS flux of 3 μmol m −2 day −1 , a total SO 2 sink of 1.9 × 10 4 molec cm −3 s −1 and an SO 2 production efficiency of 85%. The dashed  Figure 2a is a best fit to the data. An estimate of the anomalous DMS loss rate X (s −1 ) was obtained by evaluating the following finite difference expression for the observed DMS over the 24 hour cycle: Here, F DMS is the sea-air DMS flux, V ex is the entrainment rate and BLH is the boundary layer height. DMS i , OH i , and NO 3i are the concentrations of the different species at time i and DMS i+ t is the DMS concentration at time i + t. X is the first order rate constant for the anomalous DMS loss process. The resultant diel variation in X is shown in Figure 3. X reaches a maximum of ∼6.5 × 10 −5 s −1 in the early morning then decreases to insignificant rates at night. If the early morning DMS loss is caused by an unknown radical oxidant, then X can be expressed as a second order rate coefficient and a radical concentration (k (radical)). Of the halogen radicals Cl, ClO, Br, BrO, only Cl and BrO seem likely to have an impact on the lifetime of DMS in the boundary layer. Br reacts with DMS to form a DMS-Br adduct which decomposes thermally too rapidly for this pathway to be important in the boundary layer (Nakano et al., 2001). Similarly the DMS + ClO reaction rate is believed to be too slow for ClO to be important unless ClO levels are very high (Finlayson-Pitts and Pitts, 1999).
The dashed line in Figure 2b is obtained assuming that all the DMS lost via the oxidation process X produces SO 2 . Overall the agreement with observations is reasonably good suggesting that if the differences between model and observation are a result of reaction with an unknown oxidant then the reaction produces SO 2 with near unit efficiency. Neither BrO nor Cl is expected to do this (Pitts and Pitts, 2000 and references therein). BrO+DMS produces DMSO + Br with a yield of almost 1 (Bedjanian et al., 1996). The Cl+DMS reaction is expected to proceed via both an abstraction pathway to produce CH 3 SCH 2 +HCl and an addition pathway to produce the CH 3 SCH 3 Cl adduct with equivalent yields (Stickel et al., 1992) and the CH 3 SCH 2 pathway is not expected to produce SO 2 with unit efficiency (Yin et al., 1990). While there are numerous assumptions and adjustable parameters in these calculations it is difficult to produce a yield significantly lower than 100%.
If the early morning decrease in DMS is caused by air being mixed down into the boundary layer at sunrise then the loss rate constant X can be expressed as an entrainment velocity, where X = V ex [(DMS f t /DMS)-1]/BLH and DMS f t is the mixing ratio of DMS above the boundary layer. Assuming that DMS f t = 0, a minimum V ex of 6.6 cm s −1 is required to support the observations. For DMS f t > 0 a larger entrainment velocity would be required to explain the observations. The required entrainment velocity is inversely proportional to the assumed boundary layer height. A boundary layer of ∼100 m would be required to obtain exchange rates typically associated with the trade wind regime. It is difficult to determine conclusively whether an early morning mixing event is responsible for the observed changes in dimethyl sulfide and sulfur dioxide mixing ratios with out an independent chemical tracer of atmospheric dynamics. However, the above calculations suggest that it is the most likely explanation. It is also possible that some combination of the above two processes is responsible but an assessment of the likelihood is beyond the scope of this paper.

Summary
DMS and SO 2 mixing ratios were measured in the tropical boundary layer by isotope dilution GC/MS. Average DMS and SO 2 levels were 22 ± 7 (n = 488) pmol/mol and 23 ± 7 (n = 471) pmol/mol respectively. Anti-correlated DMS and SO 2 diurnal cycles, consistent with DMS + OH oxidation were observed on most days. Photochemical box model simulations suggest that the yield of SO 2 and total SO 2 sink are ∼85% and ∼2 × 10 4 molec cm −3 s −1 respectively. On several days the rate of decrease in DMS and increase in SO 2 levels in the early morning were larger that predicted by the model. The rate and timing of the early morning loss are consistent with chemical oxidation of DMS by halogen radicals, but the quantitative yield of SO 2 is not, suggesting a dynamical explanation.