Dimethyl sulfide (DMS) and its oxidation products, which have been proposed to provide a climate feedback mechanism by affecting aerosol and cloud radiative properties, were measured on board the Canadian Coast Guard ship Amundsen in sampling campaigns in the Arctic in the fall of 2007 and 2008. DMS flux was calculated based on the surface water measurements and yielded 0.1–2.6 μmol m−2 d−1 along the Northwest Passage in 2007 and 0.2–1.3 μmol m−2 d−1 along Baffin Bay in 2008. DMS oxidation products, sulfur dioxide (SO2), methane sulfonic acid (MSA), and sulfate in aerosols were also measured. The amounts of biogenic SO2 and sulfate were approximated using stable isotope apportionment techniques. Calculating the threshold amount of SO2 needed for significant new particle formation from the formulation by Pirjola et al. (1999), the study suggests that instances of elevated biogenic SO2 concentrations (between 8 and 9 September 2008) derived using conservative assumptions may have been sufficient to form new aerosols in clean air conditions in the Arctic region.
 Sulfate in the atmosphere is derived from both anthropogenic and biogenic sources. The largest source of biogenic sulfate in the remote marine atmosphere is the oxidation of dimethyl sulfide (DMS), a compound released by the breakdown of dimethylsulfoniopropionate (DMSP) produced by phytoplankton. The suggestion that DMS could provide a feedback mechanism to stabilize the Earth's climate in the event of a warming episode has been known as the CLAW hypothesis, after the initials of its authors [Charlson et al., 1987]. This hypothesis has fueled numerous studies on DMS to ascertain its effects on climate [Ayers and Cainey, 2007]. However, the feedback mechanism has yet to be confirmed.
 Studies show that DMS flux and concentrations are related to the number of cloud condensation nuclei (CCN) [Hegg et al., 1991; Gabric et al., 2005] and that DMS is a contributor to new aerosol formation [Cainey and Harvey, 2002]. However, oxidation of DMS does not necessarily produce new aerosols. DMS oxidation can proceed via addition of the hydroxyl radical (OH) and halogen oxides, forming an adduct, or via hydrogen abstraction by nitrate radical (NO3), OH and halogens [Barnes et al., 2006; von Glasow and Crutzen, 2004]. The oxidation pathway followed by DMS would depend on the oxidants present and the temperature at which the reaction took place. In the absence of halogens in the polluted atmosphere, NO3 and OH radicals are the major oxidants of DMS. In the case of DMS oxidation with OH, the branching ratios between the two pathways was shown to be temperature dependent and was estimated to be 0.75 for abstraction and 0.25 for addition at 300 K [Hynes et al., 1986]. At lower temperatures (<285 K), the addition pathway dominates [Barnes et al., 2006; Yin et al., 1990]. The reaction with NO3 proceeds via abstraction and occurs in polluted regions and during nighttime when the concentrations of NO3 are high. The addition pathway leads to the production of dimethyl sulfoxide (DMSO) and methane sulfonic acid (MSA), while abstraction leads mainly to sulfur dioxide (SO2) [Osthoff et al., 2009; Barnes et al., 2006]. MSA does not form new aerosols under ambient conditions [Kreidenweis and Seinfeld, 1988] and instead may condense onto preexisting aerosols [von Glasow and Crutzen, 2004]. In the remote marine atmosphere, halogens could be an additional sink for DMS by oxidation. DMS oxidation with BrO (via the addition pathway) could be an important sink for DMS [Barnes et al., 2006] and could result in a smaller SO2 yield [von Glasow and Crutzen, 2004].
 The production of sulfur dioxide (SO2) and its oxidation to sulfate could form new aerosols. SO2 oxidizes via gas phase or aqueous phase reactions in the troposphere [Seinfeld and Pandis, 2006]. Aqueous phase reactions would occur in or on the surface of preexisting aerosols and cloud or fog droplets. Thus, it is the gas phase reaction of SO2 with OH which could form H2SO4 vapor which may contribute to new aerosol formation. The branching between the gas phase and heterogeneous oxidation of SO2 is on average at 24% for the gas phase reaction and 76% for heterogeneous reactions [Faloona, 2009].
 It is important to understand DMS oxidation pathways because it helps ascertain the yield of MSA and biogenic sulfate. This could lead to a better understanding on how much of this biogenic sulfate may form new aerosols. Also, H2SO4 and its precursor SO2 may aide in the growth of preexisting aerosols instead of contributing to aerosol formation. Pirjola et al.  derived an expression to determine the amount of SO2 that would yield significant new aerosol formation in relation to the condensation sink (CS), which is determined by the distribution of preexisting aerosols. Pirjola et al. defined significant new aerosol formation as a 10% increase in total aerosol concentration for a model simulation time of 4 hr. They derived the expression using a size-sectioned aerosol dynamic model that includes nucleation, condensation, coagulation, aerosol deposition and sulphuric acid formation in the gas phase and ran their model for clean air marine conditions to polluted urban conditions valid for temperatures between −20°C to 20°C and relative humidity between 0.5 and 0.9 [Pirjola et al., 1999]. The Pirjola et al.  expression has been used by Cainey and Harvey  to determine if biogenic SO2 in the Antarctic, clean marine air and free troposphere for the Southern Hemisphere was sufficient to produce new aerosol formation.
 This study aims to investigate one step in the proposed DMS feedback mechanism; whether DMS sea to air flux and DMS oxidation products are present in sufficient amounts to produce aerosol nucleation. The study conducted simultaneous measurements of atmospheric and surface water DMS and its oxidation products in the atmosphere in the Canadian Arctic in the fall of 2007 and 2008. Surface water measurements of DMS were used to calculate DMS flux to the atmosphere. Atmospheric DMS, SO2 concentrations together with MSA and sulfate in aerosols, and aerosol size distributions were measured simultaneously. Stable isotope apportionment techniques were employed to ascertain the amount of SO2 and sulfate from DMS oxidation as described in earlier studies by Nriagu et al. , Li and Barrie , and Norman et al. . The condition for new particle formation derived by Pirjola et al.  was used to determine if the amount of biogenic SO2 obtained in this study was sufficient to produce new particle formation.
2. Field and Laboratory Methods
 DMS and aerosol samples were collected on board the icebreaker CCGS Amundsenduring fall of 2007 and fall of 2008. The fall season is an interesting period for measurements as DMS and its oxidation products can be observed during the transition from open to ice-covered waters. Atmospheric samples were taken on top of the navigation bridge to minimize contamination from the ship. The route of the CCGSAmundsen for 27 September to 8 November 2007 and 4 September to 12 October 2008 is shown in Figure 1.
2.2. DMS Sampling
 Measurements of atmospheric and sea surface DMS were conducted in this study. Atmospheric DMS samples were collected every hour for 24 hr throughout the sampling period. Surface water DMS samples were taken from 8 to 13 October 2007, 23 to 24 October 2007, and 23 to 25 September 2008. Atmospheric and surface water DMS samples were taken within half an hour of each other.
2.2.1. Surface Water DMS
 DMS concentrations in water were obtained from samples taken from the ship's seawater pumping system with the intake located approximately 5 m below surface. Water was drawn from a tap in the ship's engine room. Care was taken to maintain a slow flow so as to minimize stress that could break the cell walls of DMSP-producing species and release DMS into the water. The outlet pipe was slowly flushed to remove stagnant water before sampling. Analysis of surface water DMS was conducted on board following the purge and trap, and GC method described byLuce et al. . Comparison between DMS water concentrations, obtained from the pump and those obtained from bucket and rosette measurements, were performed to ensure reliability of measurements. Water temperature and salinity were also measured for the pump samples and compared with bucket and rosette samples to serve as an indicator of the validity of the measurement.
2.2.2. Atmospheric DMS
 Atmospheric DMS was collected using a custom-made DMS sampler at a calibrated flow rate (approximately 200 mL min−1) for 5 min through a KI-treated 47 mm QA Whatman filter fitted on the exposed end of cartridges made of deactivated borosilicate liner packed with Tenax and silanized wool. The DMS samples were analyzed within 24 h after collection to prevent deterioration using a Hewlett Packard 5890 gas chromatograph (GC) fitted with a Sievers Model 355 sulfur chemiluminescence detector (SCD). The instrument was calibrated using 50 ppm and 1 ppm DMS standards from Praxair. Linearity was checked by injecting different volumes of the DMS standards. Standards were injected before and after running DMS samples. Blanks were analyzed during the initial startup of the GC-SCD. No DMS peaks were observed in the blanks. These methods were based on those described bySharma et al. .
2.3. Aerosol and SO2 Measurements
 Size-segregated aerosols were collected using a high-volume sampler (H1) with a PM10 head and a cascade impactor fitted with slotted quartz filters for collecting size-segregated and fine aerosols. The cascade impactor segregated the aerosol samples into 6 size bins with a range of aerodynamic diameters: A (7.2–10μm), B (3–7.2 μm), C (1.5–3 μm), D (0.95–1.5 μm), E (0.49–0.95 μm), and F (<0.49 μm). H1 was run for approximately 3 days for each sample at a flow rate of ∼1.10 ± 0.03 m3min−1. Daily measurements of SO2and total aerosols were conducted using a second high-volume sampler (H2) fitted with a 8 inch × 10 inch (20.3 cm × 25.4 cm) quartz filter and a cellulous filter pretreated with potassium carbonate (K2CO3) glycerol solution. H2 was run approximately 24 h for each sample with a flow rate of 1.00 ± 0.14 m3min−1. The high volume samplers were turned off when contamination from the ship's exhaust was evident, for example, when the ship stopped during periods where ocean water and sediment samples were collected and (in 2008) during waste incineration episodes as well.
 Field blanks were collected for each new lot of filters by loading clean filters on the high volume samplers while the samplers were turned off.
2.4. Isotope Analysis
 Sulfate from aerosols captured on particulate filter paper was dissolved by immersion in distilled deionized water. A 0.45 μm Millipore filter was used to remove the particulate filter residues and separate the dissolved sulfates. The filtrates were collected and treated with BaCl2 to precipitate dissolved sulfate as BaSO4. A similar process was performed for the SO2 filters but with the addition of hydrogen peroxide to oxidize the SO2 to sulfate. BaSO4 precipitate was filtered through a Nucleopore filter, dried, and packed into tin cups with Nb2O5 catalyst and loaded into an elemental analyzer to convert the BaSO4 to SO2. Sulfur dioxide was then analyzed using a PRISM II continuous flow isotope ratio mass spectrometer (CF-IRMS) to obtainδ34S values. δ34S values expressed in units of parts per thousand (permil = ‰) are calculated as
where Rx is the ratio of the heavier versus lighter isotope of sulfur (34S/32S) of the sample and Rs is the 34S/32S of the international standard, Vienna Canyon Diablo Troilite (V-CDT). The instrument was calibrated using standards with knownδ34S values (STB = −2.0‰ and SW = +20.8‰). The standard deviations of the δ34S values of the standards for each run were taken as the uncertainty for δ34S values.
2.5. Ion Chromatographic Analysis
 MSA, SO2, sulfate and other major ion concentrations (Cl−, Br−, Na+, Mg+2) were obtained using ion chromatography. Cation measurements were performed using a DX500 ion chromatograph (IC) with a CD20 conductivity detector at the Glacial Hydrochemistry Laboratory, University of Alberta. Anion measurements were obtained using a Dionex IC-2500 ion chromatograph with an ED50 electrochemical detector also at the Glacial Hydrochemistry Laboratory. For anion concentrations from SO2, a Dionex integrated IC 1000 at the Environmental Science Laboratory, University of Calgary, was used.
2.6. CO2 Analysis, Aerosol Size Distribution, and Meteorological and Radiation Data
 A LICOR CO2 gas analyzer was used to measure instantaneous CO2 at 1 min intervals. CO2 peaks were used as marker of the ship's smoke stack emissions. The size distributions for 10 to 500 nm diameter aerosols were determined using a scanning mobility particle sizer, SMPS (3080, 3081, 3010, TSI Inc.). Details of the sampling method are given by Chang et al. . The SMPS aerosol size distribution data were used to calculate the condensation sink used in section 4.5. Meteorological data measured every minute were obtained from the ship's meteorological station. Wind measurements used for flux calculations were measured from a meteorological tower installed on the foredeck of the ship, 14 m above sea level (asl). Winds were scaled to 10 m using a power law relationship given by
where the subscripts 1 and 2 indicate two different levels, u is the wind speed, and z is the height of the corresponding levels. P is the power exponent that is dependent on the stability and surface conditions. We assume P = 0.11 based on studies done by Hsu et al. in determining power law coefficients in sea conditions for a near stable atmosphere above deep and shallow ocean. Infrared radiation (IR) was measured using an Eppley Model PIR pyrgeometer. Shortwave radiation was measured using an Eppley PSP pyranometer. Ultraviolet radiation (UV-A and UV-B) was measured using a Kipp and Zonen UVS-AB-T. A more detailed description of the meteorological equipment is provided byElse et al. .
2.7. DMS Flux
 DMS flux was calculated using the expression
where Cw is the concentration of DMS in water, Ca is the concentration of DMS in air, KH is Henry's law constant or solubility constant and kw is the exchange rate or transfer velocity which can be expressed as
 The above parameterizations of kw were obtained for CO2 which has a Schmidt number Sc = 600. Using
from the work of Saltzman et al.  where SST is the sea surface temperature, the transfer velocity for DMS, kwDMS, was calculated.
 Water concentrations of DMS (100–3,000 nmol m−3) obtained in this study were three orders of magnitude higher than atmospheric concentrations (<0.3 to 4.1 nmol m−3). To simplify calculations for flux, the second term of the gas transfer equation (equation (3)) involving the atmospheric concentrations was neglected.
2.8. Stable Isotope Apportionment
 The biogenic component of SO2 and sulfate was apportioned using stable isotope analysis. The fraction of biogenic SO2was calculated using a two-source mixing given by
while the fraction of biogenic sulfate was calculated using a three-source mixing
where the quantities in brackets are concentrations.
 The amount of sea salt sulfate was determined from the mass ratio of Na+ and SO42− in seawater
 Anthropogenic sources from long-range transport originating from North America and Europe were found to haveδ34S values of +3.8‰ to +6.3‰ [Nriagu et al., 1991; Barrie et al., 1992]. Norman et al.  obtained δ34S values of +2.3‰ and +5.8‰ for nonmarine sources of sulfate measured during the fall season in Alert from air parcels originating from Eurasia and North America, respectively. During the sampling period, the ship's smoke stack emissions, with an average δ34S value of +3‰ based on measured δ34S values of ship fuel samples, were also a source of anthropogenic sulfate. A δ34S value of +3‰ was used to represent anthropogenic sulfate from both long-range transport and local ship emissions for isotope apportionment.
 An additional source of sulfate was also found and attributed to a local terrestrial source. Highly negative δ34S values for both SO2 and aerosol sulfate samples were observed in the vicinity of the Smoking Hills on Cape Bathurst, Northwest Territories. The Smoking Hills is an area of spontaneously burning sea cliffs facing Franklin Bay in the Northwest Territories that are fueled by pyrite and shale materials. Radke and Hobbs  showed that the Smoking Hills are a regionally significant source of SO2 with estimated emission of 0.3 kg s−1. δ34S values for shale from the Smoking Hills samples were measured for this study and found to be between −30‰ and −40‰. This study presents the first reported measurement of δ34S values for the Smoking Hills. The low δ34S values of SO2 and SO42− in aerosol samples obtained near the Smoking Hills region indicate that the hills were a large source of SO2 and sulfate in the area influencing the δ34S values obtained on board the ship. Thus, a range of δ34Sbackground was used varying from δ34Sbackground = −30‰ due to influence from emissions from the Smoking Hills to δ34Sbackground= +3‰ from ship emissions and long-range transport.Figure 2 shows the δ34S values for different sources of atmospheric sulfur in this study.
3.1. DMS Flux
 DMS flux was calculated using the Nightingale et al.  parameterization. Huebert et al.  showed that direct measurements of DMS flux by eddy correlation fall between the predicted values of Liss and Merlivat  and Wanninkhof  parameterizations as suggested by Nightingale et al. . Calculated values for the fall of 2008 flux ranged from 0.2 to 1.3 μmol m−2 d−1. Calculations of the fall 2007 flux ranged from 0.1 to 2.6 μmol m−2 d−1. DMS water concentrations were obtained in the presence of ice cover during the fall of 2007. Flux for 2007 was corrected with observed ice cover by multiplying fluxes with the fraction of open water. We observed the presence of greasy ice on open water surface that is not accounted for in the flux parameterizations. This is likely to reduce the gas transfer [Guest and Davidson, 1991]. Leck and Persson [1996a] assumed that there was no transfer of DMS to atmosphere when greasy ice was present. Thus our calculations for flux would be overestimated in the presence of greasy ice. It is useful to examine the flux based on the Liss and Merlivat  and Wanninkhof  parameterizations for comparison with flux reported in earlier studies. Table 1 shows the flux calculated for this study using Liss and Merlivat , Wanninkhof  and Nightingale et al.  along with values reported in the literature for the Arctic region for comparison. Fluxes for both years were less than those obtained in earlier studies conducted in the spring and summer using the same parameterization. This may be attributed to lower biological productivity during the fall.
 The total sulfate is a combination of biogenic (+18‰), sea salt (+21‰), and a background with varying δ34S values (+3 to −30‰). The distribution of δ34S values for sulfate versus the percentage sea salt sulfate (Figure 3) was obtained to aid in identifying the source mixing of sulfate. Figure 3shows the mixing line for sea salt sulfate and biogenic sulfate (solid line) and the mixing line for sea salt sulfate and anthropogenic sulfate from long-range transport and fuel emissions with a +3‰δ34S value (long-dashed line). Points falling within the region enclosed by the two lines would indicate a mix of biogenic sulfate, sea salt sulfate and sulfate from fuel emissions or long-range transport. Values falling below the sea salt sulfate and fuel mixing line indicate mixing with a more negativeδ34S source. The mixing line for sea salt sulfate and −30‰ δ34S value from the Smoking Hills is shown as a dotted line. However, δ34S values for the samples fall above this mixing line suggesting that the background is not well represented by a −30‰ δ34S value. This shows that δ34S value of background air was not constant throughout the study. Within the vicinity of the Smoking Hills the δ34S value of the non–sea salt sulfate would approach −30‰: without the influence of the Smoking Hills it would approach the δ34S value of anthropogenic sulfate at +3‰. For a mixture of the two sources of background sulfate it could be anywhere between the two values and there is a need to approximate this value for the background.
 Most of the δ34S values for sulfate versus percentage sea salt sulfate fall above a −5‰ δ34S value mixing line (Figure 3, short-dashed line).δ34S values for SO2 also mainly lie above −5‰ for both years. Fall 2008 δ34S values for SO2 when sampling was performed along Baffin Bay far from the Smoking Hills remain consistent with a minimum δ34S value of −5‰ for background sulfate since 30% of the samples have δ34S values in the range of −2‰ to −5‰. Thus a δ34S value of −5‰ was used as a minimum value for the isotopic signature of the background atmospheric sulfur except for samples taken between 20 to 25 October 2007 close to the Smoking Hills. Between 20 and 25 October 2007, −30‰ was chosen as a background δ34S value as suggested by the very negative δ34S values for SO2 and non–sea salt sulfate. The actual δ34S of the background could be lower than −5‰ as we approached the Smoking Hills, due to more influence of the −30% source. Therefore, the use of −5% as the δ34S value for background sulfate is a conservative estimate. This unexpected contribution from the Smoking Hills results in a large uncertainty in the estimation of the biogenic component of the non–sea salt sulfate. Thus, for the apportionment calculations of the concentrations of biogenic SO2 and sulfate, a δ34Sbackground of +3‰ for minimum and −5‰ for maximum possible concentrations were used with the exception of samples influenced by Smoking Hills emissions (20 to 25 October 2007) wherein δ34Sbackground is taken as −30‰ for maximum possible biogenic sulfate concentration. This gives a range of possible concentrations of biogenic sulfate and SO2 for both 2007 and 2008. Minimum biogenic SO2 and biogenic aerosol sulfate concentrations range from 0 to 97 nmol m−3 and 0–1.2 nmol m−3, respectively, for 2007 and 0–11 nmol m−3 and 0–0.15 nmol m−3, respectively, for 2008. The maximum possible biogenic SO2 and biogenic aerosol sulfate concentrations were obtained and found to range from approximately 0–110 nmol m−3 and 0–2.0 nmol m−3, respectively, for 2007 and 0–17 nmol m−3 and 0–0.5 nmol m−3, respectively, for 2008. Minimum possible percentages of biogenic SO2 and sulfate are 0–75% and 0–35%, respectively, using minimum biogenic concentrations. Table 2 shows the range of median biogenic SO2 and sulfate concentrations along with averaged daily DMS, MSA and meteorological and radiation measurements.
Table 2. Summary of Daily Meteorological and Atmospheric Measurementsa
Daily average temperature, relative humidity, and PAR were calculated by averaging per minute measurements over the daily sampling period. Daily DMS was calculated from hourly measurements of DMS concentrations averaged over the sampling period. The sampling period is the time period for total daily particulate and SO2 sampling. Daily median biogenic is the midpoint for the range of biogenic SO2 and sulfate approximated using stable isotope apportionment. We show here the minimum, maximum, average, or median of the daily values.
Wind, m s−1
Relative humidity, %
PAR, μmol m−2 s−1
DMS, nmol m−3
MSA, nmol m−3
SO2, nmol m−3
Sea salt sulfate, nmol m−3
Non–sea salt sulfate, nmol m−3
Minimum biogenic SO2, nmol m−3
Minimum biogenic sulfate, nmol m−3
Maximum biogenic SO2, nmol m−3
Maximum biogenic sulfate, nmol m−3
3.3. Size-Segregated Aerosol Measurements
 Sulfate and MSA concentrations for aerosols were obtained for 3 day sampling periods in different size bins, which are defined with the range of aerodynamic diameter as follows: A (7.2–10 μm), B (3–7.2 μm), C (1.5–3 μm), D (0.95–1.5 μm), E (0.49–0.95 μm), and F (<0.49 μm). As the size bins cover different size ranges, the concentrations were normalized by dividing by the size of the bin. The mass concentrations for sulfate and MSA associated with the size distributions, normalized over the size of the bin for fall 2007 and 2008, are shown (Figure 4). Sea salt sulfate (SS) mass concentrations mostly peaked in the B bin. A normalization of the mass concentration with the bin size resulted in values of SS mass concentrations that peaked mainly in the F bin in 2007 but peaked in the D and B bin in 2008 and is consistent with aged sea salt aerosols in 2007 in contrast to fresh sea salt in conjunction with more open water in 2008. The existence of high SS concentrations in the F bin is also reported at lower latitudes [Seguin et al., 2011]. Seguin et al.'s  values for SS concentrations normalized over the bin size showed peaks in the <0.49 μm size range consistent with our findings. Sea salt sulfate in the supermicrometer modes, D and B, were not coincident with higher wind speeds. The difference in the size distribution of SS, specifically where peak concentrations occur, may be due to a change in the air mass source region. Air parcel back trajectories in 2007 show variable source regions while air parcel back trajectories in 2008 show air parcels originate mainly from north of Baffin Bay and Greenland. However, 2007 samples with air parcels originating from Baffin Bay and Greenland did not exhibit peaks in the D and B bin. Another thing to consider is that more open water was present in fall 2008 than in 2007 so that winds swept off Greenland across Baffin Bay may have additional local SS possibly from bubble breaking.
 MSA peaked mainly in the F size bin (seven out of nine samples in 2007 and all five samples in 2008) while non–sea salt sulfate (NSS) peaked in the E and F size bins for both years. Less than half of the NSS peaked in the F bin in 2007 while the majority of the samples displayed NSS that peaked in the E bin in 2008, and one sample showed a peak of NSS in the D size bin. Peak MSA and NSS concentrations in the F bin is consistent with the work of Leck and Persson [1996b] and Hillamo et al.  that showed MSA and NSS were found primarily in submicron particles. The F, or fine size fraction, is of primary interest because newly formed aerosols from biogenic sulfate and subsequent growth by accumulation would fall within this size range. The peak in MSA and NSS in this size range suggests that MSA and NSS or their gaseous precursors are condensing or oxidizing on preexisting aerosols in the fine size fraction. High 3 day average particle count means most of the MSA and NSS or gaseous precursors would condense or oxidize on preexisting aerosols rather than form new aerosols. Instances in fall 2007 and most of fall 2008 when MSA peaked in the F bin and NSS sulfate peaked in the E bin may indicate growth of aerosols due to NSS or its gaseous precursors condensing or oxidizing onto preexisting aerosols. Due to the low time resolution of the sampling, it is impossible to determine growth by just looking at the NSS concentrations. Evidence of aerosol growth in aerosols <30 nm in diameter were reported for the fall of 2008 by Chang et al.  (September 7, 10, and 19–25). However, because the cutoff for Chang et al.'s  measurements were at 50 μm, growth in the E bin is outside this range and was not detected.
 The temporal distribution of MSA and NSS for 3 day samples obtained from size F particulate filters (<0.49 μm) with the 3 day averaged DMS for 2007 and 2008 (Figures 5a and 6a) and the corresponding biogenic component of NSS for 2007 and 2008 (Figures 5b and 6b) are shown. DMS ranged from below detection limit to 0.9 nmol m−3 (2007) and 1–2.4 nmol m−3 (2008). MSA ranged from 0.004 to 0.01 nmol m−3 (2007) and from 0.01 to 0.026 nmol m−3 (2008). Non–sea salt sulfate ranged from 0.3 to 1.7 nmol m−3 (2007) and from 0.1 to 0.4 nmol m−3 (2008). It is interesting to note atmospheric DMS, MSA, and NSS closely follow a trend of decreasing concentrations further into the season in 2007. Of particular interest is the fact that NSS follows the temporal trend of DMS and MSA despite it not being fully biogenic in origin. Sea salt sulfate does not follow this trend. MSA, NSS, and atmospheric DMS showed a decrease approaching 20 October 2007 along the Northwest Passage and Beaufort Sea which may be due to reduced flux due to increased ice cover (Figure 1) in combination with decreased productivity in the sampling area as indicated by decreased chlorophyll a measurement reported by Luce et al. . MSA and atmospheric DMS concentrations were higher in 2008 than in 2007, while NSS concentrations were lower in 2008 in comparison to 2007. Using apportionment techniques in the fine fraction using minimum and maximum δ34S values, biogenic sulfate concentrations ranged from approximately 0 to 0.27 nmol m−3 (2007) and 0 to 0.18 nmol m−3 (2008). The biogenic component of NSS in the fine fraction ranged from 0% to 50%.
4.1. Comparison of Flux and Atmospheric DMS Concentrations
 The temporal distribution of DMS fluxes were compared to atmospheric DMS concentrations and wind speeds for 2007 and 2008 and are shown in Figure 7. DMS flux and concentrations in 2007 increased with wind speed (Figure 7, left). DMS flux in 2008 peaked on 24 September after 1:00 and on 25 September after 15:00. Increased flux means more DMS is released into the atmosphere which could potentially increase atmospheric DMS concentrations. This is opposite to what was observed in 2008. Increased DMS flux from 23 September 21:00 to 24 September 3:00 corresponded to decreased atmospheric DMS concentration. A few hours after 3:00, DMS concentrations in air increased while flux decreased. A possible explanation for this could be the collapse of the boundary layer at night followed by a shallow inversion. Otherwise, it could be due to the combination of factors such as movement of air parcels or a change in oxidation mechanism that controlled the atmospheric concentrations.
4.2. Comparison of Biogenic SO4 and SO2, DMS, and MSA Concentrations
 The major products of DMS oxidation, MSA and SO2, are dependent on the oxidation pathway. The addition pathways (via OH, or possibly BrO) lead to the production of DMSO and MSA, while abstraction leads mainly to SO2 [Osthoff et al., 2009; Barnes et al., 2006]. Thus, determining the amount of MSA and biogenic SO2 gives insight into the oxidation mechanism of DMS. Comparison of atmospheric DMS, MSA and biogenic SO2 and sulfate for both years show that atmospheric DMS and MSA were on average higher in 2008 than in 2007 (Table 2). This is consistent with the generally higher DMS surface water concentrations measured in 2008 (0.52 to 4.75 nmol L−1; J. Motard-Côté et al., Distribution and phylogenetic affiliation of dimethylsulfoniopropionate (DMSP)-degrading bacteria in northern Baffin Bay/Lancaster Sound, submitted toJournal of Geophysical Research, 2011) compared with later in the season in 2007 (0.05 to 0.80 nmol L−1 [Luce et al., 2011]). However, biogenic SO2 and sulfate were on average higher in 2007 than 2008. There were instances where biogenic SO2 concentrations were higher relative to atmospheric DMS for both years even when the minimum value of SO2 from isotope apportionment was used. This may indicate that DMS was confined to a localized source region while SO2 was transported from a wider source region. Due to the uncertainty in the value of the background δ34S (from +3‰ to −30‰), the calculation of biogenic SO2 concentrations might be overestimated if the choice of the δ34S value of the background sulfur (+3‰, −5‰, −30‰) was underestimated. However, the choice of δ34S values was very conservative and there are instances of nonzero minimum biogenic SO2 which are greater than the measured atmospheric DMS for those days. This supports the postulate that biogenic SO2 came from a wider source region. Sharma et al.  showed that atmospheric DMS in the high Arctic has a lifetime of 2.5 to 8 days. On the other hand, SO2 lifetime is more than a week in Arctic conditions [Thornton et al., 1989], which would allow for long-range transport of both biogenic and anthropogenic SO2.
 Long-range transport and the ship's emissions, two possible causes of high SO2, were investigated. Ship's emissions were accounted for using isotope apportionment and validated by comparing CO2 measurements to SO2 concentrations. There was no marked increase in SO2 concentration with increased CO2 indicating that there was no dependence of SO2 concentration on ship stack emissions. Thus, it is more likely that the higher biogenic SO2measured was due to the influence of long-range transport.
 Daily 3 day back trajectories for fall 2007 and 2008 cruises were obtained using the NOAA Hysplit model (R. R. Draxler and G. D. Rolph, 2010, http://ready.arl.noaa.gov/HYSPLITT.php/). The location of the ship was averaged over the sampling period of 1 day. Air parcels in 2007 have more varied source regions than in 2008, while back trajectories for fall 2008 show air parcels mainly coming from north of Baffin Bay or coming from the east passing over Greenland, with occasional winds coming from the northwest passing over the Canadian Archipelago. Back trajectory analysis for 7–8 October 2007 and 9–10 October 2007, sampling periods displaying with the highest SO2 concentrations, are associated with air parcels coming from the south and therefore may have been influenced by air passing over the Hudson's Bay which may provide an additional source of SO2 and sulfate. The Hudson's Bay area has been reported as a source of DMS and thus is expected to contribute to biogenic SO2 and sulfate [Richards et al., 1994].
4.3. Correlations of DMS, MSA, and NSS in Fine Aerosols
 It is interesting to note that in Figures 5 and 6 NSS and MSA in fine aerosols and atmospheric DMS all followed a decreasing trend in concentration further into the season despite the fact that NSS can be derived from other sources and is estimated to be no more than 50% biogenic in origin. A plot of DMS averaged over the 3 day sampling period versus MSA in the F mode aerosols (<0.49 μm) for 2007 and 2008 yields a Spearman rank correlation, rho = 0.88 (Figure 8a). This correlation may indicate that MSA in the <0.49 μm aerosol size range comes primarily from oxidation of DMS from a local source. It shows that atmospheric DMS is rapidly oxidized to MSA under the Arctic conditions present during the studies. No significant correlation was found between DMS and minimum or maximum biogenic SO2. This may be due to the large uncertainty in the approximation of biogenic SO2concentrations and possible long-range transport of SO2 as discussed in section 4.2.
 Despite SO2 oxidation contributing to NSS from a wider source area, the NSS and MSA in fine aerosols exhibited a linear correlation with rho = 0.95 (p value = 0.001) and 0.90 (p value = 0.08) for 2007 and 2008, respectively (Figure 8b). This correlation between MSA and NSS suggests a common process or source. However, the source of MSA and NSS is not identical since less than 50% of the NSS is biogenic in origin. More likely a common process such as the condensation of MSA and NSS could explain the results. Their concentration in aerosols are more likely dependent on the presence of existing aerosols, primarily those in the fine fraction, which provides a large surface area for reactions to take place. It is possible that SO2 produced by DMS oxidation leads to both nucleation and further growth of aerosols. However, in most cases in this study, the high number of preexisting aerosols and thus high condensation sinks would suppress nucleation as will be shown in section 4.5. This demonstrates the importance of preexisting aerosols in the fine fraction to new aerosol formation and growth from DMS oxidation.
4.4. MSA/Biogenic Sulfate Ratio
 MSA/NSS ratios in clean air have been shown to have an inverse linear relationship with temperature [Bates et al., 1992]. In the case of clean air, NSS has been reported as representative of biogenic sulfate. Our study is one of the first to be able to define MSA/biogenic sulfate rather than MSA/NSS. We used nonzero maximum values of biogenic sulfate concentrations using a conservative δ34S value for background sulfate to obtain an approximate value for MSA/biogenic sulfate. We obtain an MSA/biogenic sulfate ratio in the total aerosol ranging from 0.02 to 0.21 for 2007 and 0.2 to 0.3 for 2008. Bates et al.  showed that MSA/NSS ratios are affected by the aerosol size fraction. We obtained the MSA/biogenic sulfate in the fine fraction (<0.49 μm) for comparison with Bates et al.  who reported a <0.6 μm size fraction. MSA/biogenic sulfate using maximum nonzero biogenic sulfate was 0.03 to 0.17 (2007) and 0.14 to 1.2 (2008) in the fine size fraction. Extrapolating the expression obtained by Bates et al.  for the temperature dependence of MSA/NSS ratios, a ratio of 0.57 for a temperature of −10°C would be expected in 2007 and 0.50 for −5°C in 2008. This study observed a maximum ratio in 2008 higher than expected indicating instances where there is a strong preference for DMS oxidation to MSA or an underestimation of the amount of biogenic sulfate. Due to the uncertainty in the approximation of biogenic sulfate and using a modest estimate for biogenic sulfate concentrations, the higher MSA/biogenic sulfate ratio may be due to an underestimation of biogenic sulfate and the ratios may be consistent with Bates et al. within the range of uncertainty of the apportionment technique. MSA/biogenic sulfate ratios were lower in 2007 in comparison to 2008 which may be due to more sulfate produced from long-range transport of SO2 rather than a preference of local DMS to oxidize to SO2.
4.5. Condition for New Aerosol Formation
 Sulfur dioxide produced by DMS oxidation could result in the growth of existing aerosols or the formation of new aerosols. Formation of new aerosols is key to the feedback mechanism proposed in the CLAW hypothesis as this could form new CCN and increase the cloud albedo. To determine if biogenic SO2 from DMS oxidation produces a significant amount of new aerosols, the condition given by Pirjola et al.  was used. The condition is given by
and T and RH are the temperature and relative humidity, respectively. Concentrations for SO2 and OH are given in molecules cm−3 and the unit for CS is cm−2.
 The condensation sink, CS, is calculated from
where r is the radius in cm and N is the number concentration in cm−3 of each size bin obtained from the SMPS. βm is from Fuchs and Sutugin  and is given by
where α is the dimensionless sticking coefficient and it is assumed α = 1 based on the work of Clement et al.  and was the value used by Pirjola et al. , and Kn is the dimensionless Knudsen number equal to the mean free path over the aerosol radius (Kn = 2 λ/D).
 Calculated OH concentrations were based on measured UV-B irradiance and existing measurements of OH in the Arctic.Rohrer and Berresheim  observed a linear correlation between OH and the photolysis frequency of ozone J(O1D) (s−1) which is dependent on the intensity of UV-B. For simplicity, the irradiance, I (Wm−2), was assumed to be proportional to the photolysis frequency based on the linear relationship observed by Palancar et al.  for the photolysis frequency of O3and UV-B irradiance. From this, the following assumption is made.
 Median OH radical values of 6.4 × 106 molecules cm−3 were reported over the Greenland ice pack during summer daylight [Sjostedt et al., 2007]. In an earlier study, OH radicals were reported to be less than 3 × 106 molecules cm−3 measured above the boundary layer from flights over the Canadian Arctic during early spring [Mauldin et al., 2003]. The amount of solar radiation in the fall is similar to that in spring, thus 3 × 106 molecules cm−3is used to represent the maximum OH concentration for the fall. From this, the daily OH concentration for fall 2008 was scaled according to the daily average UV-B irradiance. For fall 2007, where UV-B was not measured, the average OH calculated for October 2008 was used to give an approximate value of [OH]. [SO2], [OH] and CS were averaged over the daily sampling period. These concentrations were used to calculate daily F values, which were compared to the condition given by Pirjola et al. . Using this condition the threshold SO2 concentration, [SO2]T, that would yield significant new aerosol formation was also calculated. CS ranged from 2.0 × 10−3 cm−2 to 3.4 × 10−1 cm−2 with an average of 5.6 × 10−2 cm−2 for 2007 and 1.8 × 10−4 cm−2 to 5.3 × 10−2 cm−2 with an average of 1.4 × 10−2 cm−2 for 2008. With the exception of the minimum value in 2008, the CS is significantly larger than the observations of Cainey and Harvey  for the Southern Hemisphere (e.g., 3.7 × 10−4 cm−2 for Baring Head, New Zealand; 5.6 × 10−5 cm−2 in the free troposphere above New Zealand; and 8.1 × 10−6 cm−2 in the East Antarctic Plateau). No significant new aerosol formation would theoretically occur for most of the atmospheric conditions experienced on board the Coast Guard ship based on the daily average CS calculated.
 However, it is possible to examine new aerosol formation for instances where clean air free of ship stack emissions was sampled. Note that daily averages are presented since SO2 measurements were taken daily. Despite high average daily CS, there were significant periods where N was in the order of 102 cm−3 comparable to those observed by Cainey and Harvey  for clean air in the Southern Hemisphere. From 8 September 18:10 and 9 September 00:10 clean air conditions were exhibited with an average N = 294. The average CS for this period is 1.0 × 10−4 cm−2, which is in the same order of magnitude observed in clean marine environments (e.g., Baring Head). A CS value of 1 × 10−4 cm−2 was used to approximate clean air conditions and a [SO2]T of 7 × 109 molecules cm−3 or 11 nmol m−3 was obtained using average OH concentrations.
 Minimum possible values of biogenic SO2 in this study ranged from 0 to 97 nmol m−3 (2007) and 0–11 nmol m−3 (2008). A minimum daily averaged SO2 of 11 nmol m−3 was observed from 8 to 9 September. This value equals the threshold SO2 for clean atmospheric conditions from 8 September 18:10 to 9 September 00:10. Thus, it is possible at this instance that for clean Arctic air, the SO2 derived from DMS oxidation is sufficient to lead to new aerosol formation. This is consistent with aerosol nucleation events observed by Chang et al.  for fall 2008 in the Arctic.
5. Summary and Conclusion
 DMS oxidation products MSA and SO2 can change aerosol radiative properties by condensing onto aerosol surfaces, or in the case of SO2, may also oxidize to form H2SO4 vapor that could produce new aerosols. To ascertain the effects of DMS and its oxidation products under Arctic conditions in the fall, concentrations of DMS, SO2, MSA and SO4 were obtained for 2007 and 2008. DMS flux was calculated based on parameterizations by Nightingale et al.  and yielded 0.1 to 2.6 μmol m−2d−1 in 2007 and 0.2 to 1.3 μmol m−2d−1in ice-free conditions of 2008. This is slightly lower than what is reported in previous studies and is believed to be due to the lower biological productivity during the fall season. Atmospheric DMS measurements yielded concentrations ranging from below detection limit (0.3 nmol m−3) to 1.3 nmol m−3 in 2007 and 4.1 nmol m−3 in 2008. MSA concentrations ranged from 0.03 to 0.07 nmol m−3 in 2007 and 0.02 to 0.14 nmol m−3 in 2008. Higher MSA in 2008 is coincident with higher atmospheric DMS concentrations and is likely due to the earlier sampling period in 2008 (September) as compared to 2007 (October).
 The MSA concentrations were less than 1% of the atmospheric DMS concentrations. MSA concentrations in aerosols <0.49 μm in diameter were shown to be correlated with the DMS averaged over the aerosol sampling period yielding a Spearman rank coefficient, rho = 0.88. This good correlation indicates that the MSA was likely derived from locally emitted DMS. However, there is no correlation between biogenic SO2 and DMS in this study. Concentrations of biogenic SO2 in some instances exceeded those for DMS. This suggests that SO2 came from a wider source area due to its longer residence time in Arctic conditions. Back trajectories of two instances with high biogenic SO2 showed air parcels originating from the Hudson's Bay area, which may provide additional sources of DMS.
 Non–sea salt sulfate in <0.49 μm diameter aerosols was estimated to be at most only 50% of the sulfate present in this size bin. Despite NSS originating from sources other than DMS oxidation and the fact that SO2,the precursor of NSS, may have come from long-range transport, the NSS was found to have a linear correlation with MSA that yielded a Spearman rank coefficient, rho = 0.95 in 2007 and rho = 0.90 in 2008. This correlation suggests that both MSA and NSS or its gaseous precursors condensed onto preexisting aerosols and the concentrations of MSA and NSS are likely dependent on existing aerosols in the fine fraction.
 Finally, this study has shown that during clean air conditions present between 8 and 9 September 2011, the minimum amount of biogenic SO2 derived from conservative assumptions using isotope apportionment techniques reached the threshold SO2 concentrations needed to produce new aerosols. This indicates that SO2 from DMS oxidation could result in new particle formation in clean air conditions present in the Arctic.
 This research is part of the Arctic SOLAS program and was supported by ArcticNet and CFL and by funding from NSERC and IPY. The authors would like to thank fellow scientists and the crew of the CCGS Amundsen.