The distribution and biological cycling of the climate active trace gas dimethylsulfide (DMS) and its algal precursor dimethylsulfoniopropionate (DMSP) were characterized at 20 stations across the Canadian High Arctic during fall 2007. Transformation rates of DMSP and production rates of DMS from dissolved DMSP (DMSPd) were measured during 3 h onboard incubations with radioactively labeled 35S-DMSP. Particulate DMSP (DMSPp) in surface waters varied between 2 and 39 nmol L−1 and increased with chlorophyll a (Chl a) concentrations (r = 0.84). DMS concentrations in surface waters ranged from 0.05 to 0.8 nmol L−1 and were positively correlated with DMSPp (r = 0.89) and Chl a (r = 0.74). The DMSPd loss rate constant varied from 0.01 to 0.14 h−1 and was also positively correlated with Chl a concentrations (r = 0.67). The turnover time of the DMSPd pool varied between 0.3 and 3.4 days (mean = 0.96 day). Bacterial DMS production varied between 0.01 and 0.51 nmol L−1 d−1 (mean = 0.14 nmol L−1 d−1). Assuming local steady state conditions at the time scale of a day, the turnover time of the DMS pool based only on production from DMSPd was ∼6 days at the sampling stations. This long turnover time suggests that DMS production was dominated by nonbacterial processes during our study. Our results show that DMS production could persist at low rates in late fall under ice-free conditions. The magnitude of this production appears to be limited by the low algal and bacterial production prevailing at that time.
 Dimethylsulfide (DMS) is a biogenic sulfur gas that is the single most important natural source of sulfur to the atmosphere, accounting for up to 80% of global biogenic sulfur emissions [Kettle and Andreae, 2000]. Once in the atmosphere, DMS is oxidized to sulfate aerosols that exert a climatic cooling effect directly by scattering solar radiation, and indirectly by forming small-radius cloud condensation nuclei, thereby increasing the albedo of low-altitude clouds. DMS oceanic emissions are important globally, but their impact on local climate may be particularly important in the Arctic, where warming is magnified by lower albedo due to the reduction in the extent of the summer ice cover [Perovich et al., 2007; Zhang et al., 2008]. Modeling experiments suggest that the interaction between a larger ice-free surface available for gas exchange and a potential stimulation of DMS biological production could partly offset the warming caused by the loss of ice albedo [Gabric et al., 2005].
 Due to its remoteness and difficulty of access, little of the Arctic Ocean has been intensively studied. As a consequence, DMS measurements in Arctic waters are scarce, with data mostly reported from the Bering Sea [Barnard et al., 1984; Bates et al., 1987; Sharma et al., 1999], the Barents Sea [Leck and Persson, 1996; Matrai and Vernet, 1997], the North Water polynya in northern Baffin Bay [Bouillon et al., 2002], and Resolute Passage in the Canadian Arctic Archipelago [Levasseur et al., 1994]. Those studies reported DMS levels ranging from non detectable to greater than 10 nmol L−1, with maximum concentrations generally measured at the marginal ice zone or during blooms of strong dimethylsulfoniopropionate (DMSP) producers such as the prymnesyophytes Emiliania huxleyi (Lohmann) Hay and Mohler in the Bering Sea and Phaeocystis spp. in the Barents Sea.
 DMS is produced by the enzymatic cleavage of intra and extracellular DMSP, a compatible osmolyte synthesized by many microalgae [Keller, 1989]. The few available DMSP measurements in the Arctic show concentrations ranging from 0.5 nmol L−1 in low-biomass waters to ∼27 nmol L−1 during diatom and Phaeocystis blooms in the marginal ice zone of the Barents Sea [Matrai and Vernet, 1997] and North Water polynya [Lee et al., 2001]. The activity of the enzymes responsible for the conversion of DMSP into DMS has been measured in microalgae, notably in dinoflagellates and prymnesiophytes, and in several bacterial clades [Stefels et al., 1995; Niki et al., 2000; Ansede et al., 2001; Steinke et al., 2002]. Few studies have successfully quantified the respective roles of photosynthetic and heterotrophic organisms in DMS production [Slezak et al. 2007; Vila-Costa et al., 2008]. However, in many cases, the bulk of DMS production is thought to be carried out by bacteria consuming the DMSP released in the water column by microalgae upon senescence, grazing or viral lysis [Kiene et al., 2000; Zubkov et al., 2002]. The DMSP consumed by bacteria can then be routed through either one of two competing pathways: (1) the cleavage pathway (DMSP-lyases or DddD/acyl coenzyme A) leading to the formation of DMS [Ansede et al., 2001; Todd et al., 2007] and (2) the demethylation/demethiolation pathway leading to the production of methanethiol and other products [Kiene et al., 2000]. Experiments conducted with radiolabeled 35S-DMSP showed that bacteria switch from the demethylation/demethiolation pathway to the production of DMS when their sulfur demand is satisfied [Pinhassi et al., 2005]. In spite of their key role on DMS dynamics, the relative importance of bacterial processes in controlling DMS production in the Arctic is still unknown.
 The objective of this study was to expand the limited data on DMSP and DMS concentrations in the Canadian and Greenlandic Arctic waters and to quantify the microbial metabolism of dissolved DMSP (DMSPd) using radiolabelled 35S-DMSP. The study was conducted during the fall of 2007 in conjunction with the International Polar Year and coincided with a record ice minimum [Comiso et al., 2008].
2.1. Study Area
 Measurements were carried out onboard the CCGS Amundsen at 20 stations in the Canadian High Arctic from northern Baffin Bay (76° 22.136′N, 71° 18.93′W) to the Beaufort Sea (71° 3.876′N, 130° 36.37′W) through the Northwest Passage, between 1 October and 2 November 2007 (Figure 1). Sampling was conducted at 2–3 m depth with 12 L Niskin-type bottles (OceanTest Equipment) mounted on a rosette, except at station B, where surface water was collected with a bucket from the ship's deck. Temperature and salinity were measured with a CTD (911plus, Sea-Bird Electronics) attached to the rosette. All samples were collected during daylight, which varied from 8 h on 1 October to 5 h on 2 November. Incident photosynthetically active radiation (PAR, 400 to 700 nm) was measured at 15 min intervals with a LI-COR cosine sensor (LI-190SA) placed on the foredeck in an area protected from shading. New ice was abundant in the Beaufort Sea after 25 October, but was generally absent east of 120°W, with the exception of icebergs in northern Baffin Bay and ice floes at station 309.
2.2. Chlorophyll a, Eukaryotic Cells >2 μm, and Bacterial Counts
 For the determination of chlorophyll a (Chl a), 500 mL samples of surface seawater were filtered onto Whatman GF/F filters and processed following the acidification method of Parsons et al.  and analyzed on a Turner Designs fluorometer. Identification and enumeration of eukaryotic cells >2 μm were carried out on 250 mL surface seawater samples preserved in acidic Lugol's solution [Parsons et al., 1984] for ∼3 months prior to analysis with an inverted microscope [Lund et al., 1958]. For heterotrophic bacterial abundance, water samples were fixed with paraformaldehyde (final concentration 2% v/v) and kept in the dark for 24 h, and then filtered onto 0.2 μm polycarbonate filters, which were kept frozen at −80°C until subsequent analysis. Portions of the filter were stained with 4′, 6-diamidino-2-phenylindole [Porter and Feig, 1980] and bacteria were counted at 1000x under an epifluorescence microscope. In this paper, bacteria may include Archaea which could represent ∼10%–15% of total prokaryotic standing stocks in the Canadian Basin of the western Arctic Ocean [Kirchman et al., 2007].
2.3. DMSP and DMS Concentrations and Rate Measurements
 All samples were taken in triplicate. For the determination of total DMSP (DMSPt), 3.5 mL of unfiltered surface seawater was treated with 50 μL of 50% H2SO4 (1.4% final concentration). Seawater samples for the determination of DMSPd were collected following the small-volume gravity drip filtration (SVDF) technique [Kiene and Slezak, 2006] and further processed as described above for DMSPt samples. Acidified DMSPt and DMSPd samples were allowed to react for 24 h, in the dark at 4°C. For DMS analysis, 25 mL of unfiltered seawater was gently poured into serum vials leaving no headspace, hermetically sealed, then kept on ice in the dark and processed within 1 h of collection.
 DMSP was analyzed as DMS following its hydrolysis with a strong base (NaOH). DMS concentrations (from DMSP and DMS samples) were determined using a modification of the method of Levasseur et al.  to accommodate the SVDF technique. Samples were quantified on a gas chromatograph (GC Varian CP-3800) equipped with a pulsed flame photometric detector (PFPD). Hydrolysis of the DMSP samples was achieved directly in the sparging column of a purge-and-trap system upstream of the GC by injecting 1 mL of 5 mol L−1 NaOH followed by the 3.5 mL DMSP sample.
 Calibration was done with a DMS standard (Kin-Tek Laboratories) diffusing at a constant rate from a certified permeation tube at 40°C and diluted with helium. Detection limit was ∼0.03 nmol L−1 for samples of 3 mL. In this study, DMSPd concentrations measured using the SVDF filtration method were very low, and triplicates were pooled when necessary to get sufficient volume allowing detectable concentration of DMSPd. Levels of DMSPd remained below our detection limit at stations 309 and 1606 and these two stations were excluded from further analyses requiring calculations based on DMSPd values. Particulate DMSP (DMSPp) was computed as the difference between DMSPt and DMSPd.
 Microbial DMSP uptake and metabolism were investigated using duplicate surface water samples incubated with the radiolabeled tracer 35S-DMSP, following the protocol of Kiene and Linn  and methods described in greater detail by Merzouk et al. . Briefly, trace amounts of 35S-DMSP (0.0011 nmol L−1 final concentration) were added to surface seawater representing an addition of less than 4% of the measured in situ DMSPd concentrations. Total initial radioactivity was determined following a gentle mixing of the 71 mL polyethylene bottles and a settling period of 5 min. The subsequent 3 h incubations were carried out in the dark at in situ temperature (−1.8 to +0.05°C). After 0, 60, 120 and 180 min, subsamples were taken to determine the particulate, dissolved nonvolatile and volatile fractions following a slightly modified version of the method described by Merzouk et al. . Ellman's reagent was used to bind methanethiol (MeSH), another volatile product of DMSP degradation. Any volatile 35S remaining in the sample was thus assumed to be 35S-DMS.
 Radioactivity in all subsamples was measured directly onboard the ship using a Tri-Carb 2900 TR scintillation counter. Since the samples used for determining the unreacted 35S-DMSPd were not filtered, a fraction of the 35S-DMSP taken up into bacterial cells (the particulate fraction) might have been wrongly counted as part of the unreacted 35S-DMSP fraction. Thus, the microbial DMSPd loss rate constant (kDMSPd) might have been underestimated, although the fraction of DMSP retained in bacterial cells after several hours is generally only 5%–10% [Kiene and Linn, 2000].
 This incubation protocol allows the determination and calculation of the following four parameters: (1) the microbial (bacteria and phytoplankton) DMSPd loss rate constant (kDMSPd; slope of the natural logarithm-transformed radioactivity of unreacted 35S-DMSPd during the 3 h time course), (2) the microbial DMSPd uptake rate (DMSPd concentration x kDMSPd), (3) the bacterial DMS yield (percentage of 35S-DMSPd consumed by bacteria that was recovered as volatile 35S after 3 h of incubation), and (4) the bacterial DMS production rate (microbial DMSPd uptake rate x DMS yield). Based on the report that some diatoms can take up and accumulate DMSPd [Vila-Costa et al., 2006], the kDMSPd and DMSPd uptake rate are considered to be microbial while the DMS yield and DMS production rate are assumed to be mostly bacterial, although algal DMSP lyase activity cannot be totally excluded.
2.4. Statistical Analyses
 Normal distribution for each physical, chemical and biological variables measured along the transect was assessed using Kolmogorov-Smirnov's D statistic. Association between variables following a normal distribution with α = 0.05 was investigated with Pearson's coefficients of correlation, which were tested for statistical significance with Student's t statistic. Statistics were computed with the R software (http://www.r-project.org/), graphics were produced with SigmaPlot (Systat Software Inc.), and the map was generated with the m_map package (http://www.eos.ubc.ca/∼rich/map.html) for MATLAB (The MathWorks), using the GSHHS coastline database (http://www.ngdc.noaa.gov/mgg/shorelines/data/gshhs/).
3.1. Temperature, Salinity, Chlorophyll a, and Sulfur Compounds
 Sea surface temperatures were close to or below 0°C within the study region, with the highest values measured in northern Baffin Bay, Dease Strait and Amundsen Gulf (Figure S1a in the auxiliary material). Temperature lower than −1.5°C was recorded in the Northwest Passage and the Beaufort Sea. The highest salinities (∼32–33) were measured in Beaufort Sea and Baffin Bay with a minimum value of 27 in Dease Strait (Figure S1b in the auxiliary material).
 Surface Chl a ranged from 0.11 to 2.39 μg L−1, with maximum concentrations in northern Baffin Bay (sampled in early October), moderately low concentrations around 0.4 μg L−1 in the Northwest Passage, and concentrations less than 0.25 μg L−1 in the Beaufort Sea (sampled in early November) (Figure 2a). Bacterial abundance ranged from 0.58 to 1.39 × 109 cells L−1, was not correlated with Chl a (Table 1), and showed no consistent pattern between regions (Figure 2b). DMSPp concentrations varied between 2.0 and 39.2 nmol L−1, with the maximum value measured at station 115 in Baffin Bay, intermediate values through the Northwest Passage, and minimum values in the Beaufort Sea (Figure 2c). Levels of DMSPd were below 2 nmol L−1, except at one station in the Beaufort Sea, one in Amundsen Gulf and one in Baffin Bay where they reached 2.5–5 nmol L−1 (Figure 2d). DMS concentrations varied between 0.05 and 0.8 nmol L−1 (Figure 2e). The distribution of DMSPp was positively correlated with Chl a, while DMS concentrations followed the distribution of Chl a and DMSPp (Table 1). The distribution of DMSPd was weakly correlated to DMS but not to DMSPp (Table 1). We found no significant correlations between the sulfur pools and the variations in incident light and mixed layer depth (MLD; data not shown).
Table 1. Pearson's Coefficient of Linear Correlation (r) Between Water Temperature (Temp), Salinity (Sal), Chlorophyll a (Chl a), Bacterioplankton Abundance (Bact), DMSPp:Chlorophyll a Ratio (Ratio), DMSPp, DMSPd, DMS, DMSPd Loss Rate Constant (kDMSPd), Bacterial DMS Yield (Yield), DMSPd Microbial Uptake Rate (UR), and Bacterial DMS Production Rate (PR)a
Single asterisk indicates 0.01 < p < 0.05, two asterisks indicate 0.001 < p < 0.01, and three asterisks indicate p < 0.001.
 Flagellates numerically dominated the community at all stations except at the Chl a–rich northern Baffin Bay station (115) where diatoms were the most abundant group (Figure S2 in the auxiliary material). Phaeocystis spp. was detected only at station 309 in Viscount Melville Sound, accounting for 1.3% of the total eukaryotic cell number (data not shown).
3.2. DMSP Microbial Metabolism
 The microbial DMSPd loss rate constant (kDMSPd), which is a measure of the ability of the microbial assemblage to use DMSPd, varied between 0.01 and 0.14 h−1 with a mean value of 0.06 h−1 (Figure 3a). Maximum kDMSPd values were measured at station 115 in Baffin Bay and, in spite of high variability among stations, a general declining trend was observed as our sampling progressed westward and toward the end of the fall season (Figure 3a). Among all the variables measured, including bacterial numbers, the variations in kDMSPd were positively correlated only with Chl a concentrations, which also tended to decrease westward and as the season progressed (Table 1). The bacterial DMS yield varied between 4% and 15% with a mean value of 7.5% (Figure 3b). Variations of the bacterial DMS yield were positively correlated with Chl a and DMSPp concentrations (Table 1), suggesting a more efficient bacterial conversion of DMSP into DMS at stations rich in phytoplankton biomass and DMSP producers. Stations with higher DMS concentrations were also characterized by higher bacterial DMS yields (Table 1).
 The microbial DMSPd uptake rates spanned two orders of magnitude across the stations, ranging from 0.17 to 5.84 nmol L−1 d−1 with a mean value of 1.83 nmol L−1 d−1 (Figure 3c). Bacterial DMS production rates ranged from 0.01 to 0.51 nmol L−1 d−1 with a mean value of 0.14 nmol L−1 d−1 (Figure 3d). Excluding the expected autocorrelation with DMSPd, rates of the microbial DMSPd consumption rates showed no significant relationship with the other variables measured (Table 1). The bacterial DMS production rates were weakly correlated with DMS concentrations (Table 1). None of these two rates were correlated with incident light and the MLD (data not shown).
4.1. Late Fall Distribution of Eukaryotic Cells and DMSP
 As expected for this study region in fall, phytoplankton biomass was low, generally below 0.5 μg Chl a L−1, throughout most of the stations, with a single peak of 2.5 μg Chl a L−1 measured at the most easterly station in northern Baffin Bay. Except for that peak in Chl a, which coincided with high concentrations of diatoms, the eukaryote communities at other stations were numerically dominated by flagellates, dinoflagellates, and unidentified cells. Chl a concentrations in the Beaufort Sea and the Amundsen Gulf, our most intensively sampled regions, were comparable to the ones previously reported in the same regions at this time of the year [Forest, 2008; Tremblay et al., 2009; Brugel et al., 2009].
 The distribution of DMSPp closely followed the variations in Chl a in the region under investigation. Usually these two variables are poorly correlated in the field. The relationship probably resulted from the relatively homogeneous composition of the phytoplankton community. The mean DMSPp:Chl a ratio was 39±16 nmol μg−1 reflecting the dominance of DMSP-rich flagellates. At station 108, a low DMSPp:Chl a ratio of 13 nmol μg−1 reflected a diatom-based community. The DMSPp concentrations measured in this study were comparable to the range of values measured in subsurface waters by Matrai et al.  in the High Arctic leads in summer (4 to 25 nmol L−1) and by Matrai and Vernet  in the Barents Sea during the spring bloom (6 to 27 nmol L−1). Our maximum DMSPp value of 39 nmol L−1 measured in the North Water was however much higher than the maximum concentrations measured by Bouillon et al.  in the same region in spring and early summer (9.5 nmol L−1). It should be noted that Bouillon et al.  collected DMSPp by filtering large volumes of water which could have resulted in a loss of DMSPp into the dissolved phase [Kiene and Slezak, 2006]. In spite of the different methods used, the comparison between the observations of Bouillon et al.  and our data suggest that the spring and early summer diatom-based communities in the northern part of Baffin Bay are succeeded by flagellate-based communities richer in DMSP in late summer and fall.
 Dissolved DMSP concentrations are notoriously difficult to quantify, mostly due to the propensity of microalgae to rapidly release DMSP when manipulated or stressed [Kiene and Slezak, 2006]. DMSPd concentrations measured from the SVDF technique tend to be an order of magnitude lower than those produced by previous methods involving large volumes. We have thus restricted the comparison of our DMSPd results to the few studies using this technique. Generally below 2 nmol L−1, some of our DMSPd values were extremely low, but most values compared well with concentrations reported from DMSPp-rich waters of the coastal Gulf of Mexico (0.7 to 2.4 nmol DMSPd L−1 and 30 to 120 nmol DMSPp L−1 [Kiene and Slezak, 2006]) and of the Northeast Pacific (1.5 to 3.6 nmol DMSPd L−1 and 19 to 160 nmol DMSPp L−1 [Royer et al., 2010]). Polar waters from both the Arctic and Antarctic also had low DMSPd (<2 nmol L−1) even when DMSPp ranged up to 50 nmol L−1 [Kiene et al., 2007; Damm et al., 2008]. Similar low levels of DMSPd over such a broad range of ecosystems and DMSPp richness have been previously attributed to the high lability of this compound and its rapid bacterial turnover [Kiene and Slezak, 2006; Royer et al., 2010].
 DMS concentrations remained below 1 nmol L−1 at all stations during this study. Previous studies conducted in fall reported mean DMS concentrations of 0.17 and 1.53 nmol L−1 in the Greenland Sea and Bering Sea, respectively [Leck and Persson, 1996; Bates et al., 1987]. Our DMS levels in northern Baffin Bay were higher than the maximum values measured by Bouillon et al.  in April 1998 (0.72 nmol L−1), but lower than those measured in May (6.7 nmol L−1) and June (4.6 nmol L−1). During our study, the standing stock of DMS relative to its precursor total DMSP (DMSPt = DMSPp+DMSPd) was very low (0.01 to 0.07; mean 0.03 mol:mol), suggesting a weak DMSPp to DMS conversion efficiency. For example, with the same quantification method, Kiene et al.  measured DMS/DMSPt molar ratios above 0.05 at most stations in the Southern Ocean in spring and summer, with peak values reaching 0.38 at the northern ice edge boundary and in the Ross Sea.
4.2. Microbial DMSPd Uptake
 The capacity of the microbial community to use DMSPd (kDMSPd) measured at all stations was at the lower end of the ranges measured in other environments. Our range of 0.01 to 0.14 h−1 was similar to the one reported by Lizotte  in the Northwest Atlantic (0.04 to 0.17 h−1), but lower than the rates measured by Vila-Costa et al.  in the Mediterranean Sea (0.03 to 0.26 h−1) and by Royer et al.  in the Northeast Pacific (0.09 to 0.92 h−1). The similarity between the kDMSPd rates measured during this study at temperatures between −1.8 to 0.1°C and those reported by Lizotte  in the Northwest Atlantic at much higher temperatures (2 to 26°C) excludes temperature as a main controlling factor. Accordingly, none of these studies reported a significant relationship between kDMSPd and water temperature. The relationship found between kDMSPd and Chl a (r = 0.67) in this study suggests that the algal biomass was directly (via algal uptake) or indirectly (via bacterial uptake) limiting DMSPd uptake. A similar relationship between kDMSPd and Chl a was not reported in the three studies cited above. The dependency of kDMSPd on the algal biomass during this study may thus be specific to environments where bacterial production is strongly substrate limited.
 The DMSPd microbial uptake rates were relatively low during this study, with values ranging from 0.1 to 5.7 nmol L−1 d−1. For comparison, microbial DMSPd uptake rates varied from 0.1 to 12.5 nmol L−1 d−1 in the New Zealand sector of the Southern Ocean and in the Ross Sea [Kiene et al., 2007] and from 4 to 58 nmol L−1 d−1 in the Northeast Subarctic Pacific [Royer et al., 2010]. Since DMSPd concentrations were comparable in all three systems, the lower DMSPd microbial uptake rates measured during our study mostly reflect the measured lower kDMSPd. Again, this result suggests that the low algal biomass and probably low bacterial productivity were responsible for the low DMSPd bacterial uptake in this part of the Canadian High Arctic in late fall. Assuming local steady state conditions on the time scale of a day, the partial microbial turnover time for the DMSPd pool (DMSPd concentration/microbial DMSPd uptake rate) ranged from 0.3 to 3.4 days, with a mean of 0.96 day. For comparison, Royer et al.  reported much shorter DMSPd turnover times (0.05 to 0.5 day, mean 0.26 day) in the Northeast Pacific. The microbial turnover rate of the DMSPd pool was thus slow across the investigation area, even in the North Water region where algal biomass was maximum.
4.3. Bacterial DMS Production
 DMS production from DMSPd consumption was a minor bacterial metabolic pathway during our study. Less than 10% of the DMSPd taken up by bacteria was converted into DMS across the Canadian High Arctic. Our mean value of 7.5% compares well with the values of 8% reported by Kiene and Linn  for oligotrophic oceanic waters and of 6% to 14% reported by Lizotte  for the Northwest Atlantic during different seasons. The strong positive relationship found between the variations in DMS yields and Chl a (r = 0.85) suggests that bacteria were producing more DMS per mole of DMSP assimilated as the algal biomass increased. The mechanism underlying this relationship still needs to be investigated.
 Bacterial DMS production varied by one order of magnitude over the study region. Given that the bacterial DMS yields varied only by a factor of 3 as compared to the factor of ∼100 for the microbial DMSPd consumption rates, the large variations in bacterial DMS production were driven mainly by changes in microbial DMSP uptake. Based on the same assumption as for the microbial turnover time for DMSPd, the bacterial DMS turnover time (DMS concentration/bacterial DMS production) varied between 1 and 19 days, with a mean of 6 days. Considering that DMS sinks (bacterial consumption, photo-oxidation, and ventilation) generally operate on a time scale of hours in the ocean upper mixed layer, these long turnover times suggest that bacteria alone could not sustain the DMS pool at most stations. Other potential DMS sources include direct algal release by exudation or release stimulated by zooplankton grazing. The strong correlations found during this study between DMS concentrations, a volatile compound with a generally short residence time in the water column, and DMSPp and Chl a suggest that the net DMS production was directly related to the biomass of phytoplankton and to the size of the DMSPp reservoir.
 A reduction in the summer ice cover in the Arctic as a consequence of global warming could lead to greater phytoplankton abundance due to the alleviation of light and nutrient limitation following the retreat of ice cover and potential deeper wind mixing, respectively [Carmack and Chapman, 2003; Arrigo et al., 2008]. Model simulations also suggest that the reduction of the seasonal ice cover will lead to higher pelagic DMSP and DMS concentrations on a seasonal basis [Gabric et al., 2005; Vallina et al., 2007]. During this study, we had the opportunity to investigate DMS biological production during late fall, under exceptional ice-free conditions which could become more common in the near future. Our results show that DMS production could persist at low rates in late fall under ice-free conditions, but the magnitude of this production appears to be limited by the low algal and bacterial production prevailing at that time. On a seasonal basis, the extension of the ice-free period in fall should nevertheless result in an increase in DMS emission to the atmosphere although late fall emissions are not likely to influence cloud albedo.
 We thank chief scientist Jean-Éric Tremblay, the Canadian Coast Guard officers, and the crew of the CCGS Amundsen for assistance during the cruise. We sincerely thank Véronique Lago for operating the rosette, Geneviève Tremblay for field assistance and cell identification and enumeration, and Yvonnick Le Clainche and Jean-Éric Tremblay for providing comments on the manuscript. This project was supported by grants from the Canadian International Polar Year Federal program office (Arctic SOLAS) to M. Levasseur; the Natural Sciences and Engineering Research Council (NSERC) of Canada to M. Levasseur, M. Gosselin, and M. Poulin; and the Network of Centers of Excellence of Canada ArcticNet to M. Gosselin, M. Poulin, and Y. Gratton. Partial operating funds for the CCGS Amundsen were provided by the International Joint Ventures Fund of the Canada Foundation for Innovation and the Fonds québécois de la recherche sur la nature et les technologies (FQRNT). Myriam Luce received graduate scholarships from NSERC, FQRNT, and the Biology Department of Laval University; financial support for fieldwork from the Northern Scientific Training Program of Indian and Northern Affairs Canada; and a stipend from Québec-Océan. This is a contribution to the research programs of Arctic SOLAS, ArcticNet, and Québec-Océan.