The distribution of methylated sulfur compounds, DMS and DMSP, in Canadian Subarctic and Arctic marine waters during summer, 2015

We present seawater concentrations of dimethylsulfide (DMS), and dimethylsulfoniopropionate (DMSP) measured across a transect from the Labrador Sea to the Canadian Arctic Archipelago, during summer 2015. Using an automated ship-board gas chromatography system, and a membrane-inlet mass spectrometer, we measured a range of DMS (~1 nM to 18nM) and DMSP concentrations (~1 nM to 150 nM) that was consistent with previous observations in the Arctic Ocean. The highest DMS and DMSP concentrations occurred in a localized region of Baffin Bay, where surface waters were characterized by high chlorophyll a (chl a) fluorescence, indicative of elevated phytoplankton biomass. Across the full sampling transect, there were only weak relationships between DMS/P, chl a fluorescence and other measured variables, including positive relationships between DMSP:chl a ratios and several taxonomic marker pigments, and elevated DMS/P concentrations in partially ice-covered areas. Our high spatial resolution measurements allowed us to examine DMS variability over small scales (<1 km), and document strong DMS concentration gradients across surface hydrographic frontal features. The new observations presented in this study constitute a significant contribution to the existing Arctic DMS/P dataset, and provide a baseline for future measurements in the region.


Introduction
The trace gas dimethylsulfide (DMS), a degradation product of the algal metabolite dimethylsulfoniopropionate (DMSP), is the largest natural source of sulfur to the atmosphere, accounting for over 90% of global biogenic sulfur emissions (Simó, regional climatic patterns (Levasseur, 2013). Indeed, modeling work has suggested that cooling associated with increased DMS production and emissions in less ice-covered polar regions may help offset warming associated with loss of sea-ice albedo (Gabric et al., 2004;Cameron-Smith et al., 2011) The important climatic and biological roles of reduced sulfur compounds, combined with altered marine conditions under a warming environment, provide the motivation for a deeper understanding of the distribution and cycling of DMS and related compounds in Arctic waters.
In this article, we present a new data set of DMS and DMSP concentrations in Arctic and Subarctic waters adjacent to the Canadian continental shelf. We used a number of recent and emerging methodological approaches to measure these compounds in a continuous ship-board fashion. In particular, we used membrane inlet mass spectrometry (MIMS) to measure DMS with extremely high spatial resolution (i.e. subkm scale), and the recently developed organic sulfur sequential chemical analysis robot (OSSCAR), for automated analysis of DMS and DMSP. Our goal was to utilize the sampling capacities of the MIMS and OSSCAR systems to make simultaneous measurements of DMS/P in Subarctic Atlantic and Arctic waters, in order to expand the spatial coverage of the existing DMS/P dataset, and identify environmental conditions leading to spatial variability in the concentrations of these compounds. sampled along a ~ 10,000 km transect from Quebec City, Quebec, to Kugluktuk, Nunavut. Data collection commenced off the coast of Newfoundland, and included waters of the Labrador Sea, Baffin Bay and the Canadian Arctic Archipelago (Figure 1).
The cruise transect covered two main distinct geographic domains -the Baffin Bay/Labrador Sea region, and the Canadian Arctic Archipelago (CAA). The majority of the surface water in the CAA is from Pacific-sourced water masses, as a shallow sill near Resolute restricts the westward flow of Atlantic-sourced water (Michel et al., 2006). Flow paths through the CAA are complex. The region is characterized by a network of shallow, narrow straits that are subject to significant regional variability in local mixing and tidal processes, and strongly influenced by riverine input, which drives stratification (Carmack et al, 2011). In contrast, both Atlantic-and Pacific-sourced waters mix in the Baffin Bay and Labrador Sea regions, and this confluence drives a strong thermohaline front, leading to lower stratification than in the CAA (Carmack et al, 2011).

OSSCAR
The OSSCAR instrument consists of an automated liquid handling / wet chemistry module that is interfaced to a custom-built purge-and-trap gas chromatograph (GC) equipped with a pulsed flame photometric detector (PFPD) for sulfur analysis. custom LabVIEW program is used to automate all aspects of the sample handling and data acquisition. During analysis, unfiltered seawater (3 -5 ml) from an underway supply (nominal sampling depth ~ 5 m) is drawn via automated syringe pump into a sparging chamber. DMS is then stripped out of solution (4 minutes of 50 ml min -1 N 2 flow) onto a 1/8" stainless steel trap packed with carbopack at room temperature. Rapid electrical heating of the trap (to ~260°C), causes DMS desorption onto a capillary column (Restek SS MXT, 15m, 80 °C, 2 ml min -1 N 2 flow) prior to detection by the PFPD (OI Analytical, Model 5380). Light emitted during combustion in the PFPD is converted to a voltage and recorded by a custom built Labview data acquisition interface. Following the completion of DMS analysis, 5 N sodium hydroxide is added to the sparging chamber for 14 minutes to the time-points of sample measurements, and compared to the known standard concentration to provide a drift correction factor for every seawater data point. Six-point calibration curves were performed every two days, using DMS standards (ranging from 0 to 18nM), produced from automated dilutions of a primary DMS stock and Milli-Q water (see Asher et al., 2015). The limit of detection of the system was calculated from the calibration curve using the formula C LOD = 3s y/x ÷ b, where C LOD is the concentration limit of detection, s y/x is the standard error of the regression , and b is the slope of the regression line. With this approach, we derived a mean limit of detection of 1.4 nM. The mean linear calibration curve R 2 value, taken over all calibration curves, was 0.9887.
The OSSCAR system is designed to automate the collection of seawater for sequential analysis of DMS, DMSO, and DMSP in a single sample. During our cruise, however, we experienced problems with the DMSO reductase enzyme used to convert DMSO to DMS for analysis, and we therefore configured the instrument to run only DMS and DMSP at sea, with one cycle requiring roughly 30 minutes. We experienced general technical difficulties with the instrument during the early phases of the cruise, and no OSSCAR data are thus available for the first half of the transect.
with the membrane, seawater was passed through a 20 foot coil of stainless steel tubing immersed in water bath held at 4 °C (Tortell et al. 2011). The system pressure (as measured by the Penning Gauge) remained stable during operation (~1.3 -1.5 x 10 -6 Torr). The DMS signal was calibrated using liquid standards that were produced by equilibrating 0.2 µm filtered seawater with a constant supply of DMS (m/z 62) from a calibrated permeation device (VICI Metronics). The primary effluent from the permeation tube (held at 30 ± 0.1 °C in a custom-built oven) was split among several capillary outflows and mixed into a N 2 stream controlled at 50 ml min -1 using a pressure regulator and fixed length / diameter tubing. This system enabled us to achieve a range of DMS / N 2 mixing ratios that were bubbled into standard bottles held in an incubator tank supplied with continuously flowing seawater. Concentrations of DMS in the standard bottles were cross-validated by measuring discrete samples using the OSSCAR system. The limit of reliable detection of the MIMS is ~ 2nM (Tortell 2005).

Post-processing of DMS data
Raw data outputs (voltages) for both OSSCAR and MIMS measurements were processed into final concentrations using MATLAB scripts. For OSSCAR data, raw voltages were captured with a sampling frequency of 5 Hz. Sulfur peaks eluting off the GC column were integrated using a custom MATLAB script, with correction for baseline signal intensities. DMS concentrations were derived from peak areas using the calibration curves as described above. Shipboard salinity, temperature, wind speed, and chlorophyll a (chl a) fluorescence measurements were collected using several underway instruments. We used a Seabird Electronics thermosalinograph (SBE 45) for continuous surface temperature and salinity measurements, and a Wetlabs Fluorometer (WetStar) to measure chl a fluorescence, as a proxy for phytoplankton biomass. We note that the chl a fluorescence data are subject to significant diel cycles associated with light-dependent fluorescence quenching. All sensors were calibrated prior to and following the summer expedition.

Ancillary seawater data
Conductivity Temperature Depth (CTD) profiles were used to measure vertical profiles of salinity and potential temperature at 17 stations, from which we computed density using the Seawater Toolbox in MATLAB. The mixed layer depth (MLD) was defined as the depth where density exceeded surface values by 0.125 kg m -3 . Sea ice concentrations were obtained from the AMSR-E satellite product (Cavelieri et al. 2006) with a spatial resolution of 12.5 km. The percent ice cover along the cruise track was derived from a two dimensional interpolation of the ship's position in time and space against the daily sea ice data.
All correlation analyses (Pearson's r) were computed in MATLAB, using the corrcoef function. Sample sizes were as follows: 33,250 data points in the MIMS DMS dataset, 344 in the OSSCAR DMS dataset, and 318 in the OSSCAR DMSP dataset.

Phytoplankton biomass and taxonomic composition
In addition to underway data, samples for the quantification of photosynthetic and accessory pigments (Table 1) were collected at a number of discrete oceanographic stations (see Table 2). For each station, duplicate samples (250-500 mL) for chl a analysis were filtered onto pre-combusted 25 mm glass fiber filters (Whatman GF/F) using low vacuum pressure (<100 mm Hg). Filters were stored at -20 ºC and chl a was determined within a few days of sample collection using fluorimetric analysis following the method of Welschmeyer (Welschmeyer 1994). Duplicate 1-2 L samples were filtered onto precombusted 25 mm GF/F for pigment analysis by reverse-phase High-Performance Liquid Chromatography (HPLC). Filters were dried with absorbent paper, flash frozen in liquid nitrogen and stored at -80 ºC until analysis following the method of Pinckney et al (1994). We used several diagnostic pigments as markers for individual phytoplankton groups, as described by Coupel et al (2015) (see Table 1). Following HPLC pigment processing, data were interpreted with the chemotaxonomy program CHEMTAX V1.95, using the pigment ratio matrix described by Taylor et al (2013).

DMS Sea-Air Flux
We derived sea-air fluxes of DMS from MIMS measurements of DMS concentrations, as these data had higher resolution and spatial coverage than OSSCAR observations. We computed sea-air flux as: where DMS sw is the concentration of DMS in the surface ocean (surface atmospheric DMS is assumed to be zero) and k DMS is the gas transfer velocity derived from the equations of Nightingale et al. (2000), normalized to the temperature and salinitydependent DMS Schmidt number of Saltzman et al. (1993). The term A represents the note that this scaling does not capture all processes present in sea-ice dominated regimes, such as turbulence generated by sea ice melt.) Sea surface salinity and temperature measurements described in section 2.5 were used in the calculations. Wind speed data were obtained from the ship's anemometer (AAVOS data, Environment Canada), corrected to a height of 10 m above the sea surface.

Phytoplankton biomass and taxonomic distributions
Using measurements of accessory photosynthetic pigments, we examined spatial patterns in the taxonomic composition of phytoplankton assemblages (see Table 1  description of HPLC marker pigments and their associated phytoplankton taxa).The distribution of pigments across our sampling stations is presented in Table 2, along with measurements of mixed layer depth and ice cover, while CHEMTAX-derived assemblage estimates are shown in Table 3. In order to remove large potential differences in total phytoplankton biomass, we normalized pigment concentrations to total chl a concentrations measured using HPLC (see Methods, section 2.5).
CHEMTAX pigment analysis shows that all stations in the study area were diatomdominated, although haptophyte, dinoflagellate, and prasinophyte markers were detected in varying quantities at all stations (see Table 3). Total HPLC-measured chl a was relatively low throughout the study area, ranging from 0.11 to 0.56 μg L -1 .

Observed DMS/P concentration ranges
The DMS data shown in Figure 1 are derived from MIMS measurements, since these have wider geographic coverage and greater spatial resolution than OSSCAR data.  early August measurements (~ km 7000 cruise track, Figure 3a), when OSSCAR DMS data were consistently higher than MIMS data. Notwithstanding this offset (for which potential reasons are addressed in the discussion), the coherent spatial patterns in data derived from these independent methods is encouraging, particularly given the rather low precision of our current OSSCAR system.
The spatial distribution of DMSP concentrations (measured with OSSCAR) along the cruise track is also shown in Figure 3. Concentrations ranged from <1 nM to 160 nM, and averaged 30 ± 29 nM. DMSP:chl a ratios measured from HPLC chl a data ranged from 52.31 nmol μg^(-1) to 181.4nmol μg^(-1). Examination of the data in Figure 3 reveals that high DMS concentrations were sometimes, but not always, accompanied by high DMSP concentrations. For example, a sharp increase in measured DMSP concentrations (around 7000-7400 km) on the cruise track was accompanied by a sharp increase in DMS measured by both instruments, while low-DMS waters observed around km 9400 along the transect also showed very little DMSP. Over the portion of the transect where measurements of both DMS and DMSP were available, the OSSCARmeasured concentrations of these compounds exhibited a statistically significant positive correlation (r = 0.52, p< 0.001). There were, however, a number of regions where increased DMS concentrations were not accompanied by increases in DMSP (e.g. ~ km 10,000). μmol S m -2 day -1 , with peak sea-air flux calculated around km 5500 on the cruise track.

Sea-Air Flux
Sea-air flux is highly dependent on wind speed and sea ice cover, with the result that even high concentrations of seawater DMS yielded low sea-air flux when low wind and/or high sea ice was present (e.g. km 2100, 7200, 8300). Conversely, very high sea-air fluxes were observed when moderately high DMS concentrations coincided with high wind speeds and ice-free waters (e.g. km 5400).

Comparison of gradients in DMS data with hydrographic features
The

Correlation with ancillary oceanographic variables
We computed Pearson correlation coefficients of DMS and DMSP with underway measurements of salinity, sea surface temperature, chl a fluorescence, and sea ice cover.
We also examined the potential relationships between DMS concentrations and MIMS- derived pCO 2, and O 2 /Ar (Tortell et al., in preparation). The results can be seen in Table   4. Only correlations significant at the 0.05 level are included. Only weak correlations are seen between MIMS-measured DMS data and ancillary variables, and OSSCAR DMS data did not exhibit any significant correlations with any ancillary variables, including Arctic, situating our results in the context of the changing hydrography and phytoplankton ecology of the Arctic Ocean.

Comparability of MIMS and OSSCAR measurements
The OSSCAR and MIMS instruments have previously shown good agreement in measured DMS concentrations in the Subarctic Pacific Ocean (Asher et al. 2015).
Similarly, we observed relatively good coherence between the two methods ( Figure 3) over much of our cruise track. The largest exception to this occurred around km 7000, when DMS measurements measured by OSSCAR were significantly higher than those measured by MIMS. This region was characterized by very high DMSP measurements (often one order of magnitude higher than the DMS measurements). If small amounts of DMS remained in the OSSCAR system after DMSP analysis, sample carry-over could contribute to higher measured concentrations in the subsequent DMS analysis. In order to minimize this potential artifact, the system was thoroughly rinsed with MilliQ water after every run. The effectiveness of this rinse was tested by subsequently purging DMSP standards without NaOH, and no carryover was observed. It is possible, however, that this approach was not entirely efficient. Another potential cause of the higher OSSCAR DMS measurements may be due to cell breakage during the sparging process in OSSCAR. In this scenario, there is the potential for release of intracellular DMSP and DMSP lyase into solution, which would lead to artificially high measured DMS concentrations. It is not possible for us to quantify the magnitude of such a potential artefact, but we note that its magnitude would likely depend on the taxonomic One challenge going forward is to increase the reproducibility and sensitivity of OSSCAR measurements, and this is an area of active work in our group. The version of our system used in 2015 had a detection limit of roughly 1.4 nM, and was thus far less sensitive than many conventional GC methods, which can achieve sub-nM detection limits. Our detection limit was of only minor consequence for DMSP measurements, given that 72% of measured DMSP concentrations were higher than 10 nM, and less than 3% fell below 1.4 nM. The relatively low sensitivity was somewhat more problematic for DMS, with approximately 22% of our OSSCAR-measured DMS values below 1.4 nM. Nonetheless, as discussed below, we believe that the OSSCAR data, in combination with our MIMS data, provide useful information on the spatial distribution of both DMSP and DMS in Arctic waters. Together, the available data (Table 5 and our measurements) are consistent with a seasonal cycle in Arctic and subarctic reduced sulfur distributions. Early season diatomdominated blooms exhibit high biomass and primary productivity but low DMS/P accumulation, while mid-summer phytoplankton assemblages dominated by haptophytes and dinoflagellates display lower phytoplankton biomass but higher reduced sulfur accumulation. This pattern is similar to the summertime 'DMS paradox' reported in a number of temperate and sub-tropical waters (Simo and Pedrós-Alió, 1999). In the fall, both Arctic primary productivity and DMS/P production decrease with the onset of lower temperatures and increased ice cover. Our data are consistent with this general scenario, representing a mixed-species assemblage with moderate biomass and DMS/P accumulation. phenomenon. For example, the frontal mixing of distinct water masses, driven by currents, wind, or melting ice, may introduce nutrients into a low-nutrient water column, stimulating localized primary productivity (Tremblay et al. (2011) and potentially increasing DMS/P production. Note that this localized increase in productivity and potential DMS/P production would operate independently of the overall seasonal progression towards increased DMS/P production during the latter summer growth season. Mixing of water masses may also potentially expose water column phytoplankton to light shock or osmotic stress by mixing them upwards in the water column or introducing an abrupt salinity gradient. Both of these factors could contribute to elevated DMSP production, given its hypothesized role as an intracellular osmolyte and antioxidant (Stefels et al., 2007). Although our data do not allow mechanistic interpretation for the underlying causes of DMS variability in surface waters, the high resolution afforded by MIMS measurements enables real-time observations of DMS gradients, which may be useful in the design of future process studies examining the driving forces for elevated DMS accumulation.

Influence of phytoplankton assemblage composition and mixed layer depth
Previous work has addressed the role of phytoplankton taxonomic composition and irradiance levels (Stefels et al, 2007) in driving the cycling of DMS/P in marine waters. Here we discuss the potential influence of these factors across our survey region.
The majority of the sampled stations were characterized by very shallow mixed layer depths (MLD; Table 2)  irradiance and surface DMS concentrations. In our dataset, however, there was no overall correlation between MLD and DMSP : Chl ratios. We did, however, observe elevated DMSP concentrations at two stations (BB3 and CAA6) with shallow MLDs.
The elevated DMSP : chl a ratios measured in our study may also reflect the presence of high-DMSP producing taxa, a phenomenon previously reported by other groups (Matrai et al., 1997;Gali et al., 2010;Lizotte et al., 2012). When comparing our DMSP: chl a ratios to other measurements, it is important to note that we measured DMSP t , while many other groups present DMSP p , without taking into account the dissolved fraction (DMSP d ). As the dissolved DMSP pool typically makes up a small (though highly variable) portion of the total water column DMSP pool, the use of DMSPt does not likely have a large effect on derived DMSP:chl a ratios (Kiene et al., 2000;2006). Despite the potential caveats raised above, the DMSP t :chl a ratios we measured conclusions on the role of taxonomy in controlling DMSP:Chla, as we were unable to detect any significant correlations between DMSP:chl a and HPLC pigment markers for different phytoplankton groups.
To conclude, our observations do not permit us to establish a firm link between MLD, phytoplankton taxonomy and DMS/P concentrations. Other factors, including bacterial activity and zooplankton grazing are potential contributing factors, but we lack the data needed to examine the importance of these processes.

4.5The interaction of DMS/P and sea ice
The presence of sea ice exerts a strong control on polar phytoplankton by controlling irradiance levels in the water column (Levasseur 2013), and influencing vertical mixing, stratification and nutrient accumulation. It is thus expected that the presence of sea-ice may affect DMS/P cycling. In a 2010 study, Gali et al (2010) found that Arctic sea ice melt drove stratification of nutrient rich surface water, triggering a sharp increase in primary productivity, with associated elevated DMS and DMSP levels.
These authors also showed that experimental exposure of phytoplankton to high light conditions (mimicking those that would follow the breakup of sea ice) led to near-total release of intracellular DMSP, providing one possible explanation for elevated DMSP levels in the water column. A number of studies also show that the ice, itself, can be a potentially significant reservoir of reduced sulfur, associated with bottom ice-algae (Levasseur et al (1994).
The weak negative correlation we observed between sea ice cover and DMS/P concentration is consistent with the idea that sea ice cover limits insolation, thereby reducing primary productivity and DMS/P production. In general, the drivers of DMSP and DMS production differ -though DMSP production has been shown to be directly influenced by sea ice melt in under-ice blooms (Galindo et al., 2014), the production of DMS from DMSP is largely dependent on the metabolism of in situ bacterial assemblages (Evans et al, 2007), and may therefore be uncoupled from the influence of ice on phytoplankton activity. It is interesting to note, however, that several sharp increases in DMS (observed with MIMS) occurred simultaneously with the occurrence of small amounts of sea ice (<20% total cover) ( Figure 2, kms 3400 and 7200 on the cruise track).
Limited station data also indicate high DMSP:chl a ratios in areas with a comparatively high sea ice cover, at stations BB3 and CAA6 (Table 2). At the time of our sampling, both of these stations were characterized by very low phytoplankton biomass (0.11 μg L -1 and 0.20 μg L -1 chl a, respectively) and had particularly high DMSP: chl a ratios (129 nmol μg -1 and 182 nmol μg -1 , respectively). This suggests a potential role for ice-edge effects, either through the melt-induced stimulation of reduced sulfur production in DMSP rich phytoplankton taxa, or through the release of ice-associated DMSP into the water column. Figures 2d and 2e show decreased salinity in partially ice-covered areas (e.g. around kms 4400, 7300, and 9200), suggesting some melt-water stratification effects. Previous groups have also reported elevated DMS and DMSP concentrations in partially ice-covered water and ice-edge regions in the Arctic Ocean ( Matrai and Vernet (1997), Gali et al. (2010) and Leck and Persson (1997) ).

Conclusion
We present a high spatial resolution dataset of reduced sulfur measurements through the Canadian sector of the Arctic Ocean and Subarctic Atlantic. We demonstrate the utility of high-resolution DMS measurements for comparison with other oceanographic variables, and show the coherence of DMS gradients with fine-scale surface hydrographic structure, suggesting elevated DMS production in some oceanographic frontal zones. We also observed elevated DMS/P values in partially icecovered regions, suggesting that ice-edge effects may stimulate DMS/P production. Our data serve to significantly expand the existing spatial coverage of reduced sulfur measurements in the Arctic, providing a baseline for future studies in this rapidly changing marine environment. Future warming of surface waters and sea-ice melt could lead to increased concentrations and sea-air fluxes of DMS, though significantly more observations will be needed to substantiate this.   , 39, 1985-1992, doi:10.4319/lo.1994.39.8.1985, 1994.    Table 2).  Table 2). , and the PMEL dataset. (415 points).