Phytoplankton and dimethylsulfide dynamics at two contrasting Arctic ice edges

Abstract. Arctic sea ice is retreating, thinning and its rate of decline has steepened in the last decades. While phytoplankton blooms are known to seasonally propagate along the ice edge as it recedes from spring to summer, the substitution of thick multi-year ice (MYI) with thinner, ponded first-year ice (FYI) represents an unequal exchange when considering the roles sea ice plays in the ecology and climate of the Arctic. Consequences of this shifting sea ice on the phenology of phytoplankton and the associated cycling of the climate-relevant gas dimethylsulfide (DMS) and its precursor dimethylsulfoniopropionate (DMSP) remain ill constrained. In July–August 2014, two contrasting ice edges in the Canadian High Arctic were explored: a FYI-dominated ice edge in Barrow Strait and a MYI-dominated ice edge in Nares Strait. Our results reveal two distinct planktonic systems and associated DMS dynamics in connection to these diverging ice types. The surface waters exiting the ponded FYI in Barrow Strait were characterized by moderate chlorophyll a (Chl a, 


1 Introduction unicellular algae including osmoregulation, cryoprotection, scavenging of free radicals, and overflow of carbon and sulfur (Stefels et al. 2007). The production of DMSP by unicellular algae is highly species-specific with Bacillariophyceae and Dinophyceae/Prymnesiophyceae being lesser and greater producers, respectively (Keller et al. 75 1989). The DMSP-to-DMS conversion involves the entire microbial food web and part of the DMS is produced directly by phytoplankton while another part is produced indirectly via the release of DMSP in the aqueous environment and its subsequent degradation by bacterioplankton (Kiene et al. 2000;Simó 2001;Stefels et al. 2007).
The relative importance of these processes is unclear, however abiotic stressors involving sudden modifications in light intensity, salinity, and temperature may all contribute to the enhanced direct and indirect production of DMS by 80 plankton communities (Sunda et al. 2002;Toole and Siegel 2004).
In the Arctic, peaks in atmospheric methane sulfonic acid (MSA, a DMS proxy) have frequently been measured in spring and in mid-summer (Sharma et al. 2012). The spring peaks have been attributed to phytoplankton blooms at low latitudes while the mid-summer peaks have been related to more localized high latitude ice edge blooms (Sharma et al. 2012;Becagli et al. 2016;. This interpretation is consistent with the elevated DMS concentrations generally 85 measured at or close to ice edges in association with developing phytoplankton blooms in the North Atlantic and European sectors of the Arctic (Matrai and Vernet 1997;Galí and Simó 2010;Park et al. 2018). The high DMS concentrations measured at ice edges have been associated with a combination of factors including: 1) an increase in phytoplankton biomass and hence DMSP concentrations; 2) the selective growth of strong DMSP and DMS producers such as the prymnesiophyceae Phaeocystis; 3) a physiological stimulation of DMS production due to the increase in 90 irradiance; and 4) an increase in bacterial activity (Galí and Simó 2010). In the eastern Canadian High Arctic, only a fragmented picture of summer oceanic DMS distributions was available until recently and none of the snapshots captured the presumably most biologically productive time of July-August: average of 1.1 nmol DMS L-1 in the North Water and Nares Strait in June (Bouillon et al. 2002 October/November (Luce et al. 2011). In spite of the recurring mid-summer atmospheric MSA peak measured at Alert, evidence of high oceanic DMS concentrations associated with summer phytoplankton blooms remained scarce for this part of the Arctic until very recently (Mungall et al. 2016;Collins et al. 2017;Jarníková et al. 2018;Abbatt et al. 2019).
The rapid shifting of the Arctic icescape bears consequences for Arctic primary production and associated DMS 100 dynamics that are still poorly understood. While observations from the field are sparse (Wassmann et al. 2011) and challenging due the remoteness and harshness of the environment as well as the dynamic nature of ice and its margins (Sakshaug and Skjoldal 1989), it is critical that impacts of ongoing physical changes on the dynamics of bloomforming microorganisms and their production of DMS be better constrained. The main objective of this study was to assess and compare mid-summer (July-August) phytoplankton and DMS dynamics at two contrasting ice edges in 105 regions of the eastern Canadian Arctic: the Barrow Strait first-year ice (FYI) dominated ice edge and the Nares Strait multi-year ice (MYI) dominated ice edge. The opportunity was also taken to investigate the ice-free waters of Lancaster Sound and North Water (northern Baffin Bay) contiguous to the Barrow Strait and Nares Strait ice edge regions, respectively. Our results reveal two distinct planktonic systems and ensuing DMS dynamics related to the presence of dissimilar icescapes. The two Straits (Barrow and Nares) were characterized by distinct and well-defined ice edges at the time of sampling ( Fig. 2). In Barrow Strait, the ice edge was located at the western end of Lancaster Sound, perpendicular to the channel, 125 between Devon Island and Somerset Island (Fig. 2a). The ice pack was mostly composed of ca. 1 m thick FYI covered by melt ponds at approximately 40% of total surface (Fig. 3 picture of melt ponds). Soon after our arrival in the study area, a large lead developed south of Griffith Island (south of Cornwallis Island), pushing the detached part of the ice pack slightly eastward (Fig. 2). The BS transect was conducted along the ice edge in this lead. In Barrow Strait, the net surface circulation is predominantly eastward at 10-15 cm s-1 in mid-summer on the south shore with a mild 130 westward current of ca. 5 cm s-1 on the north shore (Lemon and Fissel 1982;Prinsenberg and Bennett 1987;Pettipas et al. 2008, Michel et al. 2015. This region stands as an important waterway for the transport of fresher Pacific waters, originally from the inflow through Bering Strait, towards the North Atlantic (Jones et al. 2003). The water sampled across this transect was thus mostly exiting the ice pack which extended several km westwards.
In July 2014, an ice arch formed in the Kennedy Channel of Nares Strait leaving Kane Basin, and the North Water 135 region to the south largely ice-free. The comparison of the position of the ice arch in July 2014 with a decade of remotely sensed data (1997)(1998)(1999)(2000)(2001)(2002)(2003)(2004)(2005)(2006)(2007), shows that it formed that year approximately 130 km north of a median historical position (near 79oN) in southern Kane Basin (Kwok et al. 2010), in line with recent trends (2006-2010) of more northern ice bridge formation in the area (Ryan and Münchow 2017). By the time of the sampling (3-6 August), it had retreated to the head of Kennedy Channel ( Fig. 2A), leaving a 350 km stretch of open water north of Smith Sound 140 (Burgers et al. 2017). As expected for this part of the Arctic Ocean, the ice pack north of the ice arch was composed of MYI (Fig. 2C). Presence of MYI (5+ years) north of Nares Strait, near Robeson Channel, was confirmed by the Ease-Grid Sea Ice Age, Version 3 data set , which compiles weekly estimates of sea ice age in the Arctic between 1978 and 2017. Data from 2014, week 31 (28 July-3 August) and week 32 (4-10 August) were consulted for the purpose of this study. Beyond the MYI and to the south, a band of thick (>1.2 m) FYI, without any melt ponds, was also present (Canadian Ice Service (CIS) analysis, Fig. 2B). Because Nares Strait represents a major outflow path for water exiting the Arctic Ocean (Jones et al. 2003;Münchow et al. 2007;McGeehan and Maslowski 2012), the water sampled along the NS transect was exiting the northern MYI edge as it flowed southbound towards Baffin Bay.

Physical, chemical and biological measurements
Water samples were collected at 5 to 9 depths from the surface down to a maximum of 100 m depth with 12-L Niskintype bottles mounted on a General Oceanics 24-bottle rosette. The rosette sampler was equipped with a Sea-Bird 911plus Conductivity Temperature Depth (CTD) probe and a sensor for the measurement of fluorescence (Seapoint).
Water for Chl a concentration analysis was collected in 1-L brown polyethylene bottles (Nalgene) and then passed onto a 25-mm filter (Whatman GF/F). Phytoplankton pigments on the filter were extracted in 90% acetone and stored at 4oC in the dark during a period of 18-24 hours. Fluorescence of extracted pigments was then measured using a Turner 165 Designs fluorometer 10-AU after the acidification method described by Parsons et al. (1984). Chl a concentrations were calculated from the equation published in Holm- Hansen and collaborators (1965).
Samples for phytoplankton taxonomy were collected at the surface and at the subsurface chlorophyll maximum (SCM) and preserved in an acidic Lugol's solution (final concentration of 0.4% v:v; Parsons et al. 1984). Identification and enumeration of cells > 2 µm were conducted with a Zeiss Axiovert 10 inverted microscope following the Utermöhl and 170 Lund method (Lund et al. 1958;Parsons et al. 1984). A minimum of 400 cells was enumerated to be statistically significant.
Samples of DMS were collected in 23-ml serum vials and allowed to gently overflow, avoiding any bubbling, before capping. Concentrations of DMS were determined onboard within 2 hours of collection using purging, cryotrapping, and sulfur-specific gas chromatography (GC, Varian 3800) as described by Lizotte et al. (2012) and further 175 modifications described here. Briefly, 15 to 20-ml subsamples of DMS were gently filtered through a GF/F syringe filter and immediately injected into a sparging vessel. The DMS was stripped from the liquid samples using a constant flow of Ultra High Purity (UHP) helium (He) prepared using a permeation tube (certified calibration by Kin-Tek Laboratories Inc.) maintained at 40oC and volatile DMS was trapped in a Teflon loop held in liquid N2. Gaseous samples were then analyzed using a Varian 3800 gas chromatograph (GC), equipped with a Pulsed Flame Photometric 180 Detector (PFPD) and a capillary column (DB-5ms, 60 m x 320 µm x 1µm). The samples were calibrated against microliter injections of DMS diluted with UHP He (certified calibration by Kin-Tek Laboratories Inc.) maintained at 40oC. Duplicate tubes for total DMSP (DMSPt) samples were filled with 3.5 mL of unfiltered water. For conservation purposes, 50 µL of 50% sulfuric acid (H2SO4) was added in each 3.5 mL liquid sample of DMSPt. All tubes were stored at 4oC in the dark until analysis in laboratory. DMSP concentrations were quantified over the course of two 185 periods using two analytical systems. A first series of DMSPt samples (stations 323, 322, 325, 301, 304, 305, 305A, 305B, 305C, 305D, and 305E) was analyzed in the laboratories of Laval University using a purge and trap system coupled to a Varian 3800 GC PFPD as described above. DMSPt samples were hydrolyzed with a 5N NaOH solution in order to convert DMSP into DMS which was purged from the samples via an Ultra High Purity (UHP) helium stream, cryo-trapped and analyzed via gas chromatography (Lizotte et al. 2012). For these DMSP samples, the GC 190 was calibrated with milliliter injections of a 100 nmol L-1 solution of hydrolyzed DMSP (Research Plus Inc.).
The analytical detection limit on the Varian GC system was 0.1 nmol L−1 for all sulfur compounds and the analytical precision (CV) for triplicate measurements of DMS and DMSP was better than 10%. After shortcomings with the aforementioned GC system, a second series of DMSPt samples (stations 300,324,346,115,111,108,105,101,KEN1,KEN3,KANE1,KANE3,314,312,310,335,210,204,200,and 120) were determined using an automated purge and Our analysis shows an average loss of 9% in the DMSPt samples between times of sampling and analysis.
MODIS images, as well as ice charts produced by CIS, were used to visually assess the presence of ice edges. CIS ice charts, based on Radarsat 2 and NOAA-18 images, show ice properties including stage of development, concentration 210 and form of the ice (Environment Canada 2005). Color schemes of the CIS ice chart were modified using Adobe Illustrator CS6. A FYI edge appears in Lancaster Sound as a curved line between Devon Island and Somerset Island on July 22 ( Fig. 2A). The presence of MYI appears at the northern extremity of Nares Strait, i.e., at the entrance of Robeson Channel between Ellesmere Island and Greenland, on August 1 (Fig. 2C). The MYI ice was contiguous to a band of thick (> 1.2 m) FYI descending into Nares Strait (Fig. 2B).

215
The surface mixed layer depth (Zm) was estimated as the depth at which the gradient in density (σt) between two successive depths was greater than 0.03 kg m-4 following the threshold gradient method of Thomson and Fine (2003) with adaptations from Tremblay et al. (2009). Oceanic vertical cross sections and contour plots were drawn using weighted averaging gridding and linear mapping using Ocean Data View 5.1.5 macx software (Schlitzer, 2018) and schematic models of FYI and MYI dynamics were constructed in Adobe Illustrator CS6. Statistical analysis was 220 conducted using SYSTAT 13.2 software, as well as JASP 0.9.2.0 computer software, an open-source project supported by the University of Amsterdam (JASP Team 2018). Variables were tested for normality using the Shapiro-Wilk test with a 0.05 significance level, and Spearman's rank correlations (rs) were used to assess the strength of association between variables. 225 3 Results

Overview of the sea surface physicochemical and biological characteristics
The main physical and chemical characteristics of the sea surface water at the sampling stations are presented in Table   1 (Table 2). At a broad scale, and considering only sea surface data from all regions under investigation in this study, Spearman's rank correlation tests (n = 33) reveal no significant relationships between DMS and abiotic or biotic variables presented in tables 2 and 3.
Beyond sea surface data, water column vertical profiles were also plotted as cross sections in order to identify key features associated with ice dynamics and bloom development in certain regions of the CAA and Baffin Bay.

245
Information is presented below and grouped as a function of targeted transects.

Barrow Strait (BS) transect
Variables measured across the BS transect are presented in Figure 4. Seawater temperatures ranged from -1.6 to -1.2oC, with the lowest values found at intermediate depths (ca. 40-60 m). Surface water temperatures were below -250 1.4°C at all stations. Salinity varied between 30.4 and 33.0 across the transect, with the lowest and highest surface values measured at the north and south extremities of the transect, respectively. Nitrate concentrations ranged from 0.6 to 11.0 µmol L-1, with lowest and highest values measured close to the surface and at depth, respectively. The nitracline was located at ca. 30 m. Close to the surface, nitrate concentrations were low at the south end of the transect (0.6 µmol L-1 at station 305B) and increased northward to reach 2.1 µmol L-1 at station 305E. Silicic acid concentrations 255 showed a similar pattern, with a positive south-north gradient ranging from 3.5 to 10.5 µmol L-1 in the upper 30 m and high values at depth (up to 29.2 µmol L-1). Chl a concentrations varied between 0.2 and 2.1 µg L-1 with highest values measured in the upper 30 m of the water column and toward the northern tip of the transect. Phytoplankton identification and enumeration were conducted at one station on the BS transect (stations 305E) and at two stations located in the vicinity under the ponded ice cover (see stations 304 and 305 in Fig. 1 and Table 2). The phytoplankton 260 assemblages at these three stations were similar, dominated by the pennate diatoms Fossula arctica and Pseudonitzschia spp. (delicatissima group), the two taxa being responsible for 29 to 71% of the total phytoplankton abundance (Table 2). Another abundant pennate species at these stations was Fragilariopsis oceanica.

270
Variables measured across the LS transect are presented in Figure 5. Surface temperatures were at least 3 times warmer than those measured across the BS transect, with values ranging between 3.0 and 4.1oC. Surface salinities varied between 30.7 and 32.4, with the highest values measured at stations 323 and 322 towards the north shore.
Concentrations of nitrate and silicic acid exhibited no particular cross-channel pattern in the surface mixed layer, with values below 0.5 and 2 µmol L-1 in the upper 20 m of the water column, respectively. Maximum Chl a concentrations 275 were in the same range as in the BS transect (between 1.5 and 2.5 µg L-1) but exhibited a different vertical distribution.
Across the BS transect, Chl a concentrations were generally highest in the surface mixed layer (SML) while they formed a SCM at ca. 30-40 m at the stations located across the LS transect suggesting a more advanced bloom stage in the LS area. The two transects also showed distinct phytoplankton assemblages (Table 2)

Nares Strait (NS) transect
Variables measured across the NS transect are presented in Figure 6. Sea surface temperatures started at ca. -1.3oC at 295 the ice edge and increased more or less regularly southward to reach 2oC at the last station (KANE5) of the transect.
In contrast, sea surface salinities were relatively constant at 30.5 along the transect. Nitrate and silicic acid concentrations in surface waters near the ice arch were ca. 1.5 µmol L-1 and 6 µmol L-1, respectively. In the upper 20 m of the water column, concentrations of nitrate and silicic acid decreased with distance from the ice arch as a first algal bloom developed (see below), reaching 0.4 µmol L-1 and 1.9 µmol L-1, respectively, at the southernmost station 300 (KANE5). The silicic acid drawdown along the transect was indicative of a strong diatom dominance (see Table 2).  Concentrations of DMSPt were highest in the first 20-30 m of the water column, ranging from 34 to 88 nmol L-1 at the near surface, with a distinct positive gradient from west to east. A subsurface peak was observed in the three most eastern stations (108, 111, and 115) with the highest concentrations of DMSPt (112 nmol L-1, station 115) measured at 335 12 m depth. DMS concentrations in the near surface waters were relatively high and stable at 4.1-5.3 nmol L-1 between stations 101 and 111 and reached 19.5 nmol L-1 at station 115, the highest value measured during this campaign.

Discussion
During the joint ArcticNet/NETCARE cruise, summertime DMS distributions were studied in two regions of the High 340 Canadian Arctic characterized by distinct ice edges: a first one featuring mainly ponded FYI, and a second one composed mainly of MYI. Both Barrow and Nares straits, as well as the contiguous regions of Lancaster Sound and North Water (northern Baffin Bay), embody significant oceanic gateways for Pacific-originating waters towards the North Atlantic (Jones et al. 2003). The results from the four transects conducted in these regions reveal distinctive features in DMS dynamics. The highlights of this study are discussed in the context of a predicted warmer Arctic, loss 345 of perennial sea ice, and increase in the prevalence of seasonal FYI (Nghiem et al. 2007;Kwok and Rothrock 2009; Overland and Wang 2013; AMAP 2017).

Broad regional sea surface distributions of DMS
Over the entire study area, the distribution of sea surface concentrations of DMS (Fig. 8) Table 3). These findings also bring further support to the hypothesis that local DMS sources explain the mid-summer peaks of atmospheric MSA, a DMS proxy, in the High Arctic (Sharma et al. 2012;Becagli et al. 2019). Not surprisingly, considering our limited sea surface dataset (n = 33) and the overall complexity of the DMS cycle, no significant relationships were found between broad regional blooms. The melt of sea ice and snow covers also influence surface water stratification and the ensuing shifts in salinity, temperature and solar radiation doses experienced by potential DMS-producing communities. The inherent 365 heterogeneity that characterizes spatial distributions of DMS in the Arctic as well as the presence of sea ice as a potentially critical driving force of these patterns warrants further investigations into underlying mechanisms.

The FYI edge in Barrow Strait and the adjacent Lancaster Sound
The seasonal sea ice zone (SIZ) in the Arctic is modulated by large interannual variability (Parkinson and Comiso 370 2013;Simmonds 2015;Comiso et al. 2017;Serreze and Meier 2019). Correspondingly, the position of the ice edge in the Barrow Strait/Lancaster Sound area during spring may vary yearly from the mouth of the sound on the east (80oW) to Lowther Island in Barrow Strait on the west (97oW) as revealed by the analysis of CIS ice charts by Peterson et al. (2008). On July 17, 2014, the ice edge was located approximately mid-way of this historical spatial range near the longitude of Prince Leopold Island (90oW, see Fig. 2A). Satellite imagery reveals that this distinct ice edge was already 375 present a month prior to the arrival of the icebreaker CCGS Amundsen in the area and that the eastern part of Lancaster Sound (east of 90oW) was already mostly ice-free by June 16, 2014 (data from CIS not shown). The ice cover in Barrow Strait, west of the ice edge, was composed mostly of FYI ca. 1 m thick covered with melt ponds at ca. 40% of its surface. On July 20, part of the ice diverged towards the east creating a small lead in the FYI near the northern tip of Somerset Island ( Fig. 2A). The opportunity was taken to sample the western border of the lead, very close to 380 the newly formed ice edge in order to capture the outflow of under-ice waters. The predominantly eastward transport of water in the southern portion of the Strait is estimated at 14 ± 4 cm s−1 annually and is strongest in late summer at 27 ± 8 cm s−1 (Hamilton et al. 2013), suggesting that the residence time of seawater in the lead was short lived.
Biogeochemical characteristics of the surface waters sampled on July 22-23 along the BS transect, particularly its southern area, thus likely reflect conditions prevailing in the ice-covered western portion of the Strait.

385
Vertical profiles from the BS transect (Fig. 4) in proximity to the newly formed ice edge indicate that an under-ice phytoplankton bloom had developed in the ice-covered Barrow Strait area and was captured during our sampling as it exited the ice. This under-ice bloom coincided with relatively low salinities (ca. 31.5) and temperatures (ca. -1.5oC) within the surface waters. These results suggest that the bloom was linked to the development of a fresher water lens below the ice, likely resulting from the melting of snow and ice covers. Events that were also likely associated with These results agree with studies emphasizing the importance of ice algal communities as a seeding source during spring over oceanic regions when algal abundance in the water column is low (e.g., Arctic 405 Ocean north of Svalbard by Kauko et al. (2018); Frobisher Bay in Davis Strait by Hsiao (1992)). The presence of species endemic to Arctic sea ice such as Nitzschia frigida, Fragilariopsis cylindrus and Fragilariopsis oceanica (Poulin et al. 2011) in the surface waters of the Barrow Strait region brings further support to the ice origin of this under-ice bloom.
The taxonomic composition of the drifting under-ice bloom at station 305E was also dominated by pennate diatoms, 410 but with lower total cell abundance (0.48  106 cells L-1 at 305E) as compared to the two other Barrow Strait stations (> 2.00  106 cells L-1 at 304 and 305, data not shown), as well as slightly different species. The phytoplankton assemblage at 305E was similar to the one previously described by Galindo et al. (2014) for the under-ice bloom developing at a shallow station (50 m) in Allen Bay in 2011, located ca. 15 km west of 305E. In both studies, the under-ice bloom was dominated by pennate diatoms, with Fossula arctica and Fragilariopsis oceanica contributing 415 8.2 % and 7.8 %, respectively, to the total protist abundance at station 305E.
In the ice-free area of Lancaster Sound, the lower Chl a (0.2 to 1.2 µg L-1) and nutrient concentrations measured in the 13-16 m depth SML as well as the presence of an SCM (Fig. 5)  observed by Galindo et al. (2014) in nearby Allen Bay. Despite seemingly varying vertical distribution patterns of in significant correlation (rs = 0.80, p < 0.001, n = 20) suggesting that the bulk of DMSPt was intimately linked to algal biomass. In contrast, across much of the LS transect, particularly towards its southern portion, concentrations of DMSPt were highest near the nitracline, deeper in the water column (peak of 96 nmol L-1 at 20 m, station 325). The role played by environmental drivers, such as nutrients, in the accumulation of DMSP-rich organisms at this depth 445 was substantiated by the significant correlation found between water column distributions of NO3-and DMSPt (rs = -0.59, p < 0.001, n = 36). However, contrary to patterns observed in the BS transect, concentrations of DMSPt bore no significantly association with in vivo fluorescence of chlorophyll in this part of the study area, suggesting that the bulk of algal biomass was not necessarily responsible for the variability in DMSPt concentrations in these waters characterized by mixed algal populations. The above results are not unexpected seeing as the nature of DMSP 450 synthesis itself is highly species-specific (Keller et al. 1989) and subject to physiological up-or down-regulation and excretion linked to environmental stressors (see review by Stefels et al. 2007). Assuming that almost all DMSPt was particulate (see Kiene and Slezak 2006) Notwithstanding the lower DMSPt:Chl a ratios in the BS transect, DMS levels were high in surface waters, ranging from 7.2 to 12 nmol L-1 and revealed two hot spots at either end of the sampled transect (Fig. 4). One in association with a peak in DMSPt (115 nmol L-1, 305E) and a second in conjunction with relatively low DMSPt (ca. 25 nmol L-1, 465 305B) and Chl a (0.83 µg L-1) at station 305B. Statistical analysis suggests that, in the waters exiting the FYI pack in Barrow Strait, variability in DMS concentrations was significantly associated with that of it's precursor DMSPt (rs = 0.76, p < 0.001, n = 20) but was most strongly associated with fluctuations in salinity. The highly significant negative correlation ( Fig. 9) found between DMS and salinity (rs = -0.91, p < 0.001, n = 20) in the upper ca. 80 m of the water column in this region suggests a strong physical control of DMS distributions associated with ice and snow melting 470 processes. The generally sunny forecast in the days prior to the sampling excludes heavy rain as a significant contributor to this signal. During the thawing season, the increase in ice permeability and basal melting may trigger important releases of DMS in the waters just below the ice cover (Trevena and Jones 2006;Kiene et al. 2007;Tison et al. 2010;Carnat et al. 2014). The formation of an upper fresher water "lens" associated with the FYI melt may also have led to a certain accumulation of DMS following its release from the sea ice. Furthermore, it cannot be totally excluded that the stratification of the upper water column ensuing from the melting ice could have entailed higher and longer exposures of phytoplankton communities to solar radiation with enhanced DMS production as a coping mechanism against light-induced stress via an antioxidant cascade (Sunda et al. 2002;Toole and Siegel 2004;Vallina and Simó 2007;Galí and Simó 2010). Indirectly, DMS production could also have been stimulated through the possible increased availability of dissolved DMSP (DMSPd) in the environment and its bacterially-mediated enzymatic 480 conversion into DMS (Kiene et al. 2000). Laboratory salinity downshock experiments with batch cultures of diatoms and dinoflagellates have shown an increase in the excretion of cellular DMSP (Van Bergeijk et al. 2003) and an increase in the production of DMS (Stefels et al. 1996;Niki et al. 2007). A DMSP-related osmo-acclimation response to shifts in salinity (Stefels 2000) could be particularly beneficial for algae developing in highly fluctuating environments, such as in the Arctic during the thaw season, a phenomenon which could ultimately strengthen DMS 485 production. The strength of the association between DMS and salinity in the waters however suggests that physical drivers exercised the greatest control over the distribution of DMS near the FYI ice edge.  Water column patterns of salinity along the NS transect were relatively uniform between stations with fresher waters 510 reaching deeper into the water column at the most northern stations (Fig. 6). This pattern is consistent with the presence of Pacific-originating waters of lower salinity and density that enter the central Arctic Basin through Bering Strait and that partly flow south through Nares Strait as a sub-surface current (Jones et al. 2003). It may also reflect the southbound flow through Nares Strait of first-year or multiyear ice floes (Münchow 2016), or icebergs originating from the glaciers of Greenland or Ellesmere Island (Burgers et al. 2017) that can partially melt in transit and thus freshen the ocean surface waters the impact of which lessens to the south as the ice melts away. Vertical patterns of temperature along the NS transect showed well mixed waters down to 58 m in the station nearest to the ice edge (KEN1) and a progressive warming of the upper layers of the water column with decreasing latitude (Fig. 6)

520
Reservoirs of nutrients throughout the water column at station KEN1, with 1.4 and 6 µmol L-1 of nitrate and silicic acid, respectively, were at the lower end of expected pre-bloom values for Pacific-derived water of the same salinity in the higher Arctic (Tremblay et al. 2002). As the sampling stations progressed to the south, a drawdown of both those nutrients, associated with the development of phytoplankton biomass, was evident at the surface of the vertical profiles (Fig. 6). At station KANE3, nutrients exhibited a swell-like pattern, associated with an increase of nutrients 525 throughout the water column. The presence of a sill (Bourke et al. 1989 In surface waters near the MYI edge, the phytoplankton community (dominated by unidentified flagellates and Prymnesiophyceae, KEN1, Table 2), showed a moderate abundance (1.3  106 cells L-1, data not shown), suggesting that the initiation of a phytoplankton bloom had not yet occurred in waters underneath the northern ice pack. The presence of sufficient amounts of nutrients in the surface waters near the ice edge points towards light availability as 540 the primary limiting factor for the proliferation of primary producers under the ice. In seasonally ice-covered seas, the growth of shade-adapted algal cells may begin once a critical incident irradiance threshold is reached at the ice-water interface (Horner and Schrader 1982;Gosselin et al. 1986). These results are in sharp contrast to the patterns observed in the waters exiting the ponded FYI in Barrow Strait where a bloom had already begun to develop underneath the ice. The drawdown of silicic acid in the following NS transect stations concurred with the development and dominance 545 of diatoms (see Table 2), notably centrics such as Chaetoceros spp. (5-20 µm) and Chaetoceros gelidus, an assemblage similar to those previously described in the LS transect (as well as in the NOW transect later discussed). Species of the genus Chaetoceros were thus widespread throughout the study area, as previously reported in the Canadian High Arctic (Booth et al. 2002;Ardyna et al. 2011;Poulin et al. 2011).
In proximity to the northern ice edge in Nares Strait (KEN1), concentrations of DMSPt and DMS were rather modest 550 throughout the water column (< 16 nmol L-1 and < 0.4 nmol L-1, respectively). These results reinforce the notion that autotrophic and heterotrophic processes associated with the production of DMSP and DMS in the waters under the thick non-ponded MYI may have only truly taken off upon reaching ice-free, light-sufficient conditions found farther south. This is again in cutting contrast with DMSP and DMS patterns observed at the Barrow Strait ponded ice edge.
Surface peaks of 27 nmol DMSP L-1 and 2.6 nmol DMS L-1 were measured in the following station (KEN3) adding 555 support to the requisite of suitable doses of solar radiation to ensure the development of microalgae in ice-covered waters of the Arctic (Horner and Schrader 1982;Gosselin et al. 1986) and the ensuing production of S compounds. In the three southernmost stations of the Nares Strait transect, a subsurface maximum of DMSPt was present at ca. 20 m depth with a high value of 59 nmol L-1 reached at KANE5 likely in association with an increase in autotrophic biomass fueled by nutrients near the sill, hitherto discussed. Maximal concentrations of DMS were, for the most part, confined 560 to the upper 20 m of the water column within or above the SCM, with a high value of 10 nmol L-1 reached at KANE5.
Along this transect, variations in the vertical distribution of DMS were significantly correlated with its precursor DMSPt, however the strongest association was found between variations in DMS and seawater temperature (rs = 0.81, p < 0.001, n = 44) likely reflecting seasonal warming of the ice-free surface waters and ensuing development of DMSproducing organisms. The significant positive correlation found between concentrations of DMS and in vivo 565 fluorescence of chlorophyll (rs = 0.64, p < 0.001, n = 44) throughout the water column in Nares Strait reinforces this suggestion (Fig. 9).
Ratios of DMSPt:Chl a (ranging from 10 to 23 nmol µg-1) averaged over the first 20 m of the water column of the NS transect were low compared to those found in the Lancaster Sound transect (max of 170 nmol µg-1). Taking into account that our DMSPt:Chl a ratios include both particulate and dissolved pools, and considering that dissolved 570 DMSP typically contributes a small fraction of DMSPt (although highly variable; Kiene et al. 2000;Kiene and Slezak 2006), these values are nonetheless similar to previously reported DMSPp:Chl a ratios with a maximum of 39 nmol µg-1 (Luce et al. 2011) and a maximum of 17 nmol µg-1 (Matrai and Vernet 1997), at diatom-dominated stations of the Canadian High Arctic and of the Barents Sea, respectively.
Along the ice-free west-east transect in the North Water (NOW), patterns of temperature and salinity (Fig. 7) revealed 575 the interactions between the southward advection of fresh and cold Arctic waters along Ellesmere Island and saltier and warmer Atlantic waters flowing northward along western Greenland via the West Greenland Current (WGC) (Curry et al. 2011;Münchow et al. 2015). Surface water concentrations of nitrate were below 0.04 µmol L-1 across the entire transect, exposing more mature blooming stage conditions similar to those found in the LS transect. As such, maximal accumulation of biomass occurred below the surface in most stations along the NOW transect in association 580 with the nitracline (Spearman's rank correlation between in vivo fluorescence and NO3-, rs = -0.86, p < 0.001, n = 42).
The phytoplankton assemblage along the NOW transect was similar to the ones observed further south at the mouth of Lancaster Sound and further north along Nares Strait. In the surface waters of stations 101, 108 and 111, the Phaeocystis is widespread across the globe, including in high boreal and arctic waters (Verity et al. 2007) and its blooming has been linked to vast amounts of DMSP in the marine environment (van Duyl et al. 1998;Stefels et al. 2007;Asher et al. 2017). In this study, the presence of a DMSP hotspot (up to 113 nmol L-1 at ca. 12 m depth) in the upper waters of the easternmost station 115 of the NOW transect may be partially explained by the occurrence of 595 Phaeocystis pouchetii as well as the numerical dominance of unidentified flagellates, including potentially DMSPrich species (Keller 1989   and 12 nmol L-1, respectively, suggesting that a bloom had already started to develop under the melt pond-covered ice through the potential seeding of autotrophic organisms from the ice. The strong negative association found between salinity and DMS points towards ice itself as an important vector for sea surface DMS, contributing to its seeding at 630 the ice-sea interface as observed elsewhere (Trevena and Jones 2006;Kiene et al. 2007;Tison et al. 2010). Halinedriven stratification of waters under the ice cover likely promoted the physical accumulation of DMS. Alternately, the surface stratification may have favored the biological production of DMS. The formation of a fresher water lens at the surface of the water could have led to the entrapment of algal cells and to an increase in solar radiation exposure with heightened DMS production as a defense strategy against light-associated oxidative stress (Sunda et al. 2002;Toole 635 and Siegel 2004;Vallina and Simó 2007;Galí and Simó 2010). The fresher water lens may also have indirectly stimulated DMS production through the possible enhancement of DMSPd availability and its bacterial conversion into DMS, following an osmotic-related excretion of cellular DMSP (Stefels 2000;Van Bergeijk et al. 2003;Niki et al. 2007). Although biological processes cannot be completely ruled out, the strength of the association between DMS and salinity near the FYI edge suggests that physical drivers most strongly shaped DMS dynamics in Barrow Strait.

640
In contrast to the FYI-dominated region described above, the waters exiting the MYI-dominated region of Nares Strait did not exhibit the same potential under-ice development of autotrophic organisms. The phytoplankton community in the surface waters of the station sampled nearest to the ice edge was dominated by flagellates and Chl a concentrations were comparatively low (< 0.5 µg L-1), as were the concentrations of DMSPt (< 16 nmol L-1) and DMS (< 0.4 nmol L-1). The development of a phytoplankton bloom, and increase in both DMSP and DMS concentrations, occurred several 645 km (ca. 100 km, Station KEN3) away from the ice edge highlighting the requirement for sufficient light to initiate the growth of primary producers. One of the distinguishing features between the two ice edges was the presence/absence of melt ponds at their surfaces. This factor likely played a major role in driving the availability of light through the ice as suggested by Nicolaus et al. (2012), leading to the earlier onset of a bloom (Fig. 10) and shaping the associated DMS cycling under the ice in the Barrow Strait region where melt ponds covered ca. 40% of the total surface. Findings 650 from this study are of particular significance in light of the suggestion that regions of the CAA (Fortier et al. 2002;Mundy et al. 2014), the Beaufort Sea (Mundy et al. 2014) and Baffin Bay (Oziel et al. 2019) may hold regular, yet under-documented, under-ice phytoplankton blooms. The occurrence of these blooms may be linked to the fact that the archipelago is characterized by narrow waterways where landfast ice tends to linger longer, allowing advanced stages of ice melt to be reached prior to break up, and where shallow waters act to enhance the supply of nutrients into 655 surface waters fueling the potential growth of under-ice blooms (Michel et al. 2006). Autotrophic biomass accumulations below the Chukchi Sea ice cover described by Arrigo et al. (2012) bring further support to the possible widespread importance of these blooms in waters of the Arctic. Furthermore, FYI has become the prevailing type of ice in the Arctic at the expense of swiftly declining MYI (Comiso et al. 2008). As such, and because FYI tends to have greater areal melt pond coverage than MYI due to a smoother topography , climate-driven 660 changes in sea ice dynamics may lead to modifications in the timing and frequency of under-ice blooms, their role in seeding ice-edge blooms in summer (Strass and Nöthig 1996) and the associated production of DMS (Galí and Simó 2010;Levasseur 2013). It is also worth noting that the highest sea surface DMS concentration measured during this expedition was associated with the presence of Phaeocystis (STN115, West Greenland current), a genus for which a few modelling studies point towards a poleward expansion in its geographical extent (Cameron-Smith et al. 2011; 665 Menzo et al. 2018) associated with the increased intrusion of warm Atlantic water masses in the Arctic (Neukermans et al. 2018). Altogether, these factors in conjunction with the projected increase in melt pond cover and their temporal span (Agarwal et al. 2011;Stroeve et al. 2014;Holland and Landrum 2015;Liu et al. 2015) and the direct role melt ponds may play in the production of DMS (Gourdal et al. 2018) suggests that there is a need to review the potential production and cycling of DMS in ice-covered areas of the Arctic during summer. As thinner, younger and more 670 dynamic icescapes may prevail in the Arctic, earlier and more ubiquitous under ice blooms may lead to earlier pulses of DMS through leads, cracks and edges of the ice with implications for climate forecasting.
Recent modelling studies predict an increase of DMS emissions in the Arctic, predominantly associated with sea ice retreat, and inducing a negative climate feedback through the influence of atmospheric DMS on cloud formation and radiative forcing Mahmood et al. 2019). Most models however consider the ice-atmosphere interface 675 to be inert. Possible diffusion of DMS through porous ice during spring (Gourdal et al. 2019), as well as potential DMS pulses venting to the atmosphere via melt ponds (Gourdal et al. 2018) and through cracks and leads in thinner ice and at ice edges (Hayashida et al. 2017, this study) could lead to a strengthening of the DMS-related "polar-

Author contributions
M. Lizotte was responsible for a large part of the sampling as well as the data analysis and processing. M. Levasseur and M. Lizotte wrote the initial version of the paper together. Several co-authors provided specific data included in 695 the paper and all co-authors contributed to the final edition of the paper.

Competing interests
The authors declare that they have no conflict of interest.
Author's response. We agree with the referee: on L148 (now L155), chlorophyll a should be written out and there were different forms of the abbreviation of chlorophyll a. Author's changes in manuscript. On L148 (now L155), the word "chlorophyll a" was added and we changed "chl a" for "(Chl a)". On L163, we changed "chlorophyll a (chl a)" to simply "Chl a".

1415
Author's response. Yes, we agree with both the comment and the suggestion.
Author's changes in manuscript. The size of the letters on the map were made larger. A new version of Figure 2 was added to the manuscript. Figure 10: For FYI diagram, the relationship between phytoplankton bloom and light availability is 1420 clearly indicated, but I'm afraid that the reader may not catch what the authors would like to show in MYI diagram. Please modify the MYI diagram to show the relationship of phytoplankton abundance and light availability. Also, the second sentence (How these physical changes. . .) may be omitted from the figure caption.
Author's response. We thank the referee for the insight and agree that the figure should be made clearer.

1425
Author's changes in manuscript.
The following modifications were made to Figure 10. 1. On the first panel (MYI) the arrow (light) going from the sun and through the thicker ice was presented as discontinued (dotted arrow) to signify reduced intensity of light reaching the surface 1430 of the water and available for phytoplankton growth. Part of the light is absorbed by the ice (one arrow ending in the ice), and another part of the light is reflected back (2 arrows pointing upwards). 2. On the second panel (FYI) the arrows (light) going from the sun and through the thinner ice and the melt ponds at the surface of the ice show scattering and an increase in the amount of light 1435 reaching the surface of the water and available for phytoplankton. Part of the light is absorbed by the ice (one arrow ending in the ice), and another part of the light is reflected back (1 arrow pointing upwards).
Furthermore, as suggested, we modified the caption as follows and took the second sentence out.
Initial version: Figure 10: Conspicuous alterations in the Arctic Ocean are underway and include reductions in snow cover, sea ice extent and thickness, and increase in melt pond areal coverage, the occurrence of which is linked to profound modifications in light availability in surface waters below the ice and at its margin. How these physical changes will impact the dynamics of bloom-forming microorganisms and their production of the 1445 biogenic climate-active gas DMS are still unknown. The conceptual diagram depicts two types of ice edges (top panel MYI and lower panel FYI) and their potential role in modulation light penetration under the ice pack and the development of phytoplankton blooms and associated DMS dynamics.   (Table 1) and biogeochemical characteristics (Table 2), but also present the same inherent structure, reason why several elements are repeated.
Author's changes in manuscript. The caption for Table 2 was modified from its original version to include a more detailed description of the biogeochemical characteristics found in the Table per  . Values that were not available are noted as 'n.a.'