Biogeosciences,15,3169–3188,2018 https://doi.org/10.5194/bg-15-3169-2018 ©Author(s)2018.Thisworkisdistributedunder theCreativeCommonsAttribution4.0License. Dimethyl sulfide dynamics in first-year sea ice melt ponds in the Canadian Arctic Archipelago MargauxGourdal1,MartineLizotte1,GuillaumeMassé1,MichelGosselin2,MichelPoulin3,MichaelScarratt4, JoannieCharette2,andMauriceLevasseur1 1Départementdebiologie,Québec-OcéanandUnitéMixteInternationale3376TAKUVIK,CNRS-UniversitéLaval, 1045AvenuedelaMédecine,VilledeQuébec,QuébecG1V0A6,Canada 2InstitutdesSciencesdelaMerdeRimouski(ISMER),UniversitéduQuébecàRimouski,310AlléedesUrsulines, Rimouski,QuébecG5L3A1,Canada 3ResearchandCollections,CanadianMuseumofNature,P.O.Box3443,StationD,Ottawa,OntarioK1P6P4,Canada 4MauriceLamontagneInstitute,FisheriesandOceansCanada,P.O.Box1000,Mont-Joli,QuébecG5H3Z4,Canada Correspondence:MargauxGourdal([email protected]) Received:15October2017–Discussionstarted:21November2017 Revised:4April2018–Accepted:7May2018–Published:29May2018 Abstract.Meltpondformationisaseasonalpan-Arcticpro- 1 Introduction cess. During the thawing season, melt ponds may cover up to90%oftheArcticfirst-yearseaice(FYI)and15to25% Meltpondsrepresentanimportantbutunderstudiedcompo- of the multi-year sea ice (MYI). These pools of water ly- nentoftheArcticseaicesystem.Snowdepositedatthesur- ing at the surface of the sea ice cover are habitats for mi- face of the sea ice progressively melts during the thawing croorganisms and represent a potential source of the bio- season and may accumulate above sea level in depressions genic gas dimethyl sulfide (DMS) for the atmosphere. Here at the surface of the ice to form melt ponds (Lüthje et al., we report on the concentrations and dynamics of DMS in 2006),likelythrougharecentlyidentifiedprocessofperco- ninemeltpondssampledinJuly2014intheCanadianArc- lationblockage(Polashenskietal.,2017).IntheArctic,melt ticArchipelago.DMSconcentrationswereunderthedetec- pond fraction over first-year sea ice (FYI) in late spring– tion limit (<0.01nmolL−1) in freshwater melt ponds and summer usually ranges from 50 to 60%, locally reaching increased linearly with salinity (rs=0.84, p≤0.05) from 90% (Fetterer and Untersteiner, 1998; Eicken et al., 2004; ∼3upto∼6nmolL−1(avg.3.7±1.6nmolL−1)inbrackish Lüthjeetal.,2006;Perovichetal.,2011).Röseletal.(2012) melt ponds. This relationship suggests that the intrusion of reported a 15% increase of the relative melt pond fraction seawaterinmeltpondsisakeyphysicalmechanismrespon- forthemonthofJuneduringthelastdecade(2001–2011)in sibleforthepresenceofDMS.Experimentswereconducted theArctic,mostlikelyattributabletoglobalclimatechange. withwaterfromthreemeltpondsincubatedfor24hwithand Thispartlyreflectstheprogressivereplacementofmulti-year without the addition of two stable isotope-labelled precur- sea ice (MYI) by FYI observed since the 1980s (National sorsofDMS(dimethylsulfoniopropionate),(D6-DMSP)and SnowandIceDataCenter,NSIDC,http://nsidc.org,lastac- dimethylsulfoxide (13C-DMSO). Results show that de novo cess: June 2017), favouring the formation of shallow melt biological production of DMS can take place within brack- ponds that spread over increasingly large areas (Agarwal et ish melt ponds through bacterial DMSP uptake and cleav- al.,2011;Ehnetal.,2011).Theimportanceofmeltpondsin age. Our data suggest that FYI melt ponds could represent the Arctic, as a water–air interface involved in heat and gas a reservoir of DMS available for potential flux to the atmo- exchanges,isthusexpectedtoincreaseinthefuture. sphere.Theimportanceofthisice-relatedsourceofDMSfor Dimethylsulfide(DMS)isthemainnaturalsourceofre- theArcticatmosphereisexpectedtoincreaseasaresponseto duced sulfur for the atmosphere (Bates et al., 1992). Be- thethinningofseaiceandthearealandtemporalexpansion tween 17.6 and 34.4Tg of sulfur is released annually from ofmeltpondsonArcticFYI. the ocean to the atmosphere (Lana et al., 2011), accounting PublishedbyCopernicusPublicationsonbehalfoftheEuropeanGeosciencesUnion. 3170 M.Gourdaletal.:Dimethylsulfidedynamicsinfirst-yearseaicemeltponds for50–60%ofthenaturalreducedsulfuremitted(Stefelset viral lysis, and zooplankton grazing. DMSP is then read- d al.,2007).DMSisalsoaclimate-relevantgaspotentiallyin- ily available to heterotrophic bacteria as carbon and sulfur volved in a feedback loop known as the CLAW hypothesis sources (Kiene et al., 2000; Simó, 2001; Vila-Costa et al., (Charlson et al., 1987) linking biology and climate through 2006). The fraction of DMSP consumed by heterotrophic d the production of DMS-derived sulfate aerosols. According bacteria and enzymatically cleaved by DMSP lyases into to CLAW, DMS emissions may affect the global radiation DMS(DMSyield)mayvarydependingonthecomposition budgetdirectlythroughthescatteringofincomingsolarradi- ofmicrobialcommunities,theirsulfurrequirements,andthe ationandindirectlyviatheproductionofcloudcondensation availability of other reduced forms of sulfur (Kiene et al., nuclei (CCN), leading to the genesis of longer-lived clouds 2000;Stefelsetal.,2007).DMSPlyasesarealsopresentin withhigheralbedo(Twomey,1974;Albrecht,1989).Inspir- severalmembersofthemicroalgalgroupsHaptophyceaeand ing three decades of research and hundreds of publications, Dinophyceaeand,toalesserextent,Chrysophyceae(Nikiet thefeedbackmechanismproposedbyCharlsonetal.(1987) al., 2000). Ultimately, between ∼1 and 40% of the DMSP remains to be demonstrated in its entirety (e.g. Ayers and producedbyalgaereachestheatmosphereasDMS(Stefelset Cainey, 2008). Although modelling results show that DMS al., 2007; Simó and Pedros-Alio, 1999a). In addition to the emissionsmayhaveanegativeradiativeeffect(e.g.Boppet DMSP enzymatic cleavage pathway, DMS production may al., 2004; Gunson et al., 2006; Thomas et al., 2010), CCN arise from dimethylsulfoxide (DMSO) reduction by various mayexhibitalowsensitivitytochangesinDMSonaglobal groupsofmarinebacteriaincludingproteobacteria(e.g.Vogt scale (Woodhouse et al., 2010). Recent studies questioning etal.,1997),membersoftheRoseobactergroup(Gonzálezet the relative importance of DMS in new particle formation al.,1999),andmat-formingcyanobacteria(vanBergeijkand have emerged, suggesting that the global CLAW feedback Stal,1996).However,theubiquityofthisDMSO-to-DMSre- maybeweak(e.g.QuinnandBates,2011;GreenandHatton, ductionpathwayamongstbacterialassemblageshasnotbeen 2014). On a regional scale, however, the response of CCN established (Hatton et al., 2012). A limited number of phy- production to change in DMS may vary by a factor of 20 toplankton species could also be involved in the reduction (Woodhouseetal.,2010).TheimpactofDMSemissionson of DMSO into DMS (e.g. Fuse et al., 1995; Spiese et al., cloud properties (through the production of CCN) could be 2009). Increasing evidence suggests that particulate DMSO particularly important in remote pristine marine areas such (DMSO )maybedirectlysynthesizedbyapotentiallywide p as the polar regions (Carslaw et al., 2013). In the Southern rangeofmarinephytoplanktonandcouldbeinvolvedinos- Ocean,DMSmayhavecontributedupto33%oftheincrease moprotection,cryoprotection(LeeanddeMora,1999),and in CCN observed south of 65◦S as a response of increased antioxidant protective mechanisms (Sunda et al., 2002). As wind speed since the early 1980s (Korhonen et al., 2010). for dissolved DMSO (DMSO ), it is ubiquitous in seawater d The summertime Arctic marine boundary layer (MBL) is and continuous improvements in analytical techniques sug- leftrelativelycleanafterseasonalwetdepositionofparticles gestthatDMSO maybeasabundantasDMSinsurfacewa- d and reduced atmospheric transport of aerosols from anthro- ters(e.g.Simóetal.,2000).DMSOisalsoaknownsinkfor pogenicsourcesatlowerlatitudes(Stohl,2006;Browseetal., DMS(Hattonetal.,2004)viabacterialandphoto-oxidation 2012;Croftetal.,2016).Suchpristineconditions,combined of DMS to DMSO. Vertical mixing and ventilation are also withthermallystableMBL,aretypicaloftheArcticsummer- majorremovalprocessesinfluencingDMSconcentrationsin time(e.g.Aliabadietal.,2016).CleanArcticairmassesal- surfacemixedlayers(Batesetal.,1994;Kieberetal.,1996; lowultrafine(5–20nmdiameter)particleformation(Burkart SimóandPedrós-Alió,1999b;delValleetal.,2007,2009). etal.,2017)andthepotentialgrowthofsecondarymarineor- Ice-associatedenvironmentssuchasbottomseaice,brine ganic aerosols (including DMS-derived particles) into CCN channels, melt ponds, under-ice surface waters, and leads (Willisetal.,2016).Hence,theArcticisafavourableterrain provide complex and dynamic habitats to diverse microor- fornewparticleformationfrombiogenicDMS(Changetal., ganism communities involved in sulfur cycling (Levasseur, 2011;Rempilloetal.,2011;Collinsetal.,2017;Giamarelou 2013). In the Arctic, the highest microalgal biomasses are etal.,2016;Mungalletal.,2016;Willisetal.,2016). found in the bottom ∼0.1m of sea ice, with chlorophyll a DMS stems mainly from the enzymatic cleavage of (Chl a) concentrations several orders of magnitude above dimethylsulfoniopropionate (DMSP) by algal and bacterial values for under-ice waters values (e.g. Legendre et al., DMSPlyases.DMSPisacellularmetabolitefoundinseveral 1992). A similar pattern of DMSP, DMSO and DMS build- phytoplankton species as particulate DMSP (DMSP ) (see up in bottom ice has been reported both in the Arctic and p the review of Green and Hatton, 2014). DMSP plays vari- Antarctica(Kirstetal.,1991;Levasseuretal.,1994;Turner p ousrolesinphytoplankton,includingosmoregulation(Lyon et al., 1995; DiTullio and Smith Jr., 1995; Lee et al., 2001; et al., 2016), cryoprotection (Karsten et al., 1996), and pre- Trevenaetal.,2003;TrevenaandJones,2006;Delilleetal., ventionofcellularoxidation(Sundaetal.,2002).Partofthe 2007; Tison et al., 2010; Asher et al., 2011; Nomura et al., DMSP produced by algae is released in the water column 2012; Galindo et al., 2015). For example, DMSP concen- p p asdissolvedDMSP(DMSP )viaseveralpathwaysreviewed trations up to 15000nmolL−1 have been documented dur- d inStefelsetal.(2007),includingactiveexudation,celllysis, ing spring in bottom FYI of the eastern Arctic (Galindo et Biogeosciences,15,3169–3188,2018 www.biogeosciences.net/15/3169/2018/ M.Gourdaletal.:Dimethylsulfidedynamicsinfirst-yearseaicemeltponds 3171 al.,2014)whilesurfacewaterconcentrationsgenerallyrange may originate from algal and bacterial metabolism. How- between 1±0.2 and 50nmolL−1 (Vila-Costa et al., 2008; ever, in situ DMS production had never been measured nor MatraiandVernet,1997).DMSP,DMSO,andDMSarealso hadkeymechanismsbeenidentified.Here,wereportonthe present throughout the ice column within the brine network DMSconcentrationsinninemeltpondslocatedintheeastern (Levasseur et al., 1994; Trevena and Jones, 2006; Asher et CanadianArcticArchipelago(CAA)andontheprerequisites al.,2011).Giventhatprimaryproducersarethesolesource and processes responsible for the presence of this climate- ofDMSP,veryhighiceconcentrationsofChla canbecor- active gas. This is the first attempt to assess the dynamics related with DMSP through a first-order relationship (Lev- of DMS in Arctic melt ponds. We identified sea ice perme- asseur,2013).ThisChlatoDMSPrelationshipmaynothold abilityasamajorcontrolofDMSproductioninmeltponds, forlowerbiomassconcentrations(Tisonetal.,2010).Inad- mediating the transport of both DMS and DMS-producing ditiontothevariabilityinducedbyinter-specificdifferences communitiestowardthesurfaceofseaice.Wealsoprovide in DMSP cellular contents (e.g. Keller et al., 1989; Stefels, the first evidence for direct in situ DMS production in Arc- 2000), environmental forcing are known to control DMSP, tic melt ponds. We propose that seasonally melting sea ice DMSO,andDMSconcentrations.Inice-associatedenviron- mightbecomeincreasinglypronetoDMSproductionasFYI ments,brinevolumefractionmightalsobekeyinexplaining becomelargelypredominantontheregionalscale. DMS cycling variability via the control of ice permeability (Carnat et al., 2014). Structural changes within sea ice dur- ingthemeltseason,namelyincreasesinbrinevolumefrac- 2 Materialsandmethods tionandicedesalination,resultinincreasedconnectivityand permeabilityinthewarmingseaice(Willisetal.,2016;Po- 2.1 Studysitesandenvironmentalmeasurements lashenskietal.,2012)andinfluenceDMSPandDMScycling (Tisonetal.,2010;Carnatetal.,2014).Also,phytoplankton Nine melt ponds distributed between four stations located blooms developing under the ice during the melting period in Navy Board Inlet (Ice1–MP1 and MP2–18 July), Bar- havebeenshowntoproducelargequantitiesofDMSP ,po- row Strait (Ice2–MP1 to MP3–20 July and Ice3–MP1 and p tentiallyleadingtoabuild-upofDMSconcentrations(Lev- MP2–21July),andResolutePassage(Ice4–MP1andMP2– asseuretal.,1994).Inspiteofthespatialimportanceofmelt 23July)weresampledduringthejointNETCARE/ArcticNet ponds, only a few studies have investigated their role as a research cruise conducted in 2014 on board the Canadian source of DMS for the Arctic atmosphere (e.g. Levasseur, CoastGuardShip(CCGS)Amundsen(Fig.1). 2013). At each station except for Ice2 (logistical constraints as- Considerable efforts have been dedicated to the under- sociatedwithshiptimeline),measurementsofseaicethick- standing of underlying process controlling the physics of ness,snowdepth,andseaicefreeboard(theheightofseaice meltpondsandtheirfeedbacksonclimatethroughthecon- above the ocean surface) were conducted within a 3m dis- trolofsurfaceenergybalanceoftheice(Lüthjeetal.,2006; tance of the melt ponds using a gauge (Kovacs Enterprise, Polashenskietal.,2017).However,littleisknownabouttheir Roseburg,OR,USA)(Table1).The3mdistancewasacom- biogeochemistry. Four studies have specifically reported on promise between maximizing the proximity of ice and melt DMS in melt ponds so far. They reveal negligible DMS pondsamplesandminimizingmeltponddisturbanceduring concentrations in MYI ice melt ponds in the central Arctic sampling operations. Ice and freeboard thickness presented Ocean and concentrations up to 2.2nmolL−1 in the High in Table 1 are averaged values of the seven (Ice1) to eight Arctic (Leck and Persson, 1996; Sharma et al., 1999). In (Ice3 and Ice4) ice cores sampled at each station between Antarctica, DMS concentrations ranging between 1.1 and the team members for their respective projects. In order to 3.7nmolL−1andbetweenbelowthedetectionlimit(d.l.)and estimate the permeability of the ponded ice, sea ice tem- 250nmolL−1 weremeasuredintwostudies(Nomuraetal., perature and bulk salinity were measured following Miller 2012,andAsheretal.,2011,respectively).Inthelatterstudy, et al. (2015) at stations Ice1, Ice3, and Ice4. Two ice cores bacterialDMSOreductionwassuggestedasapossiblemech- for sea ice temperature and salinity measurements were ex- anismresponsibleforthehighDMSconcentrationsobserved tracted using a 0.09m core barrel (Kovacs Mark II, Kovacs although no actual rates of DMS production, either from Enterprise, Roseburg, OR, USA). In situ sea ice tempera- DMSOorDMSP,weremeasured.HighDMSconcentrations ture profiles were measured directly, at 0.1m intervals, us- reportedintheAntarcticaremostlikelyrelatedtothedevel- ing a high-precision thermometer (Testo® 720; precision of opmentofasurfaceicecommunityfollowingflooding,apro- ±0.1◦C). Corresponding sea ice salinity profiles were also cesswherebyheavysnowloadpushestheicebelowthewater determined at 0.1m intervals. Each 0.1m section was cut level.FloodingiscommonintheAntarcticandresultsinthe with a handsaw, stored in a plastic container, and allowed formation of snow ice (Hunke et al., 2011). Several studies tomeltatroomtemperature.Bulksalinityofthemeltedice documentmeltpondcolonizationbymicro-,nano-,andpico- section was determined using a conductivity probe (Cond sizedalgaeaswellasbacteria(Bursa,1963;Gradingeretal., 330i, WTW™; precision of ±0.1%). Permeability to fluid 2005;Elliottetal.,2015),suggestingthatDMSinmeltponds transport was assessed with brine volume profile calcula- www.biogeosciences.net/15/3169/2018/ Biogeosciences,15,3169–3188,2018 3172 M.Gourdaletal.:Dimethylsulfidedynamicsinfirst-yearseaicemeltponds Figure1.(a)Regionalmapshowingthelocationofthefoursamplingstations(Ice1toIce4)(redcircles)duringtheNETCARE/ArcticNet 2014campaign.(b)MODISimageryabovethefoursamplingstation(redcircles)showingtheiceconditionson18July2014inthesampling area.(c)Lefttoright,picturesofstationsIce1,Ice2,Ice3andIce4withsizescale.MPFstandsforthemeltpondfractionvisuallyestimated fromthebridgeforstationsIce1,Ice2,Ice3,andIce4. tionsfrombulksalinitiesandseaicetemperaturesfollowing theextractedpigmentswasmeasuredonboardwithaTurner equationsfromLeppärantaandManninen(1988)forseaice Designsfluorometer(model10-005R;TurnerDesigns,Inc.) temperatures>−2◦C(Fig.3).Duetologisticalconstraints before and after acidification with 5% HCl. The fluorome- mentioned above, neither ice nor snow measurements were terwascalibratedwithacommerciallyavailableChla stan- conductedatstationIce2. dard(Anacystisnidulans,Sigma).Chlaconcentrationswere Melt pond depth, length, and width were determined us- calculated using the equation provided by Holm-Hansen et ingagraduatedstickandataperuler.Meltpondwatertem- al.(1965). perature was measured using a high-precision thermometer Microscopicidentificationandenumerationofeukaryotic (61220-601digitaldatalogger,VWR)andwatersalinitywas cells>2µmwereconductedineachmeltpond.Samplesof measuredusingtheconductivityprobementionedinthepre- 250mL were collected and preserved with acidic Lugol so- vious paragraph (Table 2). For each sampling location, two lution (0.4% final concentration; Parsons et al., 1984), then tothreemembersoftheresearchteamvisuallyassessedthe storedinthedarkat4◦Cuntilanalysiswasconductedbyin- pond fraction based on pictures taken from the bridge (see verted microscopy (Lund et al., 1958; Parsons et al., 1984). Fig.1cforexamples)andameanvaluewascalculated. Foreachsample,aminimumof400cells(accuracy±10%) and three transects of 20mm were counted at a magnifica- 2.2 Phytoplanktonbiomassandenumeration, tionof400×.Themaintaxonomicreferencesusedtoiden- bacterialcount tifytheeukaryoticcellsareTomasandHasle(1997),Bérard- Therriaultetal.(1999),andThrondsenetal.(2003). For Chl a quantification, 1000 to 1500mL duplicates of in Theabundanceofbacteriawasdeterminedbyflowcytom- situ pond water were filtered onto Whatman® GF/F 25mm etry (Marie et al., 2005). Duplicate 4mL subsamples were filters. Pigments were extracted in 90% acetone for 18 to fixed with 20µL of 25% glutaraldehyde Grade I (0.1% fi- 24hinthedarkat4◦C(Parsonsetal.,1984).Fluorescenceof Biogeosciences,15,3169–3188,2018 www.biogeosciences.net/15/3169/2018/ M.Gourdaletal.:Dimethylsulfidedynamicsinfirst-yearseaicemeltponds 3173 Table1.Physicalcharacteristicsoftheseaicesurroundingthemelt Table2.Physicalcharacteristicsofthemeltpondwater.Formelt ponds.Notethatonlymeltpondsampling(i.e.noicesampling)was ponddepth,mean±standarddeviationvaluesarepresented. conductedatstationIce2duetoship-relatedlogisticalconstraints. A negative freeboard height indicates that the ice surface was lo- Station Meltpondno. Meltpond Meltpond Meltpond callybelowthemeansealevel.NAstandsfornon-availabledata. depth(m) salinity Temperature(◦C) Icethicknessandfreeboardvaluesareaveragesof7(Ice1)to8(Ice3 Ice1 MP1 0.18±0.01 5.2 1.9 andIce4)icecoressampledateachstation. Ice1 MP2 0.18±0.04 4.1 1.8 Ice2 MP1 0.29±0.05 0.7 0.4 Ice2 MP2 0.19±0.03 0.4 0.3 Station Sampling Snowand Ice Freeboard Ice2 MP3 0.12±0.01 0.2 0.2 date frozensnow thickness (m) Ice3 MP1 0.07±0.01 1.1 0.2 depth(m) (m) Ice3 MP2 0.10±0.00 0.9 0.2 Ice1 18Jul2014 0 1.21±0.02 −0.01±0.01 Ice4 MP1 0.12±0.01 8.1 0.2 Ice2 20Jul2014 0 NA NA Ice4 MP2 0.11±0.02 8.5 0.3 Ice3 21Jul2014 0+0.07∗ 1.13±0.07 0.10±0.02 Ice4 23Jul2014 0 1.27±0.01 0.07±0.04 al. (2012). Briefly, DMS was stripped from liquid samples usingheliumgas(Praxair™ He,purity99.999%)flowingat nalconcentration;Sigma-AldrichG5882),thensubjectedto 50±5mLmin−1 inthePnTsystem.Oneto5mLofsample quick-freeze in liquid nitrogen for 24h, and finally stored wasinjectedinthePnT.FivemLofMilli-Q™ water(Milli- at −80◦C until analysis. Samples were analyzed using a pore filter system, Millipore Co., Bedford, MA, USA) was FACS Calibur FCB3 flow cytometer (Becton Dickinson). subsequentlypushedintothesystemtocompletelyflushthe Heterotrophic bacteria samples were stained with SYBR sample into the glass bubbling chamber. The outer walls of Green I and measured at 525nm to quantify bacteria with the bubbling chamber were heated at 70◦C with a circu- lownucleicacid(LNA;potentiallylessactive)andhighnu- lating bath. Humidity in the gas sample downstream of the cleicacid(HNA;potentiallymoreactive)content(Gasoland bubbling step was minimized using a 4◦C circulating bath delGiorgio,2000;Lebaronetal.,2001).Analysiswereper- totriggercondensation.ANafion® membraneseparatedthe formed on an Epics Altra flow cytometer (Beckman Coul- gas sample and He-carrier gas from a drying He counter- ter), fitted with a 488nm laser (15mW output; blue), using flowsetat100mLmin−1tofurtherdesiccatethegassample. Expo32v1.2bsoftware(BeckmanCoulter). FluxesinthePnTsystemweremonitoredusingaflowmeter (Varian™). 2.3 DMS(P)sampling,conservationandanalysis ForDMSP samples,3.5mLofmeltpondwaterwascol- t lected in duplicate into a 5mL Falcon™ tube. DMSP sam- d Duplicate samples for total DMSP (DMSPt), dissolved ples were obtained using the less disruptive small-volume DMSP (DMSPd) and DMS measurements were collected gravity drip filtration (SVDF) method (Kiene and Slezak, from the melt ponds using a submersible pump (Cyclone – 2006).ParticulateDMSP(DMSP )concentrationswerecal- p Aquameric™) connected to a sealed lead–acid battery and culatedbysubtractingDMSP fromDMSP.DMSPsamples d t fittedwithLDPEtubing.Thepumpwasplacedclosetothe were preserved with 50µL of 50% sulfuric acid (H SO ) 2 4 pondbottom,withouttouchingtheice.Stratificationwasre- to prevent DMSP transformation and remove pre-existing portedinopenmeltponds(i.e.meltpondsthathavemelted DMS. Samples were analyzed using the same methods as all the way to the sea surface) in Arctic FYI (Jung et al., described above for DMS samples, following mole-to-mole 2015). However, closed FYI melt pond, such as those sam- conversionofDMSPintoDMSviaNaOH(5M)hydrolysis pledduringthisstudy,arenotpronetoverticalstratification (DaceyandBlough,1987). duetoconvective-andwind-drivenmixing(Skyllingstadand Paulson,2007).Giventheirshallowdepths(lessthan0.3m), 2.4 Processstudies meltpondstratificationwasmostprobablyinexistentormin- imal during our study. Glass serum bottles were filled with InordertoexaminethepathwaysofinsituDMSproduction sampled water, temporarily sealed with a butyl cap and an in melt ponds, three 24h incubation experiments were con- aluminum lid, and kept in the dark in a cooler until analy- ducted with water from the MP1 sampled at stations Ice1, sisuponreturntotheship.Analysiswereperformedusinga Ice3,andIce4.Waterfromthemeltpondswascollectedus- purgeandtrap(PnT)systemcoupledtoaVarian™ 3800gas ing the pump described in Sect. 2.3, pooled in clean 19L chromatograph (GC), equipped with a pulsed flame photo- Coleman™coolerjugsonsite,andthentransferredintogas- metric detector (PFPD). Analytical precision of the method tight3LpolyvinylfluorideTedlar®bags.Lighttransmittance was better than 5%. Analytical d.l. was 0.01nmolL−1 for throughtheincubationbagmaterialdiminishedwithdecreas- all sulfur compounds. The protocol is a modified version inglightwavelength.Between99and92%ofthephotosyn- of the method of Leck and Bågander (1988) as described thetically active radiations (PAR, 400–700nm) were trans- in Scarratt et al. (2000) and further revised in Lizotte et mitted through the bag material. Transmittances of ultravi- www.biogeosciences.net/15/3169/2018/ Biogeosciences,15,3169–3188,2018 3174 M.Gourdaletal.:Dimethylsulfidedynamicsinfirst-yearseaicemeltponds olet A radiations (UVA, 315–400nm) and ultraviolet B ra- DMS versus time in the controls; (3) net potential DMSP d diations(UVB,290–315nm)rangedbetween92and82and changesinnaturallightandinthedarkderivedfromthere- between82and38%,respectively.Theincubationbagswere gressionslopeofDMSP versustimeinL-DMSP/OandD- d rinsedoncewith∼10%HCl,threetimeswithMilli-Q™wa- DMSP/O; (4) net potential DMS production rate in natural ter,andtwicewithmeltpondwatertoavoidcontamination. light and in the dark derived from the regression slope of The bags were custom-built and pre-closed on three sides DMSversustimeinL-DMSP/OandD-DMSP/O.Thedaily (Dalian Delin Gas Packaging Co., Ltd.). After the addition rates were obtained from the slopes between final and ini- of the melt pond water, the bags were sealed with Clip-n- tialconcentrationsover24h.Ourexperimentalsetupalsoal- seal™ Teflonclosuredevices.Avalvewasfittedtoeachbag lowstheestimationoftherelativecontributionofDMSPand toallowtheremovalofanyremainingbubbles. DMSO to the production of DMS, using the discrimination Thesamplesweresubjectedtothreeduplicatetreatments ofthedifferentisotopesofDMS(seeSect.2.5). (total of six bags): (1) two bags of unaltered melt pond water incubated under natural light (control), (2) two bags 2.5 DMSisotopicsignatures amendedwithD6-DMSPand13C-DMSO(100nmolL−1,fi- nalconcentrationeach)incubatedundernaturallight(light- The discrimination of the different isotopic forms of DMS, DMSP/O or L-DMSP/O), and (3) two bags amended with including D6-DMS and 13C-DMS stemming from D6- D6-DMSP and 13C-DMSO (100nmolL−1, final concen- DMSPcleavageand13C-DMSOreduction,respectively,was tration each) incubated in the dark (dark-DMSP/O or D- performed using GC-MS analysis following purging as de- DMSP/O). L- and D-DMSP/O bags were amended with scribed hereafter. Two sets of DMS sample duplicates were ∼100µL of freshly thawed aliquots of two D6-DMSP and taken for the incubation experiments. The first set of du- 13C-DMSO stock solutions (high purity >99%, Sigma- plicates was measured directly on board using the Varian™ Aldrich®).Thehighconcentrationsofisotopesaddedaimed 3800GCdescribedinSect.2.2.ThesecondsetofDMSdu- totriggerarapidandclearbiologicalresponse(i.e.potential plicateswaspreservedthroughcryo-trapping.Cryo-trapping DMS production rates) measurable during our 24h incuba- of DMS was conducted using glass GC liners filled with tions.DMSPandDMSOuptakearenotexpectedtobemu- Tenax-TA polymer (high sulfur affinity) (Pio et al., 1996; tuallyexclusiveandhavebeenobservedconcomitantlyboth Zemmelink et al., 2002; Pandey and Kim, 2009) kept at inlivecultures(Spieseetal.,2009)andinsitu(Asheretal., −80◦Cpriortotheiruseandmaintainedbelow−10◦Cdur- 2011). ingthe5minpurgingandtrappingprocess.TheTenax-filled Bagswereincubatedontheforedeckoftheship.Thetem- deactivatedlinersweremounteddownstreamofthePnTsys- peraturewaskeptasneartoinsituwatertemperatureaspos- tem described earlier. After gas extraction from the liquid siblebycontinuouslyflowingsurfaceseawaterintheincuba- samples,TenaxlinersandtheirDMScontentwerewrapped tor.ThetemperaturesoftheincubationwaterforIce1-MP1, individuallyinaluminumfoil,placedinaPyrex™glasstube Ice3-MP1, and Ice4-MP1 were 1.29±1.75, −0.28±0.26, sealed with a Teflon lid, and returned to the −80◦C freezer and −0.73±0.09◦C, respectively. These mean values were forseveralweeksuntilanalysisonaland-basedGC-MS. within1◦Coftheinsitumeltpondwatertemperatures(Ta- Quantification of D6-DMS and 13C-DMS was conducted ble2). via GC-MS analysis (6978 GC coupled to a 7000B Triple- DMSP, DMSP , and DMS concentrations were mea- Quad MS from Agilent). Mass spectra were collected both t d sured in duplicate every 6h during the incubation period in full scan (m/z45–100) and in selected ion monitoring as described above. DMS production from DMSP cleav- (m/z62, 63, and 68) modes. Final concentrations were cal- age and DMSO reduction were determined through gas culatedfromstandardcurvesusingknownconcentrationsof chromatograph–mass spectrometry (GC-MS) analysis as an both unlabelled DMS and labelled DMS carrying the D6- increase of D6-DMS and 13C-DMS, respectively, in the L- DMS and 13C-DMS signatures. The comparison between DMSP/OandD-DMSP/Otreatments.Discriminationbythe fresh DMS samples measured directly on board during the microorganisms toward lighter (natural) isotopes of DMSP NETCARE/ArcticNet campaign and cryo-preserved DMS andDMSOisexpectedtobeminimal(<10%)accordingto samples shows excellent agreement between the two meth- Elizabeth Colleen Asher (unpublished data). The observed ods(r2=0.96,Fig.2). ratesofchangeintheconcentrationofDMSstableisotopes are thus assumed to be representative of the potential for 2.6 Satellitedata DMScyclinginthesemeltponds. This experimental setup allows the measurement of the Distancesbetweeneachstationsandtheopenoceanwereas- following rates over 6 and 24h: (1) net changes of in situ sessed using scaled NASA’s Earth Observing System Data DMSP andDMSP innaturallightderivedfromthediffer- andInformationSystem(EOSDIS)imagery.Mapsoftheice d p ence of DMSP and DMSP concentrations versus time in cover were accessed for the sampling dates in July 2014 d p thecontrols,respectively;(2)netinsitumicrobialDMSpro- through the MODIS (Terra/Aqua) Corrected Reflectance ductioninnaturallightderivedfromtheregressionslopeof (True Color) layer combined with MODIS (Terra) Cor- Biogeosciences,15,3169–3188,2018 www.biogeosciences.net/15/3169/2018/ M.Gourdaletal.:Dimethylsulfidedynamicsinfirst-yearseaicemeltponds 3175 each treatment L-DMSP/O and D-DMSP/O and (2) assess thepotentialeffectoflightontheconcentrationsandchange rates of the reduced-sulfur compounds under study (DMS, DMSP ,andDMSP )bycomparingpaireddependentsam- d p ples(repeatedmeasures)fromL-DMSP/OandD-DMSP/O. 3 Results 3.1 Pondedseaiceandsnowproperties The physical characteristics of the sea ice surrounding the melt ponds are presented in Table 1 and in Fig. 3. All the samplingsiteswerecharacterizedbyFYI,whichwasthepre- dominant ice type throughout the region under study. Aver- agedseaicethicknessaroundthemeltpondswererelatively Figure 2. Relationship between the concentrations of fresh DMS uniform, varying between 1.13±0.07 and 1.27±0.01m at samplesmeasuredonboardtheshipviagaschromatographyduring the different sites. Average freeboard values were relatively thecampaignandtheconcentrationsofthecorrespondingpreserved morevariable.StationIce1wascharacterizedbylowicefree- duplicate samples measured via coupled gas chromatography and boards −0.01±0.01m. Station Ice3 had the highest posi- massspectrometryinalaboratorysetting.Theconcentrationsofthe preservedDMSsamplesplottedarethesumofthethreeisotopesof tive freeboards with 0.10±0.02m. Station Ice4 freeboards DMSinvestigatedinthisstudy(m/zof62,63,and68;seeMaterials were also positive and showed the greatest variability, with andMethods). 0.07± 0.04m. Brine volume fraction was calculated using sea ice salin- ity and temperature values and used as a proxy of sea ice rected Reflectance (bands 3, 6, 7). These open-source data permeability(Fig.3).Averagedvaluesforbulkseaicesalin- are accessible through the Global Imagery Browse Ser- ityoverthefullthicknessoftheicewere1.73,2.83,and3.75 vices (GIBS) (https://worldview.earthdata.nasa.gov, last ac- atstationsIce1,Ice3,andIce4,respectively.Maximumbulk cess: August 2017). The imagery had a resolution of 250m salinity never exceeded 5.00 (Ice4, 1.2–1.3m section). In onadailyscale. situtemperatures,averagedoverthefullthicknessoftheice, were−0.54,−0.52,and−0.98◦CatstationsIce1,Ice3,and 2.7 Statisticalanalysis Ice4,respectively,andreachedaminimumvalueof−1.39◦C (Ice4, 0.8–0.9m section). Brine volume fraction constantly Normality of the data was assessed using the Shapiro–Wilk exceeded 10% in the ice profiles except in the upper 0.1m test with a 0.05 significance level (R statistical software, R sectionoftheIce3station,wherewelikelyobservedtheef- Core Team, 2016), which revealed that most variables were fectsofrefreezingmetamorphosisofsnowand/orseaicere- non-normally distributed (n=9, df=8, α=0.05). Non- crystallization. Snow meltwater percolation and refreezing parametricSpearman’srankcorrelationtest(r )witha0.05 s can form such superimposed ice layers as observed at sta- significancelevelwasusedtoassesscorrelationbetweenkey tion Ice3. The resulting impermeable layer at the top of the variables since normality could not be achieved uniformly icecontributedtothehighfreeboard(0.1m)measuredatthis through standard normalization methods. Model I linear re- station,representing6%ofthetotalicethickness.Visuales- gressions(r2)wereusedtodeterminebiologicalratesduring timates of the pond fraction ranged from 30 to 60% (see theincubationexperiments(SokalandRohlf,1995). Fig. 1c) and the remaining surface of sea ice was bare ice A non-parametric Mann–Whitney U test was used to de- atstationsIce1,Ice2,andIce4. termine whether the distributions of reduced-sulfur com- pounds (i.e. DMS, DMSPp, and DMSPd) in the Ice1-MP1 3.2 Physical,chemical,andbiologicalcharacteristicsof andIce4-MP1incubationsexperimentswerestatisticallydif- themeltpondwater ferent from one another. The difference in reduced-sulfur compoundconcentrationsbetweenthetwoincubationexper- Thephysicalandchemicalcharacteristicsofthemeltponds iments was not found to be statistically significant (n=45, are presented in Table 2. All melt ponds were closed melt df=16,α=0.05). ponds,i.e.notdirectlyconnectedwiththewatercolumn(Lee Based on the results of the Mann–Whitney U test, a se- et al., 2012). The mean depth of the individual melt ponds ries of Wilcoxon signed-rank tests with a significance level ranged from 0.07 to 0.29m, with length and width varying α=0.05 were conducted on the combined data sets of sta- between 1.00 and 25.00m (Fig. 1). Melt pond water tem- tions Ice1-MP1 and Ice4-MP1 in order to (1) assess the peraturesandsalinitiesvariedbetween0.21and1.86◦Cand presence of statistical differences between the controls and between0.2and8.5,respectively.Chlaconcentrationswere www.biogeosciences.net/15/3169/2018/ Biogeosciences,15,3169–3188,2018 3176 M.Gourdaletal.:Dimethylsulfidedynamicsinfirst-yearseaicemeltponds Figure3.Insitutemperature(•)andbulkicesalinity(◦)profilesoftheseaicesurroundingthemeltpondssampledatstationsIce1(a), Ice3(b),andIce4(c).Temperatureandsalinityvaluesofeach0.1mseaicesectionwereusedtocalculatebrinevolumes(orangebars),an indicatorofseaicepermeability,throughoutthefulldepthofseaice(CoxandWeeks,1983;PetrichandEicken,2010). variable, ranging from 0.03 to 0.48µgL−1 with a mean of 3.3 Dynamicsandcyclingofreduced-sulfur 0.20µgL−1 (Table 3). The composition of the algal assem- compoundsinArcticmeltponds blagepresentinthemeltpondswillbedescribedindetailina companionpaper(Charetteetal.,2018)butissummarizedin Results from the Ice1-MP1 and Ice4-MP1 incubation ex- Table 3. The algal assemblages were dominated by uniden- periments are presented in Fig. 4 (panels a and c, and pan- tifiedflagellates,ice-associatedpennatediatoms,andchrys- elsbandd,respectively).ResultsfromtheIce3-MP1exper- ophytes. Empty diatom frustules were abundant in all melt imentsarenotpresentedsinceDMSP andDMSconcentra- ponds.AbundanceofheterotrophicbacteriawithHNAcon- d tentvariedbetween0.02and0.24×109cellsL−1(Table3). tionsshowednovariationduringthe24hincubationperiod in the controls and in the amended treatments. This will be In situ DMSP and DMSP concentrations ranged from p d 1.8 to 4.0nmolL−1 and from below d.l. (<0.01nmolL−1) discussedinSect.4.2.2. to 1.4nmolL−1, respectively. Melt pond DMS concentra- During the Ice1-MP1 incubation, initial DMSPd con- tions ranged from below d.l. to 6.1nmolL−1 (Table 3). centration was 1.30nmolL−1 in the control and slightly increased to reach 5.3nmolL−1 during the 24h incuba- Spearman’srankcorrelationcoefficientsbetweenkeyinsitu tion period (Fig. 4a). In the light (L-DMSP/O) and dark variables measured in the melt ponds are presented in Ta- (D-DMSP/O) amended treatments, DMSP concentrations ble4.DMSconcentrationssignificantlyco-variedwithsalin- d started at 102nmolL−1, decreased to ∼35nmolL−1 at ity (r =0.84, p<0.05) and Chl a (r =0.84, p<0.05). s s T , and remained stable (dark treatments) or decreased to None of the other variables measured displayed significant 6 10nmolL−1 (light treatments) until T (Fig. 4a). Con- relationshipsbetweeneachother(notshown). 24 centrations of DMS in the control of Ice1-MP1 started at 3.0nmolL−1,increasedto8.8nmolL−1 betweenT andT , 0 6 andthendecreasedregularlyto4.2nmolL−1atT (Fig.4c). 24 TheadditionoflabelledDMSPandDMSOstimulatedDMS production. In the L-DMSP/O treatment, DMS concentra- tionsincreasedto12.6nmolL−1atT ,remainedatthislevel 6 Biogeosciences,15,3169–3188,2018 www.biogeosciences.net/15/3169/2018/ M.Gourdaletal.:Dimethylsulfidedynamicsinfirst-yearseaicemeltponds 3177 Figure4.TemporalvariationsinDMSPd(a,b)andDMS(c,d)concentrationsduringtheIce1-MP1andIce4-MP1incubationexperiments. Bothlight(◦)anddark(•)treatmentswereinitiallyamendedwith100nmolL−1ofbothD6-DMSPand13C-DMSO.Controltreatments((cid:52)) mimicnaturalconcentrationchangesovertime.Inpanels(a,b),verticalbarsrepresentstandarderrorsofmeanvaluesbetweenduplicate samples. between T and T , increased again between T and T , ment,andamaximalvalueof11.5nmolL−1 wasreachedat 6 12 12 18 and remained stable at ∼19nmolL−1 between T and T T . 18 24 24 (Fig. 4c). DMS concentrations were consistently higher in Insituandpotentialchangeratesofthesulfurcompounds theD-DMSP/OtreatmentthaninL-DMSP/O(Fig.4c).They duringtheincubationexperimentsarepresentedinTables5 first reached 15.6nmolL−1 at T , increased gradually to and 6, respectively. Changes in DMSP , and to a lesser ex- 6 d reach a peak value of 24.2nmolL−1 at T , and decreased tentDMSconcentrations,weregenerallynotlinearoverthe 18 slightlyto21.6nmolL−1 atT .NotethatdissolvedDMSO 24hincubationperiod,withmorepronouncedvariationsdur- 24 was not measured during this study due to methodological ingthefirst6h.Totakeintoaccountthisnon-linearity,both issues. hourlyratesmeasuredbetweenT andT andbetweenT and 0 6 6 In the Ice4-MP1 incubation, DMSP concentrations T ,aswellasdailyrates(T –T ),arepresentedintheseta- d 24 0 24 startedat3.0nmolL−1 inthecontrolandremainedcloseto bles. this value during the whole experiment (Fig. 4b). In the L- In Ice1-MP1, the concentrations of DMSP in the con- p DMSP/OandD-DMSP/Oamendedtreatments,DMSP con- trol decreased at a rate of 2.2nmolL−1d−1 (Table 5). We d centrations started at 87 and 96nmolL−1, respectively. As measured no change in DMSP during the first 6h, but a d observed in the previous melt pond, the concentrations de- positive net increase of 4.0nmolL−1 over the full 24h in- creasedto∼45nmolL−1atT andthenslowlydecreasedto cubation period was observed. In situ DMS changes in- 6 a value of ∼30nmolL−1 at T (Fig. 4b). DMS concentra- creased by 1.0nmolL−1h−1 during the first 6h and by 24 tionsinthecontrolofIce1-MP1startedat2.6nmolL−1 and 1.2nmolL−1d−1 over 24h. Potential net DMSP change d remained at this level during the 24h experiment (Fig. 4d). ratesof−11.6and−10.2nmolL−1h−1 weremeasureddur- IntheL-DMSP/Otreatment,DMSconcentrationsincreased ing the first 6h of incubation in L- and D-DMSP/O treat- moreorlesslinearlyfrom2.6nmolL−1atT to6.7nmolL−1 ments, respectively (Table 6). These rates became −1.2 0 at T . In the D-DMSP/O treatment, the increase in DMS and −0.6nmolL−1h−1 between T and T in L- and D- 24 6 24 concentrationswasmorepronouncedthaninthelighttreat- DMSP/O, respectively. Over 24h, negative potential net www.biogeosciences.net/15/3169/2018/ Biogeosciences,15,3169–3188,2018 3178 M.Gourdaletal.:Dimethylsulfidedynamicsinfirst-yearseaicemeltponds Table3.Reduced-sulfurcompoundconcentrationsmeasuredinsituinthemeltpondsandtheassociatedbiologicalcharacteristics(abundance ofhighnucleicacid(HNA)bacteria,Chlaconcentrations,andrelativeabundancesofmajortaxonomicgroups)ofthemeltpondwater. Station Melt Insitu Insitu Insitu Abundanceof Totalabundance Chla Dominantalgal pond DMSPp DMSPd DMS bacteria(HNA) ofalgae (µgL−1) group(%) (nmolL−1) (nmolL−1) (nmolL−1) (×109cellsL−1) (×106cellsL−1) Ice1 MP1 2.2 1.3 3.0 0.24 2.00 0.48 Unidentifiedflagellates(50%) Prasinophytes(ca.25%) MP2 2.0 1.4 3.1 0.40 Unidentifiedflagellates(55%) Prasinophytes(ca.25%) Ice2 MP1 1.8 d.l. 0.0 0.04 0.50 0.03 Unidentifiedflagellates(90%) Pennatediatoms(ca.28%) MP2 2.4 d.l. 0.0 0.09 Unidentifiedflagellates(50%) Pennatediatoms(ca.28%) MP3 2.3 d.l. 0.0 0.06 Unidentifiedflagellates(70%) Pennatediatoms(ca.28%) Ice3 MP1 2.0 d.l. 0.0 0.02 0.30 0.05 Unidentifiedflagellates(45%) Chrysophytes(29%) MP2 2.3 d.l. 0.0 0.04 Unidentifiedflagellates(55%) Chrysophytes(23%) Ice4 MP1 4.0 d.l. 2.6 0.15 1.00 0.18 Unidentifiedflagellates(50%) Pennatediatoms(20%) MP2 3.7 1.1 6.1 0.20 Unidentifiedflagellates(60%) Pennatediatoms(25%) Table 4. Spearman’s rank correlation coefficients between key in ments. Potential net DMS change rates remained low in situvariablesmeasuredinthemeltponds. bothL-DMSP/OandD-DMSP/Otreatmentsduringthefirst 6h of incubation with values at 0.1 and 0.3nmolL−1h−1, DMS Salinity Temperature Chla respectively. For the complete 24h incubation, potential ∗ ∗ net DMS change rates in light and dark reached 4.2 and DMS 0.84 0.51 0.84 8.9nmolL−1d−1,respectively. Salinity 0.40 0.56 Temperature 0.60 During both Ice1-MP1 and Ice4-MP1 incubation exper- iments, the light versus dark treatment had no effect on ∗indicatesa0.05significancelevel. the net changes in DMSP concentrations between the L- d DMSP/O and D-DMSP/O treatments (Wilcoxon signed- ranktest;n=8,df=3,α=0.05),butsignificantlyimpacted DMSP changeratesof∼−91nmolL−1and−71nmolL−1 d the rates of net accumulation of DMS (Wilcoxon signed- for the L-DMSP/O and D-DMSP/O treatments were cal- rank test; n=12, df=5, α=0.05). The accumulation of culated. Positive potential net DMS change rates of 1.6 DMS over 24h in the L-DMSP/O treatments were con- and 2.1nmolL−1h−1 were measured during the first 6h sistently and significantly lower than in the corresponding of incubation in L-DMSP/O and D-DMSP/O, respectively. D-DMSP/O treatments (Wilcoxon signed-rank test; n=8, For the complete 24h incubation, potential net DMS df=3, α=0.05). Based on the difference between the L- change rates reached 15.4nmolL−1d−1 in the light and andD-DMSP/Otreatmentsafter24h,weestimatedthelight- 18.6nmolL−1d−1inthedark. associatedDMSsinksat3.2nmolL−1d−1 inIce1-MP1and In Ice4-MP1, in situ DMSPp decreased at a rate of at4.7nmolL−1d−1inIce4-MP1(Table6). 1.9nmolL−1d−1 over the course of the incubation (Ta- ble 5). Meanwhile, in situ DMSPd changes rates were be- 3.4 IsotopicdiscriminationofDMSsources low the d.l. during the first 6h and almost null over 24h (Table 5). In situ DMS change rates were close to zero af- Table7showstheconcentrationsofDMSisotopes(m/z62) ter6handbelowd.l.after24h.PotentialnetDMSP change and(m/z68)after24hincubationinthethreetreatmentsand d rates of −8.1nmolL−1h−1 were measured during the first theirrelativecontribution(%)tothetotalDMSmeasuredat 6h of incubation in both L- and D-DMSP/O (Table 6). T .Asexpected,100%ofthetotalDMSinthecontrolsof 24 Theseratessloweddownto−0.5and−0.9nmolL−1h−1be- thetwoexperiments(3.0and2.3nmolL−1)showedtheiso- tweenT andT ,respectively.Overoneday,averagepoten- topicsignatureofnaturalDMS(m/z62).IntheL-DMSP/O 6 24 tial net DMSP change rates of ∼−59 and −62nmolL−1 treatmentoftheIce1-MP1incubation,78%(14.4nmolL−1) d were calculated for the L-DMSP/O and D-DMSP/O treat- of the DMS measured at T derived from D6-DMSP addi- 24 Biogeosciences,15,3169–3188,2018 www.biogeosciences.net/15/3169/2018/
Description: