A novel biomarker-based proxy for the spring phytoplankton bloom in Arctic and sub-arctic settings – HBI T25

cGeologicalSurveyofNorway,N-7491,Trondheim,Norway dNorwegianPolarInstitute,FramCentre,NO-9296Tromsø,Norway

a r t i c l e i n f o a b s t ra c t

Articlehistory:

Received5April2019

Receivedinrevisedform19June2019 Accepted26June2019

Availableonline19July2019 Editor: L.Robinson

Keywords:

springbloom phytoplankton proxy HBI diatoms Arctic

Thespring phytoplanktonbloomisacharacteristic featureofmid-highlatitudesinmoderntimes,but canbechallengingtoidentifyinpalaeorecords.Inthecurrentstudy,weinvestigatedtheabsoluteand relativedistributions oftwodiatom-derivedtri-unsaturatedhighlybranchedisoprenoid(HBI)lipids,at least oneofwhichhaspreviouslybeensuggested tobeapossible proxyfor theproductiveregionof themarginalicezone(MIZ)inthePolarRegions.Basedonacomparisonoftheirdistributionsinsurface sedimentsfromtheBarentsSeaandneighbouringregionswitharangeofoceanographicparameters,we identify,viaprincipalcomponent analysis,astrong associationbetweenthe relativeproportionofthe twoHBIsandsatellite-derivedspringchlorophylla(chla)concentration.Further,basedonagglomerative hierarchicalclustering,weidentifytwoclustersofHBIbiomarkerratiosandspringchlatogetherwith apotentialthresholdbiomarkerratio(termedHBITR25)forthespringphytoplanktonbloom.Amodified version of HBITR25 (i.e.HBI T25) providesapotentially more straightforward binary measure ofthe spring phytoplanktonbloom. Analysis ofHBI TR25 and HBIT25 values in aseriesof short(spanning recentcenturies)andlong(Holocene)sedimentcoresfromtheregionprovidesaninitialevaluationofthe applicabilityofthisnovelproxyinthepalaeorecord.Outcomesaremainlyconsistentwiththefindings fromthesurfacesedimentsandwithotherproxy-basedreconstructions,includingestimatesofpastsea icecover,whichiswell-knowntoinfluenceprimaryproductionintheregion.Indeed,wesuggestthat thenewHBIT25phytoplanktonbloomproxymayalsorepresentanimportantnewtoolforcharacterising theMIZinpalaeorecords,especiallywhenusedalongsidewell-establishedseaiceproxies,suchasIP25 andPIP25.Despitethelargelyempiricalnatureofthestudy,wealsoprovideapossibleexplanationfor theobservedbiomarkerratio-chlarelationship.Thus,apreviouslaboratoryinvestigationshowedthatthe distributionsofthesametwoHBIsanalysedhereinintheirlikelysource(viz.Rhizosoleniasetigera)was stronglyinfluencedbyculturetemperatureandgrowthrate.Confirmationofthegeneralityofourfindings and of the causal relationshipbetween HBI T25 and the spring phytoplankton bloomwill, however, requirefurtherlaboratory- andfield-basedstudiesinthefuture.

©^{2019TheAuthor(s).PublishedbyElsevierB.V.ThisisanopenaccessarticleundertheCCBY-NC-ND} license(http://creativecommons.org/licenses/by-nc-nd/4.0/).

1. Introduction

The spring phytoplankton bloom is a particularly characteris- ticandimportantfeature ofmid- to high-latitudesettings inthe northernhemisphere(Mahadevanetal.,2012andreferencescited therein). Relatively high photosynthetic light intensity combined with eddy-driven stratification and increased nutrient levels fol- lowingwinter verticalmixing, providethe necessary stimuli and growthconditionsforrapidphytoplanktondevelopment,suchthat

*

E-mailaddress:[email protected](S.T. Belt).

growthratescan outcompetethoseofgrazing. Asa consequence, phytoplanktonbloomscan contributesignificantlytoglobalfixing ofatmosphericcarbonanditssubsequentexportfromsurfacewa- ters.Highphytoplanktonproductivityisalsocriticalforthedevel- opmentandmaintenanceofprimaryconsumersandhighertrophic levelmarine ecosystems,moregenerally(Legendre, 1990; Søreide etal.,2010;Wassmanetal.,2006).

AsaresponsetorecentandrapidclimatechangeintheArctic andsub-arcticregions,andareductioninseaicecover,inpartic- ular,variouschangesto phytoplanktondynamicsarebeginning to emerge.Forexample,springblooms insub/low-Arcticregionsare developing earlierdue to a more rapid retreat of the productive https://doi.org/10.1016/j.epsl.2019.06.038

0012-821X/©^{2019TheAuthor(s).PublishedbyElsevierB.V.ThisisanopenaccessarticleundertheCCBY-NC-NDlicense} (http://creativecommons.org/licenses/by-nc-nd/4.0/).

Fig. 1.Structuresofhighlybranchedisoprenoid(HBI)biomarkersinvestigatedinthe currentstudy.

marginalicezone (MIZ),andtheproductive period,ingeneral, is lengtheningduetobothearliericeretreatinlatewinter/springand later freeze-up in late summer/autumn (Renaut et al., 2018 and referencescitedtherein).Further,inice-marginallocationssuchas theBarentsSeaandtheKaraSea,whichexhibit greatestsensitiv- itytomodernseaicechange(Lindetal.,2018andreferencescited therein), northward expansion of phytoplankton blooms (Renaut etal.,2018) andincreasedprevalenceofunder-icephytoplankton blooms have been reported,which likely resultfrom thinning of seaice, reducedprecipitation (snow) andan increase in the fre- quencyofopen-waterleadsbetweenicefloes(Arrigoetal.,2012).

Although such observations and possible attributions can be made through contemporary in situ measurements, deducing the same within palaeo records is much less straightforward to achieve, partly due to the challenge of finding suitable proxy measures, especiallyof thespringbloom, uniquely. Severalproxy methods for estimating past changes in overall marine primary productivityexist(seeRagueneauetal.,2000foranoverview),in- cludingthosebasedonelementalcomposition,stableisotopesand microfossil assemblages, although, aswith all proxies, each have their limitations. Biogenic silica can potentially more accurately reflectthedominanceofdiatomsandradiolarianscommonlyasso- ciatedwiththespringbloom,althoughdissolutionandoftenpoor sedimentarypreservationare limitations(Ragueneauetal., 2000).

Similarly, the accumulation rates of certain benthic foraminifera, known tobe opportunistic consumersof freshphytodetritus and thusapotentialproxymeasureofthespringbloom,maybenega- tivelyinfluencedbysignificantcarbonatedissolution,especiallyin highlatitudelocations(Polyaketal.,2013;Seidenkrantz,2013).

Certain source-specific lipids in marine sediments from high latitudesettings haveemerged asusefulpaleoceanographicprox- ies over the last decade or so. For example, the mono- and di- unsaturatedhighlybranchedisoprenoid(HBI)biomarkersIP25and IPSO₂₅ (Fig. 1) have been proposed as binary measures of sea- sonalsea iceintheArctic andAntarctic,respectively,a signature based on their selectiveproduction by certain sea ice-associated (i.e.sympagic) diatoms only (see Belt, 2018for arecent review).

Further, by considering the variable concentrations of IP25 and IPSO25 alongsidethoseofsomeopen-water(i.e.pelagic)biomark- ers,eitherindividuallyorintheformoftheso-calledPIP₂₅index (Müller etal., 2011), moresemi-quantitativeestimates ofsea ice conditionshavebeenproposed(Belt,2018).

In some recentstudies, a tri-unsaturated HBI lipid biomarker (oftenreferred to asHBIIII; Fig.1) hasbeen suggestedtorepre- sent asuitable open-watercounterpart toIP₂₅ andIPSO₂₅,partly due to its source-specific production by certain pelagic diatoms (Belt, 2018). Interestingly, based on water column and sediment datafromtheArcticandtheAntarctic,ithasbeensuggestedthat HBIIIImightrepresentausefulproxyfortheMIZ,withitselevated abundanceinsuchregions (Beltetal.,2015; Schmidtetal., 2018;

Belt,2018; Baietal.,2019) reflectingthemoregeneralfeatureof higher productivity commonly observed along the retreating ice

margin (Sakshaugetal.,2009; Wassmannetal., 2006).Moregen- erally, however, the establishmentof a robust proxy forthe MIZ remainsaninterestingresearchchallenge.

Despite theseprevious reports, there have been no dedicated studies aimed at identifying any quantitative relationship(s) be- tweenHBIIIIandotherwell-recognisedmeasuresofprimarypro- ductionsuch aschlorophylla (chla) orindeedanyotheroceano- graphic feature. Inthe currentstudy, we thereforecompared the distribution of HBI III inca. 200surface sedimentswith a range of modern-day oceanographic parameters, including sea surface temperature, salinity, water depth, sea ice concentration, photo- synthetically active radiation (PAR) andchl a. Here, we focus on the Barents Sea and neighbouring regions on the basis of well- documentedandcontrastingspringbloomdynamics,togetherwith the availability ofsuitable surface anddowncore sediment mate- rial. We also considered biomarker-based estimatesof springsea ice concentration (SpSIC; Smik et al., 2016) due to its influence over seasonal phytoplankton dynamics, andthe distribution ofa geometricisomerofHBIIII(HBIIV;Fig.1),notleastbecauseHBIs IIIandIVareoftenco-producedby certaincommondiatoms(e.g.

Rhizosolenia setigera; Rowland et al., 2001), with HBI IV having been shown recently to be a usefulpredictor of sea ice classifi- cation inthe Barents Sea whenused alongsideIP25 (Köseo˘gluet al.,2018a).

Having identified a strong relationship between the relative proportionsofHBIsIII andIV(butnotthe individualbiomarkers) andspringchla,butnoothermeasuredparameter,wethenmea- suredthesamerelativebiomarkerdistributioninaseriesofshort cores spanning recent centuries and longer (early-late Holocene) downcore recordsfrom theregion. Our findings suggest that the proportion of HBIs III and IV in marine archives may provide a proxymeasureofthepastoccurrence(orotherwise)ofthespring phytoplanktonbloom,atleastfortheBarentsSeaandneighbour- ingregions.Onthebasisofanearlierlaboratoryinvestigationinto thedistributionsofHBIs(includingIIIandIV)inthecosmopolitan pelagicdiatomR. setigera,we alsosuggest apossibleoriginofthe proxyrelationshipbetweenHBIsIIIandIV, andthespringphyto- planktonbloom.

2. Regionalsetting

DetaileddescriptionsofBarentsSeaoceanographycanbefound inLoeng (1991).Inbrief,theBarentsSeaischaracterisedbythree distinct water masses (Fig. 2a): northward inflow of warm and saline Atlantic Water (AW) via the North Atlantic Current (NAC), which continuesfurther northasthe North CapeCurrent (NCaC) andtheWestSpitsbergenCurrent(WSC),fresherandcolderArctic Water (ArW) flowingsouthwest via the EastSpitsbergen Current (ESC) andthePerseyCurrent(PC),andbrackishcoastal waterto- pographicallysteeredalongtheNorwegiancoastbytheNorwegian CoastalCurrent(NCC)(Sakshaugetal.,2009).Thenorthernregion oftheBarentsSeaalsoexperiencesseasonalseaicecover,reaching itsmaximumextentinMarch–April;however,inter-annualfluctu- ationscanbelargeduetovariableinflowofAW(Smedsrudetal., 2013).Overall,seaiceintheBarentsSeahasdecreasedby>50%in thelast40yr orso(Fettereretal.,2016),anegativetrendthathas likelyexistedsince1850AD(DivineandDick,2006).Theregionis almost entirelyice-freeattheSeptember seaiceminimum,while the positionof themaximum wintericemargin is importantfor defining the highly productive MIZ (e.g. Wassmann et al., 2006).

TheadvectionofAWalsocontributestolongerproductiveseasons comparedtootherArcticareas,makingtheBarentsSeaoneofthe mostproductive areasofthe ArcticContinental Shelf (Wassmann etal.,2006andreferencescitedtherein).

Fig. 2.MapsoftheBarentsSeashowing:(a)Labelledcentennial(blackdiamonds)andmillennial(whitesquares)downcorerecords,aswellasasimplifiedrepresentationof AtlanticWater,ArcticWater,andCoastalWatersurfacecurrentsshownbyred,blue,andwhitearrows,respectively.Abbreviationsdenote:WSC–WestSpitsbergenCurrent;

NAC–NorthAtlanticCurrent;NCaC–NorthCapeCurrent;NCC–NorwegianCoastalCurrent;ESC–EastSpitsbergenCurrent;PC–PerseyCurrent;(b)Surfacesediment locations.Forbothmaps,thesolidblacklineillustratestheaveraged(1988–2017)April–Juneseaiceextent,definedbya15%SpSICthreshold.Mapsweregeneratedwith OceanDataView(http://odv.awi.de/).(Forinterpretationofthecoloursinthefigure,thereaderisreferredtothewebversionofthisarticle.)

3. Materialsandmethods 3.1.Surfacesedimentmaterial

198 surface sediment sub-samples were taken from a range of grab samples, multicores, box cores and gravity cores reflect- ing regions of variable sea ice cover and seasonal primary pro- ductivity (Fig. 2b). All surface sediments are assumed to repre- sent recentdeposition, asdescribed previously (Belt etal., 2015;

Smiketal., 2016; Köseo˘glu etal., 2018aand referencestherein).

Sampling locations, coretypes,biomarkerdata andvariousphys- icalparameters usedfor thecalibration component ofthisstudy canbefoundinSupplementaryTable 1.

3.2.Downcoresedimentmaterial

Downcoredata spanning recent centurieswere obtained from six short sediment cores (Fig. 2a) described in detail elsewhere (Vare et al., 2010; Dylmer, 2013; Cabedo-Sanz and Belt, 2016;

Köseo˘glu et al., 2018a). In brief, cores BASICC 1, BASICC 8, and BASICC 43, hereafter referred to as cores 1, 8, and 43,were re- covered aboard the RV IvanPetrov as part of the ‘Barents Sea IceEdgeina ChangingClimate’ (BASICC)project(Cochraneetal., 2009).Weusedtheagemodelsgivenelsewhere(Vareetal.,2010).

CoreMSM5/5-712-1(hereafter,core712)wascollectedaboardthe RVMariaS.Merianduring the MSM5/5cruise andtheagemodel isbased on five ¹⁴C Accelerated Mass Spectrometry (AMS) dates (Spielhagenetal., 2011).MulticoresR248MC010 andR406MC032 (hereaftercores10and32,respectively)wereretrievedwithinthe framework of the MAREANO programme (www.mareano.no) on- board F/F G.O.Sars, with chronologies based on ²¹⁰Pb data (see Dylmer,2013andreferencescitedtherein).

Longer timeframe data were obtained from gravity cores de- scribed previously (Laberg et al., 2002; Dylmer, 2013; Berben et al.,2014,2017)(Fig.2a).GravitycoreWOO/SC-3(hereaftercore3) was retrieved fromthe Norwegian continentalmargin (Laberg et al.,2002).Theagemodelisbasedonthree¹⁴CAMSdates(Laberg etal., 2002; Dylmer, 2013) andthe analysed section corresponds tothelast ca.3.0cal kyrBP.Core JM09-KA11-GC(hereafter, core

11), was obtained fromthe Kveithola Trough,south ofSvalbard, aboardRV JanMayen.Weusethe agemodelpresentedinBeltet al. (2015),basedon ¹⁴CAMSdatesfrompreviousstudies(Berben et al., 2014 and references therein). Gravity Core NP05-11-70GC (hereafter, core70)was collected from theOlgaBasin, EastSval- bard,aboardtheRVLance.Corechronologyisbasedonthree¹⁴C AMS dates(Berben etal.,2017).Forcores11and70,wepresent data coveringlast ca. 9.5cal. kyr BP.See Table 1fora summary ofallcoresandSupplementaryTable2formoredetails regarding corechronologies.

3.3. Biomarkerdata

Biomarker data were obtainedin two ways. Forcores not in- vestigated previously (i.e.cores 3, 10and 32), lipid analysiswas carried out according to Belt et al. (2012), but with a slight modificationtotheextractionmethod.Thus,freeze-driedsubsam- ples (ca.1.5–2.5g)were saponifiedinamethanolic KOHsolution (ca. 5 mL H₂O:MeOH (1:9); 5% KOH) for 60 min (70^◦C). Hex- ane(3 × ^{2mL) was addedto the saponified content,withnon-} saponifiablelipids(NSLs)transferredtocleanvialsanddriedover N₂. NSLs were then re-suspended in hexane (0.5 mL) and frac- tionated using column chromatography (SiO2; 0.5 g). Non-polar fractionscontainingHBIswereelutedwithhexane(6mL)andpu- rified further using silver-ion chromatography(Belt et al., 2015).

Saturated compounds were eluted with hexane (2 mL) and un- saturatedcompounds,includingHBIs IIIandIV,were collected in a subsequent acetone fraction (3 mL). Prior to extraction, sam- pleswerespikedwithaninternalstandard(9-octylheptadec-8-ene, 9-OHD, 10 μL; 10 μg mL⁻¹) to permit quantification. Analysis of purified fractions containing HBIs III and IV was carried out us- ing gas chromatography–mass spectrometry(GC–MS) intotal ion current (TIC) and selected ion monitoring (SIM) modes (Belt et al., 2012). HBIs were identified based on their characteristic GC retention indices (RIHP5MS = 2081, 2044 and2091 for IP25, HBI III and HBI IV, respectively) and mass spectra (Belt et al., 2000;

Belt, 2018). HBI quantification was achieved by comparison of massspectral responsesofselectedions(e.g.IP₂₅,m/z 350;HBIs III and IV, m/z 346) in SIM mode with those of the internal

Table 1

Summaryofcorelocations,waterdepthsandagemodelmethodsforallthecoresdescribedinthestudy.Furtherinformationabouttheindividualagemodelscanbefound inSupplementaryTable2.

Core ID Short ID Time interval Latitude

(^◦N)

Longitude (^◦E)

Waterdepth (m)

Age model method

R248MC010 10 Recent centuries 70.31 12.88 1254 ²¹⁰Pb (Dylmer,2013)

R406MC032 32 Recent centuries 72.32 15.38 1035 ²¹⁰Pb (Dylmer,2013)

BASICC 1 1 Recent centuries 73.10 25.63 425 ²¹⁰Pb (Vare et al.,2010)

BASICC 8 8 Recent centuries 77.98 26.79 135 ²¹⁰Pb (Vare et al.,2010)

BASICC 43 43 Recent centuries 72.54 45.74 285 ²¹⁰Pb (Vare et al.,2010)

MSM5/5-712-1 712 Recent centuries 78.92 6.77 1491 ¹⁴C AMS (Spielhagen et al.,2011)

WOO/SC-3 3 Last ca. 3.0 kyr BP 67.40 8.52 1184 ¹⁴C AMS (Dylmer,2013)

JM09-KA11-GC 11 Last ca. 9.5 kyr BP 74.87 16.48 345 ¹⁴C AMS (Belt et al.,2015)

NP05-11-70GC 70 Last ca. 9.5 kyr BP 78.40 32.42 293 ¹⁴C AMS (Berben et al.,2017)

standard(9-OHD,m/z 350)andnormalizedaccordingtotheirre- spectiveinstrumentalresponsefactors(Beltetal.,2012).Forcores analysedpreviously,weusedthedatareportedbyKöseo˘gluetal.

(2018b). Theproportions ofthe twotri-unsaturated HBIs (IIIand IV)intheformofanHBItrieneratio(HBITR25)andare-arranged version of this (HBI T25) were calculated according to Eqs. (1) and (2).

Biomarker-based spring sea ice concentration (%SpSIC) esti- mates(andtheirroot-mean-squareerrors(RMSE))wereeitherob- tainedfromthenewbiomarkerdatasets (i.e.forcores3,10and 32)basedontherelativeconcentrationsofIP₂₅ andHBIIII anda regional calibration(Eqs. (3) and (4);Smik etal., 2016), or have beenreportedpreviously usingthesameapproach(Berben etal., 2017; Köseo˘gluetal.,2018b).SquarebracketsdenoteabsoluteHBI concentrations(ng g⁻¹ dry sed.).All downcore biomarker related datacanbefoundinSupplementaryTable 3.

HBI TR₂₅

=

(

+

)

HBI T25

=

0

.

PIIIIP25

=

(

+

×

.

)

(

) = (

−

.

)

0

.

3.4. Oceanographicdata

Sea ice concentration data were obtained from Nimbus-7 SMMR and DMSP SSM/I-SSMIS databases on a 25 × ^{25 km} grid (Cavalieri et al., 1996). Data from the Aqua satellite (NASA, https://oceancolor.gsfc.nasa.gov/data/aqua/) equippedwithaMod- erateResolution ImagingSpectroradiometer(MODIS)was usedto retrieve chlorophyll a (chl a; mg m⁻³), particulate inorganiccar- bon (PIC; mol m⁻³), photosynthetically available radiation (PAR;

E m⁻²d⁻¹), and sea surface temperatures (SST;^◦C). Sea surface salinity(SSS;psu;0–30 mwaterdepth)was obtainedfromWorld Ocean Atlas 2013 (https://www.nodc.noaa.gov/OC5/woa13/) on a 25 × ^{25 km grid. Monthly aggregates throughout} April–August were created (chl a only), as well asthose spanning April–June and July–September (all data). Daily-resolution chl a time series spanning 2003–2017 were also created to showcase differences betweenareasofcontrastingspring(ca.April–June)phytoplankton productivityinthe BarentsSea.Temporally-averaged(2003–2017) annual maximum concentration of chl a, and the timing of its occurrence(dayofyear),werealsoderived.Thepercentagediffer- encesbetween successive8-daily averaged chl a (mg m⁻³) span- ningyears2003–2017werecalculatedusingEq. (5),wherechla istherelativedifference(in%)betweenan initialandsubsequent

8-day chl a compositeatthe samelocation,labelledchl a_reference andchlacurrent,respectively.

(

) = (

−

)

chl a_reference

×

3.5. Statisticalanalysis

To explore associations between the various datasets and be- tweentheHBIdistributionsandsatellite-derivedchladata,inpar- ticular, PrincipalComponent Analysis (PCA)andcomplete-linkage Agglomerative Hierarchical Clustering (AHC) using squared Eu- clideandistancewerecarriedout usingXLSTAT(Addinsoft,2018).

Morespecifically,PCAwasusedtoreducethehigh-dimensionality dataset of HBI concentrations,PIIIIP25, TR25, satellite-derivedand other variables in surface sediments for visualisation on a two- dimensionalgrid,wheretheproximityandmagnitudeofvariables indicatedtheirdegreeofassociation.Thus,satellite-derivedparam- eters strongly associated with TR25 according to PCA were cho- sen andindividually processedvia AHCtodetermine theoptimal numberandcompositionofclusters,aswell astheir similarityto thoseobtainedusingTR25data.TheAHChelpeddetermineasin- gle satellite-derived parameter mostcloselyassociated withTR₂₅ insurfacesediments.

4. Results

4.1. DistributionofHBIsIIIandIVinsurfacesediments

HBI IV could be quantified in virtually all surface sediments, consistentwiththepreviousidentificationofnear-ubiquityofHBI III in the same sediments (Köseo˘gluet al., 2018a) and their co- production bycertain marine diatoms (Rowlandetal., 2001;Belt etal.,2000,2017).ThedistributionsofIIIandIV,whenexpressed asindividualbiomarkerconcentrations,werebothsomewhat het- erogeneous (Fig. 3a, b); however, although spatial variability in therelative amountsofthetwoHBIs (i.e.HBITR25 (Eq.(1)))was alsoevident,generallyhighervalueswereobservedforsitesinthe easternregioncomparedtothoseinthewest(Fig.3c).

Based on PCA (Fig. 4), we found no associations between the sedimentary concentrations of HBIs III or IV with any of the oceanographicparameters considered,includingchla.Incontrast, HBITR25exhibitedastrongassociationwithchla,butmainlydur- ing April and May (i.e. during the spring phytoplanktonbloom).

AHC analysisbetween HBI TR₂₅ and chl a resulted in two clus- terswithinareasofwell-definedspringbloomseasonalityandless productive regions characterised by strong Atlantic Water inflow and continuous upwelling (Fig. 5). Clustering was dependent on the month(s) selected for chl a data (i.e. April, May, April–May, April–June),withtheApril–Mayaggregateexhibitingtheleastmis- matched cluster memberships (n = ^{28) relative to those of HBI} TR25(Fig.5).Incontrast,thenumberofmis-matchesfortheother

Fig. 3.DistributionsofabsoluteandrelativeHBIbiomarkerconcentrationsinBarentsSeasurfacesediments:(a)HBIIII;(b)HBIIV;(c)HBITR25.The15%SpSICcontour (1988–2017)isshownbyablackline.MapsweregeneratedwithOceanDataView(http:/odv.awi.de/).(Forinterpretationofthecoloursinthefigure,thereaderisreferred tothewebversionofthisarticle.)

Fig. 4.Scaledfactorloadingsofprimary(greenmarkers)andsecondary(bluemark- ers)variableswithHBITR25 atsurfacesedimentlocations(Fig.2b).Greenlabels denotemonthsofaveraged(2003–2017)satellite-derivedchla(mg m⁻³).Bluela- belsrepresentsurfacesedimentwaterdepths(m),average(2003–2017)seasurface temperature(SST;^◦C), photoavailableradiation(PAR;E m⁻²d⁻¹),particulateinor- ganiccarbon (PIC;mol m⁻³),1955–2012seasurfacesalinity (SSS;psu), aswell as1988–2017seaiceconcentration(SIC;%);prefixes“Su”and“Sp”denotesum- mer(July–September)andspring(April–June).AbsoluteconcentrationsofIIIandIV (Fig.1)andHBITR25arehighlightedinred.(Forinterpretationofthecoloursinthe figure,thereaderisreferredtothewebversionofthisarticle.)

months ranged from 30 to 57 (Supplementary Fig. 1). Averag- ingtheAHC centroidsusingthe April–Mayaggregatedchla data yieldedan approximate thresholdvalue forHBI TR₂₅ of ca. 0.62

±^{0.02 toseparateregionsofhigh(i.e.HBITR}25≥^0.62)andlow (HBI TR₂₅ < 0.62) April–May chl a delineated by a 1.5mg m⁻³ boundary.

4.2.HBIbiomarkersindowncorerecords

Forcoresrepresenting recentcenturies, thesea icebiomarker IP₂₅ was absent (or below detection limits) in cores 1, 10 and 32 (Supplementary Fig. 2; Köseo˘glu et al., 2018a) as expected

Fig. 5.Mapofaverage chladuringApril–May(2003–2017).Blackandwhitecir- clemarkersrepresentthetwoAHCclustersofTR25insurfacesediments.Diagonal crossesdenoteHBITR25 clustermembershipswhichmis-matchthoseofthechla data(n=^{28).The1.5mg m−3contourforchla}(2003–2017)isshownasawhite line,andthe15%April–Mayseaiceconcentrationcontour(1988–2017)isindicated byablackline.MapsweregeneratedwithOceanDataView(http:/odv.awi.de/).(For interpretationofthecoloursinthefigure,thereaderisreferredtothewebversion ofthisarticle.)

due to their ice-free settings in modern times (Fig. 2a). Fur- ther, HBIs III and IV were present in virtually all horizons in each core (with the exception of the early part of the record in core 32; Supplementary Fig. 2), consistent with our findings from proximal surface sediments described herein (Fig. 3a, b).

In contrast, IP25 was identified in the three cores from sites of seasonal sea ice cover (i.e. cores 8, 43, 712; Vare et al., 2010;

Cabedo-Sanz andBelt, 2016; Köseo˘gluet al.,2018a), andHBIsIII andIVwereagainpresentthroughout,albeitinvariableconcentra- tions(SupplementaryFig.2).HBITR25alsoexhibitedsomespatial variability, with values broadly reflecting those found in nearby surface sediments (Figs. 3c, 6a). Thus, relatively low (i.e. <0.62) HBITR₂₅valueswereobservedthroughouteachofcores8,10,32 and 712,all of which are located in regions of low spring chl a inmodern times.Similarly, consistentlyhighHBITR25 values(i.e.

Fig. 6.HBI-derivedproxydatafromsixshort-corerecordswithinthestudyregion,spanningrecentcenturiesandexperiencingcontrastingseaiceandphytoplanktonbloom occurrenceinmoderntimes:(a)HBIT25andHBITR25.Thebinarythresholdforthemodern-dayspringphytoplanktonbloom(i.e.HBIT25>1)(Figs.5,9),isrepresentedby asolidhorizontalline.Theshadedarearepresentstheestimatederrorinthisthreshold(HBITR25ca.±^0.02;HBIT25ca.±^{0.03).Thecoloursofeachdatapointrepresent} theproposedoccurrenceofaspringphytoplanktonbloom(orange)versusnobloom(blue)ateachcoresite/timeslice;(b)springseaiceconcentrationestimates(%SpSIC).

(Forinterpretationofthecoloursinthefigure,thereaderisreferredtothewebversionofthisarticle.)

>0.62)characterisecore43,locatedinaregionofhighspringchl aadjacentto themodernwinterseaicemargin (Fig.2a). Incon- trast,HBITR25 valuesbothabove andbelow0.62wereevidentin core1(Fig.6a).

In the longer timeframe records(i.e. cores 3, 11 and 70), in- dividualbiomarkerconcentrationsandHBITR25 valueswere also variable.Forexample,IP25 wasnot identified incore3,although HBIsIIIandIVwerepresentthroughoutthelastca.3.0cal kyrBP (SupplementaryFig. 3).Consistentwithitslowspringchlasetting inmoderntimes,HBITR25wasalso<0.62throughout(Fig.7a).For core11IP₂₅andSpSICwere low duringthe early–midHolocene, withincreasesinbothtonear-modernvaluessinceca.1.1calkyr BP,as reportedpreviously (Berben et al., 2014; Belt etal., 2015) (Fig.7b).HBITR₂₅was low(<0.62)intheearly Holocene,before increasing to ca. 0.62around 6.0 cal. kyr BP, andthen to values consistently >0.62 after ca. 1.1 cal. kyr BP, coincident with in- creases to IP25 andSpSIC estimates (Fig. 7b). Finally, progressive increasestoIP25 andSpSICfromthe earlyto lateHolocenechar- acterisethecore70site,asdescribedpreviously(Beltetal.,2015;

Berben et al., 2017) (Fig. 7c). HBIs III and IV were also present throughout the Holocene (Supplementary Fig. 3), with HBI TR₂₅ values mainly greater than 0.62; however, slightly higher values

were observed ca.9.0–6.0 cal kyr BP,whilesome values closeto the0.62thresholdwereevidentthereafter(Fig.7c).

5. Discussion

5.1. UseofHBITR25andHBIT25asproxiesforthespringphytoplankton bloom

The spatially variable proportion of HBIs III and IV, albeit on a somewhatsmallersampleset,was previouslysuggestedtopos- sibly reflect the spatial distribution of Atlantic Water (AW) and Arctic Water (ArW) inthe region (Navarro-Rodriguez, 2014). The defining characteristicsofbothwatermassesincludetemperature andsalinity(e.g.Loeng,1991; Sakshaugetal.,2009).However,we observednoassociationbetweenanyofthespring-summersatel- lite sea surface temperature (SST),photoavailable radiation (PAR) orseasurface salinity (SSS)recordsandHBITR25 insurface sed- iments presentedherein (Fig. 4), which suggeststhe influenceof these is either absent or obscured by competingeffects. In con- trast,chla data,asanindicatorofstandingphytoplanktonstocks, showed a strong correlation with HBI TR₂₅, but only during the springbloom(i.e.April–May).Thiswasfurthersupportedbysimi-

Fig. 7.HBI-derivedproxydataforHolocenerecords:(a)3;(b)11;(c)70.Left-handaxes:HBITR25andHBIT25representedbysolidlineprofileswithcolouredmarkers.The solidhorizontallineindicatesthethresholdsforthespringphytoplanktonbloom(Figs.5,9),whiletheshadedarearepresentstheestimatederrorinthisthreshold(HBI TR25ca.±^{0.02;HBIT25 ca.}±^{0.03).Thecolourofeachdatapointrepresentstheproposedoccurrenceofaspring}phytoplanktonbloom(orange)versusnobloom(blue)at eachcoresite/timeslice.Right-handaxes:%SpSICestimatesrepresentedbyadash-dottedlinetogetherwithRMSEestimates(ca.±^{11%;Smiketal.,2016).(For}interpretation ofthecoloursinthefigure,thereaderisreferredtothewebversionofthisarticle.)

larclustering(AHC)ofHBITR25andchlaforApril–Mayand(toa lesser extent) May only, likely due to the high spatio-temporal variability of phytoplankton bloom development in the Barents Sea.

Driven mainly by the spring phytoplankton bloom in April- May(Fig. 8a), maximum annualchl a (Fig. 8c) is higheston the highly-productive south-eastern and central shelves, reaching its maximumgenerallyca.1–2monthsearlier(Fig. 8d)comparedto thewestern andnorthernBarents Sea(Fig. 8c).Thehighestrates ofchange inchl a,a further characteristic ofa bloom event, are also most apparent along the south-eastern and central shelves (Fig. 8b). Along the western margin, slower (thermally-induced) verticalstratificationandcontinuousAWupwellinghinderphyto- planktonaccumulation,whileinsufficientlightpenetrationthrough thick icecoverlowerspelagic productionatthe northern margin (e.g.Dalpadadoetal.,2014).Thus,inApril,onlytheice-freesouth- eastern Barents Sea shows significant increases in chl a (Sup- plementary Fig. 1a), followed by a propagation, north-eastwards along the retreating sea ice edge, by early May (Supplementary Fig.1b).Phytoplanktonbiomass sharplydeclinesbyJune (Fig. 8a) due to nutrient (e.g. nitrate and silicate) depletion and limited replenishment through the meltwater-established pycnocline in the marginal ice zone (MIZ) (e.g. Signorini and McClain, 2009;

Leu et al., 2011), with subsequent summer blooms dominated by coccolithophores (Hopkins et al., 2015), which are not HBI- producers.

Thus,HBITR₂₅appearstobemostrepresentativeofthepelagic spring bloom throughout April–May. More specifically, HBI III is most prevalent (HBI TR25 ≥ ^{0.62) in the} eastern/central Barents Sea, wherechl a is mainly in excess of 1.5 mg m⁻³. In contrast, relatively increasedIV(HBITR25 = ^{ca.0.4–0.45) generallyoccurs} inthewesternBarentsSea,wherechla concentrationsaregener- allyintherange0.5–1.5 mg m⁻³.Furthermore,bloomseasonality isnotaspronouncedinthewestern BarentsSeacomparedtothe eastern Barents Sea (Fig. 8a). Similarly, low HBI TR25 is alsoevi- dent in extensivelyice-covered areasnorth andeast ofSvalbard, wheretheproductiveseason istime- andnutrient-limiteddueto thelateseasonalseaiceretreatthroughoutJuly–August(Signorini andMcClain, 2009).ThisfurthersupportsoursuggestionthatHBI TR25 ispredominantly influencedby springphytoplanktonbloom developmentintheBarentsSea.

Finally,we suggest that theHBI TR₂₅ thresholdforthe spring phytoplanktonbloom (i.e.HBI TR25 ≥ ^{0.62)is most}conveniently expressed as a simple binary measure using a slightly modified ratioofthetwoHBItrienes.Thus,HBIT₂₅≥^{1(Eq. (2))providesa} proxymeasureforthespringphytoplanktonbloom(Fig.9).

Fig. 8.Average(2003–2017)satellite-derivedchladatawithinthestudyarea.Upperpanelshowsthetemporalevolutionofmeanchlaforregionswherethereisapresence (greendiamonds)orabsence(redcircles)ofsignificant,diatom-dominatedspringbloomsaccordingtoa1.5mg m⁻³April–Maychlathreshold:(a)dailychlaconcentration;

(b)relativechangesinchla.Lowerpanelshowsthetemporally-averaged(2003–2017)annualchlamaximum(c)andthedayofitsmaximumoccurrencewithintheannual cycle(d).MapsweregeneratedwithOceanDataView(http:/odv.awi.de/).(Forinterpretationofthecoloursinthefigure,thereaderisreferredtothewebversionofthis article.)

Fig. 9.DistributionofHBIT25 insurfacesedimentsoverlaidontoremotely-sensed April–May(2003–2017)chla.Thewhitelinerepresentsa1.5mg m⁻³chlacon- tour and delimits zones of high and lower pelagic phytoplankton productivity during the spring bloom (Fig. 5c). Mapwas generated with Ocean Data View (http:/odv.awi.de/).(Forinterpretationofthecoloursinthefigure,thereaderisre- ferredtothewebversionofthisarticle.)

5.2. HBITR₂₅andHBIT₂₅inrecordscoveringrecentcenturies

The HBITR₂₅ andHBI T₂₅ datafor the sixshortcores (i.e. 1, 8,10, 32,43 and712) representingrecent centuriesreflect their

respective locations and the occurrence of spring phytoplankton blooms (orotherwise) withinthemodern context(note:we refer onlytoHBIT25valuesfromhereon).Thus,relativelylow(<1)HBI T25 values prevail throughout the 10 and32 records, consistent withlowchlaattheseice-freelocations(Figs.2a,6a).

Similarly,HBIT25incore712wasconstantlybelowthethresh- oldforaspringphytoplanktonbloom.Thecoresiteischaracterised bylowchlainmoderntimes,andislocatedatthelargelyice-free westernSvalbardmargininfluencedbythestrongestinflowofAW with the North AtlanticCurrent (NAC; Fig. 2). The contemporary iceedgedurationatsite712islimited,andstratificationnecessary forrapidspringbloom developmentis weakerduetocontinuous AW overturning (Smedsrud et al., 2013). However, instrumental recordsshow thaticecoverattheSvalbardmargin wasmoreex- tensivepriortoca.1850AD(DivineandDick,2006),supportedby the PIIIIP25-basedestimatesofSpSICreportedpreviously (Fig. 6b;

Cabedo-Sanz andBelt, 2016). Interestingly, a gradual decrease in HBI T25 at site 712, possibly indicative of a lower frequency of spring phytoplankton blooms, also coincides withthe recent sea icedecline(Fig.6a,b).Accordingly,increasedphytoplanktonstocks atsite712priorto1850ADcouldbeattributabletolongerannual seaiceduration,whenincreasedstratificationpotentiallystabilised phytoplankton in the photic zone, facilitating the type of rapid growthnormallyassociatedwiththecontemporaryMIZinthecen- tral Barents Sea. Recentincreases in AWinflow andatmospheric temperatures (e.g. Årthun et al., 2012) subsequently shifted the Barents Seatowards less productive, predominantlyice-free con- ditions dominated by continuous upwelling, with lower HBI T25 (Fig.6a).Previously,Pathiranaetal. (2015) alsolinkedreducedMIZ durationintheBarentsSeatodecreasingprimaryproductivityover thelastca.500 yr.

Core 8 exhibits similar HBI T25 valuesto core 712,but is lo- cated in a significantly differentsetting of increased (>80%)Sp- SIC north of the central Barents Sea MIZ, and influenced pre-

duringthemeltseason.Conversely,theabsenceofseasonalseaice atsite1mayhavereducedproductivitysomewhat,asindicatedby slightlylowerHBIT₂₅values(Fig.6a).Infact,thesomewhatoscil- latory(eithersideof1)patternofHBIT25likelyreflectstheclose proximityofthecoresitetotherecent(2003–2017)springphyto- planktonbloomboundary,withshort-termvariabilityoverdecadal (orshorter) timeframesduring recentcenturies.This possiblyre- flects the variable influence of the North Cape Current (NCaC), sinceintensifiedAWupwellingcouldhavereducedthestabilityof thewatercolumnatthecoresite,resultinginlowerproductivity.

5.3.HBIT₂₅inHolocenerecords

InordertomakeafirstassessmentofthereliabilityoftheHBI T25proxymeasureofthespringphytoplanktonbloomoverlonger timeframes,we measured it in three early-late Holocenerecords fromregions ofcontrastingsea iceandphytoplanktonbloomoc- currence in modern times,and for which evidence for temporal changesin oceanographyhadalreadybeen established frompre- vious proxy-based investigations. The shortest of these records, obtained from core 3, located in the SW Barents Sea, adds to thefindingspresentedearlierforsites32and10spanningrecent centuries(Fig.2a).Thus,consistentlylowHBIT25incore3charac- terisesthisperenniallyice-freeregion withlowspringchl a over thelastca.3.0cal.kyrBP(Fig.7a).

Thecore 11site isproximal to the modern maximum winter seaice margin (Fig. 2a). During the Holocene, HBIT25 gradually increasedsinceca.9.5cal kyrBP,withvaluesgenerallyexceeding thespringbloomthresholdsincetheonsetoftheNeoglacialatca.

6.0cal kyr BP (Fig. 7b).According to previous investigations, the core11 site was relatively ice-free during the Holocene (Berben et al., 2014; Belt et al., 2015); however, a highly-productive ice edgelikely remainedcloseto theKveithola Troughfollowing the Neoglacial ice advance, as previously suggested for Storfjorden (Knies et al., 2017), located slightly further north. High produc- tivity fuelled by seasonal sea ice-induced stratification and AW upwelling could havepropagated towards thecore 11 siteatca.

6.0 cal kyr BP, thus promoting the occurrence of spring phyto- planktonbloomsasamorefrequentfeature.Finally,high(>1)HBI T25 inthe late Holoceneis consistent withspringphytoplankton blooms associated with the productive ice edge having reached the core site at ca. 1.1 cal kyr BP (Fig. 6b; Berben et al., 2014;

Belt et al., 2015), a conclusion supported further by a produc- tivityincrease inferred from higherbenthicforaminiferal content (Berbenetal.,2014).

Incontrasttocores3and11,core70issituatedatasiteofex- tensivewinterseaicecoverinmoderntimes.However,conditions duringtheearlyHolocenewerelesssevere,suchthatthesitewas proximal tothe wintericemargin untilca. 6.0cal kyrBP (SpSIC

biomarker-based HBI T₂₅ parameter described herein provides a qualitativeproxyindicatorfortheoccurrenceofspringphytoplank- tonbloomsacrossthestudyregionfromrecenttoHolocenetime- frames.Intuitively,thisassociationisperhapsnotsurprisinggiven that HBIs are produced by the main constituents of the spring phytoplanktonbloom (i.e. diatoms),although thisalone doesnot provide adequate explanation for the observed relationship. For example, HBIs III and IV are present at virtually all study sites, irrespective of theoccurrence (or not)of aspring phytoplankton bloom.Further,whileabsolutesedimentaryconcentrationsofHBIs IIIandIVshowasignificantenhancementwithintheMIZ,theyare relativelylow insome otherregionsofhighchla (e.g.intheice- free SE Barents Sea; Fig.3) andare poorly associated withchl a (Fig.4),moregenerally.Thelatterispotentiallyattributabletothe increasedprevalenceofdiatomsrelativetoothermicroalgaecloser to the well-stratified waters near the ice edge (e.g. Wassmann et al., 2006; Sakshaug et al., 2009). Absolute biomarker concen- trations are also influenced by sediment accumulation rates and export efficiency from the water column, both of which can be variable,spatially,andthusmaynotaccuratelyreflectproduction, moregenerally.Ontheotherhand,suchinfluencesareoftenmuch lessimportantwithratio-basedmeasures.

In any case, in order to rationalise the association between HBIT25 andthespringphytoplanktonbloom,we briefly consider threepossiblecontributingfactors: (i)differentsources ofHBIsIII andIV;(ii)differentialbiomarkerdegradation;(iii)variablephyto- planktongrowthrates.

First,HBIsIIIandIVareamongstthemostcommonHBIsfound inmarinesediments(Beltetal., 2000),yetveryfewsources have been identified.Ofthese, Rhizosoleniasetigera andrelatedspecies are byfarthe mostcosmopolitanandabundant,andHBIsIII and IV haveindeed been identified in such speciesfrom the current studyregion(Beltetal.,2017).Incontrast,althoughHBIsIIIandIV havebeenreportedinthebenthicdiatomPleurosigmaintermedium (Belt etal., 2000), they have not beenidentified inother marine Pleurosigmaspp.,whichare,inanycase,generallyveryloworab- sentintaxonomicinventories.ThemorecommonBerkeleyarutilans (Brown etal., 2014andreferencescited therein)hasbeenshown toproduceHBIIV(butnotHBIIII)inculture,althoughwearenot awareofanyreportsto indicatethatthespatial distributionofB.

rutilans(orRhizosoleniaspp.)wouldresultinthevariabilityofHBI T25describedherein.Likewise,additionalcontributiontosedimen- taryHBIIVmaypotentiallyalsooccurinicecoveredregionssince B.rutilanshasbeenreportedinseaice(vonQuillfeldt,2000);how- ever,thiswouldresultinareductionin HBIT25 insuch settings, which isnot the case(Figs. 3,9). Thus, we suggest that R.setig- eraandrelatedspeciesarelikelytobethemainsourcesofHBIsIII andIVintheBarentsSeaandneighbouringregions.

Ofcourse, as yet unidentified sources of HBIs III and IV may also contribute to the observed sedimentary variability in HBI T25,although someofthe mostcommonspeciescharacteristicof colder, nutrient-repletewatersintheBarents Sea,such asThalas- siosiraor Fragilariopsisspp. (vonQuillfeldt,2000), arenot known tobeHBI-producers.

Second,changestoHBIT₂₅mayresultfromdifferentialdegra- dation of HBIs III and IV in the water column or sediments, or by selective removalthrough grazing. However, previous food web studies have shown no significant change to HBI composi- tion prior to,andafter, consumption (e.g. BrownandBelt, 2017;

Schmidtetal., 2018).Insomelaboratoryexperiments,HBIIVwas shown to be slightly more reactive than HBI III towards photo- oxidationandautoxidation(Rontanietal.,2014),althoughwhether thisis true underin situ environmentalconditions is asyet un- known.Inthemeantime,wenotethatahigherrateofdegradation of HBI IV would result in higher HBI T25 values for regions of increasedwaterdepth,yettheoppositeistrueinmostcases.Fur- ther, we observe no significant association between HBI T25 and waterdepthinthePCA(Fig.4).

Third, we consider whether variable HBI T25 is controlled by changesingrowthratesofR.setigera(andpotentially relatedRhi- zosolenia spp.), with higher relative production of HBI III under conditionsofmorerapidgrowth(Fig.8b).In supportofthissug- gestion,Rowlandetal. (2001) demonstratedasystematicincrease intheamountofHBIIIIrelativetoHBIIVwithincreasing growth rateofR.setigeraculturedatdifferenttemperatures.Sincewe ob- serve no relationship between HBI T25 and SST in the current dataset,wethereforesuggestthat thevariabilitydescribed herein results from regional differences in phytoplankton growth rates (Fig. 8b).Such a workinghypothesis will, ofcourse, requiretest- ingthrough furtherinvestigations intothecontrolsover HBIpro- ductionby R.setigera andother HBI-producingdiatoms, including laboratory-basedstudies andtime-seriesmonitoring oftheir pro- ductionintheBarentsSeaandotherareasofwell-definedprimary productivity.

IncreasedgrowthratesofR.setigera,inparticular,mayalsohelp explain some ofthe anomalies in oursurface sediment data. For example,anumberofmismatchesinAHCclustermembershipsof chl a andHBI T25 occurred alongthe south-western Barents and Norwegian Seacoastlines (Fig. 5), which could be a consequence of local effects associated withcoastal watermasses flowingin- shore oftheNAC andwithin theNCC.In suchsettings R.setigera hasthepotentialtoovertakeotherspeciesunderstrongupwelling and nutrient-replete conditions,as seen at the western Svalbard shelf(Beltetal.,2017).Moreover,theNCC(Fig. 2a)carriesbrack- ishcoastalwatersfromtheBalticSea,whereincreasingdominance ofR.setigeraandothercold-waterspeciesduring springandearly summerbloomshasbeenreported(e.g.Wasmundetal.,2008).In- terestingly,severalofthehigherHBIT25 valuesfromnear-coastal locationsarealsoproximaltosomechlahotspots,despitethegen- erallylowerchlaforthisregion(Fig.9).

Apartfrom the binary division betweenspring phytoplankton bloom(HBI T25> 1) versusbloom-free (HBIT25 < 1) conditions (Fig. 9), we note some further variability in HBIT₂₅ inboth the surfacesediment anddowncoredatasets(Figs.6,7) eitherside of thisthreshold.Suchvariabilitymightpotentially reflectthemean frequency(orintensity)ofspringphytoplanktonbloomoccurrence ateachsite/timeslice,especiallysincethesedimenthorizonsinves- tigatedherein(1-cm)typicallyrepresentca.20–50 yrofaccumula- tion(e.g.Dylmer,2013;Berbenetal.,2014,2017;Beltetal.,2015;

Köseo˘glu et al., 2018a). Such an interpretation would likely im- prove the value of the HBI T25 proxy in palaeo records beyond a simplebinary measure,includingthe identificationoftemporal shiftsinthefrequencyofspringphytoplanktonblooms,moregen- erally;however,thisaspectalsorequiresfurtherinvestigation.

Finally,when usedalongsideIP25 asa binarymeasure ofsea- sonalseaice(Belt,2018) andPIP25asasemi-quantitativetoolfor spring sea ice concentration, the newly proposed HBI T25 proxy for the springphytoplankton bloom has the potential to provide a more robust indicator of the MIZin northern highlatitude lo- cations, and its spatial andtemporal variation within the palaeo record.

6. Conclusions

Based on their distribution in surface sediments from across the Barents Sea and neighbouring regions, the relative amount of two tri-unsaturated HBI (III and IV; Fig. 1) lipids (HBI TR25) appears to provide proxy evidence for the spring phytoplankton bloom. Further, by re-expressing the HBI TR₂₅ ratio in a simpli- fied binary format, a thresholdfor thespring bloom isproposed (i.e. HBI T25 ≥ ^{1). HBI T}25 values in short (decadal-centennial) and long (Holocene) recordsfrom theregion are consistent with the surface sediment calibration dataset, with some changes to the occurrence/frequencyof thespringbloom linked to temporal changes inseaiceconcentration identifiedpreviously.The identi- fication of a novelproxy forthe springphytoplanktonbloom for northernhighlatitudes(atleast)could potentiallyprovideimpor- tant insights into characterising the marginal icezone, especially when usedalongsideestablished seaiceproxies suchasIP25and PIP₂₅.

Acknowledgements

This work was supported by a Research Project Grant (RPG-2015-439) from The Leverhulme Trust (UK), the Research CouncilofNorwaythroughitsCentreofExcellencefundingscheme forCAGE,projectnumber223259,andtheUniversityofPlymouth.

We are particularly grateful to Dr. Jacques Giraudeau(Université de Bordeaux) forprovidingus withsome ofthe shortcore sedi- mentmaterialandtoDr.SuzanneMaclachlanattheBritishOcean Sediment Core Research Facility (BOSCORF),UK, forsome of the surface sediments. We thank two anonymous reviewersfor pro- vidingusefulfeedbackontheinitialversionofthispaper.

Appendix A. Supplementarymaterial

Supplementarymaterialrelatedtothisarticlecanbefoundon- lineathttps://doi.org/10.1016/j.epsl.2019.06.038.

References

Addinsoft,2018.XLSTAT:DataAnalysisandStatisticalSolutionforMicrosoftExcel.

Paris,France.

Arrigo,K.R.,Perovich,D.K.,Pickart,R.S.,Brown,Z.W.,vanDijken,G.L.,Lowry,K.E., Mills,M.M.,Palmer,M.A.,Balch,W.M.,Bahr,F.,Bates,N.R.,Benitez-Nelson,C., Bowler,B., Brownlee,E., Ehn,J.K., Frey, K.E.,Garleym, R.,Laney, S.R.,Lubel- czyk,L.,Mathis,J., Matsuoka,A.,Mitchell,B.G.,Moore,G.W.,Ortega-Retuerta, E.,Pal,S.,Polashenski,C.M.,Reynolds,R.A.,Schieber,B.,Sosik,H.M.,Stephens, M.,Swift,J.H.,2012.MassivephytoplanktonbloomsunderArcticseaice.Sci- ence 336,1408.

Årthun,M.,Eldevik,T.,Smedsrud,L.H.,Skagseth,Ø.,Ingvaldsen,R.B.,2012.Quanti- fyingtheinfluenceofAtlanticheatonBarentsSeaicevariabilityandretreat.J.

Climate 25,4736–4743.

Bai, Y.,Sicre, M.-A.,Chen, J.,Klein, V.,Jin, H.,Ren,J., Hongliang, L., Xue,B., Ji, Z.,Zhuang,Y.,Zhao,M.,2019.Seasonaland spatialvariabilityofseaice and phytoplanktonbiomarkerfluxintheChukchisea(westernArcticOcean).Prog.

Oceanogr. 171,22–37.

Belt,S.T.,2018.Source-specificbiomarkersasproxiesforArcticandAntarcticsea ice.Org.Geochem. 125,277–298.

Belt,S.T.,Allard,W.G.,Massé,G.,Robert,J.-M.,Rowland,S.J.,2000.Highlybranched isoprenoids(HBIs):identificationofthemostcommonandabundantsedimen- taryisomers.Geochim.Cosmochim.Acta 64,3839–3851.