Oxygen and carbon isoscapes for the Baltic Sea: Testing their applicability in fish migration studies

Ecology and Evolution. 2017;1–13.   | 1 www.ecolevol.org Received:28June2016 | Revised:18January2017 | Accepted:7February2017 DOI: 10.1002/ece3.2841 ORIGINAL RESEARCH Oxygen and carbon isoscapes for the Baltic Sea: Testing their applicability in fish migration studies Jyrki Torniainen1 | Anssi Lensu2 | Pekka J. Vuorinen3 | Eloni Sonninen4 | Marja Keinänen3 | Roger I. Jones2 | William P. Patterson5 | Mikko Kiljunen2 ThisisanopenaccessarticleunderthetermsoftheCreativeCommonsAttributionLicense,whichpermitsuse,distributionandreproductioninanymedium, providedtheoriginalworkisproperlycited. ©2017TheAuthors.Ecology and EvolutionpublishedbyJohnWiley&SonsLtd. 1NaturalHistoryMuseum,Universityof Jyvaskyla,Jyvaskyla,Finland 2DepartmentofBiologicalandEnvironmental Science,UniversityofJyvaskyla,Jyvaskyla, Finland 3NaturalResourcesInstituteFinland,Helsinki, Finland 4LaboratoryofChronology,Universityof Helsinki,Helsinki,Finland 5SaskatchewanIsotope Laboratory,DepartmentofGeological Sciences,UniversityofSaskatchewan, Saskatoon,SK,Canada Correspondence JyrkiTorniainen,NaturalHistoryMuseum, UniversityofJyvaskyla,Jyvaskyla,Finland. Email:[email protected] Funding information MajandTorNesslingFoundation,Grant/ AwardNumber:2010150,2011105,2012506 and2013040;AcademyofFinland,Grant/ AwardNumber:34139 Abstract Conventionaltagsappliedtoindividualshavebeenusedtoinvestigateanimalmovement,butthesemethodsrequiretaggedindividualsberecaptured.Mapsofregional isotopicvariabilityknownas“isoscapes”offerpotentialforvariousapplicationsinmigrationresearchwithouttaggingwhereinisotopevaluesoftissuesarecomparedto environmentalisotopevalues.Inthisstudy,wepresentthespatialvariabilityinoxygen ( δ18 OH 2O )anddissolvedinorganiccarbon(δ13CDIC)isotopevaluesofBalticSeawater. Wealsoprovideanexampleofhowtheseisoscapescanreveallocationsofindividual animalviaspatialprobabilitysurfacemaps,usingthehigh-resolutionsalmonotolith isotopedatafromsalmonduringtheirsea-feedingphaseintheBalticSea.Aclearlatitudinalandverticalgradientwasfoundforboth δ18 OH 2O andδ13CDICvalues.ThedifferencebetweensummerandwinterintheBalticSea δ18 OH 2O valueswasonlyslight, whereasδ13CDICvaluesexhibitedsubstantialseasonalvariabilityrelatedtoalgalproductivity.Salmonotolithδ18Ootoandδ13Cotovaluesshowedcleardifferencesbetween feedingareasandseasons.Ourexampledemonstratesthatdualisotopeapproachoffersgreatpotentialforestimatingprobablefishhabitatsonceissuesinmodelparameterizationhavebeenresolved. KEYWORDS isotopiclandscape,micromilling,modelevaluation,Salmo salar,spatialassignment, spatialinterpolation 1 | INTRODUCTION Several marking approaches have been employed to address questionsinmigrationecology.Untilrecently,conventionalextrinsicmarkers(i.e.,tags)appliedtoindividualshavebeenusedtoinvestigatetheir movement(Lucas&Baras,2000),butthesemethodsrequiretagged individualsberecapturedtoacquirespatialinformation.Duringrecent yearssomeinvestigationshavebeenconductedtostudylong-term movementsofindividualadultAtlanticsalmon(Salmo salarL.;Figure1) intheseausingtagsthatrecordenvironmentalcharacteristicsalong themigrationroutes(e.g.,Chittenden,Ådlandsvik,Pedersen,Righton, &Rikardsen,2013).However,duetothepresentsizeofthetags,the studiedfishhavetobelargeand,therefore,donotnecessarilyrepresentthemajorityofthepopulation. Intrinsicbiochemicalmarkerssuchasstableisotopescanprovide analternativeapproachtotrackindividualmovementsoverlargegeographical distances such as between continents (Hobson & Norris, 2008).All animals are isotopically marked by the environment they liveinandbytheirdiet.Theassignmentofanindividualtoacertain areaworksbyestimatingprobabilitiesofoccurrenceforanimalindividualsbycomparingvaluesobtainedfromtissuesamplestoisotopic landscapes(i.e.,isoscapes;Wunder,2010).Agreatadvantageinusing 2 | TORNIAINEN ET Al. biochemical markers is that they can be linked to those individuals that actually survived the migration to their breeding habitats, and therefore,betterrepresentthepopulation.Asfishotolithsarealmost completely mineralized from the carbonate of the environmental water(Kim,O′Neil,Hillaire-Marcel,&Mucci,2007;Patterson,Smith, &Lohmann,1993;Solomonetal.,2006),analysisandcomparisonof otolithandwaterstableisotopescanrevealthelocationswherethe otolithofanindividualfishisformed.However,ifthechosentissues/ materials are dissimilar, spatial assignment (matching of tissue and sourceisotopevalues)needsfractionationequationsbetweenthechosentissueandsourceisotopevaluesduetothedifferentfractionation oftheelementisotopesviaenvironmentalandphysiologicalfactors. Isoscapesofferpotentialforvariousapplicationsinenvironmental andecologicalresearch(Bowen,Wassenaar,&Hobson,2005;Dawson &Siegwolf,2007;Hobson&Wassenaar,2008),wherebytheisotope valuesofselectedtissuesarerelatedtoenvironmentalisotopevalues, andnotjustoneelement,butseveralelements(i.e.,multi-isotopeisoscape;e.g.,Hobsonetal.,2012;García-Pérez&Hobson,2014).The mostambitiousapproacheshavebeentheconstructionsofisoscapes onaglobalscale(e.g.,Amundsonetal.,2003;Bowen&Revenaugh, 2003;LeGrande&Schmidt,2006),whereseveralstudieshaveledto convincingresultsregardinganimalmigrationamongmanytaxainterrestrialandaquaticenvironments(Best&Schell,1996;Chamberlain, Bensch, Åkesson, & Andersson, 2000; Hanson, Wurster, EIMF, & Todd,2010,2013;Hobson&Wassenaar,1996;Wassenaar&Hobson, 1998).However,althoughinformationaboutanimalmovementsand spatialusageoftheirhabitatsattheintercontinentalscaleisimportant inecologicalresearchandconservation,manycrucialeventsoccuralso atsmallerscaleswithinthedistributionofsmallerdistancemigrants. Unfortunately,theavailabilityofisoscapedataforstudiesatamore localspatialscaleappearstobesparseorthedistanceofthesurvey stationsofglobalisoscapedatamaybetoolargeforadequatelocal isotopicdiscrimination(seeBowen&Revenaugh, 2003).Therefore, additionalisotopicdataareneededtoincreasetheresolutionofglobal isoscapestoenablemoreprecisereconstructionofanimallocations andmovement. Theaimsofthisstudywere(1)toprovidehorizontalandvertical isotopicgriddeddatasets(i.e.,isoscapes)ofoxygen( δ18O H 2O )anddissolvedinorganiccarbon(δ13CDIC)forthewateroftheBalticSea.(2) Asanexamplewedemonstratethepotentialoftheseisoscapesusing twoAtlanticsalmon individuals fromthe River Simojoki.Combined withtheisotopedatafromthesalmonandspatialprobabilitysurface maps,weshowprobablelocationsofindividualfishinvarioustime pointsduringtheirsea-feedingmigrationphaseintheBalticSea.We alsodemonstratehowparameterizationofthemodelsinfluenceson locationestimates. 2 | MATERIAL AND METHODS 2.1 | Sampling and isotope analyses of water BalticSeawatersampleswerecollectedduringthreedifferentcruises bytheR/VArandaoftheFinnishEnvironmentInstitute(SYKE).To evaluateapossibleseasonalimpactonseawaterisotopevalues,we collectedtwosummersetsandonewintersetofsamples.Thedates ofthecruiseswere(1)24Mayto11June2010,(2)9Augustto27 August2010,and(3)17Januaryto3February2011.Thefirstcruise coveredthewholeBalticSeafromtheBothnianBaytotheSouthern BalticProper,excludingtheGulfofFinland(Figure2),whichwasthe onlyareasampledduringthesecondcruise.Thethirdcruise,inwinter,coveredtheBalticSeaexceptareassouthofGotland.Seawater wassampledatadepthof10mfromeverysamplingstation(altogether316samplesfrom134stationvisits,blackdotsinFigures3 and4).From25stations,waterwasalsosampledverticallyat5–50m FIGURE1 Atlanticsalmon(Salmo salar) FIGURE2 MapoftheBalticSeawiththelocationoftheRiver Simojoki(uppermostrightcorner).Themarkedpathwaysshow transectsusedforcreatingverticalinterpolationsinFigure4 | 3 TORNIAINEN ET Al. intervalsdependingonwaterdepthatthesite(Figure4).Maximum distancesbetweensamplingstationswerelessthan100km.Water samples were taken using a CTD/Rosette sampler (Rosette 1015, Seabird,SBE911/GeneralOceanics,SIS:Plus500). Rosettesamplerrecordsconductivity,fromwhichsalinitywasautomaticallycalculated.Fromeverysamplingoccasionofwaterforisotopeanalysispurpose,salinitywasrecorded.Samplewatercollected for δ13CDICanalysis,wereinjectedinto12-mlborosilicateExetainer vials(cat.no438B;LabcoLtd.,HighWycombe,UK)preparedinthe laboratory,where0.2mlof85%orthophosphoricacid(H3PO4)was addedintoeachvial,whichwasthensealedwithacap(containinga rubberseptum)andflushedandfilledwithaheliumatmosphere.In thefield2–4mlofseawaterfromeachstationandsampledepthwas injectedthroughtherubberseptumintothevial.Samplewater for δ18O H 2O analysiswerecollectedin20-mlglassscintillationvialsand filledfullensuringnoairbubbles.Allsampleswerestoredinanunlit FIGURE3 Interpolatedmapsforthe BalticSeaof δ18 OH 2O values(a,b)and δ13CDICvalues(c,d)from10metersin summer(a,c)andwinter(b,d).Blackdots representsamplinglocations (a) (c) (b) (d) Longitude(E) Latitude(N) –7 –8 –9 –10 2 1 0 –1 –2 510 15 20 25 30 35 510 15 20 25 30 35 66 64 62 60 58 56 54 66 64 62 60 58 56 54 Summer Winter δ18O(‰)δ 13C(‰) FIGURE4 Interpolatedcross-sectional surfacesofvertical δ18 OH 2O andδ13CDIC values. δ18 OH 2O valuesfromsouthtonorth in(a)summerand(b)winter,andfromwest toeastin(c)summerand(d)winter.δ13CDIC values,fromsouthtonorthin(e)summer and(f)winter,andfromwesttoeastin (g)summerand(h)winter.Blackdots representsamplinglocationsanddepths (a) (b) (c) (d) (e) (g) (f) (h) δ18O (‰)δ 13C (‰) Depth(m) 0 100 200 300 0 50 100 150 –5 –6 –7 –8 –9 –10 –11 –5 –6 –7 –8 –9 –10 –11 20 22 24 26 28 20 22 24 26 28 56 58 60 62 64 56 58 60 62 64 Longitude(E) 2 1 0 –1 –2 –3 –4 –5 2 1 0 –1 –2 –3 –4 –5 0 100 200 300 0 50 100 150 20 22 24 26 28 20 22 24 26 28 56 58 60 62 64 56 58 60 62 64 Latitude(N) Longitude(E) Latitude(N) Summer Winter 4 | TORNIAINEN ET Al. refrigerator(+4°C)inadarkroompendinglaboratoryanalysis.Inthe laboratory,similarvialsasusedforδ13CDICsampleswerefilledwith 0.5mlseawaterandequilibratedwithCO2foratleast24hrat25°C. Analysesofsamplesstartedwithinaweekafterarrivaltothelaboratory,firsttheδ13CDICsamplesimmediatelyafterarrival.Theδ13CDIC valuesareexpressedrelativetoVPDB(ViennaPeeDeeBelemnite), measured and calibrated/normalized against international IAEA (InternationalAtomicEnergyAgency[IAEA])standardsNBS19(orTS- Limestone;Calciumcarbonate,δ13C=+1.95‰)andLSVEC(Lithium carbonate,δ13C=−46.6‰);±0.14‰.Valuesareexpressedrelativeto VSMOWandcalibrated/normalizedtotheVSMOW(ViennaStandard Mean Ocean Water [VSMOW], δ18O=0‰)—SLAP (Standard Light AntarcticPrecipitation,δ18O=−55.5‰)scale;±0.1‰.Allwatersampleswereanalyzedforbothδ13CDICand δ18O H 2O attheLaboratory of Chronology, Finnish Museum of Natural History, University of Helsinki, using a GasBench II and Delta Plus XL (Thermo Fisher Scientific,Bremen,Germany). Statisticaltestingofseasonaldifferencesin δ18O H 2O andδ13CDIC values(pairedsamplesttest)wasperformedusingPASWStatistics18 forWindows(SPSSInc.,Chicago,IL,USA).Interpolatedisoscapesfor δ13CDICand δ18O H 2O wereconstructedusingOceanDataView(ODV) softwareversion4.5.5(Schlitzer,2002,2011). 2.2 | Otolith sampling, micromilling and stable isotope analysis In order to test applicability of Baltic Sea isoscapes in fish migrationstudies,weanalyzedotolithsfromtwoexamplefemaleAtlantic salmon(hereafterBalticsalmonorsalmon)originallycaught,asthey werereturningtotheRiverSimojoki(Figure2)tospawnin2008,by theNaturalResourcesInstituteFinland(LUKE)aspartoftheprogram tomonitoryolk-sacfrymortality(M74syndrome:e.g.,Keinänenetal., 2012; original LUKE code numbers SS5447 and SS5461, hereafter FISH1and2,respectively).Bothsalmonhadspenttwoyearsfeedinginthe sea.Theorigin (wild orhatchery-reared) andage of the salmonwasdeterminedfromthescalenucleusandscalegrowthpattern(Hiilivirta,Ikonen,&Lappalainen,1998).FISH1wasofwildorigin (5,300g,80cm)andFISH2wasofhatchery-rearedorigin(6,300g, 81cm). Both sagittal otoliths were removed from the head of the salmon,cleanedindeionizedwatertoremoveanyremainingorganic tissue,anddriedovernightat60°C. Bothotolithsweresampledforδ13Candδ18Oanalysisusingthe custom-built three-dimensional computer-controlled micromilling systemintheSaskatchewanIsotopeLaboratoryattheUniversityof Saskatchewan following the procedure of Wurster, Patterson, and Cheatham(1999).Thissystemallowed31–32samplingpathstobe followed concordantwith growth banding in both otoliths. Isotope ratios of samples were determined using a Finnigan MAT 253 directlycoupledtoaKiel-IVautomatedcarbonatepreparationdevice (Thermo–Fisher Scientific).Accuracy and precision were monitored byroutineanalysisofNBS-19standard,yieldingastandarddeviation forreplicatestandardsthatwasconsistentlylessthan0.09forboth δ13Oandδ18Ovalues.Allotolithisotopemeasurementsarereported inthestandarddeltanotation(permil)relativetotheVPDBstandard asδ18Ootoandδ13Coto.Fromtheotolithdataofeachsalmon,three samplemillingpathswereselectedforcloserinspection,namely1st seawinter(1SW),thefollowingsummer(2SS),andthe2ndseawinter (2SW)toassignthesalmontoBalticSeaareasforeachperiod.During theseamigrationphase,Balticsalmonexperiencehighseasonalfluctuations in ambient temperature.These periods are clearly seen in theotoliths,asotolithcarbonatewiththe highestδ18Ootovaluesis accretedduringcoldwinterperiod,whilethelowestδ18Ootovalues representhighesttemperaturesinthesummer(Figure6;Wurster& Patterson,2003). 2.3 | Creating isotopic (isoscapes) and temperature maps for the Baltic Sea Isoscapesforδ13CDICand δ18O H 2O aswellastemperaturemapswere constructedusingODVsoftware.UsedmapdataforhorizontalisotopemapswerereceivedfromthedatabaseoftheLeibnizInstitute for Baltic Sea Research Warnemünde (IOW; Seifert, Tauber, & Kayser,2001)andforbothhorizontalmapsandverticalprofilesfrom the General Bathymetric Chart of the Oceans (GEBCO_08 Grid). GEBCO_08GridhadtobeconvertedintoNetCDFformatcompatiblewithODVwithRStatisticssoftwarev3.0.1(RCoreTeam,2013) using package RNetCDF (Michna, 2012). Interpolated maps were producedusingDataInterpolatingVariationalAnalysis(DIVA)griddingsoftware(Troupinetal.,2012)includedinODV(DIVAparameters:scalelengthschosenautomatically;signal-to-noiseratio=40; qualitylimit=3.0;excludingoutliers).VerticalprofileswerealsocreatedusingDIVAgridding(scalelengthschosendependingonused data;signal-to-noiseratio=40;qualitylimit3.0;excludingoutliers). ThecoastlinesinthemapsarebasedontheGlobalSelf-consistent Hierarchical High-resolution Shorelines database v 2.1 (Wessel & Smith,1996). Asthespatialcoverageofcollected δ18O H 2O wasrelativelysparse forassignmentmodels,andsomelargerareaslackedmeasurements, wecreatedpredictivemodels to estimate δ18O H 2O valuesfromthe BalticSeawatersalinity(S)datawhichisanexcellentpredictorfor δ18O H 2O andiscommonlyusedtoestimateoceanseawaterisotope values(e.g.,LeGrande&Schmidt,2006).The δ18O H 2O –Srelationship isfrequentlylinearlyrelated(LeGrande&Schmidt,2006),butinthe BalticfreshwaterinflowfromlargeriversintoGulfofBothniaandGulf ofFinlandhaveaneffectontherelationships.Therefore,wecreated separatemodelsfortheGulfofBothnia,theGulfofFinlandandfor therestoftheremainingareaoftheBalticSea.Shapeofthepredictive modelforeachareawasselectedbasedongoodnessoffitandmodel residualswereevaluatedusingquantile–quantileplots.Modelswere furtherappliedtoseawatersalinitysamplingsurveysobtainedfrom BalticMarineEnvironmentProtectionCommission(HELCOM)database(Andersson,2014)tocreateseasonal δ18O H 2O mapsfortheBaltic Sea.Dataonlyforsummer(July–August)andwinter(February–March) fortheperiodthestudysalmonhadspentinthesea(2007–2008) wereselected.Aswefoundstatisticallysignificantseasonaldifference in measured δ18O H 2O  values, three separate δ18O H 2O  maps [winter | 5 TORNIAINEN ET Al. 2007(1SW),summer2007(2SS)andwinter2008(2SW)]werecreatedtobeusedinassignmentmodels.Modeledvalueswerecombined withmeasuredvaluesfromthesameseasontogainbetterspatialcoverage. Marked within-season differences were not found between HELCOMandourownmeasurementsbasedonvisualvalidationofsalinitydata,obtainedfromdepths5–15m,indicatingthatalso δ18O H 2O valuesbehavethesameway. Salmon summer temperatures were fixed at 11.5°C, calculated frompreferredtrue10mstayingdepthsofsalmoninBalticsalmon data storage tag study by Westerberg, Sturlaugsson, Ikonen, and Karlsson(1999).Forwintertemperatures(2007–2008),weusedall available temperature profiles collected from HELCOM database (Andersson,2014)coveringthewholeBalticSea(meantemperatures fromdepths5–15m)andcreatedaninterpolatedsurfaceforthepreferred10mstayingdepth.Interpolatedmetabolicallyderivedbicarbonateδ13Cdietvalueswerederivedfromδ13Cvaluesofsalmondietary speciesaroundtheBalticSea(AppendixS3). 2.4 | Creation of otolithrelated isoscapes from δ18OH2O, δ13CDIC, δ13Cdiet and water temperature Understanding the dependence of ambient water temperature and δ18O H 2O  values on δ18Ooto is necessary to obtain comparable values of δ18Ooto and δ18O H 2O  for spatial assignment of salmon (e.g., Pattersonetal.,1993).Moreover,theδ13Cotovalueisamixofbicarbonateδ13CDICfromambientwaterandmetabolicallyderivedbicarbonate δ13Cdiet (Solomon etal., 2006; Wurster & Patterson, 2003). Wethereforecorrected δ18O H 2O andδ13CDICvaluesusingliterature- derivedfractionationequations,afterwhichthecorrectedvaluesare consideredas“otolithisoscape”ofbothelements(AppendixS1).This permitsadirectcomparisonofotolithandBalticSeawaterisotopes, therebyprovidingaprobabilisticspatialassignmentofsalmonduring theirsea-feedingphase. Initially, δ18O H 2O  values are calculated relative to the VSMOW scaleandδ18OotovaluesrelativetotheVPDBscalefromtheIAEA. To enable direct comparison of δ18Ooto(VPDB) with δ18Owater(VSMOW), δ18Owater(VSMOW) values were converted to VPDB using following equation(fromClark&Fritz,1997): Oxygen isotope values of otoliths reflect those of the ambient water(Campana,1999;Farrell&Campana,1996;Thorrold,Jones,& Campana, 1997), with a temperature-dependent fractionation (e.g., Pattersonetal.,1993).Followingcommonpractice,weusedthelinear temperature-dependentfractionation(e.g.,Pattersonetal.,1993). where Tistemperature(103/K),whereKisambientwatertemperatureinKelvin,andparameterα isthefractionationfactorbetween otolith and ambient water [α=(δ18Ooto(VPDB)+1,000)/( δ18O H 2O (VPDB)+1,000)]. Model fractionation constants a and b in four previous studies covering salmonid fishes are variable (a=−41.14, b=20.43:Godiksenetal.,2010;a=−33.43,b=17.88:Hansonetal., 2013; a=−33.49, b=18.56:Pattersonetal., 1993and a=−41.69, b=20.69:Storm-Suke,Dempson,Reist,&Power,2007).Toevaluate theeffectandvariationofdifferentparameterstosalmonassignment, thenumericallymostdistantparametersfromeachother(i.e.,Hanson etal.,2013;Pattersonetal.,1993)wereselectedforsalmonassignment(i.e.,forprobabilitysurfaces),hereafterModel3andModel1, respectively.Alsoanaverageofallfourmodels(a = −37.44,b = 19.39) wascalculatedandusedaccordingly,hereafterModel2(AppendixS2). Thefollowingequationgivestheotolith-relatedδ18O(VPDB)isoscapevaluesinrelationtowatertemperature(T)and δ18O H 2O (VPDB)values,thatcanbecomparedtoδ18Ootovaluesenablingtheassignment ofsalmontoseaareasbasedontheirotolithisotopevalues: Anotolith-relatedδ13Cotoisoscapewascalculatedasfollows(e.g., Solomonetal.,2006;Wurster&Patterson,2003)toallowcomparison betweenδ13CDICandδ13Coto: where δ13Cdietisameanδ13Cvalueofsalmonprimarypreyspeciesin theBalticSea(sprat[Sprattus sprattus],Balticherring[Clupea harengus membras]andthree-spinedstickleback[Gasterosteus aculeatus])ina particularlocation(AppendixS3)andMistheproportionofmetabolic carboninthesalmonotolith(Sherwood&Rose,2003): where Kcaud(Atlanticsalmoncaudalfinratio;Minns,King,&Portt, 1993)is2.4. 2.5 | Salmon assignment using otolith and water isotope values Toestimatethelocationsofsalmonindividualsintheir1SW,2SS and second 2SW from δ18Ooto and δ13Coto values, we calculated probabilitydensitysurfacesforeachsalmonbyusingadeterministicgridcoveringtheBalticSeafollowingtheapproachpresented inWunder(2010)withRStatisticssoftwarev3.0.1(RCoreTeam 2013).WefirsthadtoreinterpolateallDIVAinterpolationresults (waterand diet-basedisotopes,temperatures,etc.which werein differentkindsofnondeterministicgrids)fromODVintoadeterministicgridtoenableustocalculatethefractionationequations forallgridlocations.Thisreinterpolationwasperformedwithlocal (Nmax =4) inverse-distance weighting interpolation available in theRpackagegstat(Pebesma,2004).Theprobabilitydensitysurface calculation approach assumes water-temperature-dependent δ18O(VPDB)valuesandDIC-diet-basedδ13Cvaluestobenormallydistributedateachgridpointandthencalculatestheprobabilitiesof obtainingthemeasuredvaluesoftheotolithfromnormaldistributionswithmeanparametersaccordingtotheVPDB-correctedinterpolatedisotopevalues.Standarddeviationsofδ18Oandδ13Cvalues wereestimatedbasedonobservedvariationsof δ18 O H2O (ɛ=0.019) andδ13CDIC(ɛ=0.216)valuesobtainedfromthesamelocationand depth,andthevariationofobservedδ18Ootovaluesintheotoliths δ18 O H2O(VPDB) =0.97002×δ 18 O H2O(VSMOW) − 29.98. 1,000×ln α=a+b∕T, δ18 O (VPDB) =e ( a+b∕T ) ∕ 1,000 ×(δ 18 O H2O(VPDB) +1,000)− 1,000. δ13 C oto = M ×δ13 C diet + (1 − M) ×δ13 C DIC, M=0.025+0.066×Kcaud, 6 | TORNIAINEN ET Al. ofseveralfishindividuals(ɛ=0.207estimatedfromGodiksenetal., 2010). The resulting probability surfaces do not represent two- dimensionalprobabilitysurfaces(totalprobabilityineachmapsisnot scaledtobe1)andtheminimumandmaximumprobabilityvaluesare notthesameinallmapsasiscommoninthiskindofapproach(see e.g.,Wunder,2010).Instead,thesesurfacesvisualizewheretheprobabilityofobtainingthemeasuredvalueoftheotolith(ortheprobability ofpresenceofparticularindividualatacertaintime)isthegreatest, andwhereitis(much)lower.Inthosemapswherebothoxygenand carbon isotopes have been taken into account, probabilities were calculated by multiplying isotope-wise probabilities, assuming independence,whichmaynotbestrictlytrueforthiskindofphenomenon.However,ourdatadidnotallowforfullestimationofcovariance betweentruedatavaluesduetopartiallydifferingmeasurementor observation locations. Nevertheless, the covariance of interpolated surfaceswasabout10-foldsmallerthanthevariancesofindividual isotopes.Therefore,weexpectthatanypossibleerrorintheresults, duetodependenceinisotopevalues,issmall. Tostudy,howsensitiveourapproachistomeasurementerrors relatedtothestableisotopes,weconductedasimplesensitivityanalysisbyaddingand/orsubtractingtheobservedstandarddeviationsof watersamples(basedonliterature)ofδ13Candδ18Oto/fromtheactualmeasuredvalues,andperformedthecalculationoftheprobability ofpresenceforFISH2onthefirstwinter(1SW)usingthemodelby Hansonetal.(2013)forfractionation.Theamount±SDofwaterisotopemeasurementvariation(basedonliterature)waschosenbecause italsoreflectsabout±2SDofourownestimateofmeasurementaccuracy(0.09)ofotolithisotopevalues.So,theselimitsarealsorather closetobeingabout95%confidenceintervalofthemeasuredisotope values. 3 | RESULTS 3.1 | δ18OH2O, and δ13CDIC values in the Baltic Sea Differencein δ18O H 2O valuesbetweenseasonsat10mdepthatthe same locations was statistically significant (Paired samples T test: N = 31,p < .001).However,themean(±SD)valuesof δ18O H 2O were very similar (meansummer: −7.9‰±0.83, meanwinter: −7.6‰±0.74). Inaddition, δ18O H 2O valuesshowedaclearsouthtonorthlatitudinal decreasefromaround−6‰toaround−10‰(Figure3a,b).Aweaker longitudinaldecreasefromaround−6‰toaround−7.5‰occurred fromtheeasternendoftheGulfofFinlandtothewesternsideofthe NorthernBalticProper(Figure3a,b).Distinctdifferencesin δ18O H 2O valuesbetweenadjacentbasinswerealsoobserved(BothnianBay— BothnianSea—BalticProper—GulfofFinland),buttheBothnianSea andtheGulfofFinland,eventhoughnotdirectlyconnected,hadvery similarvalues(Figure3a,b)duetotheinfluxoffreshwaterfromrivers. Mean (±SD) δ13CDIC values were significantly different (Paired samplesTtest:N = 31,p < .001)betweensamplingtimes(meansummer:0.3‰±0.81,meanwinter:−0.9‰±0.54).δ13CDICvaluesdecreased withlatitudeinsummerfromaround2‰toaround−2‰(Figure3c). The differencewas smaller in thewinter (Figure3d).An increasing trendwasobservedinδ13CDICvaluesfromtheGulfofFinlandtothe BalticProperinbothseasons,fromaround−1‰toaround2‰in summerandfromaround−2‰toaround0‰inwinter(Figure3c,d). Verticalinterpolationofthetransectsshowedthat δ18O H 2O valuesin theCentralBalticProperwerehigherbelow~50mfromaround−6.5‰ toaround−5.5‰atthebottomandashighas−4‰intheSouthern BalticProper(Figure4a,b),whereasacleardecreasingtrendinsummer δ13CDICvaluesfromaround−1‰toaround−5‰atthebottomwas observed (Figure4e,f). Vertical interpolation also showed stability of δ18O H 2O valuesbetweensummerandwinter,whereasδ13CDICvalues exhibitedvariation between summer and winter in the upper water layer (up from ~50m; Figure4e–h). Below ~50m bothvalues were ratherstableovertime,exceptinthelesssalineareas(theBothnianSea, theBothnianBayandtheGulfofFinland)whereδ13CDICvaluesshowed minordecreasesduetoautumnwatercolumnturnover(Figure4). Therewasa strong δ18O H 2O –Srelationshipandthe coefficients ofdeterminationforallthemodelswerecloseto1.RelationshipappearedtobelinearonlyforSouthernBalticSea(R2=.98),whilein GulfofFinland(R2=.97)andGulfofBothniarelationshipswhereof logarithmicshape(R2=.99;Figure5). 3.2 | Carbon and oxygen isotope values of salmon otoliths Stableisotopeanalysisviamicromillingfromtheotolithnucleustothe otolithedgeshowedclearvariationinotolithisotopevalues.Inaddition,bothisotopevaluesofthetwootolithsanalyzedshowedsimilar life-historytrendsfromthenucleustotheotolithedge.δ18Ovaluesin theotolithnucleus(FISH1:−10.9‰;FISH2:−10.6‰)weremarkedly lowerthanthehighestvaluesobservedintheouterpartoftheotolith radius(FISH1:−5.9‰;FISH2:−6.6‰),andwiththelowestvaluesat theedgeoftheotolith(bothotoliths:−11.6‰;Figure6a,b). FIGURE5 Relationshipsbetweenseawatersalinityand δ18 OH 2O measuredin(a)GulfofBothnia,(b)GulfofFinlandand(c)Baltic Proper.Presentedmodelswereappliedwhenestimating δ18 OH 2O for eacharea –12 –11 –10 –9 –8 –7 –6 –5 –4 0246810121 41 6 a c b Salinity (‰) δ18O(‰) y = 0.263x - 8.684 R2= 0.98 y = 2.245ln(x) - 11.243 R2 = 0.97 y = 3.447ln(x) - 13.664 R2= 0.99 | 7 TORNIAINEN ET Al. Forbothotoliths,δ13Cvaluesalsoshowedclearvariationalong the otolith radius, from lowest δ13C values in the otolith nucleus (FISH1:−16.1‰;FISH2:−14.4‰)tohighestintheouterpartof theotolithradius(FISH1:−3.4‰,FISH2:−3.1‰),withintermediate valuesattheedgeoftheotolith(FISH1:−11.1‰;FISH2:−10.8‰; Figure6a,b).Bothotolithnuclei(i.e.,duringthejuvenilephase)exhibitedlowδ13Cotoandδ18Ootovalues.Thejuvenilephaseisotopevalues ofFISH1tendedtobelessvariablethanthoseofFISH2.Theotolith isotopevaluesofapparentpost-smoltmigrationfromthefreshwater tomoresalinearethemostdistinctbetweenthesetwosalmon.FISH 1otolith δ18Ovalues increasedto thehighestvaluesmoresharply thanthoseofFISH2,forwhichδ13Cvaluesincreasedinamorelinear manner.Otolithδ18OvaluesforFISH1werehigherduringthefirst seawinterthanthoseforFISH2,whilevaluesduringthesecondsea winterandapparentspawningmigration(steepdeclineinbothisotope valuesafter2SW)weresimilarforbothsalmon(Figure6a,b). 3.3 | Probability surfaces for salmon locations in the seafeeding phase BasedontheModel1(Hansonetal.,2013)assignments,themost probablefirstseawinterlocationofFISH1wasintheBalticProper (Figure7a).Duringthenextsummerandthesecondwinter,FISH1 appearedtooccupywatersclosetotheGulfofRiga,andtheareasof thesouthwestandnorthernBalticProper,respectively(Figure7b,c). TheresultsforFISH2wereverysimilar,exceptthatinthefirstwinter itappearedtobelocatedintheBothnianSeaorintheGulfofRiga (Figure7d–f).ComparedtotheModel1,assignmentsoftheModel 3(Pattersonetal.,1993)werelocatedmorenorthernandalsocolder areasortoareaswithlowersalinity(theGulfofFinland;Figure7). Model2(theaveragemodeloffourmodelsusedinthestudy)salmon assignmentswerebetweenHansonetal.(2013)andPattersonetal. (1993)models(Figure7).Someoftheassignmentsofthemodels2 and3wereconfinedtosuchasmallareathatresultsofthemodels werenotveryvisibleintheassignmentmaps(Figure7). Total of eight different results (Figure8) were obtained from thesensitivityanalysisinadditiontotheresultobtainedwithactual measuredvalues.Increasingtheotolithδ18Ovaluecausesthemost probableareaintheGulfofBothniatoshiftslightlytowardwest,and decreasingitcausesthemostprobableareatoshiftslightlytoward east.Increasingδ13Cvalueseemstomaketheassignmentmoreconcentratedatonelocationanddecreasingitcausestheresulttobeless determined. 4 | DISCUSSION 4.1 | Isotope values of the Baltic Sea Wepresentedthedistributionofδ13CDICand δ18O H 2O valuesinthe BalticSeaduringbothsummerandwinterviainterpolatedhorizontal FIGURE6 Isotopevaluesoftwo exampleRiverSimojokisalmonotoliths micro-milledandanalyzedforoxygenand carbonstableisotopes:(a)FISH1(wild origin)and(b)FISH2(hatchery-reared). Opencirclesrepresentδ13Cvaluesand filledcirclesδ18Ovalues.Isotopepoint valuesusedinthesalmonassignmentto theprobablelocationsintheBalticSea duringtheirfirst(1SW)andsecondsea winter(2SW)andsecondseasummer (2SS)areindicated –20.0 –18.0 –16.0 –14.0 –12.0 –10.0 –8.0 –6.0 –4.0 –2.0 –12.0 –11.0 –10.0 –9.0 –8.0 –7.0 –6.0 δ 13 C(‰)δ 13 C(‰) δ18O(‰) –20.0 –18.0 –16.0 –14.0 –12.0 –10.0 –8.0 –6.0 –4.0 –2.0 –12.0 –11.0 –10.0 –9.0 –8.0 –7.0 –6.0 020406080100 Distance from nucleus (%) δ18O(‰) (a) (b) 1SW 2SS 2SW 1SW 2SS 2SW 8 | TORNIAINEN ET Al. isotopicmapsfromadepthof10mandalsoviaverticalcross-sections. Differencesbetweenseaareasinbothisotopevalueswereobserved. Horizontalδ13CDICvaluesclearlydifferedbetweensummerandwinter,whereas δ18O H 2O valuesremainedrathersimilar.Vertical δ18O H 2O valuesremainedstablebetweensummerandwinter,whereasδ13CDIC showedmarkeddifferencesbetweensummerandwinter,especially nearthesurface. The distinctive hydrologic characteristics of the Baltic Sea are clearlyreflectedinourmeasurements. δ18O H 2O valuesintheBalticSea aremainlycontrolledbytheinfluxfromtheDanishstraitsofdenser, more saline water with higher δ18O H 2O  (see e.g., Dickson, 1973; Matthäus&Lass,1995)andthefreshwaterwithlower δ18O H 2O values runningofffromthecatchmentarea.Theobservedverticalgradientin isotopevaluesisalsoduetothepenetrationofmoresalinewaterwith higher δ18O H 2O valuesunderthelessdensefreshwaterwithlower δ18O H 2O values.ThisdenserwaterstagnatesinthedeeperbasinsisolatedbythesillsbetweendifferentBalticSeaareas(Matthäusetal., 2008)andleadstothearealsegregationof δ18O H 2O values.Asimilar phenomenoncanbeseeninmarineenvironmentsaroundtheworld (Schmidt, Bigg, & Rohling, 1999) and has been recorded from the SouthernBalticSea(Frohlich,Grabczak,&Rozanski,1988;Punning, Vaikmae,&Maekvi,1991).Temperature-relatedautumnoverturnof thewatercolumnbreaksdownthesummerstratificationandmixes thewaterabovethehalocline(downto~50m),butthedeeperwater FIGURE7 Probabilitysurfacesof locationsfortwoexampleRiverSimojoki Atlanticsalmons(FISH1and2)duringtheir feedingphaseoffirstseawinter(1SW;a andd,respectively),secondseasummer (2SS;bande,respectively)andsecond seawinter(2SW;candf,respectively)in theBalticSea.Calculationsarebasedon salmonotolithandBalticSeawaterδ18O andδ13Cvalues.Whiteandhigh-saturation colorsindicatethatallvaluesurfaces usedinthecalculationofprobabilities (temperature, δ18 OH 2O ,δ13CDICandprey isotopevalues)arereliableincontrastto lowconfidenceprobabilitiesindicatedin grayandgrayishcolors.Probabilitysurface resultsindicatedinredrepresenthigh probabilityofpresenceaccordingtothe usedmodelofHansonetal.(2013)(Model 1),greencolorindicateshighprobability orpresencewiththeaveragemodelof allfourusedmodelsinthisstudy(Model 2;Godiksenetal.,2010;Hansonetal., 2013;Pattersonetal.,1993;Storm-Suke etal.,2007)andbluecolorrepresents highprobabilityorpresenceonthemodel probabilitiesofPattersonetal.(1993) (Model3) Latitude (a) Model 1 high prob. Model 2 high prob. Model 3 high prob. Low probability Low confidence Latitude (b) Model 1 high prob. Model 2 high prob. Model 3 high prob. Low probability Low confidence Longitude Latitude (c) Model 1 high prob. Model 2 high prob. Model 3 high prob. Low probability Low confidence (d) Model 1 high prob. Model 2 high prob. Model 3 high prob. Low probability Low confidence (e) Model 1 high prob. Model 2 high prob. Model 3 high prob. Low probability Low confidence 54 56 58 60 62 64 6654 56 58 60 62 64 66 10 15 20 25 30 54 56 58 60 62 64 66 10 15 20 25 30 Longitude (f) Model 1 high prob. Model 2 high prob. Model 3 high prob. Low probability Low confidence FISH 1 FISH 2 | 9 TORNIAINEN ET Al. columnremainsratherstablemaintainingmoremarinecharacteristics (Lass&Matthäus,2008).However,inourstudy, δ18O H 2O valuesremainedstablebetweensummerandwinter,andonlyoccasionalstrong saltwaterpulses(e.g.,Dickson,1973;Matthäus&Lass,1995)could change this situation. Lack of winter isotope measurements south from Gotland might have affected the interpolatedvalues denoted inthemapswithgrayishcolors).TheFinnishEnvironmentInstitute (SYKE;http://www.itameriportaali.fi/en_GB/:2.6.2014)andHELCOM (Andersson,2014)havereportedstablehydrologyandsalinityofthe Balticduringthestudiedyears.Basedonthoseattributes,itisnotvery likelythat δ18O H 2O valuesinsummerwouldbeverydifferentfromthe valuesobservedinwinter. δ13CDICvaluesarecontrolledbysimilarmechanismsasthosethat control the water column characteristics and δ18O H 2O .The clearest distinction between the mechanisms based on temperature-related bioactivity. Preferential incorporation of 12C during photosynthesis FIGURE8 SensitivityanalysisofprobabilisticspatialassignmentModel1forFISH2in1stwinteratthesea(1SW).Originalotolithδ18Oand δ13Cvalues(figureinthemiddle)wereadjusted±1SD.Highprobabilitymeanswheretheprobabilityofobtainingthemeasured(ordeviated) valueoftheotolith(ortheprobabilityofpresenceofthisindividualat1SW)fromthecreatedotolithisoscapeisthegreatestbytheused model.Thehatched(low-saturationcolored)areasindicateregionswherethedensityofrealobservationsofatleastoneisoscape(usedinthe calculationofprobabilityofpresence)istoolowforreliableprediction 10 15 20 25 30 δ18Oat upper, δ13Cat lower limit High probability Low probability 10 15 20 25 30 δ18O at upper limit High probability Low probability 10 15 20 25 30 Both at upper limit High probability Low probability 10 15 20 25 30 δ13C at lower limit δ13C at upper limit High probability Low probability 10 15 20 25 30 Both isotopes as measured High probability Low probability 10 15 20 25 30 High probability Low probability 10 15 20 25 30 54 56 58 60 62 64 6654 56 58 60 62 64 6654 56 58 60 62 64 66 Both at lower limit High probability Low probability 10 15 20 25 30 δ18O at lower limit High probability Low probability 10 15 20 25 30 54 56 58 60 62 64 66 54 56 58 60 62 64 66 54 56 58 60 62 64 66 54 56 58 60 62 64 66 54 56 58 60 62 64 66 54 56 58 60 62 64 66 δ18Oat lower, δ13C at upper limit High probability Low probability Longitude Latitude Otolithδ13C value increases → Otolith δ18O value increases → 10 15 20 2 5 30 High pro b a bili t y L o w pro b a bil i t y 10 1 5 20 25 30 High pro b a bili t y L o w pro b a bili t y 10 1 5 20 25 30 High pro b a bili t y L o w pro b a bili t y 10 15 20 25 30 δ 1 3 C a t l o w e r l i m i t δ 1 3 C a t u pper li m it Hi g h probabilit y L o w probabilit y 10 15 20 25 30 Both isoto p es as measured H i g h probabilit y L o w probabilit y 10 15 20 25 30 Hi g h probabilit y L o w probabilit y B ot h at l o w e r limi t Hi g h probabilit y L o w probabilit y δ 1 8 O a t l o w e r l i m i t H i g h probabilit y L o w probabilit y 5 4 5 6 5 8 6 0 6 2 6 4 6 5 4 5 6 5 8 6 0 6 2 6 4 6 5 4 5 6 5 8 6 0 6 2 6 4 66 5 4 5 6 5 8 6 0 6 2 6 4 6 6 5 4 5 6 5 8 6 0 6 2 6 4 66 5 4 5 6 5 8 6 0 6 2 6 4 6 6 δ 1 8 O a t l o w e r , δ 1 3 C a t u p p e r l i m i t Hi g h probabilit y L o w probabilit y