ScienceoftheTotalEnvironment631–632(2018)1005–1017 ContentslistsavailableatScienceDirect Science of the Total Environment journal homepage: www.elsevier.com/locate/scitotenv Modelling climate change effects on Atlantic salmon: Implications for mitigation in regulated rivers L.E.Sundt-Hansena,⁎,R.D.Hedgera,O.Ugedala,O.H.Diseruda,A.G.Finstada,e,J.F.Sauterleuteb,d,L.Tøfteb, K.Alfredsenc,T.Forsetha aNorwegianInstituteforNatureResearch,P.O.Box5685,Sluppen7485,Trondheim,Norway bSINTEFEnergyResearch,P.O.Box4761,Sluppen7465,Trondheim,Norway cNorwegianDepartmentofHydraulicandEnvironmentalEngineering,NorwegianUniversityofScienceandTechnology,7491Trondheim,Norway dSWECO,ProfessorBrochsgate2,7030Trondheim,Norway eDepartmentofNaturalHistory,NorwegianUniversityofScienceandTechnology,7491Trondheim,Norway H I G H L I G H T S G R A P H I C A L A B S T R A C T • Futureclimatechangeislikelytoimpact onAtlanticsalmonabundanceinrivers • Abundancewasmodeledcombiningbi- ological, hydrological and hydraulic models • Futurejuvenileabundancewasreduced inthreeoffourfuturescenarios • Reductioninabundancewascausedby reducedwettedareainsummerperiods • Reducedfuturejuvenileabundancecan bemitigatedinriverswithreservoirca- pacity by releasing water in critical periods a r t i c l e i n f o a b s t r a c t Articlehistory: Climatechangeisexpectedtoalterfuturetemperatureanddischargeregimesofrivers.Theseregimeshaveastrong Received8December2017 influenceonthelifehistoryofmostaquaticriverspecies,andarekeyvariablescontrollingthegrowthandsurvivalof Receivedinrevisedform5March2018 Atlanticsalmon.ThisstudyexploreshowthefutureabundanceofAtlanticsalmonmaybeinfluencedbyclimate- Accepted6March2018 inducedchangesinwatertemperatureanddischargeinaregulatedriver,andinvestigateshownegativeimpacts Availableonlinexxxx inthefuturecanbemitigatedbyapplyingdifferentregulateddischargeregimesduringcriticalperiodsforsalmon survival.Aspatiallyexplicitindividual-basedmodelwasusedtopredictjuvenileAtlanticsalmonpopulationabun- Editor:D.Barcelo danceinaregulatedriverunderarangeoffuturewatertemperatureanddischargescenarios(derivedfromclimate Keywords: datapredictedbytheHadleyCentre'sGlobalClimateModel(GCM)HadAm3HandtheMaxPlankInstitute'sGCM Salmonids ECHAM4),whichwerethencomparedwithpopulationspredictedundercontrolscenariosrepresentingpastcondi- Individual-basedmodelling tions.Parrabundancedecreasedinallfuturescenarioscomparedtothecontrolscenariosduetoreducedwetted Populationabundance areas(withtheeffectdependingonclimatescenario,GCM,andGCMspatialdomain).Toexaminethepotential Hydropowerregulation formitigationofclimatechange-inducedreductionsinwettedarea,simulationswererunwithspecificminimum Mitigation dischargeregimes.Anincreaseinabundanceofbothparrandsmoltoccurredwithanincreaseinthelimitofmin- Climatescenarios imumpermitteddischargeforthreeofthefourGCM/GCMspatialdomainsexamined.Thisstudyshowsthat,inreg- ulatedriverswithupstreamstoragecapacity,negativeeffectsofclimatechangeonAtlanticsalmonpopulationscan potentiallybemitigatedbyreleaseofwaterfromreservoirsduringcriticalperiodsforjuvenilesalmon. ©2018PublishedbyElsevierB.V. ⁎ Correspondingauthor. E-mailaddress:[email protected](L.E.Sundt-Hansen). https://doi.org/10.1016/j.scitotenv.2018.03.058 0048-9697/©2018PublishedbyElsevierB.V. 1006 L.E.Sundt-Hansenetal./ScienceoftheTotalEnvironment631–632(2018)1005–1017 1.Introduction ThecurrentstudyexploreshowthefutureabundanceofAtlantic salmonmaybeinfluencedbyclimate-inducedchangesinwatertem- Climatechangeisexpectedtomodifythermalandhydrologicalre- peratureanddischargeinaregulatedriver,andhownegativeim- gimes of rivers, with uncertain consequences for aquatic species pacts of climate change may be mitigated by implementing (KnouftandFicklin,2017).IntheNorthernHemisphere,climatemodels different minimum discharge regimes during critical periods. A havepredictedanincreaseinairtemperatureandwinterprecipitation, spatially-explicitindividual-basedmodelisusedtopredictpopula- butadecreaseinsummerprecipitation(IPCC,2007;Schneider,Laize, tion abundance under future climate regimes for comparison to Acreman,andFlorke,2013).ForSouthernNorway,run-offisexpected abundanceundercontrol(past)regimes,withafocusonhowthe toincreaseinwinter,butdecreaseinsummer(Schneideretal.,2013). dischargeregimeaffectswettedareaandconsequentlycarryingca- Episodesoflowsummerdischargesareexpectedduetolongerperiods pacity.Theeffectofclimatechangeonsalmonidfreshwaterabun- withlowprecipitationandlowerlevelsofgroundwaterinsummer dancehasbeenexaminedinpreviousstudies(Battinetal.,2007; (Hanssen-Bauer, Førland, Haddeland, et al., 2015). The projected Hedgeretal.,2013b;Leppi,Rinella,Wilson,andLoya,2014),but changesintemperatureanddischarge(IPCC,2007),i.e.increased thisistotheauthors'knowledgethefirststudytoincludeminimum temperatures and changed discharge patterns, may have dischargeregimescenarios,implementedasmitigationstrategiesfor detrimentaleffectsonaquaticorganismsinhabitingriversbecause climatechange,intothemodelpathway. bothtemperature(Angilletta,Niewiarowski,andNavas,2002)and dischargeregimesinfluenceimportantlifehistorytraitsofmany aquaticspecies,suchasgrowthandmortality(Heino,Virkkala,and 2.Materialandmethods Toivonen,2009). TheAtlanticsalmonlifecycleisdividedintotwophases:thejuvenile 2.1.Studyarea phase,whichtakesplaceinfreshwater;andtheadultphase,which largelytakesplaceintheocean.Forjuvenilesalmonintheriver,the Thestudyriver,theRiverMandalselva(58.2°N,7.5°E),isoneofthe temperatureanddischargepatternarekeyparametersforsurvival largestriversinsouthernNorway.Theriveris115kmlongandisregu- andgrowth.Thewatertemperaturewillaffectthespeedofphysiologi- latedwithsevenhydropowerstationsandseveralreservoirs(Ugedal, calandbiochemicalreactionsofthispoikilothermic(cold-blooded)or- Larsen,Forseth,andJohnsen,2006).Atlanticsalmonandbrowntrout ganism(Angillettaetal.,2002).Salmongrowthisstronglyinfluencedby (SalmotruttaL.)canmigratefromthesea47kmupstreamtothenatural temperature(Forseth,Hurley,Jensen,andElliott,2001)andsizedeter- waterfallofKavfossen(Fig.1).Themeandischargeattheoutletofthe minedbygrowthisanimportantfactorforjuvenilesurvival(Einumand mostdownstreamhydropowerstation(Laudal)is88m3s−1;average Fleming,2007).Thestagewhenjuvenilesmigratefromtherivertothe lowestdailydischargesrangebetween18.6m3s−1duringsummer oceanassmoltsisalsolargelydeterminedbysize(Metcalfe,Fraser,and (Jul–Sep) and 33.1 m3 s−1 during winter (Jan-Mar) (Ugedal et al., Burns,1998)andthereforestronglyinfluencedbywatertemperature. 2006).TheMandalsystemhasatotalstoragecapacityof358million Thedischargedeterminesthewettedareaoftheriver,dependingon m3 (NVE Atlas, https://atlas.nve.no), providing the opportunity to theriverprofile,withthewettedwidthofa“U”-shapedcrosssection storewaterfromextrawinterprecipitationandtoreleasethiswater changing less with changing discharge than a “V”-shaped cross indrierperiodsoftheyear. section.Thewettedareaoftherivercontrolstheriver'scarrying Inthebeginningofthe20thcentury,thesalmonfisheriesinthe capacity,butcarryingcapacityisalsodependentonhabitatquality riverwerehighlyproductive,butacidificationduringthe1960s forjuvenilesalmon,i.e.shelteravailabilityisofgreatimportance extirpatedtheoriginalsalmonpopulation.However,limingwas (Finstadetal.,2009).Thewettedareastronglyinfluencesdensity- initiatedin1997andanewsalmonstock,resultingfromstrayers dependentmortalityinearlylife-stages,acommonbottleneckfor fromotherriversandthereleaseofeggsandfryfromastockina salmon abundance in rivers (Einum, Sundt-Hansen, and Nislow, nearby river, rapidly increased in size. The catch peaked at 2006).Thus,watertemperatureanddischargedrivefundamental 10 tonnes in 2001. The present salmon stock in the river is a biologicalmechanisms,bothattheindividual-andpopulationlevel. geneticblend,withlikelyweakornolinkstotheoriginalstock Atlanticsalmonhasbeeninalong-termdeclineinmostofitsdistri- (Hesthagen,2010). butionarea,bothintermsofthenumberofpopulationsandintermsof abundanceinfreshwateraswellasthemarineenvironment(Hindar, Hutchings,Diserud,andFiske,2011;Windsor,Hutchinson,Hansen, 2.2.Modellingprocedure andReddin,2012;ICES,2013).InNorway,approximately30%ofrivers withsalmonstocksareaffectedbyhydropowerdevelopment,withef- Water temperature and discharge (and consequently wetted fectsonsalmonstocksrangingfromextirpationtomodestreductions area)intheregulatedstretchdownstreamoftheLaudalhydropower inabundanceorevenpositiveeffects(Hvidstenetal.,2015).Environ- station–stretchingdownstream20kmfromtheoutletofthestation mentalflowpracticesinregulatedriversarecommonlydominatedby (upstreamdistance=20.5km)tothestartofthetidalzone(up- adefinedconstantminimumdischargevalueforwinterandahigher streamdistance=700m–weregeneratedforselectedclimatesce- constantvalueforsummer(Alfredsen,Harby,Linnansaari,andUgedal, nariosusingamodellinghierarchy(Fig.2)).Coarse-scalepredictions 2012).Thesevaluesareoftenexceeded,butcannotbelower.Inamajor- ofairtemperatureandprecipitationfromGlobalClimateModels ityofregulatedriversystems,watermassesarecommonlystoredinres- (GCMs)weredownscaledusingaRegionalClimateModel(RCM)to ervoirsduringperiodsofhighrun-offandreleasedinperiodswhen providefiner-scalepredictions(spatialresolution=1km2)ofair electricityisrequired,usuallyinwinter.Thusdischargeregimesmay temperatureandprecipitationacrossthecatchmentencompassing beadjusted,whichoftenleadstoincreasedwinterdischargesandre- theMandalselva.Thesedatawereusedtodeterminethehydrologi- ducedspringfloodsinNorwegianrivers,comparedtounregulatedriv- calregimeofthecatchmentusingahydrologicalmodel.Giventhat ers. In a few documented cases, hydropower developments have theMandalselvaisregulatedandthehydraulicpropertiesofthe increasedsmoltabundanceinpartsoftheriver(Ugedaletal.,2008) riverareinfluencedbyhydropoweroperationaswellasthehydro- ortotalsmoltabundance,inbothcasesduetoincreasedwaterdischarge logicalregime,outputsfromthehydrologicalmodelwereusedina duringwinter(Hvidstenetal.,2015).Activemanagementofriverdis- hydropowerproductionmodel,toprovidewatertemperatureand chargepatternsinregulatedwatercoursesmaythereforerepresenta discharge.Ahydraulicmodelwasthenusedtoderiveweeklywetted rarecaseofeffectivemitigationofnegativeclimatechangeeffectson area,acriticalcomponentoftheindividual-basedpopulationmodel, fishpopulations(PiouandPrevost,2013). fromthedischargedata. L.E.Sundt-Hansenetal./ScienceoftheTotalEnvironment631–632(2018)1005–1017 1007 2.3.GlobalClimateModel(GCM)andRegionalClimateModel(RCM) ClimatedatausedinthisstudywereprovidedbytheNorwegianMe- teorologicalInstituteDNMI(Engen-Skaugen,Haugen,andHanssen- Bauer,2008).ThesedatawerederivedfrompredictionsfromtheHadley Centre'sHadAm3HGCMandtheMaxPlankInstitute'sECHAM4GCM. Scenarios used were both for (1) future climates, the SRS A2 (high CO )andB2(lowtomediumCO )emissionscenarios(2071–2100), 2 2 and(2)acontrolclimate,SRSCN(1960–1990)(Table1).Thesescenarios areusedbytheDNMItoevaluatethefutureclimateinNorway.TheSRS A2andB2emissionscenariosaredescribedintheIPCCFourthAssess- mentReport(IPCC,2007)andrepresenttheclimateforcingthrough CO emissionsforspecificscenariosoffuturedevelopmentoftheworld 2 (IPCCSRES2000),andarerespectivelysimilartotheRCP8.5andRCP 6.0scenariosusedintheIPCCFifthAssessmentReport(IPCC,2014) (seeVanVuurenandCarter,2014).GCMdatawereata55×55kmre- gional domain (spatial resolution) for the HadAm3H model, and at boththisdomainandafinerdomain(25×25km)fortheECHAM4 model.GCMpredictionsoftemperatureandprecipitationwerethen downscaled toafinerresolution (grid of1km2cells) by theDNMI usingtheRegionalClimateModel(RGM)HIRHAM(Christensenetal., 2007;May,2007)whichwasbiasadjustedbasedonobservationsof localclimate(Engen-Skaugen,2007).Theuseofdatafromdifferent GCMswithdifferentspatialdomainsallowedtheinvestigationof how pronepredictionsofsalmonpopulationabundancewereto theclimatemodeloutputsonwhichtheanalyseswerebased. 2.4.Hydrologicalmodel DownscaledtemperatureandprecipitationpredictionsfromtheRCM wereusedasinputtoahydrologicalmodeltopredictwaterinflowsalong thecompletewatercourseoftheMandalselva.Hydrologicalmodelling wasperformedintheopensourcemodelplatformENKI(Kolbergand Bruland,2012).Thisgriddedmodelsimulatesinflowbasedonmeteoro- logicalinputs(temperatureandprecipitation)andcatchmentcharacter- istics(elevation,soilwaterstorage,distributionofsub-catchmentsand rivernetwork).Themodelinterpolatestemperatureandprecipitation acrossthecatchmentusingInverse-DistanceWeightingandKriging,re- spectively.Soil-moistureandsurfacerunoff(fromexcesssoilmoisture) are then calculated using the response function developed for the HydrologiskaByrånsVattenbalanavdeling(HBV)model.Toachievea bettermodelofsoilmoistureandsurfacerunoff,themodelincludesan evaporationroutinebasedontheroutineinLandPinemodel(Rinde, 2000)andasnowaccumulation/snowmeltroutine(seeKolberg,Rue, andGottschalk,2006).Theformerreducessoilmoistureandrunoff(par- ticularlyinthesummer);thelatterintroducesalaginsoilmoistureand runoffbystoringprecipitationassnowduringwinter,andreleasingthis inspring.Themodeliscalibratedforindividualsub-catchmentswithin theentirecatchmentbysettingthemodeltorunforoneindividual sub-catchmentatatimeandvalidatingwiththeremainingcatchments (iterating with the Shuffled Complex Evolution Method, with Nash Sutcliffe'sefficiencycriteria(NashandSutcliffe,1970)). 2.5.Hydropowerproductionmodel Thehydropowerproductiondischargeintheriverwassimulated usingthenMAGmodel(KillingtveitandSælthun,1995;Killingtveit, 2004)basedoninputsfromthehydrologicalmodel.ThenMAGmodel simulateshydropoweroperationsforwholesystemsconsistingofhy- dropowerstations,reservoirsandtransfers,andpredictsriverdischarge amongothervariables.Amodelofthehydropowersystemwithallres- ervoirsandpowerstationsintheMandalselvawasused(Fjeldstad, Alfredsen,andBoissy,2014)andtheoperationalstrategyofthepower Fig.1.RiverMandalselvawiththemodelledstretchoftherivermarkedwithathickblue systemwaskeptunchangedamongtheclimatescenarios.Watertem- line,fromtheoutletofLaudalPowerstationanddownstreamtoKrossen(startofthetidal zone).(Forinterpretationofthereferencestocolorinthisfigurelegend,thereaderis peraturedownstreamoftheLaudalhydropowersystemwasfound referredtothewebversionofthisarticle.) fromregressionequationsbasedonobservedwatertemperature,air 1008 L.E.Sundt-Hansenetal./ScienceoftheTotalEnvironment631–632(2018)1005–1017 Fig.2.ThemodelpathwayfromdownscalingoftheGlobalClimateModel(GCM),viaaRegionalClimateModel(RCM),tohydrological-,hydropower-andhydraulic-models,andfinally downtotheindividual-basedpopulationmodel,IB-salmon. 8 temperature,productiondischargeandsimulatedinflow.Watertem- < T ¼0:13 T b−6 res a peratureinthemodelledwatercourse,T ,wascalculatedbyweighting T ¼0:0567T 2þ0:5117T þ1:5617 −6≥T ≤1 ð3Þ w : res a a a thetemperaturecontributionsfromnaturalinflowandthereservoir Tres¼1:0324Taþ1:1685 TaN1 immediatelyupstream(Eq.1): Q Q Tw¼QinTinþQresTres ð1Þ 2.6.Hydraulicmodel tot tot Inordertoestimatetherelationshipbetweendischargeandwetted whereTinandTresarewatertemperaturesfromnaturalinflowandthe areainthemodelledstretch,theHEC-RAS®(2008)hydraulicmodel reservoirrespectively,QinandQresaretherespectivedischarges,and was applied. This model can simulate discharges within rivers, for Qtotisthetotaldischarge.Watertemperaturesinthenaturalinflow bothsteady-flowsurfaceprofiles,and1-Dand2-Dunsteadyflowcondi- and in the reservoir were estimated from regression relationships tions.TheHEC-RASmodelwasusedtodeterminedischarge–wetted establishedwithairtemperature,Ta(Eqs.2and3). widthrelationshipsforeachofthreestationswithdifferentchannel 8 profilecharacteristicsrepresentativeofthemodelledstretchofthe < T ¼0:5 T b−3 river:channelprofilesweremore“U”-shaped,more“V”shaped,andin- in a :Tin¼0:0658Ta2þ0:5287Taþ1:5707 −3≥Ta≤1 ð2Þ termediatebetweenthesetwo.Themodelwascalibratedwithfieldob- Tin¼0:9567Ta−0:8926 TaN1 servations of water level measurements in each station at a single Table1 OverviewofGlobalClimateModels(GCM),emissionscenarios,spatialresolutionsofGCMoutputdata,andscenariossimulationsusedwiththeindividual-basedmodel(IBM). GCM Emissionscenario Domain Spatialresolution(km) IBMscenarioname HadAm3H A2+control RegClim 55×55 Had.Reg.A2 Had.Reg.CN HadAm3H B2+control RegClim 55×55 Had.Reg.B2 Had.Reg.CN ECHAM4 B2+control RegClim 55×55 ECHAM.Reg.B2 ECHAM.Reg.CN ECHAM4 B2+control NorAcia/NorClim 25×25 ECHAM.Nor.B2 ECHAM.Nor.CN L.E.Sundt-Hansenetal./ScienceoftheTotalEnvironment631–632(2018)1005–1017 1009 discharge by varying the Manning's roughness coefficient 2.7.1.Modelfunctions (0.020–0.095)untilthemodelpredictionmatchedfieldobservations. The IBM used (IB-salmon) (Hedger et al., 2013a; Hedger et al., 1-Dsteady-statesimulationswerethenperformedatvariousdischarges 2013b;Sauterleuteetal.,2016)isaspatially-explicitmodeldesigned toestablishdischarge–wettedwidthcurves(Fig.3a).Curveswereex- forpredictingpopulationcharacteristicsofthefreshwaterstageofjuve- trapolatedfordischargesoutsideofthevalidateddischargerangeofthe nileAtlanticsalmon,butalsomodelsseasurvivalandthereturnofsur- HEC-RASsimulations.Thedischarge–wettedwidthcurvesofthethree vivingadultsfromthesea.Themodelhasatime-stepintervalofone stationswerethentransferredtothe50mlongriversectionswithcor- week,withtheriverdividedintoaseriesof50msections.Individual respondingchannelprofiles(determinedbyaerialphotography)and life-historytraits(growth,smoltificationtiming,fecundity,mortality) scaledbytheratiobetweenthemaximumwettedwidthofthe50m andothercharacteristics(location,migration)aremodelledusingem- sectioninquestion(again,determinedbyaerialphotography)andthe pirical functions (Hedger et al., 2013a; Hedger et al., 2013b). Life- maximumwettedwidthoftherespectivestation.Thisenabledtheesti- stages modelled as individual elements are parr (juveniles in the mationofthedischarge–wettedwidthrelationshipforeachsection river),smolts(parrthathavesmoltified,whichthenmigratetosea), throughouttheentiremodelledstretchoftheriver.Thewettedarea searesidentadults,andreturningadults(adultsthathavereturnedto of each section was calculated by multiplying the section wetted therivertospawn).Themaininputparametersofthemodelarewetted widthbythesectionwettedlength(50m),givingahighlynon-linear area(dependentonchannelprofileanddischarge),watertemperature, relationshipbetweendischargeandtotalwettedareainthemodelled spawninglocationandarea,andparrcarryingcapacity.Atthebeginning watercourse(Fig.3b). ofasimulation,annualeggdepositionisreadfromafile,andbinned into sections according to relative spawning habitat quality. Later whenafullage-distributionofspawningadultshasreturned,eggsare 2.7.Atlanticsalmonpopulationmodel depositedinsectionsasafunctionofspawningfemaleabundanceand bodymassinwhichthespawningfemalewasborn.Theweeklyparr Anindividual-basedmodelling(IBM)approachwasusedtopredict growth is determined using a Ratkowsky-type model (Ratkowsky, theimpactofclimatechangeandmitigationmeasuresonpopulation Lowry,Mcmeekin,Stokes,andChandler,1983)parameterizedwithex- abundances(expressedinthisstudyastotalnumberofindividuals perimentaldataongrowth/temperaturerelationshipsforNorwegian withinthemodelledstretch)andchangesinlifehistorycharacteristics Atlanticsalmonjuveniles(Jonsson,Forseth,Jensen,andNæsje,2001) ofAtlanticsalmonwithinthemodelledwatercourse.Thisapproach (Eq.4). wasusedbecauseseveralaspectsoftheprocessesaffectingthesalmon populationwouldhavebeendifficulttoparameterizeusingpopulation 8 differentialequations(seeDeAngelisandGrimm,2014).Forexample, >>>< Mt¼Mt−1 TbTLorTNTU coohpnaelnroagcteaelsacitanrarmyspionargttiaacllailtpyyalcoriectasyul,llleotevcdaell.frrLiovomecarlcpchrhoaafinnlgege,eslsoiicnnalwwheeattbtteeitddataaqrreueaaa,lditweypheainnchdd >>>:Mt¼ Mbt−1þb dðT−TLÞ(cid:2)110−0egðT−TUÞ(cid:3)!!ð1=bÞ T≥TU&T≤TU ð4Þ local parr biomass, and also the spatial configuration of different carrying capacities, channel widths and biomass. Running an IBM allowed population characteristics to dynamically emerge from whereMistheindividualbodymassfortimet,Tistheweeklymean heuristicfunctionsthatwerewellparameterizedatthespatiallylocal temperature,TLandTUareloweranduppertemperaturesforgrowth, andindividuallevel. andb,dandgareparametersofthemodel. Fig.3.Relationshipbetweendischargeand(a)wettedwidthintheHEC-RASstations,and(b)totalwettedareainthemodelledwatercourse. 1010 L.E.Sundt-Hansenetal./ScienceoftheTotalEnvironment631–632(2018)1005–1017 Bodylength,L,ispredictedfrombodymasseveryweek,usinga powerfunctionrelationship(Eq.5). (cid:4) (cid:5) 1=3 L¼ 105M=0:84 ð5Þ Theannualsmoltificationprobability(appliedinweekofyear20), SP,isestimatedforeachindividualasalogisticfunctionofbodylength, L(Eq.6). 8 < eðp1þðp2(cid:2)LÞÞ :SP¼1þeðp1þðp2(cid:2)LÞÞ if L≤250 ð6Þ SP¼1 if LN250 wherep1andp2areparametersofthemodel. Parrdensitydependentmortalityinanygivensectionisdependent onthetotalbiomassofparrwithinthatsection(thesumofindividual bodymassesofallparrinthesection)andthetotalcarryingcapacity ofthesection(thetotalbiomassthatcanbesupported,whichisthe productofthecarryingcapacityperunitareaandthetotalwetted areaofthesection).Ifthebiomasswithinasectionexceedsthetotalcar- ryingcapacity,surplusparrareforcedtomigrateoutofthesection:they mayeithermigratetoanewsectionorexperiencemortality(density dependentmortality)(Eq.7). 8>< Bs;t¼Bs;t−1 Bs;t−1bK Fdiegn.si4t.ieOsbcsoemrvpeadrepdatrorsaimbuunladtaendcpeaarrndabaugnedadniscteriabnudtiaognefdriosmtribeuleticotnroffirosmhintghejusavlemnoilne >: BDs;st;t¼¼Ksþ(cid:2)BDs;ts−þ1K;t−(cid:3)1 Bs;t−1≥K ð7Þ abundancemodel(IBsalmon). whereBisthetotalparrbiomass(gm−2)withinthesection,s,attime,t, GiventhattheIBMincludedprobabilisticfunctions,generatedpop- Disthetotalparrbiomass(gm−2)ofindividualsthatdisperseoutofthe ulationsabundanceswouldvaryaccordingtosimulation,evenwiththe sectionandsurvive,Kisparrcarryingcapacity(gm−2)ands istheparr samesetofparametervalues.However,apreliminaryanalysisofsimu- survivalprobability.Densitydependentmortalitymayincreaseifthe lationsshowedthattheeffectofstochasticityingeneratedoutputwas sectionbiomassincreases(duetobodymassgrowthoranincreasein small.Forexample,whenrunningtenseparatesimulationsforeachof abundanceviarecruitmentorimmigration)orthetotalcarryingcapac- theclimatescenarios,thecoefficientofvariation(CV)forannualsmolt ityofthesectiondecreasesduetoadecreaseinthewettedarea. abundancewasalwaysb0.2%foreachclimatescenario.ThisCVwas negligibleincomparisontothedifferenceinsmoltabundancebetween 2.7.2.ParameterizingandrunningthemodelintheMandalselva differentclimatescenarios. Themodelwasparameterizedtorunonapartoftheriverstretching 20kmdownstreamfromtheoutletofthemostdownstreamhydro- powerstation(Fig.1).Dischargeintothisstretchoftheriverisregulated 2.8.Mitigationinfutureclimates bythehydropoweroperator:withwaterenteringfromaturbineoutput andfromanupstreambypass(minimumdischargesof3ms−1and Inordertoexplorepotentialmitigationmeasuresforclimatechange 1.5ms−1insummerandwinterrespectively).Theminimumdischarge andhowthesewouldaffectjuvenileabundance,simulationswererun wasmanipulatedinfuturescenariostostudytheeffectofmitigation withtheimplementationofminimumdischargeregimesduringsum- measures.Themodelwasparametrizedandvalidatedusingavailable merweeks(week20–40).Summerdischargesinmostfuturescenarios dataonelectrofishingjuveniledensities(Fig.4)andjuvenilesizeatage were less than those of the corresponding control scenarios (see (basedonelectrofishingatsevenstationsinOctober/November,yearly Table2).Therefore,regulatingdischargessothattheyweregreater from2002to2010,NorwegianEnvironmentAgency),andsmoltabun- thanwhatwouldoccurnaturallyduringthisperiodallowedforcompar- danceatage(seeUgedaletal.,2006).Habitatqualityofeachsectionin ingpotentialmitigationmeasuresforclimatechange.Fivesummertime themodelledstretchwasbasedonfieldsurveysofsubstratetypeand minimumdischargeregimeswereexamined:2,4,6,8and10m3s−1. size,spawninghabitatandsheltermeasurements(Finstadetal.,2007). Spawninghabitatwasdeterminedfromsurveydata.Aspawninghabitat Table2 qualityindex,varyingbetweenzero(nospawning)andone(maximum Thesummarystatistics(median,min,maxandrange)ofdischarge(m3s−1)incontrol spawning),wasusedtoallocatetheinitialeggdepositionatstartofsim- (1961–1990)andfuturescenarios(2070–2100),forallyearsandforthesummerperiod ulationstothedifferentsections. (week20–40). Themodelwasrunwitha“burn-in”timeof10yearstoallowforthe Allyear Summerperiod simulationofasalmonpopulation,followedby40yearsofsimulationto (week20–40) provideoutputdataonthepopulation.Theburn-in-timeisusedtocre- Scenarios Median Min Max Range Median Min Max Range atestablepopulationprocesses(seeWilliams,O'Brien,andYao,2017)– inthecaseofIB-salmon,thisinvolvedgeneratinganage-andsize- Control 64.6 2.4 445.0 442.6 53.1 2.4 260.0 257.6 specificpopulationdistributionfromaninitialestimateofeggdeposi- (1961–1990) 67.7 1.8 490.0 488.2 52.4 1.8 311.3 309.6 66.4 2.4 462.5 460.1 55.0 2.4 267.2 264.8 tion.Predictionsfromtheburn-intimewereexcludedfromtheanalysis. Future 61.37 0.61 347.5 346.9 12.5 0.6 116.5 115.9 Foranalysisoftheeffectofchangesinclimatescenariosanddischarge (2070–2100) 60.33 1.5 521.83 520.33 18.5 1.5 197.6 196.1 regimesonpopulationabundance,onepopulationwassimulatedper 75.03 3.08 499.21 496.13 38.4 3.1 374.8 371.7 64.84 1.08 390.41 389.33 21.5 1.1 198.2 197.1 setofparametervalues. L.E.Sundt-Hansenetal./ScienceoftheTotalEnvironment631–632(2018)1005–1017 1011 Fig.5.Modelledmeanweeklywatertemperature,dischargeandwettedareainthecontrol(1961–1991)andfuture(2071–2100)scenarios.Curvesshowmeanweeklyvaluesacrossall outputyearsofthesimulation:controlscenarios(filledsquares);A2scenarios(circles);andB2scenarios(triangles).Hatchedareasshowthesummerseason. Theseminimumdischargeregimesaresustainablefromthehighstor- reducedinallfuturescenarioscomparedtothecontrol,andoccurred agecapacityoftheMandalselvasystem. afewweeksearlierintheyear.Dischargesinthesummermonthsof June,JulyandAugustwerelowerinthefuturescenariosthaninthecon- 3.Results trol scenarios (Fig. 5). Of the future scenarios, the ECHAM.Reg.B2 3.1.Climatechangeinfuturescenarios Table3 Meanwatertemperatures(°C)inthecontrol(1961–1991)andfutureclimatescenarios 3.1.1.Hydraulicpredictions (2071–2100). Watertemperatureincreasedunderallfuturescenarios(Had.Reg. Scenarios Globalcirculationmodels Temperature(°C) A2,Had.Reg.B2,ECHAM.Reg.B2andECHAM.Nor.B2)comparedtothe controlscenarios(Fig.5,Table3).TheHad.Reg.B2scenarioshoweda Control Had.Reg.CN 6.25(SD±5.16) Control ECHAM.Reg.CN 6.27(SD±5.16) slightlylowermeantemperatureduringsummerweeksthantheHad. Control ECHAM.Nor.CN 6.27(SD±5.14) Reg.A2scenario,butonethatwasstillseveraldegreeshigherthanthe Future Had.Reg.A2 9.05(SD±5.76) controlscenario. Future Had.Reg.B2 8.48(SD±5.52) Winterdischargesinallfuturescenariosweregreaterthaninthe Future ECHAM.Reg.B2 8.66(SD±5.76) corresponding control scenarios. In contrast, the spring flood was Future ECHAM.Nor.B2 8.89(SD±5.16) 1012 L.E.Sundt-Hansenetal./ScienceoftheTotalEnvironment631–632(2018)1005–1017 Fig.6.Meanparrabundance(a)andsmoltabundance(b)accordingtoage(0+,1+,2+,3+,4+)inthecontrol(1961–1991)andfuture(2071–2100)scenarios. scenariopredictedahigherdischargeinsummerthanthefutureHad. Themajorityofsmoltsinthefuturescenarioswere2+,comparedto3 Reg.A2,Had.Reg.B2andECHAM.Nor.B2scenarios. +and4+inthecontrolscenarios. Insummer,wettedareawasconsiderablyreducedinthefuturesce- Theweeklydensitydependentmortalityofparr(asaproportionof nariosHad.Reg.A2,Had.Reg.B2andECHAM.Nor.B2.Incontrast,wetted thetotalparrabundance)wasinverselycorrelatedwithwettedarea areapredictedundertheECHAM.Reg.B2scenario,didnotchangecon- andwashighestduringthesummerperiodinallfutureclimatescenar- siderablycomparedtothecontrolscenario.(Fig.5). ios,whenthewettedareawaspredictedtobesmall(Fig.7).Futuresce- narios with greater summertime reduction in wetted area (HAD scenarios)causedgreaterdensitydependentmortalitythanscenarios 3.1.2.Salmonpopulationpredictions. withsmallersummertimereductionsinwettedarea(ECHAMscenarios). Parrabundancedecreasedinallfuturescenarios,comparedtothe respectivecontrolscenarios(Fig.6a).Thisreductionwas,however, 3.2.Mitigationofclimatechange small in the ECHAM.Reg.B2 scenario. Because of changes in age at smoltificationandsubsequentemigrationtosea,theagecomposition 3.2.1.Dischargeandwettedarea inallfuturescenariosshiftedtowardsyoungerparrincomparisonto Dischargesinthefuturescenarioswerereducedduringsummer thecontrolscenarios.Inthefuturescenarios,the3+and4+juvenile weeks,particularlyinprojectionHad.Reg.A2,Had.Reg.B2andECHAM. ageclass(inyears)disappearedandaverysmallproportionof2+ Nor.B2,andlesssointheECHAM.Reg.B2scenario.Implementationof wasleft,comparedtothecontrolscenarios. minimumdischargeregimes(from2to10m3s−1)increasedsummer Smoltabundanceinthreeofthefuturescenarios–Had.Reg.A2,Had. wettedareafortheHad.Reg.A2,Had.Reg.B2andECHAM.Nor.B2scenar- Reg.B2andECHAM.Nor.B2–waslessthanintherespectivecontrolsce- ios(Fig.8).TheeffectwasstrongestfortheHad.Reg.A2wheresummer narios(Fig.6b).However,smoltabundanceintheECHAM.Reg.B2sce- wettedareasincreasedfrom≈4.5×106m2underconditionsofno nario was greater than in the respective control scenario. The age assignedminimumdischargeto≈7×106m2underaminimumdis- compositionofsmoltschangedinallfuturescenarios,withageshifting chargeof10ms−1.Incontrast,implementationofminimumdischarge towardsayearyoungercomparedtosmoltsinthecontrolscenarios. regimeshadlittleeffectonwettedareasfortheECHAM.Reg.B2scenario Fig.7.Meanweeklydensitydependentmortalityofparr(proportionofthetotalweeklyparrabundance)andthecorrespondingweeklywettedareainthecontrol(1961–1990)andfuture (2071–2100)scenarios.Barplotsandcurvesshowmeanweeklyvaluesacrossalloutputyearsofthesimulation.Hatchedareasshowthesummerseason. L.E.Sundt-Hansenetal./ScienceoftheTotalEnvironment631–632(2018)1005–1017 1013 1014 L.E.Sundt-Hansenetal./ScienceoftheTotalEnvironment631–632(2018)1005–1017 Fig.8.Meanweeklywettedareainthecontrol(1961–1990)andfuture(2071–2100)scenarioswithaminimumdischargeinweekofyear20–40of2,4,6,8and10m3s−1inthefuture scenarios.Curvesshowmeanweeklyvaluesacrossalloutputyearsofthesimulation.Hatchedareasshowthesummerseason. duringsummermonthsduetosummerdischargesin thisscenario forsalmonabundanceinfutureclimateswereidentified.Thisstudysug- beinggreaterthanthoseassignedintheminimumdischargeregimes. geststhattheabundanceofAtlanticsalmoninfutureclimateswillde- creasefortheregionmodelled,andelucidatesmechanismsimportant 3.2.2.Parrandsmoltabundance forregulationofjuvenileAtlanticsalmonindividuals.Lowwaterdis- WhenrunningtheIBMwitharangeofdifferentminimumdischarge chargeinsummerwasidentifiedasapossiblebottleneck,andsimula- regimes,parrandsmoltabundanceincreasedwithincreasingminimum tionofdifferentminimumdischargeregimesshowedthatchangesin dischargeinallscenariosexceptfortheECHAM.Reg.B2scenario(Fig.9). riverregulationmaybeapossiblemitigationmeasure.Theexactchange Theincreaseinabundanceofbothparrandsmoltoccurredwhenmin- inAtlanticsalmonabundancedependedontheGCMandGCMdomain imumdischargeinsummerwas4m3s−1orgreater.TheECHAM.Reg.B2 usedtosupplypredictionsofairtemperatureandprecipitation,sofur- scenariowasnotstronglyinfluencedbythedifferentminimumdis- theradvancesinGCMmodellingarerequiredtoincreasetherobustness chargeregimesbecausethepredicteddischargeinsummerwashigher ofthepredictionofhowanygivenAtlanticsalmonpopulationwillre- thanthatassignedintheminimumregimes. spondtoclimatechange. Reducedparrabundancewasfoundinallfutureclimatescenariosin 4.Discussion comparisontocontrolscenarios,althoughthereductionwassmallfor theECHAM.Reg.B2scenario.Threeoutoffourclimatescenarios(Had. 4.1.Effectofclimatescenariosonfuturefreshwatersalmonabundance Reg.A2,Had.Reg.B2andECHAM.Nor.B2)predictedreducedsmoltabun- dance,whereasthefourth(ECHAM.Reg.B2)predictedincreasedsmolt In this study, the effect of future climate change on an Atlantic abundance.Climateoutputsderivedfromthethreeclimatescenarios– salmonpopulationinaregulatedriverwasmodelled,andbottlenecks Had.Reg.A2, Had.Reg.B2 and ECHAM.Nor.B2 – resulted in a strong