Table Of Content

Review Climate Velocity Can Inform Conservation in a Warming World Isaac Brito-Morales,1,2,* Jorge García Molinos,3,4 David S. Schoeman,5,6 Michael T. Burrows,7 Elvira S. Poloczanska,8,9 Christopher J. Brown,10 Simon Ferrier,11 Tom D. Harwood,11 Carissa J. Klein,1 Eve McDonald-Madden,1,12 Pippa J. Moore,13,14 John M. Pandolfi,15 James E.M. Watson,1,16 Amelia S. Wenger,1 and Anthony J. Richardson2,17 Climate change is shifting the ranges of species.Simple predictivemetrics of Highlights rangeshiftssuchasclimatevelocity,thatdonotrequireextensiveknowledge Climatevelocityisasimplemetricthat or data on individual species, could help to guide conservation. We review describesthespeedanddirectionof climate movement at any point in research on climate velocity, describing the theory underpinning the concept space. and its assumptions. We highlight how climate velocity has already been applied in conservation-related research, including climate residence time, Climatevelocity is providinginforma- tionaboutclimatechangethatisrele- climate refugia, endemism, historic and projected range shifts, exposure to vant for conservation, including the climatechange,andclimateconnectivity.Finally,wediscusswaystoenhance studyofprotectedareas, novel and/ or disappearing climates, rates of the use of climate velocity in conservation through tailoring it to be more endemism,andrangeshifts. biologicallymeaningful,informingdesignofprotectedareas,conservingocean biodiversity in 3D, and informing conservation actions. Tobetterinformconservation,climate velocity can be tailored to be more biologically meaningful through the additionofdispersalcapabilities,phy- Simple Climate Metrics Could Help Conservation in a Changing Climate siological tolerance, and potential Climatechangeislikelytobecomethemostseriousthreattobiodiversitythiscentury[1,2].In routesofmovementsofspecies. fact,anthropogenicclimatechange,initiatedintheIndustrialRevolution,hasalreadyaffected ecologicalsystemsfromindividualorganismstobiomes[3,4],andhasinfluenced>80%ofall Thereisuntappedpotentialforusing climate velocity and climate-velocity biological processes [5]. Although ecological responses to climate change are numerous, trajectoriesininformingthedesignof complex,andmultifaceted,probablythemostfundamentalisthespatialredistributionofglobal protected areas and their networks, biodiversity [3]. Such species range shifts, in response to a changing climate, have been conserving ocean biodiversity in 3D, observed across terrestrial and marine ecosystems during the current warming period [6–8] andininformingconservationactions. andsincethemostrecentglacialmaximum[9,10].Understandingtheprocessesunderpinning Tostimulatefutureresearchusingcli- rangeshiftsandpredictingtheirpotentialoutcomeswillbenecessarytoinformconservation mate velocity, we introduce the R andtoreduceriskstofoodsecurity,humanhealth,andtheviabilityofnumerousindustriesthat packagevocc. dependon ecosystemservices,includingforestry, fisheries,andeco-tourism. Mechanisms underpinning range shifts are a blend of a species exposure, sensitivity, and vulnerabilitytoclimatechange,combinedwithitsadaptivecapacity[11].Ofthesecharacter- 1SchoolofEarthandEnvironmental istics, only exposure to climate change might be considered to be relatively generic across Sciences,[508_TD$DIFF]TheUniversityof species,withothertraitsbeingspecifictoindividualspeciesorpopulations.However,detailed Queensland,StLucia,QLD,Australia 2CommonwealthScientificand physiological,ecological,andevolutionarydataaremissingformostspecies,especiallyinthe IndustrialResearchOrganisation tropicsandmuchoftheglobalocean[12],andcurrentresearchprioritiesmakecollectionof (CSIRO)OceansandAtmosphere, suchdataincreasinglydifficult[13,14].Thisleavesconservationandmanagementagenciesto [510_TD$DIFF]BioSciencesPrecinct(QBP),StLucia, QLD,Australia makedecisionswithwhateveralternativetoolsareavailable.Threatstobiodiversityposedby 3ArcticResearchCenter,Hokkaido climate change have thus traditionally been quantified using rates of warming or cooling, University,Hokkaido,Japan 4[512_TD$DIFF]GlobalStationforArcticResearch, temperature anomalies, or degree heating weeks [15]. What these simple indices do not GlobalInstitutionforCollaborative conveyistherelativelikelihoodthataspeciesmightescapethethreatofclimatechangeby ResearchandEducation,Hokkaido TrendsinEcology&Evolution,June2018,Vol.33,No.6 https://doi.org/10.1016/j.tree.2018.03.009 441 ©2018ElsevierLtd.Allrightsreserved. shiftingitsdistribution.Apromisingsolutionthatretainsgenerality,butconveysmoreecolog- University,Sapporo,Hokkaido,060- icallyrelevantinformation,isthevelocityofclimatechangeor,moresimply,climatevelocity[16– 0810,Japan 5SchoolofScienceandEngineering, 18]. Climate velocity [525_TD$DIFF]is a metric that uses freely accessible environmental and climate data, UniversityoftheSunshineCoast, withouttheneedfordetailedecologicalknowledge[19],toapproximatetheobservedshiftsin Maroochydore,QLD,Australia speciesdistributions[20–23].Climatevelocitytherebyprovidesasimpleandintuitivemeasure 6CentreforAfricanConservation Ecology,Departmentof?[513_TD?$DIFF]?Zoology, ofthreatstobiodiversityposedbyclimatechange[24],andassuch,inissimplestform,isnot NelsonMandelaUniversity,Port bespoke forparticular species. Elizabeth,SouthAfrica 7Scottish[514_TD$DIFF]AssociationforMarine Science,Oban,UK Here,weexplorethemeaning,utility,andapplicationofclimatevelocity,withaparticularfocus 8TheGlobalChange[515_TD$DIFF]Institute,[508_TD$DIFF]The on the potential for its use to guide conservation under a changing climate. We begin by UniversityofQueensland,StLucia, definingtheconceptofclimatevelocitybecausethereareseveralformulationswithdifferent QLD,Australia 9AlfredWegenerInstituteforPolar conceptualunderpinnings.Thisleadstoasummaryofthemethodologicalaspectsandcaveats andMarine?[516_TD?$DIFF]?Research,Integrative that need to be considered when using climate velocity. We then describe the different Ecophysiology,Bremerhaven, Germany applications of climate velocity that have provided new insights into many areas of climate- 10Australian?[517_TD?$DIFF]?RiversInstitute,Griffith change ecology. Next, we look to the future and explore four ways to improve the utility of University,Nathan,QLD,Australia climatevelocityinconservation.Wefocusonsimplemetricsthatuserawclimatevariables,and 11CSIROLandandWater,Canberra, ACT,Australia donotconsidervelocitiesthatcanbecalculatedfromspeciesdistributionmodelsorassem- 12AustralianResearchCouncil(ARC) blage models that scale climate space by biological data (e.g., generalized dissimilarity CentreofExcellencefor modeling)[25].Thisreviewistargetedatecologistsseekingtounderstandhowclimatechange EnvironmentalDecisions,Schoolof BiologicalSciences,[508_TD$DIFF]TheUniversityof couldaffectcommunities,andatconservationpractitionerswantingtoincludeclimatechange Queensland,StLucia,QLD,Australia intheir planning. 13InstituteofBiological,Environmental andRuralSciences,Aberystwyth University,AberystwythSY233FG, What Is Climate Velocity? UK Climatevelocityisavectorthatdescribesthespeedanddirectionthatapointonagriddedmap 14CentreforMarineEcosystems wouldneedtomovetoremainstaticinclimatespace(e.g.,tomaintainanisolineofagiven Research,EdithCowanUniversity, Joondalup,WA6019,Australia variable in a univariate environment) under climate change. From an ecological perspective, 15SchoolofBiologicalSciences,ARC climatevelocitycanbeconceptualizedasthespeedanddirectioninwhichaspecieswould CentreofExcellenceforCoralReef needtomovetomaintainitscurrentclimateconditionsunderclimatechange(Box1).Forthis Studies,[508_TD$DIFF]TheUniversityof Queensland,StLucia,QLD,Australia reason, climate velocity can be considered to represent the potential exposure to climate 16WildlifeConservationSociety, changefacedbyaspeciesiftheclimatemovesbeyondthephysiologicaltoleranceofalocal GlobalConservationProgram,Bronx, population.Despitetheintuitiveecologicalrelevance,however,climatevelocityisbasedsolely NY,USA 17CentreforApplicationsinNatural on environmentalvariablesand noton speciesdata(Box1). ResourceMathematics,Schoolof MathematicsandPhysics,[508_TD$DIFF]The Two major approaches to calculating climate velocity have emerged: namely local climate UniversityofQueensland,StLucia, velocit[526_TD$DIFF] (seeGlossary)andclimate-analogvelocity (Figure1). Localclimatevelocity isthe QLD,Australia originalmetricproposedin2009byLoarieetal.[16].Tocalculatelocalclimatevelocityata location–howfarandinwhichdirectiontheisolineofanenvironmentalvariablewouldmove– *Correspondence: only the rate of change of a variable (e.g., temperature) through time (i.e., the trend, usually [email protected] (I.Brito-Morales). estimatedastheregressionslope)andthecorrespondingspatialgradientofthatvariableare needed.Thespatialgradientrepresentsthecomplexityoftheclimatelandscape–itsmagni- tudecalculatedasthelengthofavectorresultingfromtheweightedsumofthelatitudinaland longitudinalpairwisedifferencesinvaluesoftheclimatevariablebetweenafocalcellandits nearest neighbors (Figure 1A). The associated angle of the vector gives the direction of the spatial gradient. Directions of climate velocity are reversed relative to those of the spatial gradienttoreflectresponseexpectations(e.g.,inawarmingclimate,movementtowardscooler locations). It is this dependence on neighboring (local) cells for the estimation of the spatial gradient inclimatethatgiveslocal climatevelocityitsname. Climate-analogvelocity[26]emergedasanextensionoftheclimateanalogconcept[27]–in otherwordstheidentificationofpointsinspacewithclimatessufficientlysimilartothoseofthe 442 TrendsinEcology&Evolution,June2018,Vol.33,No.6 points under consideration (Figure 1). Euclidean distances are often used as measures of Glossary multivariate climatic dissimilarity, climate analogy being set by reference to a dissimilarity Climate-analogvelocity:aclimate- threshold defined either subjectively [28,29] or using regional statistics (e.g., 95th percentile velocitymetricthatconsidersthe of the minimum Euclidean distance between each future climate and all current climates) distancebetweenpointsata particularpointintimeandtheir [26,30].Importantly,theselectedthresholdisconstantandcommontoalllocalclimates.When futureclimateanalogs,dividedbythe thepointsunderconsiderationrepresentthecurrentclimate,andtheiranalogsaresoughtina timedifference(Figure1B).Thereare futureclimate,thegeographicdistancebetweenpointscanbedividedbythetimeseparating twotypes:forwardanalogvelocity, theperiodstocomputeaspeedofclimatechange.Thedirectionfortheclimate-analogvelocity whichisthestraight-linespeedand directionrequiredtoreachagiven is provided by the relative positions of the original point and its future analog (Figure 1B). climate-analogdestinationatsome Climate-analog velocity can be further conceptualized in two related but distinct ways: ‘for- pointinthefuture(usuallyasingle ward’ analog velocity, the original formulation, and ‘backward’ analog velocity, which is the destinationforanyoriginunder inverse offorwardvelocity [28]. consideration),andbackwardanalog velocity,whichconsidersa destinationandaskswhichpoints Localandclimate-analogvelocitieshavebeenusedindifferentsituations.Localclimatevelocity (usuallyseveral)oforiginmight has usually been used for exploring potential responses of biota to single variables, usually eventuallyfeedintothedestination. Climateresidencetime:the temperature[31]butsometimesprecipitation[32].Thismetrichasbeenfavoredbyecologists amountoftimenecessaryfora whengradientsaresmoothandwherethereisonemainvariabledrivingchange(e.g.,inthe climateisolinetoemergefroma openocean;FigureS1inthesupplementalinformationonline).Localclimatevelocitycanbe specificarea(usuallyaprotected constrained by species requirements for particular habitat features, such asbeing limited to area).Itisestimatedasthe (equivalent)diameterofthearea coastalmarineregionsbytheneedforlightontheseabottom,orsubstratumtypesforreef dividedbythemeanclimatevelocity formation,orintertidalzones[33].Bycontrast,climate-analogvelocityhasusuallybeenused withinthatarea[16]. with multiple variables [34]. It has greater ecological realism in complex environments with Isoline:alineconnectingpointsof contrastingclimaticgradients,andisfavoredbyecologistsdealingwithspecieswithmultiple equalvalueacrossspace.Isoline, isocline,andisoplethareall needs.Forexample,onland,temperatureandrainfallhaveoftenbeenanalyzedinmultivariate synonyms. spaceusingclimate-analogvelocity(FigureS1).Irrespectiveoftheclimate-velocitymetricused Localclimatevelocity:theoriginal anddataavailability,researchersshouldbeawareofseveralassociatedcaveats(Box2)anda climate-velocitymetric[16]thathas suiteofmethodologicalaspects,includingwhichenvironmentalvariablestouse,theirtimeand twomaincomponentsinits calculation:atemporaltrendanda spacescales,and how tocombine multiplevariables(Box3). spatialgradient,bothforthesame climatevariable(Figure1A).Local Toencouragetherobustuseofclimatevelocityintheecologicalandconservationresearch climatevelocityisanestimateofthe communities,weprovidetworesources.ThefirstisacollectionofRfunctionsaggregatedinto instantaneousclimatevelocityofan isolineatalocation. apackage,vocc,thatisfreelyavailableonGitHub.iThispackagecalculatesthelocalclimate velocityforunivariateenvironmentaldatasets,onlocaltoglobalscales(seetheSupplemental Online Material of Hamann et al. [28] for R code for climate-analog velocity). The second resourceisalistofallfreelyavailableenvironmentaldatasets(andtheirwebsites)thathavebeen used inclimate-velocityresearch (TableS1). Current Applications of Climate Velocity Figure S2 shows conceptual relationships among different applications of climate velocity, highlighting key references, and common applications between local climate and climate- analog velocity. There are six main areas where local and climate-analog velocities have provided newinsightsintoclimate-change ecology. (i) ClimateResidenceTime From itsinception, localclimatevelocity wasused to estimatethe residencetime of current climates (climate residence time) in protected areas and different biomes under climate change [16,17]. Large protected areas, especially in hilly regions, are likely to continue to provide climate space for resident species into the next century (because air temperature decreaseswithaltitude),butsmallreservesandreservesinflatterareasarelikelytofailtodoso (Boxes 1 and 3). The latter conclusion should, however, be viewed with caution: values of TrendsinEcology&Evolution,June2018,Vol.33,No.6 443 climate residence time can be alarmingly small, but might not reflect individual species residence times because the local climate might not approach critical thermal limits for a species,thethermalrangeofaspeciesmightbelarge,oraspeciesmightbeabletoadapt behaviorally(orotherwise),therebypersistinginaclimatethatmightotherwisebeinhospitable [33,35]. Nevertheless, the primary conservation-related recommendations from studies of climate residence time seem to be defensible. They include emissions reductions to slow the rate of climate change, expanding networks of protected areas, and including more mountainous terrain [36] to increase the residence time of climates (and therefore migrating species). (ii) ClimateRefugia andRatesof Endemism Areasoflowlocalclimateandclimate-analogvelocitiescanbeconsideredascandidateareasfor protection [24,37] because they are likely to contain a consistent suite of species and their ecologicalinteractionsthatevolvedtogetherinaslowlymovingclimate.Suchareasareoften calledclimaterefugia,andhavebeenlinkedwithhighlevelsofendemism[38].Forexample,Sandel etal.[9]relatedlocalclimatevelocitybetweenthepastglacialmaximumandcurrentclimates,and usedthesetoexploreendemismofamphibians,mammals,andbirds.Relationshipsbetween climatevelocityandratesofendemismwereweakestforwide-rangingspeciesandstrongestfor Box1.TheEcologicalContextofClimateVelocity Estimatesofspeedanddirectionassociatedwithclimatevelocitycanbeconceptualizedbyconsideringairtemperatureonland.Becauseairtemperaturedecreases predictablywithelevation[487_TD$DIFF]((cid:1)6.5(cid:3)Cper1000m),astheclimatewarmsanorganismatthebottomofahilltendstomoveuphillortothenearestclimate-analogareato maintainitsthermalenvironment(i.e.,short-distancedispersal).Thiswouldyieldslow(low)climatevelocities(directeduphillortotheclosestclimateanalogarea) becauseanorganismdoesnotneedtomovefartomaintainitsthermalenvironment(FigureI,bluearrow).Conversely,flatlandscapesaremorehomogenousthermal environments,andanorganismexperiencingawarminglandscapemightneedtomigratealongdistancetoremaininitsoriginalthermalenvironment(i.e.,long- distancedispersal).Thiswouldmanifestasahighclimatevelocitydirectedtowardsthenearestoccurrenceoftheoriginaltemperature(FigureI,redarrow). Howthedistributionofaspeciesrespondstoagradualchangeinitsclimatespace[488_TD$DIFF][86]requiresconsiderationoftherelationshipbetweenthephysiologicaltolerance andtherangedynamicsofaspecies.Thiscanbeconceptualizedintwoways:arepresentationofaspeciesperformancecurveacrossalatitudinalgradient (FigureIIA),andageographicalrepresentationofspeciesdistributionacrossalatitudinalgradient(FigureIIB).Asclimatewarms,theinitiallocationofthethermal performancecurvewillshiftinspacetowardscoolerenvironments,commonlyhigherlatitudes(FigureIIA).Thisshiftinclimate,whichcanberepresentedbyclimate velocity,willtendtocausegeographicrangeshiftsinspeciesdistribution(i.e.,rangeexpansionsorcontractionsoflocalpopulations)asspeciesmaintaintheiroriginal thermalenvironment(FigureIIB). Ecological meaning Warming In hilly landscapes species distribu(cid:3)ons Objec(cid:3)ve: to maintain an would need to shi(cid:2) short distances to maintain isoline of a given variable their preferred climate envelope In flat landscapes species distribu(cid:3)ons would need to shi(cid:2) longer distances to maintain their preferred climate envelope Landscape characteris(cid:2)cs The velocity of climate change 1 Hilly landscapes = = Hilly landscapes Low Heterogeneous thermal Short distance shi(cid:2) to find environment a similar climate condi(cid:3)on 2 MFlaotre l ahnodmsocagpeneesous = = Long-disFtalantc ela snhdi(cid:2)s ctoa pfiensd High thermal environment a similar climate condi(cid:3)on FigureI.UnderstandingClimateVelocityonLand. 444 TrendsinEcology&Evolution,June2018,Vol.33,No.6 Species performance (A) Range shi(cid:332) Op(cid:415)mum at t1 Op(cid:415)mum at t2 e c n a m r o f r e P Equatorwards Polewards Warm Cold Warming Geographical species distribu(cid:415)on (B) t1 t2 Leading edge (cooling) Trailing edge (warming) Coloniza(cid:415)on event Contrac(cid:415)on Expansion Ex(cid:415)rpa(cid:415)on event FigureII.UnderstandingClimateVelocityintheContextofDynamicsSpeciesRanges.(A)Simplebell-shapedcurvefortherelationshipbetweenspecies distributionandperformance(probabilityofoccurrence)acrossalatitudinalgradientunderclimatechange.(B)Thedistributionofaspeciesshowingseparate populations(darkcircles)acrossalatitudinalgradientattwotimes.Localpopulationcontractionsandexpansionsareobservedateachrangeedgeattimet2. narrow-rangedspecies,suggestingthatareasofslowclimatevelocityprovideimportantrefugia forbiodiversityunderclimatechange.Subsequentstudiesonendemicspeciesofinsectsand mammals[39],birds[40,41],andplants[42,43]confirmthesepatternsataregionalscale,andthe patternsseemtoholdevenatlocalscaleswithinfreshwaterstreams[44]. TrendsinEcology&Evolution,June2018,Vol.33,No.6 445 (A) Local climate velocity (B) Climate-analog velocity Spa(cid:415)al gradient (cell-neighborhood method) Analog climates (distance-based method) Temporal Slope (°C year¯¹) Distance between climate analog (km) = = km year¯¹ = km year¯¹ Spa(cid:415)al Spa(cid:415)al gradient (°C km¯¹) Δ Time year(s) t1 t2 e bl a vari × × e at m × Cli × × Focal cell Focal cell Cells under analog climate condi(cid:415)ons at t2 3×3 Neighborhood of cells Eight adjacent neighbors Isocline shi(cid:332) at t2 Figure1.MathematicalandGraphicalDifferencesbetween(A)LocalClimateand(B)Climate-AnalogVelocities. (iii) HistoricRange Shifts Themagnitudeanddirectionoflocalclimatevelocityexplainsrangeshiftsinmanyspecieson land[22]andintheocean[7,21,22,45–47].Forexample,onland,globalmeta-analysisofover thepast40yearsshowedthatterrestrialspeciestrackedlocalclimatevelocityinresponseto warming to higher latitudes and higher elevation [48]. In marine systems, extensive data on marine species (128 million individual fish and invertebrate records across 360 harvested species) around North America closely track local climate velocity, both horizontally and vertically in the ocean, over the past 50 years [20]. We expect greater agreement between climatevelocityandspeciesdistributionshiftsinhomogenoussystemssuchastheopenocean and continental plains. Such homogenous systems pose fewer constraints on movement becausespeciesare moreabletofollowlocalclimatevelocity, whereasheterogeneous and complex systems have barriers to dispersal and movement that can constrain distribution shifts. In such environments, estimates of climate velocity can be modified – see section ‘TailoringClimateVelocitytobeMoreBiologicallyMeaningful’.Notealsothat,eveninrelatively homogenous regions, divergence among climate variables mediating species distributions might complicateresponses. (iv) Exposureof Organismsto ClimateChange,Migration Velocities,andthe Formationof Novel Communities Becauseclimatevelocityquantifiesthespeedanddirectionofachangingclimate,italsoquantifies theexposureofaspeciestoclimatechange[19,29].Recently,Ordonezetal.[30]usedlocal climatevelocityasoneofthreemechanismsdrivingthereshufflingofspeciesandtheemergence ofnovelcommunitiesunderclimatechange,theothertwobeingclimatenovelty(openingofnew suitableenvironments)anddivergence(discrepancyinthedirectionofchangeamonggradientsof differentclimatevariablesinrelationtothenicheofaspecies).Aselsewhere[24,26,49],slowlocal 446 TrendsinEcology&Evolution,June2018,Vol.33,No.6 Box2.CaveatsAssociatedwithClimateVelocity Climate[489_TD$DIFF]VelocityIsNotSpeciesMovement Whendiscussingclimatevelocityitissometimeseasytofallintothetrapofmakingunsupportedclaimsaboutspecies movement.Arange-edgemightbemorelikelytomoveifitisnearthethermalmaximumofaspecies,butother responsestoclimatechangearepossible,includingbehavioralmodificationandgeneticselection,whicharemore importantinspecieswithlimitedcapacitytodisperse. TheFractionalNatureoftheLocalClimate-VelocityMetriccanbeMisleading Becauselocalclimatevelocityistheratioofthetemporaltrendoverthespatialgradientinclimate,smallandbiologically irrelevanttemporaltrendsovervanishinglysmallspatialgradientscanleadtohighlocalclimatevelocities.Imaginetwo differentlocationsonthesurfaceoftheEarth,oneofwhichwarmsby[490_TD$DIFF]0.1(cid:3)Coveragiventime,andtheotherby[491_TD$DIFF]1(cid:3)Cover thecorrespondingperiod.Furtherimaginethattrackingthe[490_TD$DIFF]0.1(cid:3)Cchangeexperiencedatthefirstlocationrequires moving100km,whiletrackingthe[491_TD$DIFF]1(cid:3)Cchangeatthesecondlocationrequiresmoving50km.Thefirstlocationhas twicetheclimatevelocityofthesecond,butitignoresthemagnitudeofchangeatthelocationitself,whichcan sometimesbeabetterindexoftheneedforarangeshift. ClimateVelocityCurrentlyhasNoStandardMeasureofUncertainty Therearemanypotentialsourcesofuncertaintyinestimatesofclimatevelocitythatareusuallyunacknowledged.These include(butarenotlimitedto):(i)errorinthegriddedclimatemetricsthataffectestimatesofspatialgradientand temporaltrendintheclimatevariable,and(ii)variabilitybothwithinindividualclimateprojections(modelruns)andamong climateprojections(differentgeneralcirculationmodelsandrepresentativeconcentrationpathways).Schliepetal.[492_TD$DIFF][87] gobeyondtheconventionalfinite-differenceapproachtoclimatevelocityexplainedherebymodelingtemperature(as anexampleofaclimatevariable)asafunctionofbothspaceandtimewithinastochasticBayesianframework.This allowsthequantificationofvariabilityassociatedwithsimultaneousestimatesofspatialgradientsandtemporaltrendsin temperature[i.e.,uncertaintysource(i)above].Althoughthisprocessisnumericallycomplexandcomputationally demanding,itisanimportantfirststepinquantifyinguncertainty.Accountingforremainingsourcesofuncertainty requiresfurtherresearch. ClimateVelocitydoesNotIncludeBiologicalInformation Initssimplestform,climatevelocitydoesnotincludebiologicalinformationsuchasdispersalpotentialofspecies, landscapepermeability,habitatsuitability,orspeciesinteractions.Thislackofbiologicalinformationmeansthatclimate velocitiesaregeneral;anyincreaseinbiologicalrealismreducesthisgenerality(seesectionTailoringClimateVelocityto beMoreBiologicallyMeaningful). andclimate-analogvelocitieswereassociatedwithregionsofstrongspatialgradientsinenviron- mentalconditions(e.g.,mountains)thatwereassumedtobeleastexposedtoclimatechange(i.e., requiringshorterdispersaldistancestotrackchangesinclimate).Climateexposurecanalsobe modifiedbyclimateconnectivity(seebelow)[24,29,50].Inthiscase,exposurerelatestothecostof movingthroughclimaticallyheterogeneousland-orseascapes,possiblyaccountingforother non-climatedriversconditioningdispersal[29]. (v) Climate-VelocityTrajectoriesandClimate Connectivity To address the caution of Loarie et al. [16] that local climate velocity is discontinuous, Burrows et al. [24] developed climate-velocity trajectories by moving climate ‘tracers’ betweenneighboringgridcellsbasedonthelocalclimatevelocity.Climate-velocitytrajecto- riesthustrackspecificclimateconditionsthroughtimeascontinuouspaths(seeFigureIin Box 4). Spatially aggregated patterns of climate-velocity trajectories suggest changes in speciesrichnesswithclimate,andnotablyhighlightareasthatmightreceivefewornoclimate migrantsthroughlackofconnectionstowarmerplaces(climate‘sources’:locallywarmareas suchasequatorward-facingcoastlinesonlandorpoleward-facingcoastlinesintheocean), andareaswheretheremightbelocalextirpationsthroughlackofconnectionstocoolerareas TrendsinEcology&Evolution,June2018,Vol.33,No.6 447 Box3.MethodologicalConsiderationsWhenApplyingClimateVelocity WhichEnvironmentalVariables? Most analyses of climate velocity have used temperature because it influences species distributions on land, in freshwater,andintheocean.Temperatureisaparticularlystrongenvironmentaldriverintheoceanbecauseitis correlatedwithnutrientavailability,therebyalsocontrollingsystemstructureandfunction[14].However,climatevelocity canbeappliedtoanyenvironmentalvariable.Forexample,climate-velocityanalysesonlandhaveoftenincludedrainfall becausethedistributionandproductivityofplantcommunitiesisregulatedbywateravailability. Whenapplyingclimatevelocitytoanewenvironmentalvariable,oneshouldconsiderthefunctionalrelationshipbetween theenvironmentaldriveranditsbiologicalresponse.Climatevelocitymighthaveecologicalrelevanceforavariable wheretherelationshipwithbiologicalperformanceissymmetrical(Box1),butmightnotifitisastepfunction.For example,mostmarinelifecannotsurviveoxygenconcentrations<2mg.l(cid:4)1,andtrackingthis‘threshold’oxygenisoline mightbemoreinformativethanestimatingclimatevelocityforallisotherms,mostofwhicharenotecologicallyrelevant. Technicallythisisonlytheclimate-analogvelocityofasingleisoline. Finally,mostenvironmentalvariablesarerepresentedinclimate-velocityanalysesusingsummarystatistics,andtheir selectionwarrantscarefulconsideration.Forexample,annualmeanvaluesmightbetterpredictshiftsovertheentire speciesranges,whileextremevaluesmightbemoreappropriateatrangeedges.Similarly,bottomtemperaturesare moreappropriatethansurfacetemperaturesforbottom-dwellingmarinespecies[21].Theoftenunacknowledged uncertaintiesassociatedwithdataproductsshouldalsobeconsidered(Box2). WhatTimescales? Climatevelocityisbestsuitedtostudiesofclimate-changeimpacts,whichbydefinitionimpliestimescalesofdecadesor longer. WhatSpaceScales? Climatevelocityhasbeenappliedtogriddedenvironmentaldataatspatialscalesfrom(cid:1)1kmto(cid:1)110km.Onland, mostapplicationshaveusedafinespatialresolution(e.g.,afewkilometers[26,32]),reflectingtheimportanceofterrain onmicroclimatesandorganismdispersal[29].Bycontrast,analysesintheoceanhaveusedacoarserspatialresolution (e.g.,100km)notonlybecausefine-scaledataarenotalwaysavailablebutalsobecausetherearefewerdispersal barriers[494_TD$DIFF][88],suchthatorganismsdispersefurther,andbecausemicroclimatesmightbelessimportant[495_TD$DIFF][89].However, shallow-water and seafloor communities are structured more by biological than environmental processes [496_TD$DIFF][90], suggestingtheneedforfiner-scaleanalyses.Itmightbedesirableinsomeinstancestomatchthespatialresolution toclimateturnoversuchthatthespatialresolutionmightbefineraroundmountainsthaninplains,andcoastallythanin theopenocean.Irrespective,coarserspatialresolutionleadstogreaterclimatevelocitybecauseitaveragesoverfine- scalevariation[32]. CombiningEnvironmentalVariables? Climatevelocityhasusuallybeenappliedtoanindividualvariable.Whenconsideringmultiplevariables(e.g.,temperatureandrainfall), thesehave generallybeentreatedseparately asindependentdrivers ofspeciesmovement [17,26,32].However,Hamannetal.[28]developedamultivariateapproachtoclimate-analogvelocitybasedona principalcomponentsanalysisofmultiplemetrics(e.g.,minimum,maximum,mean)oftemperatureandrainfall.This approachhasthebenefitofconsideringthemultivariatemovementofclimatespace,butatthecostofcomplicating interpretation.Moreover,multivariateclimate-analogvelocitiesarelikelytobehigherthancorrespondingunivariate estimates[28,34]becausefindingsimilarmultivariateclimateswilloftenrequirealargesearchradius(i.e.,similarrainfall islikelytobefoundcloserthansimilarrainfallandtemperaturecombined).Themagnitudeofthiseffectcanbemitigated byrelaxingassumptionsdefininganalogclimates(e.g.,expandingbandwidthtoincorporatemoreclimatevariability [67]).Multivariatelocalclimatevelocitycouldbecalculatedbyapplyingvectoralgebratomultipleunivariateestimatesof localclimatevelocity.Forexample,ifthereweretwounivariateclimatevelocities(e.g.,temperatureandrainfall)in opposingdirectionsthatareequalinmagnitudetheywouldcancel.However,ingeneral,thenewmultivariateclimate spacewouldnotbethesameastheoriginal.Thisdivergenceinanglesofsuchunivariateestimatescanbeconsidered as a measure of climate stress on an organism, and has provided insight into potential ecological responses to multivariateclimatechange[30]. 448 TrendsinEcology&Evolution,June2018,Vol.33,No.6 (climate ‘sinks’: locally cool areas such as mountain tops on land and equatorward-facing coastlines in the ocean) (e.g., [2,22]). (vi) ProjectedRangeShifts withClimateChange Becauseclimatevelocityisanindicatorofthespeedatwhichtherangeshiftsofspeciestrack climatechange–potentiallythemaximumpossiblerateofrangeshiftwhendispersalisnota limitingfactor–climate-drivenchangesinthegeographicaldistributionofspeciescanbesimply predictedbyforward(orbackward)projectionoftheirclimateenvelopesfollowingthespeed and direction of local or analog climate velocities. This approach has been combined with speciesthermaltolerancesanddepthpreferencestopredictchangesindistributionofmarine species.Applyingthisapproachfor>13000marinespecies,GarcíaMolinosetal.[33]found thatbiodiversitywoulddecreaseinequatorialregions,butincreaseinothers,andtherewould beaspatialhomogenizationofbiodiversityby2100.Recentobservationsofmarinecommu- nitiesconfirmthoseresultsinresponsetoclimatechange[51,52].However,thelikelihoodofa response,andofasubsequentshiftinrangemirroringclimatevelocity,isspecies-specific.For example,opportunitiesfortheexpansionandriskofcontractionofageographicalrangewill depend on changes in the local climate space relative to the physiological tolerances of a species(seeFigureIIinBox1).Evenifageographicalshiftistriggeredbychangesinclimate, differentdispersalcapacitiesofspeciesresultinrangeshiftsthatkeeppacewith,lagbehind,or even exceed rates of climate displacement [53–60]. Range shifts will also depend on the interaction between climatechange and externaldirectionalforces. Ina recent globalmeta- analysis[61],statisticalmodelscombiningtheeffectofclimatevelocityanditsalignmentwith Box4.ACaseStudyApplyingClimateVelocity,ResidenceTime,andClimate-VelocityTrajectoriesto theUKMarineProtectedAreaNetwork Toillustratetheutilityofclimatevelocityfornetworksofmarineprotectedareas(MPA),weexamineclimateconditions acrossthenetworkinUKterritorialwatersforpast(1960–2009)andfuture(2006–2050)climateat1(cid:3)spatialresolution. Pastandfuturelocalclimatevelocitieswerecalculated,respectively,fromannualmeanseasurfacetemperatures (SSTs)fromtheHadleyCentredatasetHadISST1.1andamultimodelensemblefortheIntergovernmentalPanelon ClimateChange(IPCC)RCP8.5climatepathway[497_TD$DIFF][91].Climatevelocitieswerecalculatedforbothperiodsascellratiosof thelocaltemporaltrend(slopefromthelinearregressionofannualSSTovertime)tothe(3(cid:5)3)spatialgradientbased onaverageannualmeanSSTs[18].LocalclimatevelocityassociatedwiththeMPAnetworkoverthepast50yearsinUK watersshowsstrongcontrastsbetweenwesternandeasternhalvesoftheUKExclusiveEconomicZone(FigureIA). However,bothsidesareprojectedtohavesimilarmagnitudesoflocalclimatevelocityby2050becauseofageneral decreaseinlocalclimatevelocityintheNorthSeaandlocalincreasesonthewesternside(FigureIB).Thelargespatial variabilityinlocalclimatevelocitywillrequirespeciesrespondingtoclimatechangetoshifttheirdistributionupto10-fold fasterorslowerdependingonthelocationoftheMPAwithinthenetwork. Bycontrast,climateresidencetimeshowshighvariationacrosstheUKMPAnetworkforbothperiods(FigureIC,D). MPAsalongthewestcoastofScotlandarepredictedtoregisterlargestreductionsinresidencetime,whilethosewithin theIrishSeaandnorthoftheStraitofDoverarepredictedtoincrease.Reductionofresidencetimesuggestsreduced viabilityofaprotectedareaastherateofchangeinconditionswithintheareaincreases,potentiallycompromisinglocal adaptationtoclimatechange,especiallyofrange-restrictedspecies,whilefacilitatingtheestablishmentofimmigrant andinvasivespecies[498_TD$DIFF][92]. Climate-velocitytrajectoriesoverthepast50yearsaregenerallydirectedpolewardalongtheEnglishcoast(FigureIE), suggestingthatthecoastalnetworkcurrentlyexhibitsgoodconnectivity(MPAsinthenorthshouldreceiveclimate migrantsfromthoseinthesouthastemperaturewarms).However,climate-velocitytrajectoriesuntil2050,asprojected fromRCP8.5,showadifferentpatternontheeastcoastoftheUK,wherethermalnichesmoveoffshoreintotheNorth SeatowardsScandinavia(FigureIF).Thisscenariosuggeststhatlittoralspeciesonthiscoastmightbeforcedtoadaptin situbecausetheybecomedisconnectedfromtheircurrentthermalniches.Thiscouldhavemanagementimplications, especiallyforsmallerprotectedareasontheeastcoastofScotlandwhereresidencetimeswillcontinuetobeshort. Herethepossibilityofassistedmigrationandtranslocationsofspeciesofconcernmightbeconsidered. TrendsinEcology&Evolution,June2018,Vol.33,No.6 449 VoCC (A) (B) 25–5020–2515–20 10–158–106–82–41–2 0–1–2 to 0–10 to –2 <–10 (km year−1) (C) (D) Residence (cid:415)me (year) 0–1 1–2 2–5 5–10 10–20 20–35 (E) (F) FigureI.ACaseStudyIllustratingtheApplicationof(A,B)LocalClimateVelocity,(C,D)ResidenceTime,and(E,F)Climate-velocityTrajectories.(A, C,E)Past(1960–2009)and(B,D,F)future(2006–2050)climateconditionsacrosstheMPAnetworkinUKterritorialwaters(brokenline).ForeachMPAcentroid(points onthemaps),weshowtheexpectedthermalshiftbyprojectingitsSSTintimefollowingthespeedanddirectionoflocalclimatevelocities(VoCC)ateachcell. 450 TrendsinEcology&Evolution,June2018,Vol.33,No.6