Towards threshold-based management of freshwater ecosystems in the context of climate change

Towards threshold-based management of freshwater ecosystems in the context of climate change

Junguo Liu

*, Giri Kattel

, Hans Peter H. Arp

, Hong Yang

aSchoolofNatureConservation,BeijingForestryUniversity,Beijing,100083,China

bSelf-SustainingRegionsResearchandInnovationInitiative,CollaborativeResearchNetwork(CRN),FederationUniversityAustralia,Mt.HelenCampus, Victoria3350,Australia

cNorwegianGeotechnicalInstitute,Sognsveien72,NO-0855,Oslo,Norway

dCEES,DepartmentofBiosciences,UniversityofOslo,Blindern,Oslo0316,Norway

ARTICLE INFO

Articlehistory:

Received2April2014

Receivedinrevisedform4September2014 Accepted15September2014

Availableonline23September2014

Keywords:

Ecologicalthreshold Ecosystemmanagement Ecosystemservices Watermanagement Climatechange

ABSTRACT

Climatechangeisanincreasingthreattofreshwaterecosystemgoodsandservices.Wereviewrecent researchregardingthedirectandindirectimpactsofclimatechangeonfreshwaterecosystemsandthe severityoftheirundesirable effectsonecosystemprocessesandservices.Appropriatemanagement strategiesareneededtomitigatethelong-termorirreversiblelossesofecosystemservicescausedby climatechange.Toaddressthis,thisreviewputsforwardathreshold-basedmanagementframeworkasa potentialplatformforscientists,decisionmakersandstakeholdersoffreshwaterecosystemstowork togetherinreducingrisksfromclimatechange.Inthisframework,thesusceptibilityoflocalfreshwater ecosystemstochangebeyondthresholdsiscontinuouslyinvestigatedandupdatedbyscientists,usedto design policy targets by decision makers, and used to establish mitigation measures by local stakeholders.

ã2014TheAuthors.PublishedbyElsevierB.V.ThisisanopenaccessarticleundertheCCBYlicense (http://creativecommons.org/licenses/by/3.0/).

1.Introduction

Freshwater ecosystems, defined as aquatic systems with averagesalinitieslessthan0.5partsperthousand(Moss,2009), provideadiverserangeofessentialservicessuchasfoodproducts, clearwater, wasterecycling,nutrientcycling,carbonstorage,as wellasculturalandrecreationalamenities.However,freshwater ecosystemsarechangingrapidlyduetofactorslikeurbanization and climate change. Over the past 50 years, economic and populationgrowth have resultedin more rapid changes in the structure and function of freshwater ecosystems than in any other comparable time period of human history (MEA, 2005).

Populations of freshwater species in North America, Europe, Australia,andNewZealandareestimatedtohavedeclinedonan averageby50%between1970and2000(MEA,2005).Freshwater ecosystemsinurbanareasareamongthemostaffected(Kozlowski

and Bondallaz,2013), in that urbanization has ledto dramatic changesinfreshwaterecosystemsthroughouttheglobe(Alberti etal.,2007;KozlowskiandBondallaz,2013).

Adding to the ongoing burdens of intensive agriculture, industrialization andurbanization, climate change is an addi- tionalseriousthreattofreshwaterecosystemsandbiodiversity worldwide.Climatechangecanalterfreshwaterecosystemsvia various direct and indirect mechanisms (Chu et al., 2005;

Vörösmarty et al., 2010). Rising temperature and changing precipitation directly influence shifts in habitats and seasons, andalsothephysiologicaladaptationandphenologyoffreshwa- ter species,therebyalteringfoodweb structureandecosystem dynamics(DoakandMorris,2010;Waltheretal.,2002).Climate change can also indirectly affect freshwater ecosystems via geomorphologicalalterationsoflakeandriversystems,changes in nutrient and ionic loads (leading towards alteration of photosyntheticrates, eutrophication, acidification,salinization) aswellasenhancingtheimpactsofprevalentdiseases,chemical pollutants, biological invasions, and changes in predation and competitionamongspecies(IPCC,2007).Therehavebeenalarge number of discussions on the challenges and solutions facing humaninterventionstofreshwaterecosystems(e.g.,Chenetal., 2014), considering experimental and theoretical ecological thresholds(e.g.,Horanetal.,2011),andhowtobestimplement

* Correspondingauthor.Tel.:+861062336761;fax:+861062336761.

E-mailaddresses:[email protected],[email protected](J.Liu).

1 Junguo Liu and Giri Kattel have equal contributions to the conceptual framework;GiriKattelpreparedthefirstdraft,andJunguoLiufollowedupto thepaperpreparation;GiriKattel,HansPeterH.ArpandJunguoLiucreatedall figures;alltheauthorscontributedtoTable1;HongYang,HansPeterandJunguoLiu contributedtoTable2;alltheauthorssynthesizedtheresultsandwrotethepaper.

http://dx.doi.org/10.1016/j.ecolmodel.2014.09.010

0304-3800/ã2014TheAuthors.PublishedbyElsevierB.V.ThisisanopenaccessarticleundertheCCBYlicense(http://creativecommons.org/licenses/by/3.0/).

ContentslistsavailableatScienceDirect

Ecological Modelling

j o u r n a lh o m e p ag e : w w w . e l s e vi e r . c o m / l o c a t e / e c o l m o d el

riskmanagementof aquaticecosystems(e.g., Chenetal., 2011, 2013). However, it remains rare to explicitly discuss the freshwaterecologicalthresholdsassociatedwithclimatechange.

Thereisanurgentneedforintegratingthescientificunderstand- ingof thediverse,complexandinterrelatedimpactsofclimate change on the thresholds of freshwater ecosystems. Here we reviewfreshwaterecosystemthresholdsinthecontextofclimate change, and suggest the need for collaborative efforts across scientists,decisionmakersandstakeholdersatalllevels.

2.Whatareecologicalthresholds?

Thereareseveraldefinitionsfortheterm“ecologicalthreshold”. Mostofthesedefinitionscommonlyemphasizethenon-linearity ofecologicalorbiologicalresponsestopressurescausedbyhuman interventionsornaturalprocesses.AsdefinedbyGroffmanetal.

(2006),anecologicalthresholdisthepoint,or“tippingpoint”,at whichthereisanabruptchangeinanecosystemquality,property orphenomenon,orwheresmallchangesinenvironmentaldrivers canleadtodramaticchangestoanecosystem.Thresholdsandtheir associatedstability towardsdifferentenvironmental driverscan be conceptualized within a coupled socioeconomic–ecological system(Horanetal.,2011).Anyrestorationoflossesinecosystem servicesaftera threshold is crossedcouldbedifficult orcostly (Groffmanetal.,2006).

Identifyingecologicalthresholdsrelatedtoclimatechangeis complex, as various climate-change control variables (e.g., atmospheric and surface water temperature) can be related differently tochanges in theecosystem service responses(e.g., fishsupply,volumeofpotablewater)(Rockströmetal.,2009).

Fig.1presentsageneralschematicdiagramonhowtoviewthe interactions between climate change control variables and

Fig.1.Conceptualizationofthreshold-basedecosystemchange.Climatechangecontrolvariablesareonthex-axisandtheresponsevariablesonecosystemservicesareon they-axis.Theclimatechangevariablescouldbeoneoranycombinationofparametersthataredirectlyorindirectlyaffectedbyclimatechange(suchastemperature,number ofinvasivespecies,nutrientconcentration).Theecosystemservicecouldrefertoanyrequiredbythelocalpopulation(suchaspotablewater,fishpopulation).Conceptualizing ecosystemchangeinthiswaycouldserveasabasisforcommunicationinecosystemmanagement,wherescientistsinformofsusceptibilitytothresholdchanges,decision makersmanageappropriately,andstakeholdersgetinvolvedthroughenactingmitigationmeasurestopreventlossofservices.

ecosystem service response variables using a threshold-based approach. Ecosystem services can be linearly dependent, non- linearly dependent, or independent of the climate control variables. When a shift in a climate control variable causes a substantial depletion in a dependent ecosystem service, a

“threshold”, point F1, may be crossed, after which the drop in thisservicecanbecomedepletedrapidlywithslightincreasesin theclimatecontrolvariable.Oftenitisdifficulttopredictwhena thresholdoccurs,orevenascertaintherelationshipbetweenthe climatechangevariableandtheecosystemservice.Thus,arange can be assigned as a “zone of uncertainty”. The “ecosystem boundary”,pointF₀inFig.1referstothelowestvalueofaclimate changecontrolvariable(comparedtothepresentvalue)inwhich theprobabilityofpassinga“threshold”becomeshigherthanan assigned marginof safety (Rockström et al.,2009).Sometimes, ecosystems exhibit resilience (Kattel et al., 2013). After the thresholds are crossed, ecosystem services can recover either duetotheclimaticandotherenvironmentalvariablereversingto anearlierlevelorsomedelayedfeedbackorbufferingmechanism.

Thisrestorationisusuallynotinstantaneous.Itoftenrequiresthe control-variable to decrease to a level substantially below the threshold, which is a phenomenon, referred to as ecosystem hysteresis(Schefferetal.,2001).However,ifecosystemthresholds arecrossedandrestoredperiodically(e.g.,fromforestfires),this behavior would be “cyclical”. However, if the change in a climatically sensitive and controlledvariablebecomes extreme, a“pointofnoreturn”canbereached,indicatedaspointF2inFig.1, i.e.,a‘catastrophic’transitionwheretheecosystemcannolonger returntoitsoriginalstate(e.g.,Katteletal.,2013).Thecondition between F1 and F2 can have profound implications for ‘early warningsignals’forecosystemmanagement(Wangetal.,2012).

However,thisisoftenunnoticed(Schefferetal.,2001;Folkeetal., 2004).

Akey goalofecosystemmanagement is toavoidthresholds withslowrecoveryor points of noreturn. Though difficultfor scientists to anticipate, boundaries and thresholds are highly usefulformobilizingpolicymakersorstakeholders.Forinstance, theuseof350ppmCO₂beingasafeleveltoavoidatippingpoint forglobal ecosystemchangehasbeensuccessful formobilizing stakeholdersandpolicyefforts(Hansenetal.,2008),ashasbeen theprediction that a warming of 1–2.5C above pre-industrial

levelscouldbeathresholdvaluewheresubstantiallossofArctic summericeortheGreenlandicesheetcouldoccur(IPCC,2007).

3.Effectsofclimatechangeandthresholdtendenciesin freshwaterecosystems

The mostpractical thresholdsfor use bypolicy makers and stakeholdersarethosebasedonmeasurablecontrolvariablesthat haveastraightforwardlinktoanecosystemresponse(Table1).The easiertheclimatecontrolvariableistoquantify,aswellasitseffect onecosystemservices,themorepracticalitistoimplementpolicy and effectivemonitoringmeasures. However,in practice direct links between climatechange control variables and ecosystem responsesaredifficulttoestablish.Anthropogenicclimatechange isquantifiedbyseveralparametersthatcouldbeusedascontrol variablesfordefiningthethresholdsoffreshwaterecosystems,e.g., atmosphericandsurfacewatertemperature,thelengthandonset of seasons,andprecipitation(Table1).However,inmostcases, multiplecontrolvariablesactingtogethermustbeconsidered.For example, riverine ecosystems are strongly and simultaneously influenced byvariableslikestreamflow,erosionrates,tempera- ture, and concentrations of micro-pollutants; though how to accountforallthevariablesinathresholdtypemodelisnotalways straightforward (Groffmanet al., 2006).Othercontrol variables thatareindirectlyinfluencedbyclimatechangeaswellasother anthropogenicprocessesincludewatervolume,salinity,dissolved oxygen(DO) andpH.Allof theseparameters,in isolationor in combination,couldinfluenceecosystemservices.InSection4,we reviewthestate-of-the-artunderstandingofhowshiftstowards ecologicalthresholdsforfreshwaterecosystemscanbeinfluenced directlyorindirectlybyclimatechange.Aschematicoverviewof thispresentationisprovidedinFig.2.

4.Thresholdsassociatedwithdirecteffectsofclimatechange

Overthepastdecades,freshwaterecosystems haveexhibited direct responses to changes in temperature, precipitation, and atmospherichumidity(Haderetal.,2007),particularlyinregardto thephenology and physiologyof freshwaterspecies,as well as theirhabitats.Theseresponsesbecomemorepronouncedwithan increasingglobalsurfacetemperature(Schippersetal.,2004).

Table1

Examplefreshwaterecosystemmonitoringapproacheswithpossiblescientific,policyandstakeholderinvolvementtoassessorpreventpossiblethresholdsandlossofsocio- economicservices.

Environmentalparametermonitoring Potentialaffectedthresholdresponse monitoring^a

Potentiallyaffectedsocio-economicservicesmonitoring^a

Climate:surfacetobottomwatertemperature,

evapotranspirationrates,precipitation,humidityseason onsetandlength

Populationofalgaeandplants, migrationoffishandotherspecies, habitatrange

Changesinlocaleconomyduetochangeinproducts(pricesof food,water,plants,realestate),aswellasqualityoflife(health, opinionpools,recreationallandarea)

Hydrology:waterlevel,watervolume,waterresidenttime Populationofalgaeandplants,areasof excessflooding,drowning,lossofwater habitat

Light:UVradiation,visiblewaterdepth Populationofalgaeandplants (planktonicandbenthic)

Oxygen:dissolvedoxygen(DO)levels,lakestratification. Phytoplanktonblooms,aerobicbenthic species,verticallymigratingfishspecies, oxygensensitivespecies,

Geomorphology:erosionrates,sedimentation,siltation Riparianfishhabitat,formationofanoxic zones

Waterquality:pH,micropollutants,dissolvedorganic carbon(DOC),nutrientsincludingtotalnitrogen(TN),total phosphorus(TP)andchlorophylla(chla)

Lossofsensitivespecies,changein speciesdynamics,populationofalgae andplants

Speciesdiversity:presenceofinvasivespecies,changesin foodwebstructure,biodiversitylosses

Lossoflocallyimportantspecies

aThethresholdlikeresponsesandsocio-economicservicemaybeaffectedbyacombinationofclimateandlocalanthropogenicimpactparameters(e.g.,pollutionlevels, wateroverusage,urbanization).

4.1.Phenologicalthresholds

Climatechangehasdirectlyinfluencedthephenologyofaquatic biota.Changesinairtemperatureinfluencewatertemperatureand thetiminganddurationofthestratifiedperiod,whilechangesin precipitationcaninfluencewaterresidencetime.Thesechangesin turninfluencetheseasonaldynamicsofthebiotapresent(Peeters etal.,2007).Phytoplanktonphenologyinlenticsystemsisdirectly affectedbyanincreasedwatertemperature,changesintheonset andthedurationofthermalstratification,andearliericebreakup (Meisetal.,2009).Forinstance,phytoplanktonbloomsintheUS lakesandTaihuLake,thethirdlargestlakeinChina,havebegunto occurearlierduetothewarmerspring(Qinetal.,2010;Winder andSchindler,2004).Inaddition,highairtemperaturesandlow wind speeds have directlyled to an early onset of the spring phytoplankton bloomin Upper Lake Constance of the western EuropeanAlps(Peetersetal.,2007).

Changesintemperatureandseasonsinfluencemultiplelevels ofbiologicalorganization,includingforagingbehavior,phenology andmetabolicrates,potentiallyleadingtoprofoundshiftsinthe stoichiometry of elemental fluxes between consumers and resourcesatthebaseofthefoodweb(Woodwardet al.,2010).

Arise insurface-water temperatureand a regionaldecrease in windvelocityincentralAfrica’sLakeTanganyika,whichprovides 25–40% of theanimal proteinfor surrounding populations, has leadtoareductioninprimaryproductivityby20%,implyinga 30%decreaseinfishyields(O’Reillyetal.,2003).Arcticlakeand riversystems,however,experiencingtheoppositeeffect,aslonger growingseasonshaveledtoincreasedprimaryproductivity(Smol etal.,2005).Suchdifferencesinphenologyandlifecyclecuescan inducestrongvarianceincommunitycomposition(Burgmeretal., 2007), trophic levels and fish production, ultimately changing ecosystemservices(Fickeetal.,2007).

4.1.1.Physiologicalthresholds

Some species that are exposed towarmer temperatures for longerperiods will showbiochemical variations atthe cellular level.Heatwaves,whicharelikelytoincreaseinseverityduetothe changing climate(Scharetal.,2004), mayinfluencefreshwater gastropods and mussels by affecting the release of the stress- associated enzymes (e.g., protein kinases), altering metabolic activityandcausinghypoxiaintheseanimals(Anestisetal.,2010).

Physiologicalprocesses in aquaticorganisms,including ratesof oxygen uptake, movement, feeding, developmental rate, and immunefunctionarestronglydrivenbytemperature(Helmuth, 2009).Warmersurfacetemperaturesoffreshwaterlakescanlead directly(throughwatercurrents)andindirectly(throughaffecting plantand algaegrowthasdiscussed above)tothealterationof thermal stratification in temperatelakes, which can leadto an increaseddissolved organiccarbon(DOC) concentrationsin the hypolimnion, limiting theDO levelfor vertically migrating fish species, potentially causing hypoxia (Hader et al., 2007). A mismatchbetweenthe demandforoxygen and thecapacityof oxygen supply to the tissues can deviate the physiological adaptation of aquatic species and restrict tolerance of these animalstoarangeofclimaticextremes(PörtnerandKnust,2007).

Decreasedoxygeninlakesalsodecreasestheprocessingoforganic matter, recyclingof nutrients, and the microbial breakdownof pollutants(Carpenteretal.,2011).

Toplacethisinthecontextofafreshwaterecosystemthreshold,a previousmodelingstudysuggeststhatadoublingofCO2concen- trationswouldcauseearlieronsetoflakestratificationanda1–7C increaseinsummerepilimnetictemperature,consequentlylimiting theverticalmigrationofzooplanktoninadditiontophysiological andrespirationfailuresincoldwaterfishspecies(DeStasioetal., 1996).Thisexampleandotherexamplesoffreshwaterecosystem thresholdsrelatedtoclimatechangearepresentedinTable2.

Fig.2.Overviewofecosystemservicesthataredirectlyandindirectlyimpactedbyclimatechangeandlocalanthropogenicimpacts.Thefigureisanillustrationofthe complex,cyclicalnatureofhowtheuseofecosystemservicescan,throughdirectandindirectmechanisms,affectthosesameecosystemservices.Appropriateglobalactionto reduceharmfulclimatechangeimpactsandlocalactiontoreduceharmfuldisturbancesinwaterqualitycanhelpmitigatelossesofecosystemservices.

4.1.2.Habitat-associatedthresholds

Climate change also directly influences fish populations, as changesinclimateleadtochangesinthehydrologicalregimesand habitatnichesthatendemicfishhaveevolvedtocopewith(Ficke etal.,2007).Globalwarminghasbeenattributedwithdecreasing thethermalnicheandhabitatavailabilityforcoldwaterfish,such as galaxid and salmonid fishes (Carpenter et al., 2011). A 3C change in mean annual temperature corresponds to a shift in approximately 300–400km latitudinal and 500m altitudinal isotherms (Smol et al., 2005). One studyestimates that global warming could shift natural habitat ranges by approximately 6.1kmperdecadetowardsthepoles(ParmesanandYohe,2003).

Thischange,thoughsubtle,isenoughtoimpactfishcommunities andecosystemadaptationinthehigherlatitudes.Toplacehabitat changes in the contextof climatechange related thresholds,a coupledCO2–climatemodel suggeststhat anincrease in2–6C summermeantemperaturewillreduce50%ofthesuitablehabitats forcoldwaterfishintheUS(EatonandScheller,1996).Afurther 4Cincreaseinmeanairtemperaturewillextendthehabitatof smallbassmouthandyellowperchasfaras500kmtowardsthe north(ShuterandPost,1990).

5.Thresholdsassociatedwithindirecteffectsofclimatechange

Climatechangecanaffectfreshwaterecosystemsindirectlyvia arangeofpositiveandnegativefeedbacks.Theseindirecteffects

may be gradual and subtle in the short term but could be catastrophic in the long term, leading to slowly recovering or irreversible threshold type losses. Changes in geomorphology, waterdepth,flowregimes,siltationandsedimentfluxinlakeand riversystemscanoccurasa resultofindirecteffectsofclimate change (Lake et al.,2000). Biotic and abioticinteractions with climatecanfurtherleadtounprecedentedecologicalconditions, includingthe alteration ofphotosynthetic rates,eutrophication, acidification andsalinization(Schindler,2001;Woodwardetal., 2010). Widespread invasion from exotic flora and fauna, and changes in predation and competition among species are also exacerbatedbyindirectinteractionsbetweenclimatesandhuman activities(Baronetal.,2002).

Below,wetakethetrophicstatethresholdasanexampleof indirect effects of climate change. Other thresholds related to indirecteffectsaredescribedinSupplementaryinformation,and theyincludemorphological,photosynthetic,acidification,salinity, pollutant thresholds, disease, biological invasion, as well as predationandcompetition-associatedthresholds.

Decreased precipitationand lowerwaterflowsimplysmaller dilution volumes and thus, higher concentrations of nutrients downstream frompoint pollutionsource, particularly untreated sewageindevelopingcountries.Ontheotherhand,theincreased occurrenceofstormeventscancauseanoverflowanddischargeof nutrient into receiving rivers or lakes (Whitehead et al., 2009). Severe stormscanentrainlargeamountofdeeperandanoxicsediments,

Table2

Thresholdsforfreshwaterecosystemregimeshift

Regime-shift Parameter Threshold Location References

Physiologicalrelated

Earlylakestratification,limitingvertical migrationandcausingphysiological failureincoldfish

AtmosphericCO2concentrations AdoublingofCO2

concentrationfromthe 1980slevel

Northtemperatelakes DeStasio etal.(1996)

Habitatrelated

Reductioninfishhabitatsby50% Summermeantemperature 2–6C USlakes Smoletal.

(2005) Shiftofsmallbassandyellowperch

habitat500kmnorth

Meanairtemperature 4C USlakesandrivers Shuterand

Post(1990)

Substantiallossoficesheet Temperature 1–2.5Cincreaseabove

pre-industriallevels

Arctic,Antarctic,Greenlandandother mountainareas

IPCC(2007)

Waterqualityrelated

Shiftindiatomcommunity Nitrogendepositionloads 1.4kgNha ¹y ¹ LakesaroundtheeasternSierra NevadaandtheGreaterYellowstone NationalPark,US

Sarosetal.

(2011) Shiftinbioticassemblages,waterquality

andcarbonbudget

Dissolvedorganiccarbon 91%increase 22uplandlakesandstreams,UK Evansetal.

(2005) Fromcleartoturbidstate¹⁾ Orthophosphatephosphorus

concentrationinthewater

10–12mg/L 26temperatelakes,NorthernItaly Chiaudani andVighi (1974) Fromcleartoturbidstate Totalphosphorusinwater 150m^{g/L LakeVelume,Netherland}

Fromturbidtoclearstate Totalphosphorusinwater 100m^{g/L LakeVelume,Netherland}

Fromcleartointermediatestate²⁾ Totalphosphorusinwater 61m^{g/L GehuLake,China Taoet al.}

(2012) Fromintermediatetoturbidstate³⁾ Totalphosphorusinwater 115mg/L GehuLake,China Taoetal.

(2012) Fromclear-toturbid-state Totalphosphorusinwater 70–100mgm ³ 46smallandmediumlakesalongthe

mid-lowerYangtzeRiver

Wang(2007) Fromturbid-toclear-state Totalphosphorusinwater 20–30mgm ³ 46smallandmediumlakesalongthe

mid-lowerYangtzeRiver

Wang(2007) Fromturbidtoclearstate⁴⁾ Phosphorusloading 0.3gm ²y ¹ LakeMendota,Wisconsin,USA Carpenter

andLathrop (2008) Fromcleartoturbidstate Dynamiclinearmodelstandarddeviation

(DLMSD)ofphosphorusinsoil,waterand sediment

Around1gm ² LakeMendota,Wisconsin,USA Carpenter andBrock (2006) Note:Althoughdifferentwordsareusedforthestatesoffreshwaterecosystems,weusedthetermsclearandturbidstates(widelyacceptednow);theoriginalwordsforthe stateswerelisted:(1)frommeso-toeutrophic;2)fromgrass-stabletograss-algaeintermediatestate;(3)fromgrass-algaeintermediatetoalgae-stablestate;(4)from persistenteutrophytooligotrophy.

particularlywhenthewaterlevelislow.Asanexample,astormin 1996thatoccurredatLakeVõrtsjärv,Estonia,resultedinagreater releaseof sediments onone stormy day comparedto thetotal nutrientinflowsover1y(NõgesandKisand,1999).Theeffectsof wind-mixingontheverticaltransferofnutrientsfromsedimentare moreimportantinsmalllakesthanthelargelakes.Therelationship between North Atlantic Oscillation Winter Index (NAOw) and phosphatehasbeenobservedinsomeEuropeanlakes(e.g.,George etal.,2004;Weyhenmeyer,2004).IntemperateEurope,positive valuesofNAOwimplymilderwinter,higherwaterlevelsandlower springconcentrationsofphosphate,whilenegativevaluesindicate drierwinters,lowwaterlevelsandlargeramountsofphosphate releasedfromsediment.Incoldareas,theextensionoftheice-free periodcancausemarkedincreasesinphosphorusconcentration,as hasbeenobservedinthehypolimnionandepilimnioninLakeErken, Sweden(Pettersson et al.,2010). Thisadditional phosphorus is releasedannuallyfromthesedimentzonetothewatercolumnasa resultofincreasedtemperatureandlowoxygenconcentrationin thebottomwater(MalmaeusandRydin,2006).Increasedtempera- tureandprecipitationintensityandchangesinwinterprecipitation are expected to enhance phosphorus loading in the temperate freshwaterlakesandreducetheloading inMediterraneanlakes (Jeppesenetal.,2009).

Nutrientloadsarepredictedtoincreaseunderclimatechangein many areas such as Denmark and UK (Andersen et al., 2006;

Whiteheadetal.,2006).Highertemperatureincreasestherelease ofnitrogenfromsoil,andlowerstreamflowsdecreasethedilution capacityofrivers(Whiteheadetal.,2006).However,theimpacton eutrophication is not straightforward due to the complex interactionbetween nutrient,light, temperature,residencetime andflowconditions(Jeppesenetal.,2005;YangandFlower,2012).

Increased surface water temperature can indirectly lead to strongerstratificationandregenerationofwatercolumnnutrients, whichcanintensifyeutrophication(Rabalaisetal.,2009).Climate changehasexacerbatedeutrophicationthroughnutrientdynamics of the north temperate lakes due to a longer growing season, dependingonepilimnionandmineralmixing(Smoletal.,2005).

The analyses of 103 Chinese lakes indicated that the mean precipitationwasoneofthemainpredictorsofeutrophication(Liu etal.,2010).A41-yeardataseries(1968–2008)fromBlelhamTarn UK,whereeutrophicationhasreducedhypolimneticDO,indicates negativeeffects that arelikelytobeexacerbated bychanges in climateandthethermalstructureoflakes(Foleyetal.,2012).The interactiveeffectsoffutureeutrophicationandclimatechangeon harmfulcyanobacterialbloomsaredifficulttounderstand,though muchofthecurrentknowledgesuggeststhatclimatechangewill likely enhance the magnitude and frequency of these events (O’Neiletal.,2012).Recentlyeutrophicationinlacustrineenviron- mentshasbeenadoptedasawaytodefineecologicalandsocietal thresholdsinspaceandtime(Carpenteretal.,1999).Experimental managementprogramshavebeenreplicatedamongseverallakes basedonthesethresholds(Carpenteretal.,1999).

Thephosphorus thresholds for theregimeshiftbetweenclear and turbid states have beensuggested for different lakes (Table 2).

Comparedtotheprimarydriversofeutrophicationsuchasnutrients, theroleofclimatechangetomodulateeutrophicationthresholdsis notwell understood.However,increasingtemperature,changing precipitationandatmosphericcirculationcanindirectlyinfluence thenutrientdynamics,and increasethelikelihoodof exceeding thresholdsofnutrientsinfreshwaterecosystems(Battarbee,2000).

6.Identifyingthresholdresponsesformanagingfreshwater ecosystems

Identifying and predictingthreshold responses in ecological systemsisachallenging,thoughusefultask.Thresholdsrelatedto

lakeeutrophication causing regime shiftis a well studiedarea (Scheffer etal., 2001)and one that canbeused tounderstand thresholdsfromotherclimate-changecontrolvariables.Excessive nutrient levels, particularly phosphorus, have been widely accepted as the leading factor of eutrophication (Schindler, 1974; Carpenter et al., 1999). The orthophosphate phosphorus concentration 10–12

m

thresholdfromcleartoturbidstatesinthelakesofNorthernItaly (ChiaudaniandVighi,1974)(Table2).Higherconcentrationswere also foundto be thresholds for the regime shiftin Dutchand Chinese lakes (Tao et al., 2012). Although the phosphorus concentration in water can be controlled to some extent by lowering nutrients emissions, such as by controlling fertilizer usage; internal sediment release of phosphorus in addition to temporalchangesinexternalcatchmentinputcancausethresh- olds to be exceeded e.g., the level of 0.3gm ²y ¹ external phosphorus input in Lake Mendota, Wisconsin for threshold crossingbenchmark(Carpenterand Lathrop,2008).Withmany dynamicinfluentialfactors,changingnon-linearlywithtime,the useofBayesianforecastingthroughDynamicLinearModels(DLM), whichallowsinfluentialparameterstohavea priordistribution (West and Harrison, 1989), is regarded as a more reasonable approachformodelingchangesinreleaseofphosphorusfromthe lakesediment(Cottinghametal.,2000),thansteadylinearmodels.

DLMstandarddeviation(DLMSD)ofphosphorusloadinsoil,water andsedimentwasalsosuggestedasawaytoanticipateathreshold oflakeregimeshift(CarpenterandBrock,2006).

Thresholds from other climate-control variables, similar to thosecausedbyphosphorusinducedeutrophicationarenotonly relatedtoexternaldrivingforces,butarealsomediatedbyinternal changesintheecologicalregimes(WalkerandMeyers,2004).The directandindirectimpactslistedabovedonotoccurinisolation butsimultaneously,andareoftenindicatedbyachangeinseveral quantifiableparameters(Fig.2).Oftenthesechangesareinconsis- tentandmaybehavenon-linearlyacrossandwithinecosystems.

Evena minordisturbancemaymoveasystemtoanewregime, for instance, a gradual increase in nutrients can transform an oligotrophic lake into an eutrophic one (Limburget al., 2002).

Approachestoidentifyingthethresholds offreshwaterecosys- tems should involve the use of continuous monitoring,paleo- limnological data and simulations, and thereby account for changesinthepresent,pastandfuture.

6.1.Implementingeffectivemonitoringprograms

Environmentalpoliciesdesignedtomaintainfreshwaterquality are often dependent on identifying “threshold” dose–response relationships(Groffmanetal.,2006).Effectivemonitoringisuseful for trackingongoing ecologicalsystem shiftsthat aredrivenby externalenvironmentalforcesorinternally-mediateddrivers,and thus for predicting threshold type responses, and devising prioritiesandpracticalstrategiesforbiologicalconservation(Doak andMorris,2010;LindenmayerandLikens,2009).Forexample,a timeseriesanalysisofDOCin22uplandlakesandstreamsofthe UKobtainedduringa 15ymonitoringprogram indicateda 91%

increase in DOC concentration has triggered a shift in biotic assemblages,waterqualityandthecarbonbudgetintheregion (Evans et al., 2005). A 20-year monitoring study of water conductivity of shallow lakes across the Arctic region (in combinationwithpaleolimnologicaldata)hasprovidedadescrip- tionofthresholds relatedtowetlandpermanencyasa resultof climatewarming,inwhichevaporation/precipitationratiosarea suitableclimate-controlvariable(SmolandDouglas,2007).Forall monitoring programs, it is essential to establish an archive of environmentaldata,suchastheVannmiljøsystemusedinNorway

(http://vannmiljo.miljodirektoratet.no/)foranalyzingtrendsand predictingthresholdresponses.

Keyvariables,suchasprecipitation,temperature,totalnitrogen (TN), total phosphorus(TP) and chlorophyll a(chla)should be continuouslymonitored basedonfieldsamplingandlaboratory measurement.Inaddition,advancementoftechnologiesincluding remotesensingwillimprovetechniquesof monitoringenviron- mental forces over a regionalscale. A few examples of further monitoringapproaches toobtainrelevantdataarepresentedin Table1.

6.2.Reconstructinghistoricaldata

Palaeolimnologicaltechniquescanhelpidentifythresholdsof ecosystemchange.Forexample,criticalloadsofnitrogendeposi- tionweredeterminedforalpinelakeecosystemsintheWesternUS usingfossildiatomassemblagesinlakesedimentcores(Sarosetal., 2011).A transferfunctiontechnique, wherecalibration training sets are constructed to show relationships between individual species and one or more quantifiable (climate change related) parameters,followedbyareconstructionofchangesinecosystem functionsand services based ona biological data,is useful for identifying ecological regime shifts caused by environmental perturbations (Battarbee, 2000). As an example, sub-fossil assemblagesofdiatomsinlakesaroundtheeasternSierraNevada and theGreater Yellowstone National Park suggeststhat 1.4kg Nitrogenha ¹y ¹ can cause a regime shift in the diatom community (Saros et al., 2011). The advancement of dating techniques andtheuseof multiplespecies approachesfor lake sedimentscanfurtherhelpimproveclimatereconstructionand define climatethresholds of freshwater ecosystems (Battarbee, 2000),butitisstillnecessarytodisentangletherelativeroleofall influentialfactorsusinga combinationofbothlimnologicaland paleolimnologicaldata(Battarbeeetal.,2012).Italsoneedstobe pointed out that the historically caused perturbations result primarilyfrominternalshiftsandclimatechange,whereascurrent human induced perturbation (e.g., land use, changed nutrient budgets) is now superimposed on internal shifts and climate- relatedchanges,makingitdifficulttousepastchangesasdirect analogsofcurrentandfuturethresholds.

6.3.Developingappropriatenumericaltoolsandmodels

Precisioninecosystemprojectionswillincreasewhenquanti- tativeestimatesofthresholdsarebackedbyreliablymeasureddata (Woodward et al., 2010). Scenario-based models can provide decision-makers a range of possibilities, as well as integrate naturalandsocial sciencebasedstrategiesforadaptation(IPCC, 2007).TheIntergovernmentalPanelonClimateChange(IPCC)has implementedavarietyofscenario-basedclimate-changemodels, includingmodelsassociatedwithhistoricalfreshwaterecosystems (IPCC, 2007). However, models to describe thresholds that influence individual freshwater ecosystems have not yet been developed,duetotheuncertaintyinfutureclimateprojectionsas well as scale issues, as small lakes cannot be adequately represented using coarse resolution climate models (Groffman et al., 2006). Some attempts have been made for improved modelingoftheecologicalthresholdssuchasthetrophiccascade modelforgamefish(CarpenterandBrock,2004)andthewater quality model for DO concentrations in North American lakes (Stefan et al., 1993). Advancement of computer software and numericalmodelingwillcontinuetoimprovetheknowledgeof thresholdbasedecologicalmanagementstrategies.

Several types of models, e.g., statistical, rule-based, the previously mentioned DLM models,and process-based models, have beenused to guideecosystem management under global

change. Process-based models are based ona scientific under- standingofrelevantecologicalprocesses,andprovideavaluable framework toinclude responses tothe changedenvironmental conditions(Cuddingtonetal.,2013).Thecurrentunderstandingof theroleofthresholdsisoftenbaseduponconceptualmodelsof how ecosystems work (Dennison et al., 2007). Process-based threshold models can be specifically designed to anticipate ecological consequences of human activities on freshwater ecosystems, and should playa keyrole in setting conservation targetsbywaterresourcesmanagers.

6.4.Testingclimatologicalhypotheses

Laboratory or field-based experimental designs have the potential to characterize ecological thresholds (Adler et al., 2009).Forinstance,alaboratoryexperimentonburrowingmayfly, Hexagenialimbata,fromtheLowerMobileRiver,Alabama,suggests that thenymphsof this animal cansurviveatelevated salinity levels(6.3%)onlyfortemperaturebelow18C,however,whenthe salinity level is reduced to 2.4% they can survive at higher temperaturesupto28C, whichwas furthersupported byfield observations (Chadwick and Feminella, 2001). Similarly, an experiment which reduced the pHof a small lakefrom 6.1 to 4.7wasfoundtocauseasubstantialdeclineinspeciesrichnessof manyfreshwatertaxa,reducingtheecosystem’sadaptivecapacity (Hogsdenet al.,2009).These studiesassistwithunderstanding which species within ecosystems are the most vulnerable to climatechange,andhowthedisappearanceofthesespecieswill influencethebroaderecosystemservices.

6.5.Relatingthreshold-associatedcoststoecologicalandsocietal adaptation

Ecosystemservicechangesalsoinfluencethesocietalstructure ofthepopulationdependentonthoseservices.Lossofecosystem services,includingwaterquality,waterquantity,fisheryresources, and recreationalamenities, can lead tothe crossing additional thresholds associated with social systems and unacceptable societalcosts.Integratingkeycomponentsofsocietaldevelopment canpromoteecologicalresilience,includinginformationmanage- ment,culturalintegrity,technologicaldevelopment,institutional responsibility,andeducation(Falkenmark,2003).Considerationof aprobabilisticframeworkwitharangeofcomponentscanoffer policy and management responses to emerging crisis from freshwater ecosystem changes worldwide (Vörösmarty et al., 2010).Specifyingthethresholdsrelatedtotime,location,species, disturbance and scale, and understanding their economic and societalimplicationscanhelpsetupregulations,improveadaptive capacityoffreshwaterecosystems,andfacilitatesocietaladapta- tion (Naiman and Turner, 2000). The socio-economic costs of crossing ecologicalthresholdsinfreshwaters needtobeplaced clearlyintothecontextofthreshold-basedmanagementprogram (Table 1). This involves a careful cost-benefit analysis of the servicesaffectedbeforeandaftercrossingthresholds.

Certain threats to freshwaterecosystems maybe addressed throughinvestmentsandadvancesininfrastructure,scienceand technology(Vörösmartyetal.,2010).Thoughthismaybemainlya questionofpoliticalwillinrichercountries,poorercountriesoften lacktheresourcestoinvestinfreshwaterinfrastructure(UNFCCC, 2007).ItisestimatedthattheUnitedStatescurrentlywouldhave toputin$60billiondollarsfortherestorationofitsecosystems.In regionsliketheMiddleEastandNorthAfrica,waterpollutionand over exploitation of water is expected to incur environmental damage costs equivalent to 2.1–7.4% of their gross domestic product(GDP).Similarly,thecostofwaterrelatedcrisesinChina was2.3%ofChina’sGDPin2003.

Assigning an economic value to the crossing of ecological thresholdsisaneffectivewayofcommunicatingthesocietaland economicalcoststodecisionmakersandstakeholders(Balmford etal.,2002).Forexample,theuseofartificialdrainageinSouth Ontario marshes for the purpose of increasing agricultural productivity was foundto beless beneficialeconomically than notdraining(vanVuurenandRoy,1993).Thecost-benefitanalysis suggestedthatthenaturallyintactmarshecosystemswouldcost less (3700USDha ¹) than the artificially drained marshes (8800USDha ¹).Thustheuseofscenario-basednumericalmodels offreshwatersystemscoupledwitheconomicalmodelsisstrongly encouraged,asitcanhelp implementthreshold-basedmanage- ment approaches of freshwater ecosystems effectively, and therebyunderscoretheservicesthatintactecosystemscanprovide tosociety(deGrootetal.,2002).Riskanalysisshouldbeconveyed togetherwith thresholdanalysis, in which models are used to predicttheriskandbenefits(environmental,economical,societal) from thresholds being crossed under different climate change scenarios,aswellastheriskandbenefitsfrompotentialmitigation efforts.

7.Threshold-basedmanagementinfreshwaterecosystems

Thegoalof ecosystemmanagement istorestoreorpreserve some main attributes of an ecosystem that are desirable for humans (Mayer and Rietkerk, 2004). Due to the inherent complexityof ecosystems, this canonly beachievedthrough a framework (e.g., Fig. 1) involving long-term collaborations between scientists, policy makers and stakeholders. The main goal of such collaborationwould be to set up a warning and preventionsystemfromapproachingthresholds(nearF0orF1),as wellas mitigation strategies to prevent such thresholdsbeing reached. In this effort, scientists with the assistance of local stakeholderswould monitorrelevantenvironmentalparameters and potential threshold responses (examples are presented in Table1).Decisionmakerswouldsetupaprogramtoencourage suchmonitoring efforts, as well as be prepared in advance to enforceappropriatemitigationstrategiesasneeded(e.g.,lowering fish quotas, capping contaminated areas, preventing excess nutrients from entering the local area). Mitigation strategies wouldideallyhave tobeplannedfarwellinadvancebased on previouslyconductedsimulations and risk analysis. Clearcom- munication with stakeholders regarding risks from crossing thresholdsis critical,particularly incaseswhere longtermand short term interests collide (e.g., over fishing, urbanization).

Further,stakeholdersshouldbeencouragedtonotifyauthoritiesin case a rapid loss in ecosystem services occurs (potentially indicating a threshold F₁ beingmet)(Fig.1), toget involved in monitoring programs (e.g., species counting, environmental parameterlogging), aswellasbe presentedwithopportunities toactinawaytomitigaterisksfromboundariesF₀beingreached.

Disagreement mayoften occuramongst actorsonpractices, remediationandlogisticissues.Scientistsbeingtoocautiousabout thepositionofF₀,mayrisklosing trustbydecisionmakersand stakeholders,particularlyifnochangein aresponse variableis evidentafterF0iscrossed(Fig.1).Ontheotherhand,ifscientists areundercautious,predictinganF0thatisclosetoF2,thentrustin scientistswouldalsobelostalongsidetheecosystemservice.

Ecological systemsarenotsimpletounderstand,neitherare thresholdresponses.Afterathresholdiscrossed,thesituationmay becomerestoredbyvariousfeedbackloops(Fig1).Theprudent optionis,however,toassumethatthisdoesnothappen(Briske etal.,2010).Avoidingdisagreementsamongstactorsisessential whenecosystemisatarisktocrossthresholds.Disagreementsare resolved through regular interactions by organizing public involvement (e.g., through fairs, festivals, workshops, panel

discussions,newsmedia),aswellasresearchprojectsandsocietal awarenessprogramsimplementedamongstecologists,modelers, economists, sociologists, resource managers, stakeholders and decision makers,activistsandreporters(Carpenteretal.,2009;

Lindenmayer and Likens, 2009). Bringing these actors in a common, formerly established platform can be a successful approach for freshwater ecosystems globally (Carpenter et al., 2009; MEA, 2005). This reflects an adaptive management approach, where solutions to the problems are proposed and implemented, and the management strategies are constantly reviewedoverthecourseofecologicalresponsetoclimatechange (Williams,2011).Duetomultiplespatialandtemporalscalesof ecosystems,Côtéand Darling(2010)arguedthattheecosystem management to control local anthropogenic disturbance, for examplenutrientinput,andtoreversetheecosystemdegradation willinadvertentlylowerresiliencetoclimatedisturbance.Ongoing research on freshwater ecosystem management is needed, especially in regards to integrating rare or extreme events in freshwaterecology (e.g.,see Fuentes etal., 2006;Denny et al., 2009), and evaluating evidence and uncertainty in threshold conceptsandmodels(KatharineandRichard,2009).Establishing programsthatseektoidentifyorpreventthresholdsfrombeing crossed,andwhichhavetheinvolvementofthescientific,policy and stakeholder communities, are a wayto effectively manage threatsoflocalimpactsandglobalclimatechangeonfreshwater ecosystemservices.

Acknowledgements

ThisstudywassupportedbytheInternationalS&TCooperation ProgramfromtheMinistryofScienceandTechnologyofChina(No.

2012DFA91530),Special FundforForestryScientificResearchin the Public Interest (No. 201204204), Projects of International Cooperationand ExchangesNSFC(41161140353, 91325302),the 1stYouthExcellentTalentsProgramoftheOrganizationDepart- ment of the Central Committee of the CPC, the Fundamental ResearchFunds fortheCentral Universities(TD-JC-2013-2),and theNGIGBVFund(H.P.H.Arp).WealsothankPotsdamInstitutesof ClimateChangeImpacts(PIK)inGermanyforsupportingJ.Liu’s visits,and theUniversityof Leedsfor providingJ.Liu aCheney SeniorFellowship.WethankSteveCarpenterfromtheUniversity ofWisconsinandElenaBennettfromMcGillUniversityforsharing discussionsthroughthepreparationofthemanuscript.

AppendixA.Supplementarydata

Supplementarydataassociatedwiththisarticlecanbefound,in the online version, at http://dx.doi.org/10.1016/j.ecolmo- del.2014.09.010.

References

Adler,P.B.,Leiker,J.,Levine,J.M.,2009.Directandindirecteffectsofclimatechange onaprairieplantcommunity.PLosOne4,e6887.

Alberti,M.,Booth,D.,Hill,K.,Coburn,B.,Avolio,C.,Coe,S.,Spirandelli,D.,2007.The impactofurbanpatternsonaquaticecosystems:anempiricalanalysisinPuget lowlandsub-basins.Landsc.UrbanPlann.80(4),345–361.

Andersen,H.E.,Kronvang,B.,Larsen,S.E.,Hoffmann,C.C.,Jensen,T.S.,Rasmussen,E.

K.,2006.Climate-changeimpactsonhydrologyandnutrients inaDanish lowlandriverbasin.Sci.TotalEnviron.365,223–237.

Anestis,A.,Portner,H.O.,Michaelidis,B.,2010.Anaerobicmetabolicpatternsrelated tostressresponsesinhypoxiaexposedmusselsMytilusgalloprovincialis.J.Exp.

Mar.Biol.Ecol.394,123–133.

Balmford,A.,Bruner,A.,Cooper,P.,Costanza,R.,Farber,S.,Green,R.E.,Jenkins,M., Jefferiss,P.,Jessamy,V.,Madden,J.,Munro,K.,Myers,N.,Naeem,S.,Paavola,J., Rayment,M.,Rosendo,S.,Roughgarden,J.,Trumper,K.,Turner,R.K.,2002.

Economicreasonsforconservingwildnature.Science297,950–953.

Baron,J.S., Poff,N.L., Angermeier,P.L.,Dahm,C.N.,Gleick,P.H.,Hairston,N.G., Jackson, R.B., Johnston, C.A., Richter, B.D., Steinman, A.D., 2002. Meeting ecologicalandsocietalneedsforfreshwater.Ecol.Appl.12,1247–1260.