GLOBIO-Aquatic, a global model of human impact on the biodiversity of inland aquatic ecosystems

GLOBIO-Aquatic,

a

global

model

of

human

impact

on

the

biodiversity

of

inland

aquatic

ecosystems

J.H.

Janse

*

,

J.J.

Kuiper

,

M.J.

Weijters

,

E.P.

Westerbeek

,

M.H.J.L.

Jeuken

,

M.

Bakkenes

,

R.

Alkemade

,

W.M.

Mooij

,

J.T.A.

Verhoeven

a_{PBL,NetherlandsEnvironmentalAssessmentAgency,Dept.ofNatureandRuralAreas,P.O.Box303,NL-3720AH}

Bilthoven,TheNetherlands

b_{NetherlandsInstituteofEcology(NIOO-KNAW),Dept.ofAquaticEcology,P.O.Box50,NL-6700ABWageningen,}

TheNetherlands

c

EcologyandBiodiversity,InstituteofEnvironmentalBiology,UtrechtUniversity,Padualaan8,NL-3584CHUtrecht,

TheNetherlands

d_{WageningenUniversity,EnvironmentalSystemsAnalysis,P.O.Box47,NL-6700AAWageningen,TheNetherlands}

e_{WageningenUniversity,AquaticEcologyandWaterManagement,P.O.Box47,NL-6700AAWageningen,}

TheNetherlands

a

r

t

i

c

l

e

i

n

f

o

Landusechange

Eutrophication Hydrologicaldisturbance Catchment Scenarioanalysis Cyanobacteria

a

b

s

t

r

a

c

t

Biodiversityinfreshwaterecosystems–rivers,lakesandwetlands–isundergoingrapid

globaldecline.Majordriversarelandusechange,eutrophication,hydrologicaldisturbance,

climatechange,overexploitationandinvasivespecies.Wedevelopedaglobalmodelfor

assessingthedominanthumanimpactsoninlandaquaticbiodiversity.Thesystemconsists

ofabiodiversitymodel,namedGLOBIO-Aquatic,thatisembeddedintheIMAGEmodel

framework,i.e.linked tomodelsfordemography,economy,landusechanges, climate

change,nutrientemissions,aglobalhydrologicalmodelandaglobalmapofwaterbodies.

Thebiodiversitymodelisbasedonarecompilationofexistingdata,therebyscaling-upfrom

local/regionalcase-studiestoglobaltrends.Wecomparedspeciescompositioninimpacted

lakes,rivers andwetlandstothatincomparableundisturbedsystems.Wefocussedon

broad categoriesofhuman-inducedpressuresthatarerelevantattheglobalscale.The

driverscurrentlyincludedarecatchmentlandusechangesandnutrientloadingaffecting

waterquality,andhydrologicaldisturbanceandclimatechangeaffectingwaterquantity.

Theresultingrelativemeanabundanceoforiginalspeciesisusedasindicatorfor

biodiver-sityintactness.Forlakes, weuseddominanceofharmfulalgalbloomsasanadditional

indicator.Theresultsshowthatthereisasignificantnegativerelationbetweenbiodiversity

intactness and thesestressors in all typesof freshwater ecosystems. In heavily used

catchments,standingwaterbodieswouldloseabout80%oftheirbiodiversityintactness

andrunningwatersabout70%,whileseverehydrologicaldisturbancewouldresultinlosses

*Correspondingauthorat:PBL,NetherlandsEnvironmentalAssessmentAgency,Dept.ofNatureandRuralAreas,P.O.Box303,NL-3720AH

Bilthoven,TheNetherlands.Tel.:+31650834609.

E-mailaddress:jan.janse@pbl.nl(J.H.Janse).

1 _{Presentaddress:B-ware,Nijmegen,TheNetherlands.}

2 _{Presentaddress:MangroveInfo&KayakCentre,Bonaire,Caribbean,TheNetherlands.}

3 _{Presentaddress:Deltares,Delft,TheNetherlands.}

Available

online

at

www.sciencedirect.com

ScienceDirect

journalhomepage:www.elsevier.com/locate/envsci

http://dx.doi.org/10.1016/j.envsci.2014.12.007

1462-9011/#2014The Authors. Publishedby ElsevierLtd.This is anopen access articleunder theCC BY-NC-ND license (http://

1.

Introduction

An estimated 11–13 million km2_{, or 8–9% of the earth’s}

continentalsurfaceconsistsofinlandaquaticecosystems,of

which about 21% are lakes, 3% reservoirs, 3% rivers, 33%

floodplain marshes and swamps, 6% coastalwetlands, and

35%otherwetlands(LehnerandDo¨ll,2004).Thesesystemshost

ahighanduniquebiodiversityanddeliverimportantecosystem

serviceslikefreshwateravailability,waterpurification,climate

regulation,foodandrecreationalvalues(MEA,2005a).

Globalfreshwaterbiodiversityisdecliningandisexpected

tofurtherdecline(MEA,2005b;Revengaetal.,2005;CBD,2014),

possiblyat evenhigherratesthaninterrestrialandmarine

habitats(Lohand Wackernagel,2012). Aquatic systemsare

especiallyvulnerable becausehuman populationdensity is

on average higher near lakes, rivers and estuaries, and

becausewaterbodiesaccumulatetheeffectsofdevelopments

in their catchment (Williamson et al., 2008). Population

increase,economicdevelopment,foodandfueldemandand

urbanization are the main indirect anthropogenic drivers

causingthisdeclineofbiodiversityattheglobalscale.These

leadtomanifolddirectdriversofchangewhichcanbeassigned

toseveral broadcategories:land-use changes,hydrological

disturbance (both leading to loss of habitats), pollution,

climatechange,overexploitationand exoticspeciesarethe

most-mentionedones(Salaetal.,2000;Revengaetal.,2005;

MEA,2005b;Dudgeonetal.,2006).

Oneofthemostprominentdirect driverscontributingto

the declineof aquaticbiodiversity at a global scaleis land

use change, which involves both the direct conversion of

wetlandsaswellasindirecteffectsofland-useinthecatchment

(WatzinandMcIntosh,1999;Allan,2004;Revengaetal.,2005;

Verhoevenetal.,2006).Recentestimatesstatethatover60%of

wetland area has been converted worldwide since 1900

(Davidson,2014).Indirecteffectsofland-usechangesinclude

elevated suspended solid concentrations resulting from

in-creasederosionafterdeforestation(Wissmaretal.,2004;Cohen

etal.,1993),eutrophication(nutrientconcentrationsgenerally

stronglycorrelatewiththeintensityoflanduseintheupstream

catchment(CrosbieandChow-Fraser,1999;Harper,1992)),and

increasedpollutionbyother(toxic)substances.Acombination

oftheseandotherfactorsrelatedtoland-usechangesleadsto

changes in river channels and floodplains that disturb the

naturalhabitatsofaquaticbiota(Allan,2004).

A second important category contributing to a global

decline in aquatic biodiversity is hydrological disturbance

resulting from water withdrawal for e.g. irrigation and

publicwatersupply,andfromregulationofwaterflowsby

infrastructure for e.g. hydropower generation, protection

against flooding, navigation or water storage (Rosenberg

etal.,2000;BunnandArthington,2002).By2010therewere

about 50,000dams (higherthan 15m),creatingatotalarea

ofabout300,000km2_{ofreservoirsandimpactingsome70%}

oftheworld’srivers(Lehneretal.,2011).Damsaffectbiota

via disruption of the natural flow regime or the seasonal

flood pulse to which organisms are adapted (Ward, 1998;

Keddyetal.,2009)andbyblockingmigrationroutes(Poffetal., 1997; Poff and Zimmerman, 2010). In wetlands and lakes,

hydrological alternation maycause changesinwaterlevel,

floodingordesiccation(NilssonandBerggren,2000;Wantzen

etal.,2008).

Recently global climatechange has been identifiedasa

dominantdriver ofchangeaffectingaquatic ecosystemsin

severalways(e.g.Palmeretal.,2008;Mooijetal.,2005;Moss

et al., 2009; Vescovi et al., 2009), including rise in water

temperature and hydrological changes (such as increased

peakdischargesorlongperiodsoflowflow).Thelattermay

alsoleadtoincreasednutrientloadingand,insomeregions,

salinization. In streams, temperature increase mayleadto

extinctionofcharacteristicspecies.Instandingwaters,biotic

communities willbeaffected by a range ofprocesses, like

increased frequency of stratification periods, productivity

increasesandalgalblooms.Climatechangecanaggravatethe

effectsofeutrophication(Mooijet al.,2005;Jeppesen etal.,

2009;Mossetal.,2011).

Many other factorshave beendescribed as influencing

biodiversity atvarious scales,such asinvasions ofexotic

species (e.g.Salaetal.,2000;Leprieuret al.,2009)andthe

exploitation of aquatic biota (FAO, 2012) including the

harvest of food (fish, crustaceans and other organisms)

and fibre (reed, papyrus). In addition, water bodies are

increasinglyusedforaquaculture,aboomingsector,which

is already responsible for about half of the world’s fish

production for human consumption, of which 80% takes

place in freshwaters,mainly inAsia (FAO, 2012). Impacts

includeeutrophication,pollution,escapeofcultured

organ-ismstothewildandspreadofdiseases. Additionalfactors

influencing biodiversity atvarious scalesare local habitat

changes,acidification,salinization,organicpollution,

genet-icdisruptionandtoxicstress.

The worldwide losses of biodiversity and ecosystem

services areamajorconcerntopolicymakersat thelocal,

national andinternationallevel.Examplesofthe latter are

ofabout80%inrunningwatersandmorethan50%infloodplainwetlands.Asan

illustra-tion,ananalysisusingtheOECD‘baselinescenario’showsaconsiderabledeclineofthe

biodiversityintactnessinstillexistingwaterbodiesin2000,especiallyintemperateand

subtropicalregions,andafurtherdeclineespeciallyintropicalregionsin2050.Historical

lossofwetlandareasisnotyetincludedintheseresults.Themodelmayinformpolicy

makersatthegloballevelinwhatregionsaquaticbiodiversitywillbeaffectedmostand

bywhatcauses,andallowsforscenarioanalysistoevaluatepolicyoptions.

#_{2014TheAuthors.PublishedbyElsevierLtd.ThisisanopenaccessarticleundertheCC}

international forums like the Convention on Biological

Diversity(CBD),RamsarConventiononWetlands,EU,OECD

andotheragencies,internationalNGO’sand companies.To

evaluatepolicyoptions, thereisagreatneedforintegrated

modelsthatdynamicallydescribethedriversofchangeand

theirimpactonbiodiversity(MEA,2005b;PereiraandCooper,

2006;Dudgeon,2010;CBD,2014).

Several global models exist that describe important

driversofchange, suchasclimate,hydrology(Vo¨ro¨smarty

etal.,2000;Do¨lletal.,2009;Biemansetal.,2011;VanBeek

et al., 2011) and nutrient loading (Seitzinger et al., 2010;

Bouwman et al., 2011). Vo¨ro¨smarty et al. (2010) made an

integrated model ofthreatsto biodiversity inrivers.Most

genericsimulationmodelsonecologicalprocesses(seefor

instance overview for lakes by Mooij et al. (2010)) do not

specificallyaddressbiodiversity,althoughcorrelativemodels

have been developed for species composition related to

abioticfactors(seee.g.overviewforstreamsbyVerdonschot

(2000)). These models, however, are generally confined

to smaller spatial scales or specific water types. Direct

correlation between species distribution data and abiotic

dataatlargerscaleshaveproducedmuchinsight,butmay

failtodiscernnaturalandanthropogenicfactorsormaygive

misleadingresultsduetocovariationbetweenfactors(e.g.

Xenopoulosetal.,2005).Inconclusion,abroad,overarching

and consistent model of the human impacts on aquatic

biodiversity in inland waters is not yet available. This

paperdescribestheoutlinesofsuchaglobalmodel,called

GLOBIO-Aquatic.

GLOBIO-Aquaticmodelsthedominanthumanimpactson

inlandaquaticbiodiversityusingameta-analysisofexisting

information. Itconsistsofaset ofempirical relationships

betweenenvironmentaldriversandtheirimpacton

biodi-versityindifferentaquaticecosystems.Thefocusisonbroad

categoriesofhuman-inducedpressuresthatstillholdwhen

scaled up from a local/regional level to the global level.

Currentlythedriversland-usechange(including

eutrophica-tion),hydrologicaldisturbanceandclimatechangeare

investi-gated. The severity of impacts is expressed as a biotic

intactnessindexrelativetotherespectivereference

compo-sition (i.e. inthe undisturbedstate; see Section 2.2). This

allowsstudyingtheimpactsofdifferentdriversinconcert,

and comparing their impacts among different types of

aquaticecosystems.Hence,wedonotaimto‘explain’the

biodiversitypatternsinallkindsofaquaticecosystems,but

rather to ‘extract’ the impact of the main anthropogenic

pressures on the natural species pattern. The model is

embedded in the IMAGE model framework for land use

andglobalenvironmentalchange(Stehfestetal.,2014)and

is complementary to the GLOBIO model for terrestrial

ecosystems(Alkemadeetal.,2009).

We first describe the chain of models to estimate the

magnitudeofthemaindriversofchange.Wethenpresentthe

biodiversity intactness indicators. We document how we

linkedtheabove-mentioned driversandthe biodiversityof

rivers, lakes and wetlands and compare the biodiversity

impactamongdifferentecosystems.Finally,wepresentthe

implementationandapplicationofthemodelchain,andusea

globalbaselinescenario(OECD,2012)fortheperiod2000–2050

asanexample.

2.

Methods

2.1. Driversandmodelchain

Theenvironmentaldriversareevaluatedthroughachainof

globalmodelsandmapsconsistingofalanduseandclimate

changemodel,ahydrologicalmodel,anutrientmodelanda

mapofthewaterbodies.Thecatchmentapproachisapplied

by includingthe spatialrelationsbetweenpixels, basedon

flow direction.Fig.1shows schematicallytherelationships

betweenthemodelsforthedriverscurrentlyaddressed.

Projections of land use and climate change arederived

from the IMAGE model (Stehfest et al., 2014) that uses

projections on human population size, economic growth,

foodandenergyrequirements,andfoodtrade,tomodelfuture

agriculturallanduse.Basedontherequirementsandsources

ofenergy,IMAGEalsomodelstheworld’scarbonemissions

andclimaticchangessuchastemperature,precipitationand

potential evapotranspiration. The Global Nutrient Model

(Beusen,2014)translatesfuturepopulationsizeand

agricul-turalland-usepatternsintosoilnutrientbudgets(Bouwman

et al.,2011)andnutrient loadingstoaquaticsystems,from

both diffuse and point sources. Nitrogen and phosphorus

leachingand runofffromtheland tothesurfacewatersis

modelled based on agricultural area, the application of

fertilizerandmanure,precipitationandspatialcharacteristics

ofslope,soiltextureandgroundwatercharacteristics.Urban

nutrientemissionsaremodelledbasedonpopulation,

afflu-ence(GDP),sanitationandtheuseofdetergents(VanDrecht

etal.,2009).Retentionofnutrientsintheglobalsurfacewater

networkisincluded,basedonslopeandretentiontime.

Water discharge iscalculatedby theglobal hydrological

model PCR-GLOBWB (Van Beekand Bierkens, 2009) or the

hydrological moduleof the global vegetation modelLPJmL

(Biemans et al., 2011). The discharge is based on a water

balanceperpixel,includingprecipitation,evapotranspiration,

snowmelt, infiltration to groundwater and human water

abstraction. In these models, the discharge is affected by

climatic variables, land use, waterabstraction, and by the

presenceandwayofmanagementofdamsandreservoirs.The

twomodelsdifferintimescale,intheschematizationofriver

floodplainsandwetlandsandinthedefinitionofvegetation

andcroptypes.ThemodelPCR-GLOBWBalsocalculatesthe

watertemperature(whichiscurrentlyusedforthealgalbloom

indicator only;see Section 2.5). Data on existing dams are

takenfromtheGRANDdatabase(Lehneret al.,2011)anda

projection of future dams made according to Fekete et al.

(2010).

Thedeviationbetweennaturalandimpactedflowpattern

is derived from the modelled discharges as the ‘amended

annual proportional flow deviation’ (AAPFD; Ladson and

White, 1999 and implemented as described by Biemans etal.,2011): AAPFD¼ X 12 i¼1 QiQi0 ¯ Qi0  2 2 4 3 5 1 2 (1)

Thisdeviationisaveragedovertheyearsofrecord.Inthe

naturalrunoffintheithmonthand ¯Qi0fortheyear-averaged

natural runoff. Thevalue ofAAPFD may range from0 for

unregulatedriversto+1;ingeneral,valuesabove3denotea

strongdeviation.

Driversarecurrentlymodelledinaspatialresolutionof

300_300_{(approx.50km50kmattheequator).Allfluxesare}

accumulated downstream according to the water routing

routine,whichisbasedonadigitalelevationmap(DEM).The

locationandtypeofwaterbodiesisbasedontheGlobalLakes

andWetlandsDatabasemap(LehnerandDo¨ll,2004),whichis

availableatdifferentresolutions.Thismapdiscernsthemain

inlandwatertypes:lakes,reservoirs,riversandseveraltypes

ofwetlandsnamelyriverinemarshesandswamps,isolated

wetlands (bogs, etc.), intermediate, brackish and coastal

wetlandsaswellaswetlandmosaics. Fromthe GLWDwe

calculatedthefractionalareaofeachtypeofsurfacewaterin

each grid cell. The routing model and GLWD map are

combined to estimate the nutrient loadings to the water

bodies of the GLWD categories 1–6 (i.e. lakes, reservoirs,

rivers,floodplainwetlands,swampsandcoastalwetlands)

and the fraction of human land use in their upstream

catchment. Theother wetland types (categories 7–12) are

assumed to be more isolated and hence to have their

catchmentsconfinedtoonlythegridcellinwhichtheyare

located.Dataonlakedepthsare(ifavailable)derivedfrom

the ‘FLAKE’ dataset (Kourzeneva, 2010). Lakes aredivided

intothecategories‘shallow’and‘deep’basedonaboundary

valueof3mmeandepth.Incaseofmultiplevalueswithina

cell,afrequencydistributionofthesecategoriesiscalculated.

Missing values are estimated by (in this order): (a) the

elevation map:lakesinmountainousregionsareassumed

to be deep; (b) expert judgement based on regional

char-acteristics;(c)nearestneighbour(onlywithinabiome);and

(d)theworldaverage(percategory).

Allaforementioned driversareconfinedtothe existing

waterbodies(i.e.definedintheGLWD-2004,basedondata

fromthe1990s).Astherearenohistoricallakeandwetland

mapsavailabletoestimatehistoricalwetlandconversions,a

firstattempt was made toderive suchamapfrom model

calculations(Brolsmaetal.,2012).ThemodelPCR-GLOBWB

(VanBeeketal.,2011;VanBeekandBierkens,2009)wasrun

with onlynaturalhydrologicalinput,excludingallhuman

interventions, and all permanently inundated areas were

selected. This gives an estimate of all potential natural

wetlands(disregardinghistoricalclimatechange).Forfuture

projections of wetland conversion to human land use, a

model was made by VanAsselen et al. (2013)basedon a

meta-analysisofconversionsthathaveoccurredinthelast

century. The conversion risk can be calculated from a

number of physicalandsocio-economic drivers, ofwhich

agricultural demand appeared to be prominent. But this

studydidnotcoverthefactorsdeterminingtheprecedence

of wetlands or other land cover types for conversion. As

these modulesare stillindevelopment,they havenotyet

been included in the version documented in this paper.

Instead, as a conservative guess, a minimum estimateof

wetlandconversionwasmade,basedontheareaofGLWD

wetlandsminimallyrequiredtomeettheprojectedincrease

inagriculturallanddemandifallnon-wetlandareas(such

asforests)havebeenused.

Fig.1–Modelchainforfreshwaterbiodiversity.Rectanglesdenotevariablesorprocesses,ovalsdenotemodels,rounded

rectanglesdenotedata,blackarrowsdenotemodelinputoroutput,bluearrows(inwebversion)orgreyarrows(inprint

2.2. Biodiversityindicators

Toexpressbiodiversity,weusedindicatorsthatallowedusto

quantifyandcomparetheecologicalimpactforhighlydifferent

studiesandecosystemtypes.Themostimportantindicatorused

inthisstudyis‘biodiversityintactness’or‘naturalness’ofthe

bioticcommunity,denotedas‘MSA’(MeanrelativeAbundance

oforiginalSpecies)(Alkemadeetal.,2009).Thisindicator,also

referred toas‘relativetaxonrichness(RTR)’(Verboom etal.,2007;

Weijtersetal.,2009),isrelatedtotheBiologicalIntactnessIndex

(BII)(ScholesandBiggs,2005)andisalsousedintheterrestrial

GLOBIOmodel.TheMSAcalculatesastheaverageremaining

abundance of originally occurring species, relative to the

correspondingnaturalabundance,ona0–1scale:

MSAs¼ P iRis N (2a) Ris¼ Aisd Aisc (2b)

whereMSAs isthemeanrelative abundanceoftheoriginal

speciesestimatedinstudys,Nthenumberofspeciesinthe

study,andRistheratiobetweentheabundanceofspeciesiinthe

disturbed(Aisd)andthecorrespondingundisturbed(reference)

situation(Aisc),respectively.The‘referencesituation’maybe

thesituationofthesamewaterbodybeforethedisturbance

occurred,oranaturallycomparableundisturbedwaterbodyin

thesameregion.Theabundanceofaspeciesmaybegivenas

numberofindividualspersite,thenumberofsitesatwhichthe

speciesis found,thepooled abundanceover the year, or a

comparablemetricthatfulfilstheoverallaimofametricof

thedegreeofoccurrenceofthespecies.Onlythosespeciesthat

occurinthereferencesituationareincluded,andtheratiofor

eachspeciesistruncatedat1:anincreaseofaspeciesbeyondits

‘undisturbed’densityisnotconsideredasanimprovement.The

MSAconceptallowsscalingandcomparingdifferentecosystem

types.Itdiffersfromspeciesrichnessorotherdiversity

indica-torsliketheShannon–Wienerindexinthatspeciesthatonly

occurinthedisturbedandnotintheundisturbed(‘pristine’)

situation,arenotincluded.Invasionby‘exotic’speciesisnot

reflectedintheindicator,butisindirectlyaccountedforbyan

assumedlinkwithadeclineofnativespecies.

TheMSAindicatorisrelatedtothewidelyusedIndexof

BioticIntegrity(IBI),a multi-metricindex thatdescribesan

ecosystem’sbioticcommunitycomparedwithitsundisturbed

state.TheoriginalversionoftheIBI(Karr,1981),basedonfish

data,had12metricswithvaluesbetween1(disturbed)and5

(pristine),andhencecumulativerangesbetween12and60.

TheIBImethodhasbeenfurtherdevelopedand(regionally)

adaptedformanydifferentecosystemsandbioticgroups(Karr

andChu,2000;Wrightetal.,2000;Parsonsetal.,2002),with

varyingnumberandnatureofmetrics.Forstudiesinwhich

onlythe IBI valueswere reportedinstead of theraw data,

we transferred these into MSA values by rescaling them

between0and1fortheminimumandmaximumIBIvalues,

respectively,assuminglinearinterpolation.So,

MSA¼ IBIIBImin

IBImaxIBImin

(3)

In studies where species presence or abundance data

(allowingMSAcalculation)werepublishedbesidesIBIscores

(e.g.CreweandTimmermans,2005),thecorrelationbetween

the two was high (r2=0.62), justifying the use of both

indicatorsinthisstudy.

TheMSAcanalsobelinkedtotheEcologicalQualityRatio

(EQR)usedintheEuropeanWaterFrameworkDirective.Thisis

also an indicator, scaled 0–1, based on the biota (species

compositionandabundanceoffunctionalgroups)relativeto

thereferenceconditionoftherespectivewatertype.TheEQR

uses dataon macrophytes, algae, macro-invertebrates and

fish. Although the EQR calculation process is complicated

and notnecessarily linear, in thisstudy theEQRhas been

converted1:1intoMSAforthecasesconcerned.Inascoping

study in The Netherlands, the average values of both

indicators wereingood agreement(PBL, 2008), but amore

rigorouscomparisonwouldbeneeded.

Complementary to the MSA, the occurrence ofharmful

algalblooms(primarilycyanobacteria)hasbeenincludedas

anindicatorfortheecologicalstatusoflakes.Algalblooms

areoftenusedasadisturbanceindicator,generallynegatively

related to MSA, as phytoplankton dominance excludes

othernativespecies.Thealgalbloommoduleisincludedto

covertheimpactsofclimatechangeintermsoftemperature

rise.

2.3. Datacollection

Therelationbetweentheselectedenvironmentaldriversand

thebiodiversityinrivers,lakesandwetlands,wasbasedon

meta-analyses of literature data. Studies were selected in

which biodiversity data in impacted systems had been

compared with those in undisturbed reference systems,

eitherintime(before–after)or space(providedthesystems

were comparable concerning natural factors). Case studies

werederivedfromliteraturepublishedinscientificjournals,

reportsorbooks,disclosedbyonlinesearchengines(Scopus,

Google Scholar and Web of Science) and/or referenced in

reviewpapersand,occasionally,ondatasetsobtainedthrough

personal communication. Grey literature has not been

surveyed. Search terms were grouped in four categories,

comprisingtheecosystemtypeunderconcern,thedriverof

interest, the effect parameters, and the type of study

(comparator), respectively (Table 1). Search terms were

combinedtoselectpapersthatcontainedat leastoneterm

fromeachgroupintheirtitle,keywordsorabstract.Fromthe

hits,weselectedpapersthatmetthefollowingcriteria:

 The studies compared disturbed systems with reference

systems(intimeorspace).

 Thestudiesclearlydefinedthenatureandthedegreeofthe

disturbance.

 Thestudiesreportedonspeciesrichness,species

composi-tion,IndexofBioticIntegrity(IBI)orEcologicalQualityRatio

(EQR).

Thesecriteriawereappliedwithsomeflexibility,inthatthe

definitions ofthe reference stateor the description of the

degreeofdisturbancemaysometimesdifferamongstudiesor

requiredsometranslation(explainedbelow)or,aspointedout,

thatsometimesderiveddatawereusedasaproxyforMSAif

Forlandusechangesinthecatchmentand

eutrophica-tion,weanalyzedtheeffectonthebiodiversityof(1)rivers

and streams, (2) wetlands and (3) lakes. Accordingly, we

analyzed the effect of hydrological disturbance on the

biodiversity of (4) rivers and streams, and (5) floodplain

wetlands.Table1showsthesearchtermsusedinthesearch

query.

Theeffectoflandusechangesonriversandstreams(Weijters

etal.,2009)wasbasedonstudiesonbiodiversityinriversand

streamswithdifferentcatchmentlanduse(forest,

agricultur-al,urban,etc.)and/ordifferentnutrientconcentrationsinthe

water. Aswater nutrient concentrations generally strongly

correlate with land use in the upstream catchment (e.g.

Harper,1992;Johnes et al.,1996; Crosbieand Chow-Fraser,

1999) these variables might be used as interchangeable

disturbanceindicators.Aslandusegradientsinriver

catch-mentsoftenparallelnaturalgradients(WatzinandMcIntosh,

1999; Allan, 2004), we only compared sites located in

comparableriversegmentswithsimilarstreamorder.

Forthe effectsofland usechanges on wetlandswe based

landuse intensity onland use fractions, nutrientlevels or

disturbance rank, whatever reported. The results were

categorized in the wetland classes defined in the GLWD

(LehnerandDo¨ll,2004).

Forlakes,the analysiswasmainly basedonphosphorus

and nitrogen concentrations as an indicator for land use

intensityinthecatchment,partlyfordatareasonsandalso

becauseeutrophicationbynutrientsisoftenthedirectdriver

foreffectsonbiodiversityinlakes.

We derived impacts of hydrological disturbance on rivers

comparing data on biota in rivers at different degrees of

regulation(mostlybydams,insomecasesbycanalizationor

waterabstraction)to thesituation beforetheimpact, orto

neighbouring unregulated river stretches. Not all studies

reportedthedegreeofflowdisturbanceinauniformway. If

reported, theAAPFD was used, inother casesanestimate

wasmade.

Inthestudyonhydrologicaldisturbanceonfloodplainwetlands

(Kuiperetal.,2014),theflowdisturbancehasbeenexpressed

inthecategorieslow,mediumorhigh.Thisstudycoversthe

GLWD wetlandtypes 4(floodplainmarshes) and 5(swamp

forests).

2.4. Dataprocessingandcombination

Biodiversity intactness (MSA) values were calculated or

estimated fromthe primary speciescomposition datathat

wereextractedfromthecasestudies(Eq.(2)).IBIscoreswere

translatedintoMSAaccordingtoEq.(3)andEQRvalueswere

setequaltoMSA.Fromeachstudy,allavailabledatapoints

wereusedanddepictedinthegraphs(seeSection3).

Forthemes1–4,aregressionanalysisbetweenthe

distur-bancevariables andtheMSAvalueswasdone,usingthepackage

stats version3.2.0available inRversion 3.1.1(R-Core-Team,

Table1–Searchtermsfortheliteraturesurveys.

Theme# Ecosystem Drivers Effectparameters Comparator

11 _{Stream Landuse Biodiversity Impactassessment}

River Landcover Speciescomposition Pristine

Catchment Community Reference

Watershed Speciesrichness

Urbanlanduse Bioticintegrity

Agriculturallanduse IBI

Anthropogenic disturbance

Macro-invertebrates

Humanimpact Fish

Deforestation Macrophytes Eutrophication Vegetation Phosphorus Phytoplankton Nitrogen Zooplankton Amphibia Reptiles Birds Mammals

2 Wetland Idemas1 Idemas1 Idemas1

3 Lake Idemas1 Idemas1 Idemas1

4 River Riverregulation/alteration Idemas1 Idemas1

Stream Flowregulation/alteration

Flowregime Alteredhydrology Alteredflooding Hydrologic(al)change/alteration/regime Flowdisturbance Floodpulse/regime Inundationperiod/frequency

52 _{Floodplainwetland Idemas4 Idemas1 Idemas1}

Riverinewetland Riparianwetland

1 _{Weijtersetal.(2009).} 2 _{Kuiperetal.(2014).}

2014). In accordance with the definition of MSA, the

regressionswereforcedto1atzeroorminimumvalue of

thedisturbancefactor.Inthemes1and2alinear

relation-shipbetweendisturbancelevelandMSA wasassumed.In

theme 3, logistic regressionshave been performed, based

on assumed non-linear (sigmoidal) relationships between

nutrientconcentrationsandecologicaleffects(e.g.Scheffer

et al., 1993). The regressions were done for shallow and

deep lakes separately. In theme 4 we used the log10 of

thedisturbancelevel,i.e.theAAPFD,asthisvariablevaries

between0 and+1.In theme 5data points were grouped

intothreelevelsofdisturbance.Weusedmeta-analysesto

calculate mean effect sizes using the package metafor in

R(fordetailsseeKuiperetal.,2014).Thedatasetsdidnot

allow to calculate possible interactions between the

different categories of drivers, which were therefore not

included.

2.5. Algalbloommodule

The algal bloom module calculates the probability of the

dominance of harmful algal blooms of cyanobacteria in

lakes. Several empirical models already exist that relate

cyanobacterial biomass to total phosphorus (TP) and total

nitrogen(TN)concentrationsandwatertemperature,

devel-opedby Smith (1985), Watson et al. (1997), Downing et al. (2001), Ha˚kansonet al.(2007) and Kostenet al.(2012). The

modelbyHa˚kansonetal.(2007)(seeTable2)(slightlymodified

by addinga cut-off at TPbelow 0.005mgL1)was used in

thisstudy,asitwasjudgedasthemostcomprehensiveand

easy-to-useonaglobalscale.

2.6. Implementationandapplication

The model chain has been implemented, parallel to the

GLOBIO-Terrestrialmodel,inDelphiinthesoftwarepackage

Arisflow to control a correct handling of input data and

calculations. Spatial data on land-use, water discharge,

nutrient concentrations and flow deviation (calculated as

explained in Section 2.1) were read from GIS files and

combined with the regressions from the meta-analyses,

applied for the water types present in each300_300 _pixel

accordingtotheGLWDmap.Astheimpactsofthedifferent

driversareassumedtobeindependent,theMSAvalueper

waterbodyhasbeenobtainedbymultiplyingthevaluesforthe

relevantdrivers.Thefinalindicator‘aquaticMSA’ perpixel

hasbeencalculatedbyarea-weightedaveragingoftheMSA

valuesforrivers,lakesandwetlandsasfarastheyoccurin

theparticularpixel.

Toillustrateitsapplicabilityatthegloballevel,themodel

chain including GLOBIO-Aquatic has been applied for the

2000situationandfor2050accordingtotheOECDbaseline

scenario(OECD,2012).Themaindifferencesbetweenthose

years area 40% global population increase (from6.5 to 9

billionpeople),a60%increaseoffoodproductionandenergy

demand(with80%fossilfuel),a4degreesincreaseofaverage

airtemperature,a50%increaseinhydropowercapacity, a

55% higher freshwater use and an increase in urban

(doubling) anddiffuse (20–50% higher) nutrient emissions

towater.

3.

Results

3.1. Biodiversityrelations 3.1.1. Landusechanges

Riversandstreams:Theliteraturesearchgaveabout240papers,

fromwhich only12 paperspresented dataapplicablefora

quantitative analysis of MSA, resulting in 18 relationships

(Weijtersetal.,2009).Themostcommonlyreportedgroups

were macro-invertebrates and fishes and all data were

basedonspatialcomparisonofsites.Inthemajorityofthe

cases,totaltaxonrichnessdecreasedwithincreasinghuman

land-use (urban oragricultural)inthe catchmentandwith

increasingnutrientconcentrations(Fig.2,Table2),butthere

werealsoquitesomeoppositeorindifferentexamplesandthe

variability betweenthe studieswas large(Fig.2).Subgroup

analysis revealed that fish tend tobe more sensitive than

macroinvertebrates(fordetailsseeWeijtersetal.,2009).

Wetlands: The search resulted in nearly 400 articles, of

which 35 reported on qualitative relations with species

richness,butonly12paperswithquantitativedataforMSA

calculation (24 relationships). All studies involved spatial

comparisonsorgradientstudies.Avarietyofbioticgroupswas

represented in the dataset: plants, mosses, fishes,

amphi-bians, macro-invertebrates, birds, mammals. The GWLD

classes 4(floodplainmarshes),5(swampforests),6(coastal

wetlands) and 9(intermittentor isolatedwetlands) arethe

wetlandtypesincluded.Theclasses7(brackishwetlands)and

8(bogs,fensandmires)wereunderrepresentedinthedataset.

Ourliteraturestudyrevealsthatspeciesrichnessinwetlands

is always negatively related to human land use in the

catchment (Fig. 3, Table 2) and in most cases positively

related toforest cover(data notshown). Thedataset was

toosmallforconclusionsonsubsetsorcofactors.

Lakes:Inthissurvey17paperswerefoundfromwhichMSA

valuescouldbederived,mostlyfromcomparisonsbetween

lakes,somefromtimeseries,andsomefromalreadycompiled

0.0 0.2 0.4 0.6 0.8 1.0 100 80 60 40 20 0 MSA

Non-natural land use (%)

Fig.2–MSAinriversandstreamsinrelationtolandusein

thecatchment(adaptedfromWeijtersetal.,2009),

includingtheregressionline(blackline;R2_=0.33),

confidenceinterval(greydashedline)andprediction

data.SomepapersgaveIBIorEQRvalues.Thedatacovereda

widerangeoftaxonomicgroups:algae,macrophytes,

macro-invertebrates, fishes and zooplankton. From the dataset,

significantnegativerelationships couldbederivedfor MSA

asafunctionofnutrientconcentrations.Logisticregressionon

the logarithmically transformed total phosphorus (TP)

con-centrations(afteradditionofaminimumvalueof0.001mgL1₎

wasused,asconcentrationsarezero-boundedandasigmoidal

response expected. The regression has been performed for

shallowanddeeplakesseparately;thelimithasbeenchosen

at an average depth of 3m, crudely based on frequent

dominanceofsubmergedmacrophytesandinaccordancewith

thetypologyoftheEuropeanWaterFrameworkDirective.The

regressionlinefordeeplakesisbelowtheoneforshallowlakes,

indicatingthattheoriginalbioticcommunityinthelattergroup

islessvulnerabletoeutrophicationthanthefirstgroup(Fig.4

and Table 2). This could partly be explained by stabilizing

feedback mechanisms of the submerged macrophytes that

oftendominateinnon-eutrophicshallowlakes(Schefferetal.,

1993). Analogous data for TN were much more scarce, in

conjunctionwiththegeneralnotionofPbeingconsideredas

the main limiting nutrientfor algal growth infreshwaters.

There are indications, however, of a negative effect of N

loadingonbiodiversityinsomeinstances,mainlyintropical

waters butalso insome temperatelakes. Weperformedan

analogouslogisticregressionontheNdata(Table2).Finally,

the two relations were combined by selecting the highest

value,inaccordancewiththelimitingnutrientconcept:

MSAnut¼MAX½MSAP;MSAN (4)

3.1.2. Hydrologicaldisturbance

Rivers:Thesearchqueryresultedin20studiesthatcontained

usablequantitativedata.Themostfrequentlystudiedgroups

were fishes and macro-invertebrates.The resultsgenerally

revealed a clear decline of MSA in response to the flow

deviations (Fig. 5). A linear regressionon the log10 ofthe

reportedorestimatedAAPFD(withasmallvalue(0.1)added)

hasbeenperformed,forcedto1atzerodeviation.Fromthe

graphitappearsthata(moderate)flowdeviationof1would

resultinanMSAvalueofabout0.6andaflowdeviationof3

inanMSAofabout0.4.Inthemodelapplication,theequation

hasbeencut-offataminimumMSAvalueof0.1forveryhigh

flowdeviations.

Floodplainwetlands:Fortheflooddependentwetlands,19

suitablepapers(outofaninitial686)werefound,fromwhich

29 data-sets could be extracted to calculate MSA. In the

majorityofthecasesdammingwasthemainimpactandplants

themajorbioticgroupdescribed(fordetailsseeKuiperetal.,

2014).Thecaseswithalow,mediumandhighflowdisturbance

had weighted average MSA values of 0.60, 0.53 and 0.46,

respectively (Fig. 6). This indicates that alreadya moderate

disturbancehasadrasticimpact.Toincorporatethisintothe

GLOBIO model,these valueswereusedtofit anasymptotic

exponentialrelationoftheformy=aexp(b/(x+c)),assigning

the threecategories an AAPFDvalue ofabout 0.3,1 and3,

respectively(Table2).Theresponsetoflowdisturbancewas

somewhat influenced byother factorslike land use in the

upstreamcatchment,biomeandtaxonomicalgroup,butthe

datasetwastoosmallforconclusionsonsubsets.

0.0 0.2 0.4 0.6 0.8 1.0 MSA 20 0

Land use intensity

40 60

in the catchment (%)

80 100

Fig.3–MSAinwetlandsinrelationtocatchmentlanduse

intensity,includingtheregressionline(blackline;

R2_{=0.23),confidenceinterval(greydashedline)and}

predictioninterval(greydottedline).

Fig.4–MSAindeeplakesandshallowlakesinrelationto

nutrientconcentrations;regressionlines(solidlines)and

95%confidenceintervals(dashedlines).

0.0 0.2 0.4 0.6 0.8 1.0 1 0.5 0 -0.5 -1 MSA

Flow deviation (log(AAPFD+0.1))

Fig.5–MSAinriversandstreamsinrelationtoflow

disturbance,includingtheregressionline(blackline;

R2_{=0.1),confidenceinterval(greydashedline)and}

3.1.3. Combination

Ageneralremarkisthatforallthemes,onlyalimitednumber

ofpaperspresentedthedatainsuchawaythatMSAvalues

couldbecalculated.ThesepapersarelistedinWeijtersetal.

(2009),Kuiperetal.(2014)andonthewebsitewww.globio.info.

Themajorityofthesepapers(over90%ofthepapersonland

usechangesandover80%oftheonesonhydrologicalchanges)

describedstudiesinthe‘developed’partoftheworld:North

America,EuropeandAustralia/NewZealand.

Thederivedrelationshipswerecombinedbymultiplying

the appropriate MSA factors per water type(Table 2). The

combined MSA value per pixel was calculated by

area-weightedaveragingofthesevalues.

3.2. OECDbaselinescenario

Asanillustrationofanapplicationofthemodelchainwith

globalscenariodata,theaverageaquaticMSAprojectedforthe

OECDbaseline is showngeographically forthe years2000

and2050(Fig.7a andb)aswell asthedifference (Fig.7c).

Pixelswithoutaquatic ecosystemsaccording tothe GLWD

(Lehner and Do¨ll, 2004) are shown in white on the map.

Accordingtothemodel,theaquaticbiodiversityintactness

in2000hasalreadydeclinedconsiderablyinmanypartsof

the world, especially in western, central and southern

Europe,theUSA/Mexico,southandeastAsia,thesouthern

SahelandpartsofSouthAfrica,ArgentinaandBrazil(Fig.7a).

Areas like northern Europe, Canada, Russia, Australia,

centralAfricaandlargepartsofSouthAmericahavemuch

less beenaffected. In general,the boreal biomehas been

affectedleastandthepopulatedtemperate,mediterranean

andsubtropicalbiomes most.Theworld averagedaquatic

MSA (the average for all pixels with water bodies) has

decreased toabout0.75; aboutthree-quartersofthedecline

can beattributedtoland-usechanges (Fig. 8). Asexpected,

thelargestimpactsappearinthoseworldregionsthatarethe

most densely populated and the most cultivated. Rivers

and floodplain wetlands are affected in some of the less

populatedcatchmentsaswellasaconsequenceofdamming.

Theoccurrenceofalgalbloomsgenerallycorrelatesnegatively

withtheMSAforlakes,ascanbeseenfromacomparisonof

bothmaps(Fig.9aandb),whichislogicalastheyarelargely

basedonthesamedriversinthemodel.

In the OECD baselinescenario, the MSA is expectedto

declinefurtherinthefuture(Fig.7bandc).Amajordeclineis

projectedforAfrica,inlinewithpredictedchangesinlanduse

inthisscenario.InAsia,LatinAmericaandEasternEurope

furtherdeclinesarealsoprojected.Amodestimprovementis

projectedinpartsoftheUSA,centralAsiaandEurope,dueto

an assumed stabilization of agricultural area and/or some

resultofeutrophicationabatement.AllprojectedMSAlosses

of wetlands and shallow lakes should be regarded as

minimumvalues,astheyarebasedontheareaspresented

in the GLWD (Lehner and Do¨ll, 2004); historical wetland

Table2–Summaryoftheempiricalrelationshipsinthe

model.

MSAcalculations Landuseandnutrients

Rivers fLU=10.0070xwithx=humanland

useincatchment(0–100)

Wetlands fLU=10.0081xwithx=landuse

intensityinthecatchment(0–100) Shallowlakes fP=exp(x)/(1+exp(x))with

x=2.089–1.048LN(TP+0.001); TPinmgPm3

fN=exp(y)/(1+exp(y))with

y=0.2640–0.9975LN(TN+0.01); TNinmgNm3

fnut=MAX(fP,fN)

Deeplakes fP=exp(x)/(1+exp(x))with

x=4.002–1.176LN(TP+0.001); TPinmgPm3

fN=exp(y)/(1+exp(y))with

y=0.145–4.768LN(TN+0.01); TNinmgNm3 fnut=MAX(fP,fN) Hydrologicaldisturbance Rivers fHy=0.3985x+0.60with x=10_{log(AAPFD+0.1)} Floodplain wetlands fHy=0.3519exp(0.5885/(x+1.5636))

withx=10_{log(AAPFD+0.1)}

Combination MSAperwater

type(‘wt’)

MSAwt=(fLUjfnut)fHy1

TotalMSAaqua MSAaqua=Sum

(AreawtMSAwt)/WaterArea

Cyanobacterialbiomass (Ha˚kansonetal.,2007)

Lakes B=0.001[5.8510_{log(1000MAX(TP,}

0.005))4.01]^4fTN/TPfT; IFTN/TP15:fTN/TP=1;ELSE: fTN/TP=13[(TN/TP)/151]; IFT>15:fT=0.86+0.63 ((T/15)^1.51);ELSE: fT=(1+1(T/15)^31));

TPandTNinmgm3_{;B,cyanobacteria(mgL}1_{);T,mediansurface}

watertemperatureingrowingseason(8C).

0.0 0.2 0.4 0.6 0.8 1.0 0.5 0 -0.5 -1 MSA

Flow deviation (log(AAPFD + 0.1))

Fig.6–MSAinfloodplainwetlandsinrelationtoflow

disturbanceforthreedifferentclassesofhydrological

alteration(MeaneffectW_{StandardError;highdisturbed,}

N=15;mediumdisturbed,N=10;weaklydisturbed,

N=4).SeeKuiperetal.(2014)foradetaileddescriptionof

Fig.7–MapsofthemeanfreshwaterMSAfor(a)2000and(b)2050(OECDbaselinescenario).(c)Differencebetween2000and

conversions are not accounted forin the calculationsand

futureconversionsonlyasaminimumestimate.

4.

Discussion

Thisstudyshowsthatbiodiversityintactnessinfreshwater

ecosystems, measured as MSA, is negatively related to

twodominantcategoriesofanthropogenicstressors,i.e.(1)

land-use and eutrophication in the catchment (affecting

‘water quality’) and (2) hydrological disturbance by dams

and/or climate change (affecting ‘water quantity’). This

conclusioncanbedrawnqualitativelyfromtheensembleof

casestudies, and underpinned quantitativelyby the

meta-analyses on the data of a subset of these papers. This

conclusion holds for the major types of inland aquatic

ecosystems:rivers,lakes andwetlands,while wetlandsare

alsodirectlyaffectedbyconversionanddrainage.Ingeneral,

standingwaterbodiesinheavilyusedcatchmentslooseabout

80%oftheiroriginalspeciescompositionandrunningwaters

about70%.Severehydrologicaldisturbancecausesadeclineof

60–80%oftheoriginalspeciescompositioninrunningwaters

andmorethan50%ofitinconnectedwetlands.

Thispatternwasderivedbyscalingupand combininga

number of local/regional case studies. The variation of

observedeffectsbetweenindividualcasesislarge–asmight

beexpectedbothfromthevariationinlocal(e.g.

morphologi-cal,geochemical)andregional(e.g.hydrological,

geomorpho-logicalandclimatic)featuresofthesites,andalsofromthe

‘composite’natureofthediscerneddrivers.Thedriver‘human

landuse’forinstanceismadeupofmanycompositefactors

(eutrophication,erosion/sedimentation,ripariansettlements

andothers)thataloneorincombinationaffectsbiota.Asin

practiceseveraloftheseunderlyingfactorswillbecorrelated,

we argue that this way of scaling up different cases is

acceptable for obtaining a broad picture. The cases were

selectedundertheconditionthattheyhadevaluatedsitesor

periodscomparablewithrespecttonaturalfactors suchas

climate,geomorphology, stream order,catchment sizeand

water chemistry. As for the hydrological disturbance, we

chosethedegreeofdeviationfromthenaturalseasonalflow

patternasthecrucialvariable(Poffetal.,1997).Thisdeviation

may have a different nature in different systems, e.g.

increased flow variation in naturally steady rivers, versus

decreasedvariationinnaturallydynamicones.

Thequantitativeresultsofthismodellingexerciseshould

beregardedasindicativeastheyarebasedononlyalimited

setofcasestudies.Manystudiescouldnotbeusedbecause

data were presented inadequately. In addition, there is a

substantial bias geographically towards case studies from

NorthAmerica,OceaniaandEurope,althoughtheabsenceof

areferencesituationwasoftenaprobleminthelatterregion.

The boreal and also some of the tropical regions were

underrepresentedbecausetheyare(untilnow)lessdisturbed,

andgenerallylessstudied.Anoverallconstraintisthatonly

(datafrom)peer-reviewedpaperswereincluded,whichalso

tendstooveremphasizestudiesfromthe developedpart of

the world. An extensionof the searchwith greyliterature

(combinedwithabasicdataqualitycheck)wouldbroadenthe

results.

Animportantcauseofthedatalimitationarisesfromthe

fact that primarydata onspeciescomposition, requiredto

compute theMSAindicator,haveoftennotbeenpublished.

Increasing journal facilitiesas wellasongoing projectsfor

internationaldatacompilationsuchastheGlobalBiodiversity

InformationFacilityGBIF(www.gbif.org)andBioFresh(www.

freshwaterbiodiversity.eu)willprobablyleadtoanincreasein

suitabledatainthenearfuture.Datanotusableasinputmight

be appropriate for validation purposes. An increase of the

numberofcaseswouldofcoursenotreducethetotalvariability

inthedataset,whichshouldbetakenasunavoidable.Still,it

wouldrevealpossibledifferencesinsensitivitybetweenbiomes

orecotypes,whichcouldthenbeusedtorefinethemodel.In

themeantime,themodelcouldservetofillthe‘datagaps’by

extrapolatingrelationsfromotherecotypes.

Despitetheselimitations,themodelhasshowntoproduce

plausibleresultsatthelevelwhereitwasmeantfor,i.e.the

impact ofbroadly-defined driversat the scale ofrelatively

largeregionsandcatchments.Itinformspolicymakersatthe

globallevelinwhatregionsaquaticbiodiversityisexpected

Fig.8–World-averagedaquaticMSAlossin2000and2050accordingtotheOECDbaselinescenarioandcontributionofthe

tobeaffectedmostandbywhatcauses.Scenariosonglobal

issues like population growth, food demand, agricultural

production,sanitationandwastewatertreatmentandenergy

mixcanbelinkedtotheecologicalintactnessofecosystems

inworldregions.Thismakesbiodiversity–onabroadscale–

‘modelable’(linkabletotheseglobalenvironmentaldrivers)

andprovidesoneofthetoolstoevaluatetheCBDbiodiversity

targets(CBD,2014).InthiswayitcomplementstheGLOBIO

sistermodelforterrestrialecosystems(Alkemadeetal.,2009).

TheGLOBIOapproachhascontributedtotheawarenessthat

the CBD biodiversity targets for 2010 were not met, by

performinga number ofglobalscenariostudies (TenBrink

etal.,2010;OECD,2012;PBL,2014).

TheGLOBIO-Terrestrialmodelcoversthedriversland-use

change,infrastructure/fragmentation, atmospheric nitrogen

deposition and climate change, by adding an important

categoryofecosystemsandspecificaquaticdrivers(suchas

the water–food–energy nexus). Both models reveal some

parallelresultsfortheeffectsofhumanland-use:thedensely

cultivatedregionsoftheworldcomeoutasthemostaffected,

but inthe aquatic modelthespatial patternin theMSAis

influencedby theconnectanceofpixelswithin river

catch-ments.Moreover,theeffectsofflowdisturbancearealsoseen

insomeless-populatedregions.TheterrestrialGLOBIOmodel

reportsanaverageworld-MSAaround0.7in2000and0.63in

2050.Ourmodelgivescomparablefigures(about0.75and0.7),

but these values are certainly an overestimation in many

regions,ashistoricalwetlandconversionwasnotaccounted

forand notalldriverswereincluded.Bothmodelslackthe

impactofexoticspeciesinvasions.Becauseoftherelatively

highernumberoflakesandwetlandsintheborealregions,

thisbiome(whichisingeneraltheleastpopulated)hasmore

influenceontheworld-averageforGLOBIO-Aquaticthanfor

GLOBIO-Terrestrial.

Many ofthe most-impacted world regions according to

GLOBIO-Aquaticalsoappearfromthemodelofglobalthreats

toriverbiodiversitybyVo¨ro¨smartyetal.(2010).Besides,these

authorsconclude thatinthedeveloping world,theregions

withahigh threattoriver biodiversityoftencoincide with

thosewherewateravailabilitytohumansisatrisk.

We considertheMSA ausefulindicatorfor thestateof

anecosystem,asitreflectstheintactnessofthenativespecies

compositionandallowscomparisonofdifferentsystemson

thesamescale.Itisalsoan‘objective’indicatorinthatituses

the same baseline for all ecosystems and regions, which

contributedtoitsacceptanceinthepolicyarena.However,

MSAisbynomeanstheonlyindicatorofbiodiversity.Other

indicatorslikespeciesrichness,Shannon–Wienerindexand

evennessprovideotherkindsofinformationandshowpartly

differentresponsestodisturbancethanMSA.Inmoderately

disturbedsituations,thedecreaseoforiginalspeciesisoften

accompaniedbytheappearanceof‘newcomers’(the

‘inter-mediatedisturbancehypothesis’),therebyincreasingspecies

richness. This is also in line with the unimodal

(‘hump-shaped’) speciesrichnesscurves oftenfoundin relationto

productivityandotherfactors(Leibold,1999;Declerck,2005).

Someofthenewcomersmaybeinvasivespecies.Thisisonly

reflectedintheMSAifthenewcomersleadtothedeclineof

nativespecies,whichisnotnecessarilythecase.

Defining the ‘undisturbed’ (or ‘pristine’) state of an

ecosystemwhencalculatingtheMSAisoftendifficult.Truly

pristineaquatic systemsarerare, butwe tookapragmatic

approach by following the definitions of ‘least disturbed

systems’ in the casedescriptions in the literatureused as

reference systems for the driver under concern. When

comparingthebiotaatnaturalanddisturbedsituations,we

implicitlyassumed thatthetimesincethedisturbancehad

beensufficientlylongtoobservethechangesinbiota.Inmany

cases,theecosystemmightstillbeinatransientstate,e.g.

speciesthatareabouttodisappear inthelong runarestill

presentinthefirstyears.Arelatedtopicisthatwewerenot

able to distinguish possible hysteresis effects between an

increasingdisturbanceandadecreasingone(restoration),as

has for instance been shown for eutrophication of lakes

(Schefferetal.,1993).

Ageneralconcern aboutscaling upspeciescomposition

dataisthattherelationshipswiththedriversmaydifferacross

scales. The correlation between a certain driver and local

diversitymaynotholdattheregionallevel.Weaccountedfor

thisbycomparingonlydatathatcoveredthesamescaleasfar

aspossible.Besides,thisproblemplayslessaroleinourstudy

becauseweusedanindicatorofintactness(ornaturalness)

basedonoriginalspeciesonly.Thisindicatorisprobablyless

sensitive for this scale issue than ‘absolute’ biodiversity

indices like species richness. Averages of indicator values

calculated per pixel over larger regions (multiple pixels)

shouldbeinterpretedwithcare.Thisholds forMSAvalues

asmuchasformore‘absolute’indicators.Forexample,ona

largerscale,localdecreaseorextinctionsofspeciesmightbe

compensated by increases elsewhere, and species may

migratewithintheregion.Theaveragedvaluewillbedifferent

fromavaluethatwouldhavebeencalculatedfortheregion.

Nevertheless, theaveragedvaluedoesgiveanindication of

how much ofthe (in this caseaquatic) ecosystemsin the

region lost their original species composition. It does not

indicate, however, the absolute number of species under

threatindifferentregions, astheMSAisscaledto1forall

studies,i.e.naturallyspecies-richandspecies-poorsystems

aretreatedinthesameway.Itispossible,however,toweigh

theMSAvaluesbythenaturalspeciesrichnessperbiome,as

hasbeendemonstratedfortheGLOBIO-Terrestrialmodel.

Ourapproachshouldhenceberegardedascomplementary

tootherapproachesandindicators.IndicatorsliketheLiving

PlanetIndex(LPI)(LohandWackernagel,2012)andtheRedList

Index(RLI)(www.iucnredlist.org)provideinformationonthe

(global)trendinselectedspeciesgroups,butaredifficulttolink

toenvironmentalmodels.Anotherapproacharetheecological

assessment methods derived at the European-scale (Moss

etal.,2003;Penningetal.,2008;VerdonschotandNijboer,2004

and others), which derived indicator species for certain

disturbance factors. Azevedo et al. (2013) related relative

speciesrichnessforseveralbioticgroupstototalphosphorus

concentrations in lakes and streams worldwide. Although

methodsand indicatorsdiffer, thesestudiesalsosupporta

decreaseofbiodiversityathighernutrientlevels.Importantis

alsotherelationbetweenthesestructuralandmorefunctional

indicators of ecological integrity, like food web structure,

boundaries for regimeshiftsof vulnerableecosystems and

deliveryofecosystemservices(Pereiraetal.,2013).However,

O¨ zkundakcietal.(2014)foundonlyweakrelationsbetween

different typesofindicatorsindeeplakes inNewZealand.

Furthermodeldevelopmentwillprobablyrequire

combina-tionsofdifferentmodellingtechniques(cf.Mooijetal.,2010);

besides meta-analyses ofcase studies these could include

process-basedmodelling(e.g.forfactorslikeexploitationand

forfunctionalindicators),orqualitativereasoningifdataare

notthere.Thesetopicswillbeaddressedinfutureversionsof

themodel.

Someother,more‘internal’,issuesneedtobeconsidered

whenevaluatingthemodel.Wehavetreatedtheimpactsof

thedifferentcategoriesofdriversasindependent(hencewe

multiplied the factors). We consider this a reasonable

assumption, as the drivers were treated separately in the

underlying case studies. Interactions between drivers can,

however,notbeexcluded,bothsynergisticallyand

antagonis-tically.Thereisaweakindicationofwetlandsin

intensively-usedcatchmentsbeingsomewhatlesssensitiveto

hydrologi-cal disturbance. A synergistic effect is sometimes found

between invasive species and other disturbances, but the

currentdatasetwastoolimitedtoshowthat.

Anotherissueiswhethertheeffectofadriverismodified

by other factors. We explicitly separated the regression

analyses per main water type (i.e. river, wetland, shallow

lake,anddeeplake)(seeTable2andFigs.2–6).Other‘effect

modifiers’wereinsomecasestaxonomicgroups;e.g.inrivers,

inwetlands,amphibianstendedtoberelativelysensitiveand

animalsgenerallymoresensitivethanplants.Inlinewiththis

is also our finding that shallow lakes, often dominated by

submergedmacrophytes,werelesssensitivethandeeplakes.

Onecouldspeculatethatthevulnerabilityofaquaticanimals

mighthavetodowiththecomplexlifecycleofmanyspecies,

orwithlimitedpossibilitiestoescapeunfavourablehabitats,

whiledispersalofplantsmightbemoreeasy.Theseaspectscan

befurtherinvestigatedwhenmoredatawillbecomeavailable.

Toconclude,wepresentedamodelapproachthatisable

to link aquatic biodiversity intactness to spatially explicit

modelsofglobalenvironmentaldrivers,andthatallowsfor

scenarioanalysestoinformpolicymakersatthegloballevelin

whichregionstheaquatic biodiversityisimpactedmostby

environmental pressures. Although still in development

andhamperedbydatadeficits,wefeelthattheapproachis

promisingandcansuccessfullybeimprovedwhendatawill

beincreasinglyavailable.

Acknowledgements

The authors thank G. van Drecht, A.F. Bouwman, A.H.M.

Beusen,H.BiemansandB.Feketeforthedisposalofinputdata

for the model application, S. Teurlincx and M. Verhofstad

for help with ‘R’, J. Meijer for his help on the GIS-based

implementationandmapdesign,himandM.deHeerforthe

designofthewebsitewww.globio.info,andtwoanonymous

reviewersfortheirvaluablecommentsonanearlierversionof

themanuscript.ThisstudyispartoftheGlobalBiodiversity

ProjectoftheNeth.EnvironmentalAssessmentAgency.This

isalsopublication5767oftheNeth.InstituteofEcology.

r

e

f

e

r

e

n

c

e

s

Alkemade,R.,VanOorschot,M.,Miles,L.,Nellemann,C.,

Bakkenes,M.,TenBrink,B.,2009.GLOBIO3:aframeworkto

investigateoptionsforreducingglobalterrestrial biodiversityloss.Ecosystems12,374–390.

Allan,J.D.,2004.Landscapesandriverscapes:theinfluenceof

landuseonstreamecosystems.Annu.Rev.Ecol.Evol.Syst. 35,257–284.

Azevedo,L.B.,Zelm,R.,Elshout,P.M.,Hendriks,A.J.,Leuven,

R.S.,Struijs,J.,deZwart,D.,Huijbregts,M.A.,2013.Species

richness–phosphorusrelationshipsforlakesandstreams worldwide.GlobalEcol.Biogeogr.22,1304–1314.

Biemans,H.,Haddeland,I.,Kabat,P.,Ludwig,F.,Hutjes,R.,

Heinke,J.,VonBloh,W.,Gerten,D.,2011.Impactof

reservoirsonriverdischargeandirrigationwatersupply duringthe20thcentury.WaterResour.Res.47,W03509.

Beusen,A.H.W.,2014.TransportofNutrientsfromLandtoSea;

GlobalModelingApproachesandUncertaintyAnalyses. UtrechtStudiesinEarthSciences,No.58..

Bouwman,A.F.,KleinGoldewijk,K.,VanDerHoek,K.W.,

Beusen,A.H.W.,VanVuuren,D.P.,Willems,J.,Rufino,M.C.,

Stehfest,E.,2011.Exploringglobalchangesinnitrogenand

phosphoruscyclesinagricultureinducedbylivestock

productionoverthe1900–2050period.Proc.Natl.Acad.Sci.

USA,http://dx.doi.org/10.1073/pnas.1012878108.

Brolsma,R.,SpernaWeiland,F.,VanBeek,R.,Winsemius,H.,

2012.Identificationofglobalwetlandareaswith

PCR-GLOBWB.Deltares,Delft.

Bunn,S.E.,Arthington,A.H.,2002.Basicprinciplesand

ecologicalconsequencesofalteredflowregimesforaquatic biodiversity.Environ.Manage.30,492–507.

CBD,2014.GlobalBiodiversityOutlook4.Secretariatofthe

ConventiononBiologicalDiversity.http://www.cbd.int/gbo4/

Cohen,A.S.,Bills,R.,Cocquyt,C.Z.,Caljon,A.,1993.Theimpact

ofsedimentpollutiononbiodiversityinLakeTanganyika. Conserv.Biol.7,667–677.

Crewe,T.L.,Timmermans,S.T.A.,2005.AssessingBiological

IntegrityofGreatLakesCoastalWetlandsusingMarshBird andAmphibianCommunities.Project#WETLAND3-EPA-01 TechnicalReport.BirdStudiesCanada,Canada.

Crosbie,B.,Chow-Fraser,P.,1999.Percentagelanduseinthe

watersheddeterminesthewaterandsedimentqualityof22 marshesintheGreatLakesbasin.Can.J.FishAquat.Sci.56, 1781–1791.

Davidson,N.C.,2014.Howmuchwetlandhastheworldlost?

Long-termandrecenttrendsinglobalwetlandarea. Mar. FreshwaterRes.65,934–941.

Declerck,S.E.A.,2005.Multi-groupbiodiversityinshallowlakes

alonggradientsofphosphorusandwaterplantcover. Ecology86,1905–1915.

Do¨ll,P.,Fiedler,K.,Zhang,J.,2009.Global-scaleanalysisofriver

flowalterationsduetowaterwithdrawalsandreservoirs. Hydrol.EarthSystemSci.Discuss.6,4773–4812.

Downing,J.A.,Watson,S.B.,McCauley,E.,2001.Predicting

cyanobacteriadominanceinlakes.Can.J.FishAquat.Sci.58, 1905–1908.

Dudgeon,D.,2010.Prospectsforsustainingfreshwater

biodiversityinthe21stcentury:linkingecosystemstructure andfunction.Curr.Opin.Environ.Sustain.2,422–430.

Dudgeon,D.,Arthington,A.H.,Gessner,M.O.,Kawabata,Z.-I.,

Knowler,D.J.,Le´veˆque,C.,Naiman,R.J.,Prieur-Richard,

A.-H.,Soto,D.,Stiassny,M.L.J.,Sullivan,C.A.,2006.Freshwater

biodiversity:importance,threats,statusandconservation challenges.Biol.Rev.81,163–182.

FAO,2012.TheStateofWorldFisheriesandAquaculture2012.

FAO,Rome.

Fekete,B.M.,Wisser,D.,Kroeze,C.,Mayorga,E.,Bouwman,L.,

Wollheim,W.M.,Vo¨ro¨smarty,C.,2010.Millennium

ecosystemassessmentscenariodrivers(1970–2050):climate andhydrologicalalterations.GlobalBiogeochem.Cycles24.

Ha˚kanson,L.,Bryhn,A.C.,Hytteborn,J.K.,2007.Ontheissueof

limitingnutrientandpredictionsofcyanobacteriainaquatic systems.Sci.TotalEnviron.379,89–108.

Harper,D.,1992.EutrophicationofFreshwaters;Principles,

ProblemsandRestoration.Chapman&Hall,London.

Jeppesen,E.,Kronvang,B.,Meerhoff,M.,Søndergaard,M.,

Hansen,K.M.,Andersen,H.E.,Lauridsen,T.L.,Beklioglu,M.,

Ozen,A.,Olesen,J.E.,2009.Climatechangeeffectson

runoff,catchmentphosphorusloadingandlakeecological state,andpotentialadaptations.J.Environ.Qual.38, 1930–1941.

Johnes,P.,Moss,B.,Phillips,G.,1996.Thedeterminationoftotal

nitrogenandtotalphosphorusconcentrationsin

freshwatersfromlanduse,stockheadageandpopulation data:testingofamodelforuseinconservationandwater qualitymanagement.FreshwaterBiol.36,451–473.

Karr,J.R.,1981.Assessmentofbioticintegrityusingfish

communities.Fisheries6,21–27.

Karr,J.R.,Chu,E.W.,2000.Sustaininglivingrivers.

Hydrobiologia422–423,1–14.

Keddy,P.A.,Fraser,L.H.,Solomeshch,A.I.,Junk,W.J.,Campbell,

D.R.,Arroyo,M.T.,Alho,C.J.,2009.Wetandwonderful:the

world’slargestwetlandsareconservationpriorities. BioScience59,39–51.

Kosten,S.,Huszar,V.L.M.,Becares,E.,Costa,L.S.,VanDonk,E.,

Hansson,L.A.,Jeppesen,E.,Kruk,C.,Lacerot,G.,Mazzeo,N.,

Scheffer,M.,2012.Warmerclimatesboostcyanobacterial dominanceinshallowlakes.GlobalChangeBiol.18,118–126.

Kourzeneva,E.,2010.Externaldataforlakeparameterizationin

numericalweatherpredictionandclimatemodeling.Boreal Environ.Res.15,165–177.

Kuiper,J.J.,Janse,J.H.,Teurlincx,S.,Verhoeven,J.T.A.,

Alkemade,R.,2014.Theimpactofriverregulationonthe

biodiversityintactnessoffloodplainwetlands.Wetlands Ecol.Manage.22,647–658.

Ladson,A.R.,White,L.J.,1999.AnIndexofStreamCondition:

ReferenceManual, 2nded.Dept.ofNat.Res.andEnv., Melbourne,Australia,p.130.

Lehner,B.,Do¨ll,P.,2004.Developmentandvalidationofaglobal

databaseoflakes,reservoirsandwetlands.J.Hydrol.296,1–22.

Lehner,B.,Liermann,C.R.,Revenga,C.,Vo¨ro¨smarty,C.,Fekete,

B.,Crouzet,P.,Do¨ll,P.,Endejan,M.,Frenken,K.,Magome,J.,

Nilsson,C.,Robertson,J.C.,Ro¨del,R.,Sindorf,N.,Wisser,D.,

2011.High-resolutionmappingoftheworld’sreservoirsand

damsforsustainableriver-flowmanagement.Front.Ecol. Environ.9,494–502.

Leibold,M.,1999.Biodiversityandnutrientenrichmentinpond

planktoncommunities.Evol.Ecol.Res.1,73–95.

Leprieur,F.,Brosse,S.,Garcia-Berthou,E.,Oberdorff,T.,Olden,

J.,Townsend,C.,2009.Scientificuncertaintyandthe

assessmentofrisksposedbynon-nativefreshwaterfishes. FishFish.10,88–97.

Loh,J.,Wackernagel,M.,2012.LivingPlanetReport.WWF,Gland.

MEA,2005a.EcosystemandHumanWell-being:Wetlandsand

WaterSynthesis.WorldResourcesInstitute,Washington, DC.

MEA,2005b.MillenniumEcosystemAssessment.Ecosystems

andHumanWell-being,StateandTrends,vol.1.Island Press,Washington,DC.

Mooij,W.M.,Hu¨lsmann,S.,DeSenerpontDomis,L.N.,Nolet,

B.A.,Bodelier,P.L.E.,Boers,P.C.M.,Pires,L.M.D.,Gons,H.J.,

Ibelings,B.W.,Noordhuis,R.,2005.Theimpactofclimate

changeonlakesintheNetherlands:areview.Aquat.Ecol. 39,381–400.

Mooij,W.M.,Trolle,D.,Jeppesen,E.,Arhonditsis,G.,

Belolipetsky,P.V.,Chitamwebwa,D.B.R.,Degermendzhy,

A.G.,DeAngelis,D.L.,DeSenerpontDomis,L.N.,Downing,

A.S.,Elliott,J.A.,FragosoJr.,C.R.,Gaedke,U.,Genova,S.N.,

Gulati,R.D.,Ha˚kanson,L.,Hamilton,D.P.,Hipsey,M.R.,t

Hoen,J.,Hu¨lsmann,S.,Los,F.H.,Makler-Pick,V.,Petzoldt,T.,

Prokopkin,I.G.,Rinke,K.,Schep,S.A.,Tominaga,K.,van

Dam,A.A.,vanNes,E.H.,Wells,S.A.,Janse,J.H.,2010.

Challengesandopportunitiesforintegratinglakeecosystem modellingapproaches.Aquat.Ecol.44,633–667.

Moss,B.,Hering,D.,Green,A.J.,Aidoud,A.,Becares,E.,

Beklioglu,M.,Bennion,H.,Boix,D.,Brucet,S.,Carvalho,L.,

2009.Climatechangeandthefutureoffreshwater

biodiversityinEurope:aprimerforpolicy-makers. FreshwaterRev.2,103–130.

Moss,B.,Kosten,S.,Meerhof,M.,Battarbee,R.,Jeppesen,E.,

Mazzeo,N.,Havens,K.,Lacerot,G.,Liu,Z.,DeMeester,L.,

2011.Alliedattack:climatechangeandeutrophication.

InlandWaters1,101–105.

Moss,B.,Stephen,D.,Alvarez,C.,Becares,E.,Bund,W.V.D.,

Collings,S.E.,Donk,E.V.,Eyto,E.D.,Feldmann,T.,

Ferna´ndez-Ala´ez,C.,Ferna´ndez-Ala´ez,M.,Franken,R.J.M.,

Garcı´a-Criado,F.,Gross,E.M.,Gyllstro¨m,M.,Hansson,L.-A.,

Irvine,K.,Ja¨rvalt,A.,Jensen,J.-P.,Jeppesen,E.,Kairesalo,T.,

Kornijo´w,R.,Krause,T.,Ku¨nnap,H.,Laas,A.,Lill,E.,Lorens,

B.,Luup,H.,RosaMiracle,M.,No˜ges,P.,No˜ges,T.,Nyka¨nen,

M.,Ott,I.,Peczula,W.,Peeters,E.T.H.M.,Phillips,G.,Romo,

S.,Russell,V.,Salujo˜e,J.,Scheffer,M.,Siewertsen,K.,Smal,

H.,Tesch,C.,Timm,H.,Tuvikene,L.,Tonno,I.,Virro,T.,

Vicente,E.,Wilson,D.,2003.Thedeterminationofecological

statusinshallowlakes—atestedsystem(ECOFRAME)for

implementationoftheEuropeanWaterFramework Directive.Aquat.Conserv.MarineFreshwaterEcosyst.13, 507–549.

Nilsson,C.,Berggren,K.,2000.Alterationsofriparian

ecosystemscausedbyriverregulation.Bioscience50, 783–792.

OECD,2012.OECDEnvironmentalOutlookto2050.

http://www.oecd-ilibrary.org/environment/oecd-environmental-outlook_1999155x.

O¨ zkundakci,D.,Hamilton,D.P.,Kelly,D.,Schallenberg,M.,de

Winton,M.,Verburg,P.,Trolle,D.,2014.Ecologicalintegrity

ofdeeplakesinNewZealandacrossanthropogenicpressure gradients.Ecol.Indic.37,45–57.

Palmer,M.A.,ReidyLiermann,C.A.,Nilsson,C.,Flo¨rke,M.,

Alcamo,J.,Lake,P.S.,Bond,N.,2008.Climatechangeandthe

world’sriverbasins:anticipatingmanagementoptions. Front.Ecol.Environ.6,81–89.

Parsons,M.,Thoms,M.,Norris,R.,2002.Australianriver

assessmentsystem:AusRivASphysicalassessmentprotocol. MonitoringRiverHealthInitiativeTechnicalReport22.

PBL,2008.Ligtvoet,W.(Ed.),Kwaliteitvoorlater:Exante

evaluatieKaderrichtlijnWater.(ExanteevaluationWater FrameworkDirective).PlanbureauvoordeLeefomgeving, Bilthoven,PBLpublicatie(nr.50014001/2008).ISBN: 9789069602035(inDutch),p.212.

PBL,2014.Howsectorscancontributetosustainableuseand

conservationofbiodiversity.CBDTechnicalSeriesNo.79,

http://www.cbd.int/doc/publications/cbd-ts-79-en.pdf.

Penning,W.E.,Mjelde,M.,Dudley,B.,Hellsten,S.,Hanganu,J.,

Kolada,A.,VandenBerg,M.,Poikane,S.,Phillips,G.,Willby,

N.,Ecke,F.,2008.Classifyingaquaticmacrophytesas

indicatorsofeutrophicationinEuropeanlakes.Aquat.Ecol. 42,237–251.

Pereira,H.M.,Cooper,H.D.,2006.Towardstheglobalmonitoring

ofbiodiversitychange.Trend.Ecol.Evolut.21,123–129.

Pereira,H.M., etal.,2013.Essentialbiodiversityvariables.

Science339,277–278.

Poff,N.,Allan,J.,Bain,M.,Karr,J.,Prestegaard,K.,Richter,B.,

Sparks,R.,Stromberg,J.,1997.Thenaturalflowregime:a

paradigmforriverconservationandrestoration.Bioscience 47,769–784.

Poff,N.L.,Zimmerman,J.K.H.,2010.Ecologicalresponsesto

alteredflowregimes:aliteraturereviewtoinformthe scienceandmanagementofenvironmentalflows. FreshwaterBiol.55,194–205.

R-Core-Team,2014.R:ALanguageandEnvironmentfor

StatisticalComputing.RFoundationforStatistical Computing,Vienna,Austria.

Revenga,C.,Campbell,I.,Abell,R.,deVilliers,P.,Bryer,M.,2005.

Prospectsformonitoringfreshwaterecosystemstowardsthe 2010targets.Philos.Trans.RoyalSoc.B:Biol.Sci.360,397– 413.

Rosenberg,D.M.,McCully,P.,Pringle,C.M.,2000.Global-scale

environmentaleffectsofhydrologicalalterations: introduction.Bioscience50,746–751.

Sala,O.,ChapinIII,F.,Armesto,J.,Berlow,E.,Bloomfield,J.,Irzo,

R.,Huber-Samwald,E.,Huenneke,K.,Jackson,R.,Kinzig,A.,

Leemans,R.,Lodge,D.,Mooney,H.,Oesterheld,M.,Poff,N.,

Sykes,M.,Walker,B.,Walker,M.,Wall,D.,2000.Global

biodiversityscenariosfortheyear2100.Science287, 1770–1774.

Scheffer,M.,Hosper,S.,Meijer,M.,Moss,B.,Jeppesen,E.,1993.

Alternativeequilibriainshallowlakes.TrendsEcol.Evolut.8, 275–279.

Scholes,R.J.,Biggs,R.,2005.Abiodiversityintactnessindex.

Nature434,45–49.

Seitzinger,S.P.,Mayorga,E.,Bouwman,A.F.,Kroeze,C.,Beusen,

A.H.W.,Billen,G.,VanDrecht,G.,Dumont,E.,Fekete,B.M.,