Scenarios for Global Biodiversityin the 21st CenturyHenrique M. Pereira,1*† Paul W. Leadley,2* Vânia Proença,1 Rob Alkemade,3

Jörn P. W. Scharlemann,4 Juan F. Fernandez-Manjarrés,2 Miguel B. Araújo,5,6

Patricia Balvanera,7 Reinette Biggs,8 William W. L. Cheung,9 Louise Chini,10 H. David Cooper,11

Eric L. Gilman,12 Sylvie Guénette,13 George C. Hurtt,10,14 Henry P. Huntington,15

Georgina M. Mace,16 Thierry Oberdorff,17 Carmen Revenga,18 Patrícia Rodrigues,1

Robert J. Scholes,19 Ussif Rashid Sumaila,13 Matt Walpole4

Quantitative scenarios are coming of age as a tool for evaluating the impact of futuresocioeconomic development pathways on biodiversity and ecosystem services. We analyze globalterrestrial, freshwater, and marine biodiversity scenarios using a range of measures includingextinctions, changes in species abundance, habitat loss, and distribution shifts, as well ascomparing model projections to observations. Scenarios consistently indicate that biodiversity willcontinue to decline over the 21st century. However, the range of projected changes is muchbroader than most studies suggest, partly because there are major opportunities to intervenethrough better policies, but also because of large uncertainties in projections.

Quantitative estimates of the future trajecto-ries of biodiversity,whichwe broadly referto as biodiversity scenarios, are typicallybased on the coupling of several complex

components (Fig. 1). Socioeconomic scenarioswithtrajectories of key indirect drivers of ecologicalchange, such as human population growth andgreenhouse gas emissions, are developed underdifferent assumptions regarding society’s devel-opment, often associated with “storylines” (1).These trajectories are then fed into models thatproject changes in direct drivers of ecosystemchange, such as climate and land-use change, indifferent regions of the world (1, 2). Finally, theprojected drivers are used as inputs to biodiversitymodels (Table 1). In some cases, associated changesin key ecosystem services are also modeled, al-though quantifying the link between biodiversityand ecosystem services remains a major scien-tific challenge (3, 4). Here, we review recentmodel-based biodiversity scenarios, which havegrown rapidly in number over the last few yearsowing to major advances in modeling and dataavailability.

Biodiversity change has many metrics (5).Here we group these metrics into four classes:species extinctions, species abundance and com-munity structure, habitat loss and degradation,and shifts in the distribution of species and biomes.Scenarios of species extinction risk (6, 7) address

REVIEW

1Centro de Biologia Ambiental, Faculdade de Ciências da Uni-versidade de Lisboa, 1749-016 Lisboa, Portugal. 2LaboratoireÉcologie, Systématique et Évolution, UMR 8079 CNRS, UniversitéParis-Sud 11, F-91405Orsay, France. 3Netherlands EnvironmentalAssessment Agency (PBL), PostOffice Box 303, 3720AHBilthoven,Netherlands. 4United Nations Environment Programme WorldConservation Monitoring Centre, 219 Huntingdon Road, Cam-bridge CB3 0DL, UK. 5Departamento de Biodiversidad y BiologíaEvolutiva,MuseoNacional de Ciencias Naturales, Consejo Superiorde Investigaciones Científicas, calle Jose Gutierrez Abascal, 28006Madrid, Spain. 6Cátedra Rui Nabeiro–Biodiversidade, Centro deInvestigação em Biodiversidade e Recursos Genéticos, Universi-dade de Évora, Largo dos Colegiais, 7000 Évora, Portugal. 7Centrode Investigaciones enEcosistemas,UniversidadNacionalAutónomade México, Apartado Postal 27-3, Sta. Maria de Guido, C.P.58090,Morelia,Michoacán,México. 8StockholmResilience Center,Stockholm University, Stockholm, Sweden 10691. 9School ofEnvironmental Sciences, University of East Anglia, Norwich NR47TJ, UK. 10Department of Geography, University of Maryland,College Park, MD 20742, USA. 11Secretariat of the Convention onBiological Diversity, World Trade Center, 413 St. Jacques Street,Suite 800, Montreal, Quebec, Canada H2Y IN9. 12College ofNatural and Computational Sciences, Hawaii Pacific University,Honolulu, HI 96822, USA. 13Fisheries Centre, Aquatic EcosystemsResearch Laboratory (AERL), 2202 Main Mall, The University ofBritish Columbia, Vancouver, BC, Canada V6T 1Z4. 14PacificNorthwest National Laboratory, Joint Global Change ResearchInstitute, College Park, MD 20740, USA. 15The Pew Environ-ment Group, 23834 The Clearing Drive, Eagle River, AK 99577,USA. 16Centre for Population Biology, Imperial College London,Silwood Park, Ascot, SL5 7PY, UK. 17Laboratoire de Biologie desOrganismes et Ecosystèmes Aquatiques, UMR IRD 207MuséumNational d’Histoire Naturelle, 43 rue Cuvier, 75005 Paris, France.18The Nature Conservancy, 4245 North Fairfax Drive, Arlington,VA 22203, USA. 19CSIR Natural Resources and Environment, PostOffice Box 395, Pretoria 0001, South Africa.

*These authors contributed equally to this work.†To whom correspondence should be addressed. E-mail:hpereira@fc.ul.pt

impacts on biodiversity impacts on ecosystem

services

socioeconomic development pathways

Population growth, fossil fuel use, food demand, etc.

e.g., IPCC SRES scenarios, MA scenarios, GEO4 scenarios

direct drivers

Climate change, land-use change, water extraction,fish harvesting pressure

e.g., Global Climate Models (GCM), IMAGE

Habitat orfunctional group-level changes

Species-levelchanges

e.g., dynamicvegetation models,

marine trophicmodels

Provisioning, regulating,supporting and cultural

services

e.g., marine trophic model(food provisioning),

dynamic vegetation models(carbon sequestration)

e.g., niche models,species-area

curves, empiricaldose-responserelationships

Fig. 1. Overview of methods and models commonly used for constructing biodiversity scenarios. Somemodels include several components of this figure, such as the integrated assessment model IMAGE (1) orthe marine trophic model “Ecosim with Ecopath” (23). Black arrows indicate key linkages treated inbiodiversity scenarios. Dashed gray arrows indicate linkages that are absent in current biodiversityscenarios. In some cases, impacts on ecosystem services may be mediated by changes in the abioticcondition of ecosystems (thin arrow from direct drivers to ecosystem services).

10 DECEMBER 2010 VOL 330 SCIENCE www.sciencemag.org1496

on

Janu

ary

27, 2

011

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fr

om

the irreversible component of biodiversity change,but species extinctions have weak links to eco-system services and respond less rapidly to globalchange than do other metrics (e.g., the range of aspecies can decline shortly after habitat change,but that species may not become extinct as a re-sult) (8).More-responsivemetrics include changesin species abundances and community structureand, at a higher organizational level, habitat lossor biome changes. At both the species and eco-system levels, many of the projected changes canbest be described as shifts in potential distribution,with their current favorable conditions vanishingin some places, which may cause local extinc-tions, and appearing in new places, which mayresult in colonizations.

Models used to estimate global change im-pacts on biodiversity vary substantially in com-plexity and underlying hypotheses (Fig. 1 andTable 1), but can be broadly classified into phe-nomenological or process-basedmodels. Phenom-enological models rely on empirical relationshipsbetween environmental variables and a biodiversitymetric (9). One of the simplest phenomenologicalmodels is to subtract future land-cover changesfrom a species’ current distribution to estimateextinction risk (6). Species-area models use theempirical relationship between area and species

number to estimate species committed to extinc-tion after habitat loss (10). Niche-based models(or bioclimatic envelope models) use statisticalrelationships between current species distribu-tions and environmental variables, such as tem-perature and precipitation, to project the futuredistribution of a species under climate change(11). Dose-response relationships depend on ex-perimental or observational data to estimate theimpacts of drivers on biodiversity, e.g., the effectof nitrogen deposition on mean species abun-dance (12, 13). Process-based models simulateprocesses such as population growth or mecha-nisms such as ecophysiological responses (14).Dynamic global vegetation models (DGVMs),which play an important role in many scenarios,are complex ecosystem models integrating pro-cesses such as photosynthesis, respiration, plantcompetition for resources, and biogeochemicalcycles (15). Marine trophic models simulate thebiomass dynamics at different levels of the troph-ic web using mass-balance equations and canbe used to assess the population impacts of har-vesting (16).

Here we review global-scale biodiversity sce-narios for each of the four biodiversity metricsoutlined above. We analyze sources of variationwithin and between scenarios, coherence between

models andobservations, links betweenbiodiversityand ecosystem services, and relevance of scenar-ios to policy.

Species ExtinctionsScenarios for terrestrial ecosystems project thatfuture species extinction rates will greatly surpassbackground rates estimated from the cenozoicfossil record (17) and could exceed recent rates ofextinction by more than two orders of magnitude(Fig. 2 and table S1). There is great variation inprojected future extinction rates both within andbetween studies, with three factors explainingmuch of this variation. First, the degree of land-use and climate change explains a substantial frac-tion of the range of projected extinctions withinstudies [e.g., projected vertebrate extinctions are11 to 34% for 0.8° to 1.7°C global warming ver-sus 33 to 58% for >2.0°C warming in (7)], indi-cating that limitation of land-use change, especiallyin tropical and subtropical regions, and aggressiveclimate mitigation could substantially reduce ex-tinction risks. Second, an important contributionto the broad range of projections within studies isa lack of understanding of species ecology, espe-cially migration rates [e.g., the highest projectedextinction rates are 38% with unlimited migra-tions rates versus 58% with no migration in (7)]

Table 1. Examples of biodiversity scenario studies highlightingmethods usedto calculate impacts of global change on several biodiversity metrics.Socioeconomic scenarios: Millennium Ecosystem Asessment (MA), GlobalBiodiversity Outlook 2 (GBO2), Global Environmental Outlook 4 (GEO4), IPCCSpecial Report on Emission Scenarios (IPCC SRES), International Assessment

for Agricultural Science, Technology and Development (IAASTD). Directdrivers: land-use change (LUC), climate change (CC), nitrogen deposition (N),water use, and fishing effort. Projections of direct drivers: indicatesmodel thatwas used to simulate future changes in direct drivers (GCM, General CirculationModel, with specific climate model indicated in parentheses).

StudySocioeconomic

scenariosDirectdrivers

Projections ofdirect drivers

Projections of impacts onbiodiversity

Metrics of biodiversity andecosystem services Year

Terrestrial(38) MA LUC, CC IMAGE Species-area relationships Species extinctions (plants)

and habitat loss2100

(7) IPCC SRES andothers

CC GCM (HadCM2) Niche-based models. Range changesconverted to extinction risk usingspecies-area curves or IUCN status

Species extinctions (plantsand animals)

2050

(6) MA LUC, CC IMAGE Habitat loss from currentspecies ranges

Species extinctions (birds) 2100

(12) GBO2 LUC, CC, N IMAGE Dose-responsemodel (GLOBIO)

Species abundance changes 2050

(15) IPCC SRES CC GCM (HadCM3) Dynamic global vegetation models Functional group range shifts(plants) and carbonsequestration

2100

Freshwater(22) MA Water use and CC Water-GAP Phenomenological model relating

river discharge to fish speciesrichness

Species extinctions (fishes) 2100

Marine(23) GEO4, IAASTD Fishing effort Ecosim Marine trophic model

(Ecosim with Ecopath).Functional group abundancechanges and fish landings

2050

(43) IPCC SRES CC GCMs (HadCM3, PCM) Phenomenological model relatingsea surface temperature tobleaching frequencies

Habitat loss of tropical corals 2100

(52) IPCC SRES CC GCMs (GFDL CM 2.1) Niche-based models. Species range shifts(vertebrate andinvertebrates)

2050

www.sciencemag.org SCIENCE VOL 330 10 DECEMBER 2010 1497

REVIEW

on

Janu

ary

27, 2

011

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fr

om

and habitat specificity [e.g., in (18), the highestextinction rates are 7% for broad habitat specific-ity versus 43% for narrow habitat specificity],emphasizing the need for research on these fun-damental aspects of species ecology and theirincorporation into global models (14). Third, alarge fraction of variation between studies appearsto arise from differences between modeling ap-proaches [in particular, compare the two studiesof global bird extinctions (6, 19)]. The few rig-orous intermodel comparisons currently availablehave also found large differences in modelsensitivity (20, 21).

Quantitative scenarios of global extinctionsfor freshwater and marine organisms are rare.One model for freshwater ecosystems, based onthe relationship between fish diversity and riverdischarge, projects 4 to 22% (quartile range) fishextinctions by 2070 in about 30% of the world’srivers, because of reductions in river dischargefrom climate change and increasing water with-drawals (22). Models of global change impactson marine organisms have focused on local ex-tinctions, shifts in species distributions, and changesin abundance (23). The limited amount of extinc-tion scenarios for aquatic ecosystems suggeststhat quantitative data for global extinctions mod-els are still lacking (24), emphasizing the need forimproved monitoring of marine and freshwaterorganisms.

Projections of species extinction rates arecontroversial because of methodological chal-lenges and because of the lack of agreement withextinction patterns in the recent and distant past(25, 26). First, models project the fraction of spe-cies “committed to extinction,” primarily result-ing from decreases in range size, habitat area or,for freshwater taxa, river flow. However, the lagtime between being “committed to extinction” andactually going extinct may range from decades tomany millennia (13, 25), so future research mustfocus on quantifying these time lags as they con-stitute windows of opportunity for restorationefforts to prevent future extinctions. Evidencefrom recent and historical land-use change andthe paleontological record suggests that manyspecies can persist for long periods, by exploitingsecondary habitats or by surviving in small pop-ulations (25, 27). This suggests that the realizedextinction rates are likely to be lower and perhapsmuch lower than the “committed to extinction”rates shown in Fig. 2 and used in other compar-isons of past and future extinction rates (28, 29).Second, complex interactions between globalchange factors are not accounted for in models,and these interactions could decrease or increasefuture extinction risks (25). These considerationsand the range of projected terrestrial extinctionsin Fig. 2 reflect general scientific agreement thatuncurtailed rates of climate and land-use changewill increase extinction risks but that the mag-nitude of these risks is still uncertain.

Field and laboratory experiments mimickingreductions in species and functional group diver-sity have shown that species loss at local scales

can have negative impacts on ecosystem servicessuch as primary productivity, nutrient cycling, andinvasion resistance (3). Extinctions of species thatplay dominant roles in ecosystem functioning,such as large predators and pollinators, could beextremely detrimental for ecosystem services(30). However, it has proved difficult to scale thesestudies up to regional or global scales.

Species Abundances and Community StructureModels project declines in the population abun-dances of species in both marine and terrestrialsystems (12, 23). Global scenarios for marine

fisheries are based on a marine trophic model,Ecosim with Ecopath, which tracks functionalgroups of species, including multiple groups ofprimary producers, invertebrates, and fish species(16). Model parameters are estimated from his-toric trends of biomass and catches. Future eco-system dynamics are projected by optimizingfishing effort for a set of criteria, including profit,number of jobs, and ecosystem structure, with theweight given to each criterion depending on thescenario (23). The scenarios explored suggest that

future increases in landings, partially driven byfisheries subsidies, can only be achieved by in-tensifying pressure on groups that are not cur-rently fished in large quantities, often at lowertrophic levels, leading to a decline in the marinetrophic index (23, 31). In contrast, reductions infishing effort and destructive fishing practicessuch as trawling would allow rebuilding of anumber of major stocks (31, 32).

For terrestrial systems, the GLOBIO modeluses dose-response relationships to estimatechanges in mean species abundance as a functionof land-use change and other drivers (12). For

instance, the model uses a matrix of changes inmean local species abundance after conversionbetween two land-use categories, derived fromempirical studies. Scenarios developed usingGLOBIO project a decline of 9 to 17% in meanspecies abundance by 2050 relative to 2000(33, 34). The most favorable scenarios involve adoubling of protected areas to 20% of total landarea or focusing on sustainability at all levels,with limited human population and consump-tion growth. The GLOBIO model has also been

0.01

0.1

1

10

100

1000

10,000

100,000

Climate changeLand-use changeCombined drivers

0.0001

0.001

0.01

0.1

1

10

63

99.99

Fossilrecord

20thcentury

21st century

E/MSYExtinctionsper century(%)

Fig. 2. Comparison of recent and distant past extinction rates with rates at which species are “committedto extinction” during the 21st century (63). E/MSY is number of extinctions per million species years; “Fossilrecord” refers to the extinction rate of mammals in the fossil record (17); “20th century” refers todocumented extinctions in the 20th century—mammals (upper bound), birds, and amphibians (lowerbound) (17); “21st century” refers to projections of species committed to extinction according to differentglobal scenarios: vascular plants (38, 18), plants and animals (7), birds (6, 19), and lizards (64). Extinctionrate caused by each driver and total extinction rates are discriminated, when possible.

10 DECEMBER 2010 VOL 330 SCIENCE www.sciencemag.org1498

REVIEW

on

Janu

ary

27, 2

011

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fr

om

used to hindcast changes in mean species abun-dance from 1970 to 2000 (35). The modeled de-cline of 6% over this period is much smaller thanthe decline of 21% recorded by the Living PlanetIndex through direct observations of terrestrialspecies abundance (5). However, large differ-ences in how these two indicators are calculatedand potential biases in data in the Living PlanetIndex (36) and in the database of GLOBIO makedirect comparison impossible and underscore theneed to harmonize model and data indices (35).

Trade-offs between provisioning services andregulating services are apparent in both marineand terrestrial biodiversity scenarios and can be aconsequence of changes in community structure.For instance, increases in fish provisioning areachieved at the cost of changes in the food webstructure with potential impacts on the regulationof trophic cascades (37), and often at the cost ofsacrificing the long-term sustainability of the ser-vice (32). Similarly, inMediterranean ecosystems,modification of forest composition to favor rapid-growth species may lead to decreased fire resil-ience (35).

Habitat Loss and DegradationHabitat loss and degradation in terrestrial eco-systems cover a wide range of alteration of natu-ral and seminatural ecosystems by human activities.

Arguably, the conversion of forest to agriculturalsystems has been the most important of thesehabitat changes. Inmost land-use scenarios, glob-al forest area declines slightly over the next fewdecades (Fig. 3), resulting from extensive de-forestation in tropical forests and subtropical wood-lands, which is partially offset by increased forestcover in the Northern Hemisphere (33, 34, 38).Therefore, in terms of impacts on biodiversity,the overall picture is worse than the global forestprojections indicate, as the habitat losses in thetropics cannot be directly compensated by foresthabitat gains in temperate regions, and some ofthe forest gains in both regions are due to theexpansion of species-poor plantations.

A striking characteristic of some studies is thatlarge within-study divergences in socioeconomicdevelopment pathways lead to relatively smalldifferences in the projected global forest cover(Fig. 3), as well as other measures of global bio-diversity change (12). Themost recent studies, i.e.,the Intergovernmental Panel on Climate ChangeRepresentative Concentration Pathways (IPCCRCP) (2) scenarios and Wise et al. (39), includemore favorable trajectories, suggesting that theopportunities for habitat recovery may have beenpreviously underestimated. In particular, Wise et al.(39) foresee large increases in global forest coverif global carbon taxes were to include all sources

and sinks of carbon, thereby fa-voring protection of forests andimproved agricultural efficiency.However, they also project mas-sive deforestation if carbon taxeswere to focus on fossil fuels only,stimulating massive dependenceon bioenergy. This study and oth-er recent global scenarios (40)emphasize the positive or nega-tive impact that climate mitiga-tion could have on biodiversitydepending on how it is imple-mented. The ongoing land-useharmonization activity for IPCCAssessment Report 5, which con-nects land-use historical data to-gether with future scenario datafrom multiple Integrated Assess-ment Models into a single con-sistent, spatially gridded set ofland-use change scenarios, willopen new opportunities for ex-ploring the impacts of a broadrange of possible land-use tra-jectories on biodiversity and forenhancing the collaboration be-tween the climate, socioeconomic,and biodiversity communities(41).

Climate change is projectedto cause major changes in ma-rine habitats, through increasedwater temperature, ocean acidifi-cation, and expansion of oxygenminimum zones (42). Tropical

corals are vulnerable to climate change becauseincreases in sea surface temperature of 1°C formore than 8 weeks can lead to severe coral bleach-ing, with the breakdown of the endosymbiosis be-tween corals and zooxanthellae [(43); but see (44)].Phenomenological models applied to climatechange projections foresee that severe tropicalcoral bleachingmay occur on average every 2 yearsby 2050 (43). In addition, ocean acidification re-duces the availability of carbonate for calcifica-tion, slowing the growth of corals, and alongwithbleaching and other stressors, is projected to leadto widespread degradation of coral reefs and theecosystem services they provide such as fish-eries, storm surge protection, and income fromtourism (45).

In freshwater ecosystems, modeling of habitatdegradation has focused on the abiotic compo-nents of ecosystems, such as river discharge andnutrient loads, and how those changes will direct-ly affect ecosystem services such as water pro-visioning and regulation of water quality (1, 46).In some cases, the same drivers affecting eco-system services may also affect biodiversity. Forexample, scenario studies project increased wateruse as a consequence of population growth andrising water demands by agriculture and industry(33), and increased eutrophication due to agricul-ture and urban pollution (46). This will lead to

2000 2010 2020 2030 2040 2050

Year

-60

-40

-20

0

20

40

Per

cen

t ch

ang

e

MiniCAM B – 450 UCT MiniCAM C – 450 FFICT

GBO2 – sustainable meat production GBO2 – limiting climate change

MA – technogarden MA – order from strength

GEO4 – security first GEO4 – policy first

RCP 4.5 RCP 6.0 RCP 3-PD RCP 8.5

Envelope of IMAGE projections

Envelope of MiniCAM projections

GBO2

MA

GEO4

RCP

Fig. 3. Change in the extent of forests to 2050 in different global scenarios (63): MA scenarios (1), GBO2 scenarios (34),GEO4 scenarios (33), MiniCAM scenarios (39), and RCP scenarios for IPCC-AR5 (41). For each study, trajectories of the twomost contrasting scenarios are shown. By 2050, the envelope of scenarios with the IMAGE model (MA, GBO2, GEO4) isnarrower than the envelope of scenarios based on the MiniCAM model.

www.sciencemag.org SCIENCE VOL 330 10 DECEMBER 2010 1499

REVIEW

on

Janu

ary

27, 2

011

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fr

om

both water provisioning shortages and declines inbiodiversity (13).

Shifts in the Distribution of Species and BiomesDGVMs project large shifts in the distribution ofterrestrial biomes, with required velocities to ac-commodate temperature change reaching morethan 1 km year−1 in some biomes (47, 48). Theseshifts are expected to cause the rearrangement ofecosystems, including the creation of novel com-munities (49). For instance, the northern limit ofboreal forests is projected to move further northinto the arctic tundra, whereas the southern limitwill experience dieback, giving way to temperateconifer and mixed forest (15, 47). DGVM pro-jections are in qualitative agreement with the pa-leontological record, which indicates that climatechange has resulted in large shifts in the distri-bution of vegetation types in the past. However,there is uncertainty in the extent of biome changessimulated by DGVMs, even in analyses using com-mon climate scenarios, with some global vegetationmodels projecting modest shiftsand little vegetation dieback andothers projecting large-scale bi-ome shifts overmuch of the globeduring the 21st century, under-scoring the pressing need to bench-mark models against data (15).

DGVMs provide a powerfulmeans to explore the relation-ship between ecosystem servicesand shifts in the distribution ofbiomes or functional groups ofplants because vegetation typeplays a dominant role in control-ling terrestrial provisioning ser-vices, as well as supporting andregulating services. Recent sim-ulations with DGVMs suggestthat the Amazon forest may reacha tipping point due to a combi-nation of deforestation, climatechange, and fire, leading to drierconditions and an irreversible shiftto savannah-like vegetation (50).If pushed beyond this point, theAmazonian forest could releaselarge quantities of carbon into theatmosphere and modify rainfallpatterns over large areas of Southand southernNorthAmerica (50).

Bioclimatic models of theranges of marine organisms alsosuggest poleward shifts becauseof climate change (47). Averagespeeds for demersal speciesmayexceed 4 km year−1 in certain re-gions (Fig. 4A), consistent withrecent trends observed in theNorth Sea for widespread thermalspecialists (51). Projected shiftsfor pelagic species are foreseen tobe more rapid than for demersalspecies (Fig. 4B), owing to the

higher motility of pelagic species and largerchanges in ocean conditions in the surface layer.Furthermore, rates of shift can be more than dou-ble in a high-range climate change scenario (A1B)compared to a low-range scenario (committedclimate change experiment) (52), suggesting thatlimiting greenhouse gas emissionswill allowmoretime for species to adapt. The capacity of fresh-water species to move polewards in response toclimate change will be more limited owing to thelinear nature of many freshwater ecosystems. Thisproblem will be particularly acute in river basinswith an east-west configuration (35). Species mayalso respond to warming by migration to higherelevations in terrestrial systems (19) and greaterdepths in marine systems (53).

Given the rapidly growing use of bioclimaticmodels for decision support, such as studies ofthe impacts of climate change on future costs andefficiency of networks of protected areas (54) anddevelopment of adaptive forestry managementschemes (55), it is important that model projec-

tions are accurate. Bioclimatic models can, insome cases, predict the direction of range contrac-tions or expansions (56) and population increasesor declines (57) observed in the last few decades.However, insufficient treatment of key mecha-nisms, such as migration, biotic interactions, andinteractions between drivers such as climate andland use, still limits the accuracy of future rangeprojections from bioclimatic models (14).

Challenges in Improving Biodiversity ScenariosReducing uncertainty within and among modelprojections is urgent.More attentionmust be paidto evaluating model projections with indicatorsthat allow comparisons between models and be-tween models and data. Key components of thiseffort will be the development of comprehensivebiodiversity monitoring through efforts such asthe Global Biodiversity Observation Network orGEO BON (58), and the harmonization of thebiodiversity indicators used by the data and sce-narios communities.

A

B

Poleward shift(km per year)

< -0.5-0.5 – 0.5> 0.5 – 1> 1– 2> 2 – 3> 3 – 4> 4

Poleward shift(km per year)

< -0.5-0.5 – 0.5> 0.5 – 1> 1– 2> 2 – 4> 4 – 6> 6

Fig. 4. Projected rate of range shifts in marine organisms caused by climate change from 2005 to 2050 (52, 63). (A)Latitudinal shift of demersal species (excluding areas >2000m in depth because of undersampling of the deep-sea region). (B)Latitudinal shift of pelagic species. The projections are based on bioclimatic envelope models for 1066 species of fish andinvertebrates, under IPCC SRES A1B. For eachmap cell, themean shift of the range centroids of the species currently present inthat cell is given.

10 DECEMBER 2010 VOL 330 SCIENCE www.sciencemag.org1500

REVIEW

on

Janu

ary

27, 2

011

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fr

om

The importance of the drivers of biodiversitychange differs across realms, with land-use changebeing a dominant driver in terrestrial systems, andoverexploitation in marine systems, while climatechange is ubiquitous across realms (28). Availablemodels reflect these differences, but fail to accountfor the full set of major drivers of future bio-diversity change—for instance, the lack of globalmodels of the impact of dams and pollution onfreshwater biodiversity. Modeling climate changeimpacts on biodiversity is currently tractable andpopular, in part because a wide range of climatescenarios and bioclimatic envelopemodeling toolsare readily available, but it is vital to develop mod-els of other important drivers and their interactions(59). This will require the development of mech-anistic models linking changes in land use, pollu-tion levels, and biotic competition (e.g., invasivespecies) to population dynamics of individual spe-cies through changes in life-history parameterssuch as survival and dispersal, using techniquesthat are scalable across space and across speciesassemblages (10, 14). This approach has recentlybeen explored to assess how interactions betweenlife history and disturbance regime mediate spe-cies extinction risk under climate change (60).

To better inform policy, scenarios must movebeyond illustrating the potential impacts of globalchange on biodiversity toward more integratedapproaches that account for the feedbacks thatlink environmental drivers, biodiversity, ecosystemservices, and socioeconomic dynamics. Currentglobal biodiversity models rarely relate estimatesof biodiversity loss to ecosystem services, infre-quently explore policy options [but see (12, 23)],and do not account for the feedbacks from changesin biodiversity and ecosystem services to societalresponse (Fig. 1, dashed arrows). Introducing com-plex feedbacks to biodiversity scenarios will re-quire moving away from the relatively linear,noninteractive relationships between the socialand natural sciences (Fig. 1, thick downward-pointing arrows) toward a more interactive, inter-disciplinary association (61).

The likely imminent launch of the Inter-governmental Science-Policy Platform on Bio-diversity and Ecosystem Services (IPBES) opensan opportunity to develop a major effort to im-prove and evaluate biodiversity scenarios. Im-proved biodiversity models will strengthen therole of scenarios in testing policies to minimizethe impacts of human activities on biodiversityand maximize the ecosystem services providedby biodiversity. Hence, scenarios should play alarge role in IPBES and in helping to achieve thetargets set in the new strategic plan of the Con-vention on Biological Diversity (62).

References and Notes1. J. Alcamo, D. van Vuuren, C. Ringer, in Ecosystems and

Human Well-Being: Scenarios, S. R. Carpenter,L. P. Prabhu, E. M. Bennet, M. B. Zurek, Eds.(Island Press, Washington, DC, 2005), pp. 147–172.

2. R. H. Moss et al., Nature 463, 747 (2010).3. P. Balvanera et al., Ecol. Lett. 9, 1146 (2006).4. S. Díaz, J. Fargione, F. S. Chapin III, D. Tilman, PLoS Biol.

4, e277 (2006).5. S. H. M. Butchart et al., Science 328, 1164 (2010).6. W. Jetz, D. S. Wilcove, A. P. Dobson, PLoS Biol. 5,

e157 (2007).7. C. D. Thomas et al., Nature 427, 145 (2004).8. A. Balmford, R. E. Green, M. Jenkins, Trends Ecol. Evol.

18, 326 (2003).9. P. Kareiva et al., in Ecosystems and Human Well-Being:

Scenarios, R. J. Scholes, R. Hassan, Eds. (Island Press,Washington, DC, 2005), pp. 73–115.

10. H. M. Pereira, G. C. Daily, Ecology 87, 1877 (2006).11. R. K. Heikkinen et al., Prog. Phys. Geogr. 30, 751 (2006).12. R. Alkemade et al., Ecosystems (N.Y.) 12, 374 (2009).13. O. E. Sala et al., in Ecosystems and Human Well-Being:

Scenarios, S. Carpenter, L. Prabhu, E. Bennet, M. Zurek,Eds. (Island Press, Washington, DC, 2005), pp. 375–408.

14. W. Thuiller et al., Perspect. Plant Ecol. Evol. Syst. 9, 137(2008).

15. S. Sitch et al., Glob. Change Biol. 14, 2015 (2008).16. V. Christensen, C. J. Walters, Ecol. Modell. 172, 109

(2004).17. G. M. Mace, H. Masundire, J. E. M. Baillie, in Ecosystems

and Human-Well Being: Current State and Trends,B. Scholes, R. Hassan, Eds. (Island Press, Washington, DC,2005), pp. 77–122.

18. J. R. Malcolm, C. Liu, R. P. Neilson, L. Hansen, L. Hannah,Conserv. Biol. 20, 538 (2006).

19. C. H. Sekercioglu, S. H. Schneider, J. P. Fay, S. R. Loarie,Conserv. Biol. 22, 140 (2008).

20. R. G. Pearson et al., J. Biogeogr. 33, 1704 (2006).21. X. Morin, W. Thuiller, Ecology 90, 1301 (2009).22. M. A. Xenopoulos et al., Glob. Change Biol. 11, 1557 (2005).23. J. Alder, S. Guénette, J. Beblow, W. Cheung,

V. Christensen, Ecosystem-Based Global Fishing PolicyScenarios (Fisheries Centre Research Reports 15(7),University of British Columbia, 2007).

24. N. K. Dulvy, Y. Sadovy, J. Reynolds, Fish Fish. 4, 25(2003).

25. N. E. Stork et al., Conserv. Biol. 23, 1438 (2009).26. D. B. Botkin et al., Bioscience 57, 227 (2007).27. K. J. Willis, S. A. Bhagwat, Science 326, 806 (2009).28. A. Duraiappah et al., Ecosystems and Human Well-Being:

Biodiversity Synthesis (World Resources Institute,Washington, DC, 2005).

29. S. L. Pimm, G. J. Russell, J. L. Gittleman, T. M. Brooks,Science 269, 347 (1995).

30. M. A. Aizen, L. A. Garibaldi, S. A. Cunningham, A. M. Klein,Ann. Bot. (London) 103, 1579 (2009).

31. D. Pauly et al., Science 302, 1359 (2003).32. B. Worm et al., Science 325, 578 (2009).33. UNEP, Global Environment Outlook 4 (United Nations

Environment Programme, Nairobi, Kenya, 2007).34. B. ten Brink et al., Cross-Roads of Planet Earth´s Life.

Exploring Means to Meet the 2010-Biodiversity Target(MNP report 55050001/2006, NetherlandsEnvironmental Agency, Bilthoven, 2006).

35. P. W. Leadley et al., Biodiversity Scenarios: Projections of21st Century Change in Biodiversity and AssociatedEcosystem Services (Secretariat of the Convention onBiological Diversity, Montreal, 2010).

36. H. M. Pereira, H. David Cooper, Trends Ecol. Evol. 21,123 (2006).

37. M. Casini et al., Proc. Natl. Acad. Sci. U.S.A. 106, 197(2009).

38. D. van Vuuren, O. Sala, H. M. Pereira, Ecol. Soc. 11, 25(2006).

39. M. Wise et al., Science 324, 1183 (2009).40. A. M. Thomson et al., Proc. Natl. Acad. Sci. U.S.A. (2010).41. G. C. Hurtt et al., Harmonization of global land-use

scenarios for the period 1500-2100 for the 5th IPCCAssessment. iLEAPS Newsletter 7, June 2009.

42. O. Hoegh-Guldberg, J. F. Bruno, Science 328, 1523(2010).

43. S. D. Donner, W. J. Skirving, C. M. Little, M. Oppenheimer,O. Hoegh-Guldberg, Glob. Change Biol. 11, 2251 (2005).

44. J. Maynard, A. Baird, M. Pratchett, Coral Reefs 27, 745(2008).

45. O. Hoegh-Guldberg et al., Science 318, 1737 (2007).46. A. F. Bouwman, G. Van Drecht, J. M. Knoop,

A. H. W. Beusen, C. R. Meinardi, Global Biogeochem.Cycles 19, GB1002 (2005).

47. A. Fischlin et al., in Climate Change 2007: Impacts,Adaptation and Vulnerability. Contribution of WorkingGroup II to the Fourth Assessment Report of the IPCC,M. Parry, O. Canziani, J. Palutikof, P. van der Linden,C. Hanson, Eds. (Cambridge Univ. Press, Cambridge, 2007),pp. 211–272.

48. S. R. Loarie et al., Nature 462, 1052 (2009).49. J. W. Williams, S. T. Jackson, J. E. Kutzbach, Proc. Natl.

Acad. Sci. U.S.A. 104, 5738 (2007).50. D. C. Nepstad, C. M. Stickler, B. S. Filho, F. Merry,

Philos. Trans. R. Soc. B 363, 1737 (2008).51. A. L. Perry, P. J. Low, J. R. Ellis, J. D. Reynolds, Science

308, 1912 (2005).52. W. W. L. Cheung et al., Fish Fish. 10, 235 (2009).53. N. K. Dulvy et al., J. Appl. Ecol. 45, 1029 (2008).54. L. Hannah et al., Front. Ecol. Environ 5, 131 (2007).55. M. S. Mbogga, X. Wang, A. Hamann, J. Appl. Ecol. 47,

731 (2010).56. M. B. Araújo, R. G. Pearson, W. Thuiller, M. Erhard,

Glob. Change Biol. 11, 1504 (2005).57. R. E. Green et al., Biol. Lett. 4, 599 (2008).58. R. J. Scholes et al., Science 321, 1044 (2008).59. B. W. Brook, N. S. Sodhi, C. J. A. Bradshaw, Trends Ecol.

Evol. 23, 453 (2008).60. D. A. Keith et al., Biol. Lett. 4, 560 (2008).61. S. R. Carpenter et al., Proc. Natl. Acad. Sci. U.S.A. 106,

1305 (2009).62. UNEP/CBD/SP/PREP/2: Revision and Updating of the CBD

Strategic Plan: Possible Outline and Elements of theNew Strategic Plan (2009); www.cbd.int/sp/sp2010p.

63. Materials and methods are available as supportingmaterial on Science online.

64. B. Sinervo et al., Science 328, 894 (2010).65. This work was developed in the context of a scenarios

synthesis coordinated by DIVERSITAS and UNEP WorldConservation Monitoring Centre for the Secretariat of theConvention on Biological Diversity, with financial supportfrom the Department of the Environment, Food and RuralAffairs of the UK, the European Commission, and UNEP. Wethank L. Simpson for organizing a workshop. H.M.P. wassupported in part by grant PTDC/AMB/73901/2006 fromFundação para a Ciência e Tecnologia.

Supporting Online Materialwww.sciencemag.org/cgi/content/full/science.1196624/DC1Materials and MethodsTable S1References and Notes

19 August 2010; accepted 13 October 2010Published online 26 October 2010;10.1126/science.1196624

www.sciencemag.org SCIENCE VOL 330 10 DECEMBER 2010 1501

REVIEW

on

Janu

ary

27, 2

011

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fr

om

Corrected 7 December 2010; see below

www.sciencemag.org/cgi/content/full/science.1196624/DC1

Supporting Online Material for

Scenarios for Global Biodiversity in the 21st Century

Henrique M. Pereira,* Paul W. Leadley, Vânia Proença, Rob Alkemade, Jörn P. W. Scharlemann, Juan F. Fernandez-Manjarrés, Miguel B. Araújo, Patricia Balvanera,

Reinette Biggs, William W. L. Cheung, Louise Chini, H. David Cooper, Eric L. Gilman, Sylvie Guénette, George C. Hurtt, Henry P. Huntington, Georgina M. Mace, Thierry

Oberdorff, Carmen Revenga, Patrícia Rodrigues, Robert J. Scholes, Ussif Rashid Sumaila, Matt Walpole

*To whom correspondence should be addressed. E-mail: hpereira@fc.ul.pt

Published 26 October 2010 on Science Express

DOI: 10.1126/science.1196624

This PDF file includes:

Materials and Methods Table S1 References and Notes

Correction: Corrections have been made in the Materials and Methods (“Calculation of species extinction rates in Figure 2”) and in table S1.

Supplementary Online Materials

Materials and Methods

Calculation of species extinction rates in Figure 2

Rates of species extinction per million species years (E/MSY) were either directly obtained in the cited articles or calculated using the data on percentage of species committed to extinction provided in each study (Table S1). To calculate species extinction rates we assumed an exponential distribution for the species lifetimes (S1). Therefore, for a given time interval T corresponding to the scenario duration, the proportion of species going extinct (P) is:

!

P = 1" e"# $T (1)

where ! is the extinction rate in extinctions per million species years (E/MSY). Therefore we estimated ! as

!

" =ln 1# P( )#T (2)

where T is given in million years. Normalized percentages of committed extinctions per century are given in the right y-axis of Figure 2, for a period of 100 years (T = 100 ! 10-6). These are equal to the percentage of committed extinctions reported for the papers where the scenario period was 100 years. Note that there is a non-linear relationship between the logarithmic scale in the left y-axis and the scale in the right y-axis. For low extinction rates the scale on the right y-axis is approximately logarithmic but it deviates from a logarithmic scale as values come closer to 100%. The lower bound and upper bound of the projections of species committed to extinction in each study are given in Table S1 and Figure 2.

Calculation of global forest projections in Figure 3

Percent changes in forest extent were assessed from 2000 to 2050. At each point in time, t, forest percent change was calculated using:

!

% change =Forest extent t

Forest extent2000" 1

#

$ %

&

' ( ) 100 (3)

For each set of scenarios, the trajectories of the two scenarios producing the most contrasting results within the time interval of 2000 – 2050 are shown. The RCP scenarios data comes from an exercise carried out for the IPCC-AR5, by four Integrated Assessment Modeling teams: RCP8.5-Message, RCP6.0-AIM, RCP4.5-GCAM, and RCP3PD-IMAGE. The land-use data from the RCP scenarios were harmonized into a consistent dataset using the methods described in S2. Data for the MiniCAM scenarios comes from S3. The MA, GBO2, and GEO4 data come from the IMAGE team, and the scenarios have been described in S4, S5, and S6. For the MA, GBO2 and GEO 4 (all based on the IMAGE model), forest includes mature forest, regrowth forest and wood plantations (for timber or carbon sequestration), but

excludes woody biofuel crops. The MiniCAM scenarios use somewhat different land cover categories, and forest includes managed and unmanaged forests.

Calculation of marine organisms shifts in Figure 4

Data presented in figures 4A and 4B were based on projected range shift of 1066 species of exploited fishes (836 spp.) and invertebrates (230 spp.) by 2050 relative to 2005 (S7). These included all species that were reported at the species level in the fisheries statistics of the Food and Agriculture Organization of the United Nations (FAO). For each species, distribution of relative abundance on 30’ x 30’ grid for each year from 2001 to 2055 was predicted using a dynamic bioclimatic envelope model (see S7). The model was driven by scenario of ocean conditions (the Special Report on Emissions Scenarios A1B scenario) projected by the ocean-atmosphere coupled climate model developed by the Geophysical Fluid Dynamics Laboratory of the U.S. National Oceanic and Atmospheric Administration (GFDL’s CM 2.1) (S8). Distribution centroid (C) of each species was calculated from:

(4)

where Ai and Lati is the predicted relative abundance and mean latitude of each 30’ x 30’ cell i and N is total number of cells with A > 0. The average rate of shift in distribution centroid was then calculated from the mean annual change in 10-year average latitudinal centroid between 2005 (average of 2001 to 2010) and 2050 (average of 2046 to 2055). Finally, for each species, the calculated rate of range shift was assigned to cells (on a 30’ x 30’ grid) with positive occurrence in the predicted current distribution of the species. The average rate of range shift in each cell over all the demersal (857 spp.) and pelagic (209 spp.) species considered here was then calculated and presented in figures 4A and 4B.

Table S1. Studies and data used to assess extinction rates in E/MSY. Values in bold represent the data available in each study. The lower bound (LB) and upper bond (UB) projections of species extinction in each study are given. CD – combined drivers, LUC – land use change, CC – climate change.

Taxa Scenarios or data sources Species extinct

Total species

Percentage extinct (CD)

Percentage extinct (LUC)

Percentage extinct (CC)

Time Interval

E/MSY (CD)

E/MSY (LUC)

E/MSY (CC) Study

Lower estimation (LB) - - - - - - 0.21 - - Mammals – Fossil record Upper estimation (UB) - - - - - - 0.46 - -

(S9)

Amphibians (LB) 23 5743 0.40% - - 1900-2000 40 - - Birds 45 9917 0.45% - - 1900-2000 45 - -

Vertebrates – 20th century

Mammals (UB) 31 4853 0.64% - - 1900-2000 64 - -

(S9)

Technogarden (LB) - - 13% 10% 3% 1995-2100 1326 1003 290 Plants

Order from Strength (UB) - - 23% 17% 6% 1995-2100 2489 1775 589 (S10)

BIOME3/Perfect Migration/Broad Biome Definition/Broad Biome Specificity/EAR (LB)

231 133149 0.17% - - One

century - - 17 Plants

MAPSS/Zero Migration/Narrow Biome Definition/Narrow Biome Specificity/SAR (UB)

56606 133149 42.5% - - One

century - - 5534

(S11, S12)

Dispersal with minimum climate change (LB) - - 9% - - 2000-2050 - - 1886 Plants and animals No dispersal with maximum climate change (UB) - - 52% - - 2000-2050 - - 14679

(S13)

Technogarden (LB) 51 8750 0.58% 0.26% 0.32% 1985-2100 51 28 23 Birds

Order from Strength (UB) 80 8750 0.91% 0.15% 0.76% 1985-2100 79 67 13 (S14)

Adaptive Mosaic/ 1.1º/ 0% move up (LB) 75 8459 0.88% - - 2000-2100 89 - - Birds Order from Strength/ 6.4º/ 100% move up (UB) 2498 8459 29.5% - - 2000-2100 3500 - -

(S15)

Lizards Global extinction derived from projections of local extinctions

- - 20% - - 1975-2080 - - 2125 (S16)

References and notes

S1. A. R. Solow, W. Smith, Paleobiology 23, 271-277 (1997).

S2. G. C. Hurtt et al., iLEAPS Newsletter, 6-8 (2009).

S3. M. Wise et al., Science 324, 1183-1186 (2009).

S4. J. Alcamo, D. van Vuuren, W. Cramer, in Ecosystems and Human Well-Being: Scenarios, Millennium Ecosystem Assessment. S. R. Carpenter, L. P. Prabhu, E. M. Bennet, M. B. Zurek, Eds. (Island Press, Washington, 2005), vol. 2, pp. 297-373.

S5. B. ten Brink et al., Cross-roads of Planet Earth´s Life. Exploring means to meet the 2010-biodiversity target. (Netherlands Environmental Agency, 2006).

S6. UNEP, Global Environment Outlook 4 (UNEP, 2007; http://www.unep.org/geo/geo4).

S7. W. Cheung et al., Fish Fish. 10, 235-251 (2009).

S8. T. L. Delworth et al., J. Clim. 19, 643-674 (2006).

S9. G. M. Mace, H. Masundire, J. E. M. Baillie, in Ecosystems and Human-Well Being: Current State and Trends, Millennium Ecosystem Assessment. B. Scholes, R. Hassan, Eds. (Island Press, Washington, 2005), vol. 1, pp. 77-122.

S10. D. van Vuuren, O. Sala, H. M. Pereira, Ecol. Soc. 11, 25 (2006).

S11. Projections of species extinction were derived under a scenario of doubling of CO2 concentration, which may occur, according to some climate scenario projections, over a 100-year period (S12, S17).

S12. J. R. Malcolm, C. Liu, R. P. Neilson, L. Hansen, L. Hannah, Conserv. Biol. 20, 538-548 (2006).

S13. C. D. Thomas et al., Nature 427, 145-148 (2004).

S14. W. Jetz, D. S. Wilcove, A. P. Dobson, PLoS Biology 5 (2007). S15. C. H. Sekercioglu, S. H. Schneider, J. P. Fay, S. R. Loarie, Conserv. Biol. 22, 140-150

(2008). S16. B. Sinervo et al., Science 328, 894 (2010).

S17. IPCC. Climate change 2001: the scientific basis. Contributions of Working Group I to the Third Assessment Report of the IPCC, J. T. Houghton et al., Eds. (Cambridge University Press, Cambridge, 2007), p. 14.