THE YEAR IN ECOLOGY AND CONSERVATION BIOLOGY, 2009

Setting Conservation PrioritiesKerrie A. Wilson, Josie Carwardine, and Hugh P. Possingham

The University of Queensland, School of Integrative Biology, St. Lucia,Queensland, Australia

A generic framework for setting conservation priorities based on the principles of clas-sic decision theory is provided. This framework encapsulates the key elements of anyproblem, including the objective, the constraints, and knowledge of the system. Withinthe context of this framework the broad array of approaches for setting conservationpriorities are reviewed. While some approaches prioritize assets or locations for con-servation investment, it is concluded here that prioritization is incomplete withoutconsideration of the conservation actions required to conserve the assets at particularlocations. The challenges associated with prioritizing investments through time in theface of threats (and also spatially and temporally heterogeneous costs) can be aided byproper problem definition. Using the authors’ general framework for setting conserva-tion priorities, multiple criteria can be rationally integrated and where, how, and whento invest conservation resources can be scheduled. Trade-offs are unavoidable in prior-ity setting when there are multiple considerations, and budgets are almost always finite.The authors discuss how trade-offs, risks, uncertainty, feedbacks, and learning can beexplicitly evaluated within their generic framework for setting conservation priorities.Finally, they suggest ways that current priority-setting approaches may be improved.

Key words: conservation prioritization; systematic conservation planning; reserve de-sign; decision theory; biodiversity; surrogates; costs; threats; vulnerability; likelihoodof success; threatened species; hotspot; focal species; uncertainty; risk; feedbacks

Introduction

Conservation activities must be prioritizedso that scarce funds and resources are used ef-ficiently and effectively to prevent long-termloss and degradation of biodiversity and eco-logical systems. A plethora of impacts arecausing a mass extinction event (Lawton andMay 1995), including habitat destruction andfragmentation, overexploitation of natural re-sources, invasive species, global climate change,and emerging diseases (Groombridge and Jenk-ins 2002). While we can attempt to increase theresources available for conservation, at presentfunding is insufficient in the context of cur-rent threats, and conservation competes with

Address for correspondence: Kerrie A. Wilson, The University ofQueensland, School of Biological Sciences, St. Lucia Brisbane, Queens-land 4072, Australia. k.wilson2@uq.edu.au

other societal priorities, such as food produc-tion, human habitation, and resource extrac-tion (Abbitt et al. 2000; James et al. 2001; Balm-ford et al. 2004; Naidoo and Iwamura 2007).

Land-use allocation patterns around theworld are not favorable for biodiversity con-servation. Almost 500 million hectares of theEarth’s surface is allocated to agriculture:arable and permanent cropland is estimated tocover approximately 1500 million hectares andpermanent pasture approximately 3400 millionhectares (FAO 2008). By comparison, terres-trial protected areas cover only 1840 millionhectares (WDPA Consortium 2004). Globalspending on conservation is also far lower thanin other sectors. For example, in 2007 the to-tal annual revenue of the world’s largest non-government conservation organization, TheNature Conservancy (www.nature.org), wasone tenth of the net profit of Wal-Mart. Ata local level, expenditure on environmental

The Year in Ecology and Conservation Biology, 2009: Ann. N.Y. Acad. Sci. 1162: 237–264 (2009).doi: 10.1111/j.1749-6632.2009.04149.x C© 2009 New York Academy of Sciences.

237

238 Annals of the New York Academy of Sciences

protection is often much lower than expendi-ture in other sectors. In Australia, for example,local governments receive over AUS$2.6 bil-lion in revenue for environment protection ac-tivities, which equates to only 13% of their totalrevenue, and only a fraction of that is spent onbiodiversity conservation (Australian Bureau ofStatistics 2004).

People have long set conservation priorities,with subsistence communities declaring partic-ular areas or species exempt from exploitation(Wright and Mattson 1996; Chandrashekaraand Sankar 1998). Today the scientific litera-ture documents a range of priority-setting ap-proaches (Smith and Theberge 1987; Costelloand Polasky 2004; Brooks et al. 2006; Sarkaret al. 2006). In its various forms, conservationpriority setting seeks to identify where, how, onwhat, and/or when we should act first to con-serve biodiversity efficiently, on the assumptionwe cannot do everything everywhere at once.Much of the conservation priority–setting liter-ature concerns the identification of new pro-tected areas or networks of protected areas,and this is often referred to as systematic conser-vation planning (Margules and Sarkar 2007). Inthis review, we explore a broad array of priority-setting approaches that concern not only thelocations we wish to conserve, but also the bio-logical assets we are concerned about, and theconservation actions available to increase theirchance of persistence. There are also a wide va-riety of tools to assist the identification of con-servation priorities, covering a broad spectrumfrom mathematical to intuitive.

The first aim of this chapter is to pro-vide a generic framework for setting conser-vation priorities. It is based on the principlesof classic decision theory (Clemen 1996; Elster2003). While we describe the process of set-ting conservation priorities mathematically, theterminology we have employed is flexible andthe framework reflects logical decision mak-ing. This framework encapsulates the key el-ements of any problem, such as the objective,the constraints, and our knowledge of the sys-tem. The system knowledge encompasses the

key elements of our area of interest, which maybe an entire biome, a country, or a parcel ofland (herein termed locations). The key ele-ments constitute the assets we wish to protect(a single species, a group of species, habitats,or an ecological service or function), the keythreats to these assets, the actions likely to abatethe threats, and the costs associated with theirimplementation.

The second aim of this chapter is to evaluatethe benefits and risks of allocating funds withand without considering these elements (e.g.,of allocating funds to particular assets withoutconsidering the cost of the action requiring im-plementation in order to conserve the asset in aparticular location). On the basis of our review,we conclude that while the prioritization of as-sets or locations is commonplace, prioritizationapproaches that do not consider the conserva-tion actions required to protect the assets inparticular locations are incomplete.

The third aim of this chapter is to illustratehow multiple conservation objectives can beintegrated. We discuss examples of trade-offswhen there are disparate conservation objec-tives and how trade-offs can be explicitly eval-uated within the generic framework for settingconservation priorities.

We acknowledge the challenges associatedwith the rational and efficient allocation ofconservation funds, including lack of a clearconservation objective, a paucity of biologi-cal and economic data, uncertainty about thelikely success of investments, and societal val-ues or political processes that can influenceinvestment decisions. The final aim of thischapter is to identify some opportunities forreducing these challenges when setting conser-vation priorities.

Defining a ConservationPrioritization Problem: What

Are the Key Elements?

Conservation prioritization problems can bestructured as classic optimization problems,

Wilson et al.: Setting Conservation Priorities 239

with an objective function, mathematicaldescriptions of our knowledge of the systemand the state variables, control variables, con-straints, and system equations (Possingham et al.2001; Wilson et al. 2007). In systematic con-servation planning there have been two broadclass of prioritization problem—the minimum-set and maximal-coverage problems—definedmathematically in Box 1.

In all problem formulations, the objectivefunction reflects our conservation goal and hasan explicit measure of performance. In a con-servation context we may seek to maximize thenumber of species conserved to an adequatelevel, or to minimize the number of speciesthat are expected to go extinct over the next100 years. Often there is more than one ob-jective. In almost all circumstances the strategythat optimizes one objective will not optimizeother objectives, which means that compro-mises must be made. Objectives can be com-bined into a single objective using a weightedsum or some other function; however, the useof weights presents challenges that we discusslater.

What we are required to know about the sys-tem (the system knowledge) will depend on theparticular problem at hand—in a conservationprioritization context we would want to knowwhat the assets are, what the threats are to theseassets, what actions can be taken to abate thesethreats or otherwise improve the state of the as-sets, and the cost of carrying out those actions,which may vary over the area of interest andthrough time. We may also require additionalknowledge, such as the uncertainty associatedwith key parameters and whether some out-comes are random or stochastic. The state vari-ables represent the assets in the area of interest.The assets might be biological (such as habi-tat types, populations of threatened species, thedistribution of invasive species, and ecosystemprocesses) or nonbiological (such as the loca-tions of landholders that are amenable to habi-tat restoration or the distribution of water flowsthrough a catchment).

The control variables reflect the things wecould do. In the context of conservation prior-

itization we control how much money or re-sources we direct toward different conserva-tion actions in any location and at a particu-lar time. These control variables will directlyor indirectly influence the states of our assets.The constraints limit the choice of control vari-ables. In maximal-coverage prioritization prob-lems (Box 1) constraints may include a budget,how many parcels of land can be restored eachyear due to operational and seasonal limita-tions, or may reflect the area of land that canbe conserved annually. In minimum-coverageproblems (Box 2) the constraints may repre-sent the minimum amount of conservation thatwe aim to achieve for each asset. For exam-ple, we could have a target amount of eachspecies requiring habitat protection or a tar-get for the frequency of flooding required ina river basin (an ecosystem process asset). Thesystem equations reflect our understanding ofhow our state variables (and also our objectivefunction) change as a function of each other,system parameters (such as the chance an areais cleared for urbanization), and the controlvariables. The overall aim is to find a solutionthrough manipulation of the control variablesthat has the highest possible value of the objec-tive function subject to our constraints. Whileoptimal solutions might be desired, multiplenear-optimal solutions are often sought for thesake of flexibility and the ease of calculation andcommunication.

The conservation prioritization problem canbe described mathematically. A general versionof the conservation prioritization problem sub-ject to a annual budget, bt , is described below.Let xjkt be the amount of money to be spenton action k in location j in year t. For example,one possible action might be to spend $10,000removing an invasive species from a single lo-cation in any one year. Each year the cost ofall the actions across all the locations must beless than our annual budget, so we have theconstraint:

N∑

j =1

P∑

k=1

x jkt ≤ bt for every year t . (1)

240 Annals of the New York Academy of Sciences

where N is the number of locations and P isthe number of possible actions. Determiningour budgetary constraints is generally straight-forward, although obtaining data on the costsof different conservation actions in differentlocations can be challenging (Naidoo et al.2006). Formally, we have the further constraintthat xjkt is greater than zero, since we can-not spend negative dollars on a conservationaction.

Next we need a set of states in every locationand time, yijt . For example, in each location andin each year we might need to know the size ofa population of a threatened species, the per-centage of plants being adequately pollinated,the condition of a particular habitat type, orthe genetic diversity within a population. Ourchoice of state variables will depend on manyissues, such as whether our state variables areassets (e.g., the population size of a threatenedspecies) or whether the state variable influencesour assets (e.g., the population size of an inva-sive species). Our objective will be to maximizesome weighted combination of the state vari-ables. As we describe later, choosing appropri-ate weightings can be difficult, but we assumefor now that weights can be derived. The ob-jective is to maximize over a time frame, T , ofinterest:

M∑

i=1

N∑

j =1

T∑

t=1

w ijt f (y. . .) (2)

where wijt is a weighted value assigned to theamount of state variable (or asset) i, in placej, at time t. In principle the value we assignto any asset at any time and location can be acomplex function of all the assets in the systemf (y. . .). In this approach, money itself can bea state variable and we can replace the annualfunding constraint by a dynamic accountingequation. Finally, we need a series of dynamicmodels that show how our actions and all thestate variables interact to move, invariably ina stochastic way, the whole system from onestate to another through time. Ultimately thewhole problem is complex, which is why more

specific and tractable versions of the generalconservation prioritization problem exist (seeBox 1).

Box 1. Mathematical Definitionof Two Broad Class of

Conservation PrioritizationProblem: The Minimum-Setand the Maximal-Coverage

Problems

Minimum-Set Problem

The objective of the minimum-set problem isto minimize the resources expended while meetinga given set of conservation targets (Pressey 2002).Each location j has a cost cj and each asset i hasa target ri . The variable xj equals 1 if location j isselected for investment, otherwise it equals 0. Thecontribution to the conservation of asset i by theselection of location j is contained in a matrix withelements aij . The objective is to minimize the cost:

∑

j ∈P

c j x j

subject to the constraint that the targets are met:∑

j ∈P

a ij x j ≥ r i

for every asset i.There are simple and complex versions of the

minimum-set problem. All locations might be as-sumed to have equal cost or we can allow each loca-tion (or location–action combination) to reflect theactual monetary or social cost. Each asset may alsobe described in different currencies (e.g., numberof individuals, extent of occurrence, probability ofoccurrence) and individual targets can be set foreach asset. If more than one action is under con-sideration, we can set asset targets for each actionand in each location, for example, we might aim torepresent 20% of the range of a species in a strictprotected area and 30% in an area managed forsustainable-resource extraction.

Maximal-Coverage Problem

The objective of the maximal-coverageproblem is to maximize some measure of

Wilson et al.: Setting Conservation Priorities 241

“benefit” (in a simple case, this might be the num-ber of targets met for our assets), given a fixed bud-get or resources that can be expended (Church andReVelle 1974). That is, the objective is to maximize

∑

i∈I

f ( y i (x .))

subject to∑

j ∈J

c j x j ≤ b

where cj and xj are as previously defined, and yi isthe amount of feature i conserved in reserve systemx., and f is a function that turns that into a value.The maximum available expendable budget is b,

which is in the same units as cj . Like the minimum-set coverage problem, there are multiple variationsto the maximal-coverage problem. The problemmay be solved without applying targets and thebudget may or may not be sufficient for meetingall targets. While budgets are typically finite, theyare not necessarily fixed and the budget can be up-dated through time if more or fewer funds becomeavailable. In the simplest case, if the target allocatedfor asset i is achieved, yi equals 1, and otherwise itequals 0. Alternatively, the benefit can be measuredby a set of functions representing the incrementalgains in the conservation of each asset per dollarinvested. The functions relating benefit to the costsof acquiring these benefits can be linear, meaningthat benefits are continuous, or curved to repre-sent situations where benefits diminish or increasewith each dollar invested. Assets can also be differ-entially weighted to emphasize investment in thosethat we value highly (e.g., assets that are locally rareor threatened) (Arponen et al. 2005, 2007).

Assets

In systematic conservation planning, the eco-logical purpose of a reserve system is to sustainrepresentative samples of the full range of biodi-versity and ecosystem processes of the region inwhich it lies (Margules and Pressey 2000). Typi-cally, conservation prioritization analyses focuson (1) protecting particular species (e.g., threat-ened, umbrella, or flagship species; see later inthis chapter), (2) protecting areas of high speciesrichness or areas with high endemism, and/or(3) protecting functioning ecosystems and theirassociated ecosystem processes. Knowledge of

our assets is therefore a key component of aconservation prioritization problem—these arethe species or other facets of biodiversity thatwe wish to conserve. Data are often incom-plete, however, even for well-known taxa suchas birds and mammals, and worse for lesser-known taxa like insects and fungi. Our knowl-edge of ecosystem processes is perhaps the leastcomplete (Pressey et al. 2007). This paucity ofdata forces us to use biodiversity surrogates inconservation priority setting.

The surrogates (such as using birds as a sur-rogate for all vertebrates) are the state variablesin the objective function. Proposed surrogatesinclude well-known taxonomic groups, habitattypes, and ecological classifications of land andwater (Ferrier et al. 2000). The level of supportin the literature for surrogates and their abil-ity to represent other elements of biodiversityare variable (Reyers and van Jaarsveld 2000;Beger et al. 2003; Faith et al. 2004). Surro-gates have been found to perform well in somecases (Howard et al. 1998; Oliver et al. 1998)and poorly in others (van Jaarsveld et al. 1998).Rodrigues and Brooks (2007) found cross-taxonsurrogates to be more effective than surrogatesbased on environmental data, particularly en-vironmental surrogates that are based only onabiotic data.

Many conservation plans involve settingtargets—amounts of a biodiversity asset atwhich we feel the asset is conserved. The ben-efits gained for each asset through investmentin their conservation can be quantified with orwithout use of targets. Targets can be set in amultitude of ways and can, for example, reflectthe percentage of historical extent of a vegeta-tion type or a minimum viable population sizefor a species. The development of targets canbe informed by ecological theory and empir-ical knowledge in order to accommodate therequirements of species for their persistence,for example, by accounting for informationon life history characteristics, habitat connec-tivity requirements, and threatening processes(Burgman et al. 2001). Targets that are gen-eralized for sociopolitical purposes (such as,

242 Annals of the New York Academy of Sciences

protect 10% of every habitat type) have limitedscientific basis and should not be interpreted asthe minimum amount of habitat requiring pro-tection in order to ensure species persistence(Soule and Sanjayan 1998; Noss 2004). Theuse of targets is implicitly based on the assump-tion that there is a critical level of protectionrequired and until that level is achieved thereare no benefits, or that the benefits accrue lin-early (Arponen et al. 2005; Carwardine et al.In Press). The relationship between the invest-ment and the benefit derived is of course morecomplicated and can take a variety of shapes(see later). In general the per-unit benefit of in-creasing the level of a protection of an asset willdecline, that is, benefit functions should haveslopes that decrease, and hence show dimin-ishing returns (Davis et al. 2006; Wilson et al.2006).

Threats

We need to understand the threats to bio-diversity in order to evaluate the risk of losingspecies or habitats if no conservation action istaken; that is, the vulnerability of different lo-cations and/or assets (Wilson et al. 2005). It isan inefficient use of funds to invest in locationswhere the risk of exposure to threats is low orwhere there are no biodiversity assets impactedby a particular threat. Ignoring threats in con-servation prioritization is only justifiable whenthere is enough money to abate all threats atonce, which is rarely the case. Incremental con-servation investment is much more commonand during a protracted process of protection,biodiversity (or the surrogates for biodiversity)might be lost or degraded.

Threats are highly variable spatially andthere are numerous ways to evaluate the vul-nerability of different locations and the assetsthey contain (Wilson et al. 2005). Furthermore,there are numerous ways for how informationon vulnerability might be used to identify pri-ority assets or locations for conservation invest-ment (Redford et al. 2003; Brooks et al. 2006). Inthe general conservation prioritization problem

outlined earlier, threats can be included as an-other state variable that, in general, has a nega-tive impact on state variables that we value. Un-der such circumstances the state variable mightbe the size of an invasive-species populationor the rate of sea-level rise due to global cli-mate change. At a local scale, the former threatcan be mitigated but the latter cannot with-out unrealistic expenditure. Other threats thatmay be prohibitive to mitigate include diseasesfor which there is no known cure, a volcaniceruption, or an invasive species for which aneffective control mechanism has not yet beenidentified. For other threats there may, in the-ory, be an action that stops the threat, but thescale and cost of implementation might be pro-hibitive (e.g., a global reduction of fossil-fuel useto reduce sea-level rise).

Costs

For all known actions there is an associatedcost and impact on assets of concern. The costof each action might reflect the cost of land pur-chase (Ando et al. 1998; Polasky et al. 2001), thecost of management (e.g., the cost of control-ling an invasive species) (Balmford et al. 2003;Wilson et al. 2007), opportunity costs (i.e., theprofits that are forgone when a conservation ac-tion is undertaken) (Naidoo and Iwamura 2007;Carwardine et al. 2008a), stewardship costs (i.e.,the compensation that a landowner is willing toaccept to manage their land for conservation)(Carwardine et al. 2008b), social costs (e.g., thenumber of people that will be disadvantagedby the creation of a protected area (Luck et al.2004), transaction costs (i.e., the costs associ-ated with negotiating an economic exchange orestablishing a conservation program) (Naidooet al. 2006), and the costs of information acqui-sition and planning (e.g., the costs associatedwith undertaking surveys or conducting priori-tization analyses) (Balmford and Gaston 1999;Gardner et al. 2008; Grantham et al. 2008).

The number of papers presenting economi-cally grounded conservation priority setting ap-proaches has grown (Ando et al. 1998; Balmford

Wilson et al.: Setting Conservation Priorities 243

et al. 2000; Polasky et al. 2001; Stewart et al.2003; Moore et al. 2004; Stewart and Possing-ham 2005; Wilson et al. 2006; Wilson et al. 2007;Bode et al. 2008; Carwardine et al. 2008a,b;Klein et al. 2008), as conservation scientistsrecognize the benefits of explicitly consideringthe spatially variable costs of conservation. Ac-counting for the costs of conservation activitiesis essential for post hoc evaluations of whethercertain actions were a cost-effective use of re-sources (Ferraro and Pattanayak 2006; Halpernet al. 2006). There is also increasing evidenceof the importance of explicitly incorporatingeconomic data in determining priorities (Bodeet al. 2008). Studies have shown that conserva-tion plans can be up to 10 times more efficient(Polasky et al. 2001; Naidoo et al. 2006), mean-ing that we can make better use of the limitedfunds available for conservation. Conservationcosts typically vary by two to four orders ofmagnitude more than biodiversity data, andconservation outcomes have been found to bemore sensitive to cost data than biodiversitydata (Bode et al. 2008). Such issues are of pri-mary importance in a dynamic and uncertainworld, and are particularly relevant in the con-text of fluctuating property markets where costscan vary stochastically and unpredictably. Thisalso has important implications for the typesof data that are prioritized for collection—dataon costs has been identified as an immediatepriority (Naidoo et al. 2006; Bode et al. 2008).

To undertake full-cost accounting we mustconsider immediate and long-term costs. Whilethe cost of land purchase for a new protectedarea is incurred immediately, the cost of manag-ing the protected area will continue indefinitely.The simplest approach is to endow manage-ment costs in perpetuity and use discountingrates for costs that are incurred over time toreflect the fact that money spent now is worthmore than money spent in the future. Thereare also interdependencies and interactions be-tween the costs of different conservation actionsthat should be accounted for. Take for exam-ple two conservation actions: predator controland fire management. Predator-control man-

agement may cost $x/ha, and fire manage-ment may cost $y/ha; however, carrying outthese actions together may equate to less than$x + y/ha, as transportation and labor costscan be shared. Or to fit with the general ver-sion of the conservation prioritization problem[Eqs. (1) and (2)], if we spend $y/ha on firemanagement and $x/ha on predator control,the benefits are likely to be greater than under-taking them separately.

Overcoming the tendency for costs to be seenas secondary to biological data in conserva-tion priority setting represents a major hurdlefor conservation science (Odling-Smee 2005).This hurdle exists because, first, obtaining ac-curate cost data at an appropriate resolution ischallenging and a frequently neglected activity(Naidoo et al. 2006). Second, biologically fo-cused conservationists can be hesitant to allownonbiological factors to influence the allocationof scarce resources, for fear that poorer conser-vation outcomes would result from “giving up”on species or ecosystems that are expensive tosave (Pimm 2000; Bottrill et al. 2008). Finally,the general lack of problem definition can be acontributing factor: when locations are priori-tized without an action in mind it is impossibleto account for the cost of the action.

Most current conservation priority settingdoes not follow the basic structure outlined inthis section. A general tendency is for conserva-tion scientists and practitioners to seek to pri-oritize locations or assets, without consideringthe overall objective and the actions requiredin different locations to conserve the assets. Inshort, it is common to solve only part of the con-servation prioritization problem, because theproblem itself is not properly defined.

Prioritizing Assets: How MuchMoney Should We Invest in theConservation of Pandas vs. the

Conservation of Snails, and Is Thisthe Right Question to Ask?

Much conservation effort is focused onparticular species. These species might be

244 Annals of the New York Academy of Sciences

keystone, umbrella, flagship, indicator, or fo-cal species (Table 1). Species can make a log-ical target for conservation investment due totheir public appeal (and raising public aware-ness can increase the pool of funds availablefor conservation), the availability of reasonable-quality data, and the lowered monitoring andevaluation costs resulting from a focus on fewerassets (Mace et al. 2007). It has been acknowl-edged that setting priorities based on a subsetof species can be prone to bias, and the choiceof species is problematic because many ofthe assumed ecological relationships are oftenuntested (Simberloff 1998; Lindenmayer et al.2002) (Table 1). Single-species managementalso ignores the suite of ecological processesthat maintain and sustain such species, and themanagement of one species may conflict withthe management of another (Simberloff 1998).

Many countries, states, and agencies pri-oritize funds for species recovery by focusingon threat status and prioritizing investment inthose species at most risk of extinction. Thethreat status might be determined by the In-ternational Union for Conservation of Natureand Natural Resources (IUCN) red-listing cri-teria, which use quantitative rules to assign arisk of extinction (IUCN 2003). Such an ap-proach is supported in many countries by leg-islated requirements for the protection of en-dangered (and especially charismatic) speciesand has a strong moral and ethical basis. It alsohas important implications for the availabilityand use of funds for biodiversity conservation.In Australia, for example, at a national level,critically endangered and endangered speciesare given preference for conservation spending(Possingham et al. 2002). Although not explic-itly stated, this approach assumes that a focuson threatened species will result in the fewestextinctions (Possingham et al. 2002; Bottrill et al.2008).

Using a threatened species as a surrogate forother species or habitats may be inappropri-ate if the presence of a threatened species doesnot indicate habitat of good condition or that isof high value for other species (Rubinoff 2001;

Possingham et al. 2002). A sole focus on threat-ened species may also result in the inefficientuse of funds, and an inefficient investment of re-sources can entail substantial opportunity costs.Imagine, for example, if securing the mostendangered species will cost millions of dol-lars with limited collateral benefits, while sev-eral less endangered taxa could be secured bya single, comparatively cheaper conservationaction.

An example of a conservation program fo-cused on a single threatened species is that ofthe California condor, Gymnogyps californianus.Over 20 years, US$35 million was invested inthe California condor recovery and release pro-gram, increasing the number of individuals inthe wild from zero to 68 by 2002 (Alagona 2004)to more than 150 in 2008 (Walters et al. 2008).Currently US$5 million is spent annually onthe project (Walters et al. 2008). Most fund-ing for this species’ conservation was from Cal-ifornia or Oregon-based donors, particularlythose focused on “birds of prey” (e.g., Pere-grine Fund), along with the United States Fishand Wildlife Service. Given the largely non-fungible nature of the funds available for itsconservation, it is likely that this condor pro-gram is a cost-effective use of conservationfunds. It is also likely that the species wouldhave gone extinct without a dedicated conser-vation program. However, without evidence tothe contrary, we assume that the relative cost-effectiveness of this investment and the asso-ciated opportunity costs in terms of speciesthat might have otherwise benefited were notassessed a priori (Walters et al. 2008). Whilethe conservation achievements of the Califor-nia condor project should not be undervalued,it is possible that the dollars allocated to thecondor could have ensured the persistence ofmany more species, particularly if a long-termand global view is taken.

Setting priorities primarily based on howthreatened a species is reflects poor prob-lem formulation. An assessment of threatstatus reveals the urgency of conservationintervention—it does not provide information

Wilson et al.: Setting Conservation Priorities 245

TABLE

1.

Com

mon

Type

sof

Spec

ies

Use

din

Sett

ing

Con

serv

atio

nPr

iori

ties,

with

aSu

mm

ary

ofth

eSt

reng

ths

and

Wea

knes

ses

ofEa

chA

ppro

ach

Spec

ies

type

Just

ifica

tion

Key

refe

renc

esA

nex

ampl

eSt

reng

ths

Wea

knes

ses

Key

ston

eC

hang

esin

the

popu

latio

nof

the

keys

tone

spec

ies

will

have

ala

rge

impa

cton

othe

rsp

ecie

s

(Pai

ne19

69;P

ower

etal

.19

96)

Cas

sow

ary

(Cas

uari

us

casu

ariu

s)E

colo

gica

lmea

ning

The

iden

tity

ofke

ysto

nesp

ecie

sm

ayno

tbe

know

nw

ithce

rtai

nty

and

perc

eive

dre

latio

nshi

psar

ela

rgel

yun

test

edU

mbr

ella

Con

serv

atio

nof

the

umbr

ella

spec

ies

will

ensu

reth

epe

rsis

tenc

eof

othe

rsp

ecie

sdu

eto

thei

rha

bita

tre

quir

emen

ts

(Shr

ader

-Fre

chet

tean

dM

cCoy

1993

;Rob

erge

and

Ang

elst

am20

04)

Jagu

ar(T

apir

uste

rres

tris

)E

colo

gica

lmea

ning

.Pe

rcei

ved

rela

tions

hips

are

larg

ely

unte

sted

.The

requ

irem

ents

ofon

esp

ecie

sm

ayno

trefl

ectt

hato

fot

hers

Fla

gshi

pC

hari

smat

icsp

ecie

sth

atw

illga

rner

publ

icat

tent

ion

and

finan

cial

supp

ort

(Shr

ader

-Fre

chet

tean

dM

cCoy

1993

)G

iant

pand

a(A

ilur

opod

a

mel

anol

euca

)Im

prov

edpu

blic

supp

ort

for

cons

erva

tion

May

notr

eflec

tloc

alpr

iori

ties

orva

lues

and

unlik

ely

tore

flect

the

cons

erva

tion

urge

ncy

ofot

her

spec

ies

Indi

cato

rC

hang

esin

the

popu

latio

nof

thes

esp

ecie

sre

flect

sen

viro

nmen

talc

hang

e

(Lan

dres

etal

.198

8;N

oss

1990

)Fr

eshw

ater

mus

sels

Ear

lyw

arni

ngof

envi

ronm

enta

lcha

nge

Unl

ikel

yto

reve

alca

sual

impa

cts

and

acco

untf

orco

mpl

exec

olog

ical

inte

ract

ions

Foca

lspe

cies

Spec

ies

with

high

sens

itivi

tyto

the

dom

inan

tthr

eats

and

who

sedi

stri

butio

nre

flect

sth

esp

atia

l,co

mpo

sitio

nal,

and

func

tiona

latt

ribu

tes

that

shou

ldbe

pres

enti

nth

ela

ndsc

ape

(Lam

beck

1997

;L

inde

nmay

eret

al.

2002

)

Gra

yw

olf(

Can

islu

pis)

Stra

tegi

cap

proa

chth

atre

flect

sth

eco

nser

vatio

nco

ntex

t

Cho

ice

ofsp

ecie

sm

aybe

bias

edan

dda

tade

pend

ent

246 Annals of the New York Academy of Sciences

about the immediate and ongoing cost of con-serving the species, the likelihood the conserva-tion effort will work, and the associated benefitsto other assets. If, on the other hand, we identifyand prioritize actions to abate the key threatsto the assets that we value, then we are im-mediately drawn to a discussion of what thoseactions protect, what the costs and likelihood ofsuccess of these actions are, and whether the ac-tions are socially and politically feasible. Josephet al. (In Press-b) provide a rational approach forprioritizing investments in species conservation(Box 2). With some added complexity, this ap-proach could be used to evaluate the overallbenefit of undertaking each action, in termsof the provision of collateral and incidentalbenefits.

Box 2. Prioritizing Projects forThreatened Species

Conservation

Joseph et al. (In Press-b) provide a decision the-oretic approach for choosing between projects (i.e.,bundles of actions required to secure a threatenedspecies) that aim to conserve species in a region.The approach ranks each project (i) by its cost-efficiency (Ei ), calculated as the product of thebiodiversity benefit (Bi ), species value (Wi ) andthe probability of success (Si ) divided by the cost(Ci ):

E i = Bi × Wi × Si

Ci

Given a fixed budget for management of threat-ened species, the top-ranking projects are selecteduntil the budget is expended. Thus, the approachis similar to a maximal-coverage problem (with-out complementarity) (Box 1); where the set ofprojects selected maximizes the total number ofspecies secured within a total budgetary constraint(K ).

∑

i

C i ≤ K

This prioritization approach may be appliedat any scale. Species values may be economic,social, biological, or political (Faith 1994; Vane-

Wright et al. 1994; Rodrıguez et al. 2004; Isaacet al. 2007; Marsh et al. 2007), or species may beconsidered equal. The algorithm provides a nearperfect greedy solution to the Knapsack problem(Martello and Toth 1990; Weitzman 1998; Hart-mann and Steel 2006).

Joseph et al. (In Press-a) apply their “ProjectPrioritization Protocol” (PPP) to explore resourceallocation for managing the threatened species ofNew Zealand. This represents the first real appli-cation of cost-effective resource allocation for themanagement of multiple threatened species.

While the approach is quantitative, rational,and repeatable, it can explicitly explore the subjec-tive elements to conservation priority setting wherethreatened species are concerned. In the case ofprioritizing resources for New Zealand’s threat-ened species, the authors investigated the conse-quences of valuing species. Species were weightedby their taxonomic distinctiveness and comparedwith a scenario where species were valued equally.A trade-off exists between funding the manage-ment of a greater number of the most cost-efficientand least risky projects and funding fewer projectsto manage the species of higher value. Their appli-cation of the PPP to 32 of New Zealand’s threat-ened species resulted in far better outcomes (i.e., agreater number of species expected to be secure)than by the commonly used approach of rank-ing projects by the threat status of species that theprojects aim to protect.

Prioritizing Locations: How Do WeIdentify a Biodiversity Hotspot or a

New Protected Area, and Is Thisthe Right Question to Ask?

A number of approaches are available forprioritizing locations to conserve biodiversity(Table 2). These approaches have also lever-aged a substantial amount of money for con-servation, with biodiversity hotspots alone mo-bilizing at least US$750 million of funding(Brooks et al. 2006). The approaches reviewedin Table 2 are not exhaustive, although theyare representative of the range of approachescurrently in use.

Locations are typically prioritized usingsome measure of biodiversity importance (e.g.,

Wilson et al.: Setting Conservation Priorities 247TA

BLE

2.

Diff

eren

tPr

iori

tizat

ion

App

roac

hes

Use

dto

Iden

tify

Prio

rity

Loca

tions

∗

Con

serv

atio

nH

owin

form

atio

nC

hara

cter

istic

sPr

iori

tisat

ion

plan

ning

onvu

lner

abili

tyof

typi

cal

appr

oach

prin

cipl

eC

rite

ria

Key

refe

renc

esis

used

prio

rity

loca

tions

Bio

dive

rsity

hots

pots

Irre

plac

eabi

lity

and

vuln

erab

ility

Thr

esho

ldsf

oren

dem

ism

(≥1,

500

ende

mic

plan

tspe

cies

)and

habi

tatl

oss

(≥70

%of

orig

inal

habi

tate

xten

thas

been

lost

)

(Con

serv

atio

nIn

tern

atio

nal

1990

;Mitt

erm

eier

etal

.199

8,20

04;M

yers

etal

.200

0)

(+)

Frag

men

ted

The

glob

al20

0ec

oreg

ions

Rep

rese

ntat

ion

Scor

ing

appr

oach

base

don

spec

ies

rich

ness

,end

emis

m,t

axon

omic

uniq

uene

ss,u

nusu

alec

olog

ical

orev

olut

iona

ryph

enom

ena,

and

glob

alra

rity

(Olso

nan

dD

iner

stei

n19

98;

Din

erst

ein

etal

.200

0;O

lson

and

Din

erst

ein

2002

)

(+/−

)E

ither

frag

men

ted

orin

tact

End

emic

bird

area

sR

arity

Thr

esho

lds

base

don

popu

latio

nsi

ze:t

hedi

stri

butio

nsof

two

orm

ore

rest

rict

ed-r

ange

bird

spec

ies

mus

tov

erla

p

(Sta

tter

sfiel

det

al.1

998)

NA

Eith

erfr

agm

ente

dor

inta

ct

Cri

sis

ecor

egio

nsV

ulne

rabi

lity

Thr

esho

lds

base

don

area

ofha

bita

tco

nver

ted

and

area

ofha

bita

tpro

tect

ed(H

oeks

tra

etal

.200

5)(+

)Fr

agm

ente

d

Hig

hbi

odiv

ersi

tyw

ilder

ness

area

sA

dequ

acy

Thr

esho

lds

base

don

habi

tati

ntac

tnes

san

dex

tent

,and

hum

anpo

pula

tion

dens

ity(M

itter

mei

eret

al.1

998,

2003

)(−

)In

tact

Las

twild

plac

esA

dequ

acy

Thr

esho

ldof

10%

ofea

chbi

ome

leas

taf

fect

edby

hum

ande

velo

pmen

t(S

ande

rson

etal

.200

2)(−

)In

tact

Impo

rtan

tbir

dar

eas

Rar

ityT

hres

hold

sba

sed

onth

epo

pula

tion

size

ofgl

obal

lyth

reat

ened

,res

tric

ted-

rang

e,bi

ome-

rest

rict

edm

igra

tory

,and

/or

cong

rega

tory

bird

spec

ies

(Eva

ns19

94)

(+)

Eith

erfr

agm

ente

dor

inta

ct

Alli

ance

for

zero

extin

ctio

nsi

tes

Vul

nera

bilit

yT

hres

hold

sba

sed

onth

eoc

curr

ence

ofhi

ghly

thre

aten

edsp

ecie

s(R

icke

tts

etal

.200

5)(+

)E

ither

frag

men

ted

orin

tact

Key

biod

iver

sity

area

sR

arity

Thr

esho

lds

base

don

the

popu

latio

nsi

zeof

glob

ally

thre

aten

edan

dre

stri

cted

-ran

gesp

ecie

san

dco

ngre

gato

ryan

dbi

ome-

rest

rict

edsp

ecie

s

(Eke

net

al.2

004)

(+)

Eith

erfr

agm

ente

dor

inta

ct

∗ Eac

hpr

iori

tizat

ion

appr

oach

issu

mm

ariz

edin

term

sof

the

cons

erva

tion

plan

ning

prin

cipl

eem

ploy

ed,

the

spec

ific

crite

ria

used

,an

dho

wth

ein

form

atio

non

vuln

erab

ility

isus

ed[(+)

vuln

erab

lear

eas

are

prio

ritiz

ed,o

rot

herw

ise

(−)]

.The

char

acte

rist

ics

ofty

pica

lpri

ority

loca

tions

are

also

sum

mar

ized

.

248 Annals of the New York Academy of Sciences

irreplaceability, rarity, adequacy) and/or ameasure of threat or vulnerability, althoughdifferent priority locations for conservation in-vestment are typically identified (Redford et al.2003; Brooks et al. 2006). Where differencesoccur they are likely due to the application ofdifferent data, analysis at different spatial scalesor resolutions, or due to differing institutionalvalues, constraints, and associated objectives.

One important distinguishing factor iswhether the approach for prioritizing locationsincorporates the principle of complementar-ity; that is, whether it allows for the selec-tion of areas that complement one another interms of the assets conserved (Justus and Sarkar2002). Some approaches evaluate sets of loca-tions on the basis of their collective representa-tion of biodiversity assets such that the wholeis not merely the sum of the parts. Hotspot ap-proaches sidestep the issue of complementarityby focusing on endemic species (complementar-ity is not relevant when there is no overlap in as-sets between locations), and scoring approachesignore complementarity meaning that they willbe inefficient when protection of multiple assetsis sought (Smith and Theberge 1987; Presseyand Nicholls 1989; Possingham et al. 2006).

Another important distinguishing factor isthe way each approach integrates informa-tion on vulnerability (Table 2). At one endof the spectrum lies the identification of en-demic bird areas (EBAs), which does not em-ploy information on vulnerability, but focuseson rare or restricted-range species (although,see Caldecott et al. 1996). In contrast, someapproaches place high importance on areas ofboth high vulnerability and high biodiversityvalue (e.g., biodiversity hotspots), and other ap-proaches give priority to areas with low vulner-ability and high biodiversity value (e.g., highbiodiversity wilderness areas).

Whether vulnerable areas are given prefer-ence or avoided depends upon the objectiveof the priority-setting approach and whetherthe threat is perceived to be stoppable or un-stoppable. For example, the aim of the vulner-ability assessment employed in the biodiversity

hotspots approach is to highlight the urgencyfor conservation action, as without rapid actionthe habitat loss to date in these hotspots threat-ens the remaining biodiversity (Mittermeieret al. 1998). In contrast, the global 200 ecore-gions aim to achieve the greatest long-term rep-resentation of biodiversity, and hence typicallyprioritize locations that are not vulnerable (Din-erstein et al. 2000; Groves et al. 2002), althoughin some cases high-risk areas might be soughtif they are important for representation. Gen-erally, using this approach, each ecoregion isranked in terms of its suitability for conserva-tion, prioritizing areas with low human use, andconversion of natural land cover on the assump-tion that these areas will be cheaper to manageand that assets in these areas will likely havea higher probability of persistence (Abell et al.1999; Ricketts et al. 1999; The Nature Con-servancy 2000; Groves et al. 2002). A similarrationale is employed in identifying wildernessareas (Sanderson et al. 2002). Future threats areoften also analyzed in the ecoregional planningapproach to gauge the urgency of conserva-tion actions, but typically after the candidatepriority areas have been identified (Olson et al.1998; Dinerstein et al. 2000).

While it is clear that there are differencesbetween priority-setting approaches in termsof how information on vulnerability is used,there is no simple rule as to whether favoringor avoiding vulnerable areas provides the bestoutcomes for biodiversity conservation. Theremight be logical merit in avoiding vulnerableareas if the probability of success in an areaunder high threat is low if conservation invest-ment in these areas will stimulate social or polit-ical conflict and accelerate further biodiversityloss. Nonetheless, when areas of high biodiver-sity value are threatened, then vulnerable areasmust be allocated high priority in order to avoidbiodiversity loss.

Most approaches do not provide a sequenc-ing of locations for conservation investmentand simply identify locations as either a pri-ority or not. In these cases, additional deci-sions are needed to determine the order that

Wilson et al.: Setting Conservation Priorities 249

each location must be invested in. Anotherimportant distinguishing factor between ap-proaches is whether or how the costs of conser-vation are considered. No approach explicitlyaccounts for the costs of conservation in eachlocation (Brooks et al. 2006), although conserva-tion organizations are actively pursuing meth-ods and data to do so (Murdoch et al. 2007;Wilson et al. 2007; Bode et al. 2008).

The conundrum of how to prioritize invest-ments through time in the face of threats (andalso spatially and temporally heterogeneouscosts) can be solved by proper problem defi-nition. Using the general framework for settingconservation priorities outlined earlier in thechapter (Eqs. 1 and 2), we can rationally in-tegrate multiple criteria and devise schedulesof where and when to invest conservation re-sources (Costello and Polasky 2004; Meir et al.2004; Drechsler 2005; Wilson et al. 2006).

The majority of the prioritization ap-proaches outlined in Table 2 comprise the anal-ysis of scores or the use of thresholds. Scoringsystems can be broadly defined as any approachfor evaluating or ranking options (assets, loca-tions, actions) based on summing the weightedscores for a number of criteria (e.g., global 200ecoregions). Multiple Criteria Analysis is a for-mal scoring approach developed for operationsresearch in the 1940s and has been applied innatural resource management and for the con-servation of biodiversity (Fernandes et al. 1999;Hajkowicz et al. 2000; Villa et al. 2002; Men-doza and Martins 2006; Moffett and Sarkar2006). The approach integrates disparate andcompeting objectives (see the section on trade-offs later in this chapter) and captures the pref-erences of stakeholders by allowing them tocontrol the relative weights given to each crite-rion. Criteria can be diverse and may includebiodiversity importance, fishing value, and eco-tourism potential.

There are, however, challenges associatedwith weighted scoring systems. For example,there can be subjectivity associated with howmuch weight to be given to different crite-ria. As a result, scoring systems are suscep-

tible to manipulation by preferentially mod-ifying weights according to user preferencesfor specific outcomes. Because each option isscored independently of other options, the av-erage reigns—low values in a few criteria canoverwhelm high values in a single criterion.Thus, weights are not necessarily reflected inthe outcomes. There can also be challengesin combining scores for criteria that have dif-ferent currencies. Combining the scores for“apples and oranges” to obtain a final over-all score can obscure useful information held inthe individual scores obtained for the separatecriteria.

Thresholds, as opposed to targets, are mea-sured at an individual site level. For exam-ple, a threshold may specify that priority lo-cations must have at least three threatenedspecies or be at least 50% intact. Thresholdmethods, such as biodiversity hotspots (Myerset al. 2000), have the advantage of being easyto communicate, but the simplicity comes ata price. If a location meets all but one crite-rion, then it would fail to be prioritized, andlocations that are prioritized because they ex-ceed a threshold by a small degree are con-sidered of equal value to a location that ex-ceeds a threshold by a large amount. Similarto weights in scoring methods, it is difficult torationally determine the value of thresholds,even though they have such a dramatic effecton the prioritization outcome.

Conservation prioritization, regardless of theapproach employed (e.g., ecoregional planning,biodiversity hotspots, etc.), is scale-dependent.Given perfect knowledge, it would always bemore efficient to prioritize a range of conser-vation actions at a fine resolution, even on aglobal scale. While comprehensive global datasets on biodiversity values, threats, and costsare improving (Hoekstra et al. 2005; Orme et al.2005; Ceballos and Ehrlich 2006; Naidoo andIwamura 2007; Naidoo et al. 2008; Underwoodet al. 2008), data that are available at a globalscale are still typically sparse and of varyingquality. At local scales, we are more likely tobe equipped with data to guide the choice of

250 Annals of the New York Academy of Sciences

conservation actions. Rouget et al. (2003), forexample, analyzed threats at the scale of theentire Cape Floristic Province and the scale ofthe composite Algulas Plain. The coarser scaleanalysis identified the broad priorities for con-servation, whereas the finer scale assessmentallowed for the analysis of more detailed andcomprehensive data sets but did not providean evaluation of the broader conservation con-text. Therefore the identification of broad pri-ority locations, particularly in data-poor areas,may help refine the geographic extent withinwhich specific actions must be identified, eval-uated, prioritized, and applied to suit localconditions.

Prioritizing Actions: Should WeRemove Weeds, Buy Land, PlantTrees, or Install Pollution Traps,

and Is This the RightQuestion to Ask?

The creation of protected areas has tradi-tionally been the primary action to achieveconservation outcomes, and is often used im-plicitly as a surrogate for the broader suiteof conservation interventions available to pro-tect biodiversity (Margules and Pressey 2000;Chape et al. 2005). Conservation practitionersof course routinely invest in a diverse array ofactivities such as fire management, invasive-species control, and revegetation, either in pro-tected areas, or on government or privatelyowned land. In many places land acquisition isnot feasible, and neither appropriate nor cost-effective (Pence et al. 2003; Stoneham et al.2003; Possingham et al. 2006; Carwardine et al.2008a). More recent approaches to conserva-tion priority seek to prioritize between multi-ple actions (and locations) to achieve conser-vation objectives. Some approaches are alsodynamic, capable of prioritizing actions tem-porally as well as spatially. In this section wediscuss and illustrate by example some ap-proaches to prioritizing multiple conservationactions.

Multiple Zoning of Land and Sea

Much traditional conservation land-useplanning has focused on a binary decision-making framework, where the landscape orseascape is allocated to either protected or un-protected status. Realistically, there is a range ofland (or sea) uses, which contribute differentlyto biodiversity conservation and have differentcosts of implementation (Box 3). We need a lotmore information to zone an area for multi-ple uses. At a minimum, we need to know thecontribution that putting each area into eachzone will make to the conservation of everyasset. This contribution could be asset-specificor constant across all assets. We also need toknow the possible zone transitions and the costof each transition. In more advanced scenar-ios, there may be desired spatial relationshipsbetween zones.

Box 3: Land-Use Zoning inTropical Forest Regions

Tropical rain forest habitat is used for a diver-sity of land uses, ranging from protected areas toproduction forests. Each alternative land use con-tributes differently to the conservation (or destruc-tion) of biodiversity. Some land uses provide habitatthroughout all levels of forest strata, along with adiversity of food sources for fauna species occupy-ing the forest. Other land uses are more restrictedin their provision of habitat and food sources withthe resultant floristic and faunal diversity reflectingthese differences (Meijaard et al. 2006; Meijaardand Sheil 2008). Because of the relative sensitivityof species to different land uses, their contributionto the conservation of biodiversity varies (Dutton2001; Jepson et al. 2001; Bengen and Dutton 2004;Curran et al. 2004; Meijaard et al. 2006; Nakagawaet al. 2006; Meijaard and Sheil 2007; Wells et al.2007).

This situation presents a variety of challenges.The suite of possible land uses is complex and thecontrol variable is not binary. Furthermore theremay be constraints, upper and lower, on how muchland that can put into different land uses, suchas in protected areas or oil palm plantations. Wecan, however, examine different land allocation

Wilson et al.: Setting Conservation Priorities 251

scenarios by accounting for these constraints, andthe relative costs and benefits of different land uses.First, we need to know the relative sensitivity of ourbiodiversity assets to forest conversion, and thenassign targets that accounts for this. We can thendetermine the contribution of each land-use zoneto achieving the targets for each asset. With suchinformation we can evaluate not only where to actbut how to act in order to effectively and efficientlyconserve biodiversity.

Conservation Action Planning

Conservation Action Planning (CAP) is onepart of the Conservation by Design process de-veloped by The Nature Conservancy (TheNature Conservancy 2000, 2003). It was de-veloped as a stakeholder-driven planning ap-proach for landscape-scale projects, but has alsobeen used at regional and local scales. The ap-proach is designed to be iterative and adap-tive, recognizing that information is always in-complete, but should improve over the life of aproject. The CAP approach is based on “TheFive S’s.”

• Systems: The conservation assets in a land-scape and the key ecological attributes thatmaintain their viability (generally, 6–8 as-sets are employed).

• Stresses: The types of destruction or degra-dation impacting each conservation asset.

• Sources: The agents generating the stresses.• Strategies: The types of conservation actions

that can be deployed to abate sources ofstress.

• Success: Measures of biodiversity health andthreat abatement.

The aim of the CAP approach is to iden-tify actions that maintain healthy, viable oc-currences of the conservation targets. By def-inition, healthy occurrences are not signifi-cantly stressed, and measures of success caninclude indicators of the effectiveness of con-servation strategies and indicators of targetviability.

The CAP process has the benefits of be-ing stakeholder driven and iterative. The ap-proach accounts for the key elements of a con-servation prioritization problem and reflects aproject-management tool that is a mixture ofa focal species approach and a static multiple-action prioritization approach. The process en-courages consideration and articulation of whatconservation success in a landscape would looklike and how progress toward this goal will bemeasured. A key limitation of the approach isthat the consideration of the cost and likeli-hood of success of each action are not an inte-grated and explicit part of the planning process,and therefore have little influence on the initialidentification and prioritization of conservationactions.

Dynamic Multiple-Action Prioritization

In dynamic versions of the multiple-actionconservation prioritization problem our man-agement decision is how much of our budgetto allocate to each conservation action at eachtime step. For each conservation action we needto know what it costs per unit area and itsbenefit to the relevant biodiversity assets(Table 3; Box 4). We also need to know thecurrent investment in the conservation actionand the remaining area that would benefit fromfurther investment in the action. We can thengenerate dynamic investment schedules thatreflect shifts in the allocation of funds as thereturn from investing in each conservation ac-tion diminishes. The investment schedule is de-termined by the interplay of three main fac-tors: (1) the relationship between the additionalarea invested in each conservation action andbenefit to biodiversity, (2) the cost of this in-vestment, and (3) the existing level of invest-ment. We can then seek to prioritize actionswith the greatest benefit per dollar invested(i.e., to maximize gains), or if the rate of habi-tat loss is known, we can prioritize actionsthat will minimize the expected loss of biodi-versity (i.e., to minimize losses) (Sarkar et al.2006; Wilson et al. 2006; Murdoch et al. 2007).

252 Annals of the New York Academy of Sciences

TABLE 3. Examples of Three Conservation Actions Identified to Abate Key Threats to Biodiversity in TwoAustralian Ecoregions∗

Invasive predator Phytophthora cinnamomi

control management Revegetation

Eyre and Mount Eyre and Mount Eyre and MountConservation action York Lofty York Lofty York Loftyand ecoregion mallee woodlands mallee woodlands mallee woodlands

Number of assets protected 514 517 251 255 153 157Cost of the action (US$/km2) 301,154 301,005 514,592 514,443 7,091 6,942Total area of the ecoregion (km2) 60,896 23,786 60,896 23,786 60,896 23,786Area already receiving action (km2) 18,926 4,753 0 0 416 1,221Area requiring action (km2) 6,295 2,855 12,791 4,198 18,510 3,532

∗The number of assets estimated to be protected by the action (i.e., the number of species sensitive to the threatsthat each action will abate) and the cost of investing in each action is provided, along with estimates of the area alreadyreceiving the action and the additional area requiring the action.

Multiple-action conservation prioritization hasbeen shown to outperform more traditionalapproaches to conservation funding allocationthat focus solely on protected-area establish-ment (Box 4). Although fully comprehensivestudies are desired, multiple-action prioritiza-tion analyses suffer from the same limitationsas other prioritization analyses discussed in thisreview. In particular, data and computationalconstraints often deem it prohibitive to fullyaccount for complexities such as the temporaland spatial heterogeneity associated with in-formation on costs, threats, and benefits, not tomention the stochastic nature of environmentaland socioeconomic factors (Wilson et al. 2006,2007).

Multiple-action prioritization problems canbe potentially very complex. For example, therecan be spatial dependencies and the impact(and cost) of actions will likely vary betweenassets. The change in value from investmentin different actions depends on the relation-ship used to describe the objective function;common forms include threshold, linear, sig-moid, or concave (Fig. 1). Van Teeffelen andMoilanen (2008) evaluate the consequences ofusing different functional forms for the objec-tive function. A linear relationship provides a

Figure 1. Four possible objective-function shapes:linear, threshold, sigmoid, and concave.

continuous accumulation of benefit, whereas athreshold function reflects an assumption thatthe benefit peaks once the threshold is reached.A concave objective function reflects a situa-tion of diminishing marginal gains. The sig-moid objective function reflects an assumptionthat low levels of investment will deliver low re-turns and that the returns increase rapidly aftersome amount of investment is reached and thentapers off.

Wilson et al.: Setting Conservation Priorities 253

Box 4. Maximizing theConservation of Biodiversity

Through Investment in a Rangeof Conservation Actions

Through Time: A Case Studyfrom Mediterranean

Ecoregions

Wilson et al. (2007) developed a framework forguiding the allocation of funds among alternativeconservation actions that address specific threatsand applied it to 17 of the world’s 39 Mediter-ranean climate ecoregions. This subset of Mediter-ranean ecoregions covers parts of Australia (tenecoregions), Chile (one ecoregion), South Africa(three ecoregions), and California and Baja Cali-fornia (three ecoregions). In total, they evaluated51 ecoregion–conservation action combinationswhich were termed “ecoactions.” In Australia, forexample, ten Australian Mediterranean ecoregionswere analyzed and three primary threats wereidentified: habitat fragmentation and salinization(with salinization the result of altered hydrologicalregimes); introduced predators (focusing on cats[Felis catus] and foxes [Vulpes vulpes]); and the soil-borne pseudofungus Phytophthora cinnamomi. Themitigating actions selected were (for each respec-tive threat) strategic revegetation, invasive predatorcontrol and research, and a mix of precautionaryand preventive measures, including the applicationof phosphite, quarantine, education, research, andcommunication. For each of these threats severalpieces of information were obtained (Table 3).

Wilson et al. (2007) assumed that the marginalbenefit decreases for each new unit area receiv-ing investment, and therefore, the benefit dimin-ishes with cumulative investment. This relationshipwas modeled using species-area curves, which wereconverted into species-investment curves. Only 24ecoactions (of the 51 possible) received investmentduring the first five years. A mix of land protec-tion and off-reserve management in South Africawas highly rated, along with invasive-plant con-trol in Chile, California, and South Africa. Theseconservation actions yielded the greatest marginalreturn on investment over five years because thebiodiversity benefits were high and the costs werecomparatively low. However, this does not involvesimply prioritizing the cheapest actions, or themost species-rich ecoregions; rather, this informa-tion is integrated to find the most cost-effectivecombinations.

The ecoaction-specific framework was thencompared to a model of conservation that focusesonly on land acquisition. Over five years almostfour times as many species can be protected us-ing the ecoaction approach (2780 vs. 703 species).These results suggest that investing in a sequenceof conservation actions targeted toward specificthreats, such as invasive-species control and firemanagement, will deliver greater biodiversity re-turns than by relying solely on acquiring land forprotected areas.

We can also incorporate complementaritywithin this dynamic multiple-action prioritiza-tion framework. Complementarity is importantif we aim to minimize the cost of the conser-vation action under consideration, ensure thatall biodiversity assets (e.g., species, vegetationtypes) receive some level of conservation invest-ment, and identify location-and-action combi-nations that complement those that have al-ready received investment (Justus and Sarkar2002). To apply complementarity we requireinformation on the specific biodiversity con-tent of locations, as opposed to summary statis-tics such as species richness. Furthermore, loca-tions and actions must be evaluated collectivelyrather than independently.

Our objective when accounting for the prin-ciple of complementarity is to protect as manydifferent, or “distinct,” biodiversity assets aspossible. As described earlier, the shape of ourobjective function reflects the functional rela-tionship between assets protected and the levelof investment. In the case of two actions A andB, we would consider not only an “A curve’ anda “B curve’ for assets only benefited by actionsA and B, respectively, but also an “AB curve’ forassets benefited by both actions A and B (Fig. 2).For three or more actions, we would considera total of seven curves: A, B, C , AB, AC , BC ,and ABC . The number of curves thus increasescombinatorially as the number of actions in-creases. Using complementarity favors actionswhere for each dollar invested, the sum of theadditional assets benefitted is greatest. Someactions might require other actions in order to

254 Annals of the New York Academy of Sciences

Figure 2. A hypothetical example of the curvesrequired to account for complementarity between twoactions (A and B). Curve A and curve B representthe number of distinct assets benefited by investing inaction A and B, and curve AB represents overlappingassets benefited by investing in both actions A and B(Underwood et al. 2008).

be successful, and there can be negative inter-actions between actions if the action requiredfor one species is detrimental to the conserva-tion goals of another. Not accounting for theseforms of complementarity will overestimate theconservation benefit of a set of actions.

Exploring Trade-Offs and Dealingwith Multiple Objectives

A trade-off exists in conservation priority set-ting under two conditions: when two or moreassets are not perfectly aligned, and/or whenthe budget or some other constraint preventsus from reaching our targets for all biodiversityassets. For example, we may be unable to im-mediately conserve adequate representations ofeach threatened species whose distributions donot overlap. Under such circumstances we havea trade-off between those targets that we meetnow and those we may attempt to meet later.Trade-offs are unavoidable in priority settingwhen there are multiple considerations, whichis almost always the case. Even in a simplis-tic prioritization problem involving only the

identification of potential protected areas, pro-tection of multiple assets will likely be soughtand the implementation of protected areas willlikely be incremental. We therefore need tomake decisions about which areas should beprotected first. Trade-offs can become evenmore complex when benefits other than bio-diversity conservation are sought, such as thedelivery of ecosystem services or the protectionof sites of cultural significance.

In the case of ecosystem service conservation,we can assess the spatial and temporal con-gruence of ecosystem services and biodiversityassets. If such an analysis shows that areas ofhigh biodiversity and ecosystem service valuecoincide, then conservation intervention willsimultaneously benefit both (a “win-win” sit-uation). Concordance between ecosystem ser-vice delivery and biodiversity conservation hasbeen found at a variety of scales, suggesting thatsuch opportunities exist (Turner et al. 2007).However, other analyses have found limitedspatial congruence between important areasfor ecosystem services delivery and biodiversityconservation (Chan et al. 2006; Naidoo et al.2006). There are several reasons for such out-comes. It is likely that the provision of ecosys-tem services is best achieved by land-use optionsthat are suboptimal for biodiversity conserva-tion (for example, the growth and harvest ofcereal crops may be more efficient in relativelyspecies-poor ecosystems) (Cassman and Wood2005). Many forms of biodiversity do not offerknown benefit to humans, and some biodiver-sity is even detrimental to humans. In areasthat are heavily cleared, habitat loss and frag-mentation would reduce the ecosystem servicevalue of the land, while increasing the biodi-versity value of the remaining habitat (Turneret al. 2007). Finally, while patterns of richnessin different assets may coincide, patterns ofcomplementarity may not. We therefore needto explicitly account for the possibility thatsynergies between conservation and ecosystemservices will not exist. The mathematical inte-gration of multiple objectives depends on cor-rect problem formulation, and again there is a

Wilson et al.: Setting Conservation Priorities 255

distinction between maximal-coverage prob-lems and minimum-set problems (Box 1).

In maximal-coverage problems actions areselected that maximize some benefit withoutexceeding a conservation budget. When thereare multiple objectives, we can modify the rel-ative weighting of each objective: for example,the conservation of biodiversity and the provi-sion of ecosystem services. In effect, the overallobjective becomes the selection of actions thatoptimize the weighted sum off these subobjec-tives. Weighting objectives can involve somelevel of arbitrariness, as previously discussed,but also provides flexibility to explore trade-offs by investigating a range of weights (Arthuret al. 2004).

A trade-off curve can be helpful in prioritysetting for multiple objectives, by highlightingsets of options that achieve different amountsof each objective. When we vary the relativeimportance of protecting biodiversity versusprotecting ecosystem services through its fullspectrum, we define a “Pareto-frontier”: thebest landscape outcomes achievable for anytwo-criteria optimization (Nalle et al. 2004).An example of this frontier is shown inFigure 3. The highest and lowest points in thisfigure correspond to landscapes planned exclu-sively for ecosystem services and biodiversity,respectively, with the points in between rep-resenting compromises across both objectives.When aiming to achieve two disparate objec-tives, any set of actions that exists inside thePareto-optimal line is suboptimal.

In minimum-set problems actions are se-lected that meet conservation targets at a min-imal overall cost. We can include targets fora virtually unlimited number of assets. In thistype of problem formulation the achievementof targets for each asset is a kind of subobjec-tive, where the overall objective is to meet alltargets. For example, Chan et al. (2006) prior-itized places that protect variable targets for aset of assets, and that store at least 50% of thecarbon within the entire region. This target-based approach has the benefit of avoiding theneed for arbitrary weights (although targets can

Figure 3. Pareto-frontier for biodiversity andecosystem service protection for a hypotheticallandscape.

involve arbitrary elements) and of identifyingsets of priorities that can be evaluated againstquantitative targets. The target-based ap-proach itself does not explicitly address trade-offs, but trade-offs can be investigated by vary-ing the targets and evaluating the difference inthe overall cost (Egoh et al. 2007).

How Can We Better SetConservation Priorities?

Account for theSociopolitical–Economic–Ecological

Context

The probability of success of different con-servation actions is influenced by numeroussociopolitical factors, including political stabil-ity and corruption, budget continuity, gover-nance, and stakeholder willingness to be in-volved in conservation initiatives (Barrett et al.2001; Smith et al. 2003; Knight and Cowl-ing 2007). Although uncommon, we can ac-count for the probability that a conservationaction will succeed by modifying the expectedbiodiversity benefit (Hobbs and Kristjanson2003; McBride et al. 2007; Joseph et al. InPress-b).

256 Annals of the New York Academy of Sciences

In global and even regional problems, therecan be much variation in the political and eco-nomic systems and in societal needs, whichaffect the suitability of different conservationactions. Inadequate handling of such issueshas led to conservation failures. For example,the implementation of the Natura 2000 net-work in Finland was problematic due to in-adequate consideration of the needs and de-sires of landholders in the prioritization process,and the availability of sufficient funds to com-pensate landholders for the loss of land rights(Hiedanpaa 2002). Such an example points tothe importance of ensuring that the conserva-tion instruments employed are suited to the lo-cal context.

Market-based instruments take advantage ofmarket forces and social responses and areoften an efficient conservation prioritizationtool. Reverse auctions is a market-based in-strument that involves the efficient selectionof “bids” placed by landowners for the finan-cial compensation they will accept to managetheir land for conservation outcomes (Table 2).Reverse auctions have been carried out suc-cessfully in the United States (Reichelderferand Boggess 1988) and in parts of Australia(Grieve and Uebel 2003; Stoneham et al. 2003;Hajkowicz et al. 2007). Reverse auctions dif-fer from more traditional priority-setting ap-proaches because they only consider parcels ofland that are available for conservation inter-vention (only amenable landholders will placebids), and they elicit real costs, thus overcomingthe difficulty of predicting the costs of conser-vation. Reverse auctions are, however, unableto assess the potential contribution of all parcelsof land to an overall conservation goal and arenot suitable for strategic planning over largeregions.

It is also important to consider the ecologi-cal context and the extent that an asset is likelyto be benefited by a particular action. In somecases, an action implemented alone is unlikelyto achieve the conservation goal and the in-terdependencies between actions must be con-sidered. A simple example is the purchase and

management of land for the establishment ofprotected areas. If the key threat facing an areais native habitat loss due to conversion to an-other land use, then purchase of the land mightabate this immediate threat. However, if thebiodiversity values of the land are also threat-ened by weed invasion and habitat degrada-tion, then land purchase without accountingfor the costs of restoration and management toensure the persistence and health of the biodi-versity it contains will be insufficient. In suchcases the costs of all required conservation ac-tions (acquisition and management) and theirrelative benefits should be considered from theoutset.

Explicitly Account for Uncertaintyand Risk

There are many forms of uncertainty as-sociated with biodiversity conservation. Theseinclude the likelihood of success of conserva-tion actions, uncertainty about our objectives(have we been asking the right question?), anduncertainty associated with the data used toparameterize our analyses. Uncertainty (ran-domness with unknowable probabilities) differsfrom risk (randomness with knowable probabil-ities) (Knight 1921).

First, we can assess whether the uncertaintyin a particular parameter will be naturally re-duced or resolved during the process of im-plementing the conservation action. For exam-ple, there is often uncertainty associated withboth the cost of purchasing a parcel of landand whether the assets we wish to conserve arepresent at the site. However, before establish-ing a protected area, we will be exposed to up-dated information about the land. For example,we will know the price at which a landowneris willing to sell and we are likely to under-take a biological survey to ensure that the landcontains the assets we wish to conserve. Forthese kinds of uncertain parameters an adap-tive approach is essential, where priorities areupdated as new information becomes available(see below).

Wilson et al.: Setting Conservation Priorities 257

If the uncertainty will not be naturally re-solved, we might aim to reduce the uncer-tainty by collecting more information, for ex-ample, through improving our knowledge ofthe system and its dynamics (Haight et al. 2005;Armsworth et al. 2006). However, obtainingmore information can come at a cost, both fi-nancially and through opportunities that mightbe lost as we acquire further information. Forexample, it has been argued that new biodi-versity surveys lead to more efficient conserva-tion plans and therefore the surveys representgood investments (Balmford and Gaston 1999).Grantham (2008) found diminishing returns onsurveys of Protea species in South Africa, how-ever, and identified thresholds beyond whichfurther investments in surveys would not de-liver more effective protection of Protea species(despite a much better understanding of thedistribution of the species). This study indicatesthe importance of accounting for the returnon investments in different data, the benefitsthe additional data deliver to our conservationoutcomes, and the opportunity costs associatedwith the investments.

We can also explore the range of possiblevalues for an uncertain parameter, noting howdecisions may change depending upon the pa-rameter value. Such sensitivity analyses can beuseful to identify the parameters that have thegreatest impact on the outcomes, and deter-mine which parameters to collect more infor-mation for, and thus reduce uncertainty. How-ever, in many situations, risks and uncertaintycannot be removed.

We might aim to set conservation prioritiesthat are robust to risk and uncertainty (Ben-Haim 2001; Nicholson and Possingham 2007).Here we need to know (or estimate) the like-lihood that an unplanned but conservation-relevant event may occur, such as the risk ofa hurricane, fire, or coral bleaching event, orthe risk that a conservation action will not becarried out correctly (the inverse of its likeli-hood of success). We can then either priori-tize actions (or locations to carry out an action)that meet conservation targets while minimiz-

ing some combination of risk and cost (yet an-other trade-off ) (Game et al. 2008), or prioritizeactions that maximize the expected or likelyconservation benefits for a fixed budget (Josephet al. In Press-b). Note that these solutions rep-resent modifications of Equations (1) and (2),respectively.

Allow for Opportunities and Feedbacks

Like all actions, conservation actions createreactions and these reactions can be in the formof potentially beneficial opportunities or nega-tive feedbacks. In some cases, opportunities andfeedbacks can be predicted, particularly wherethere is a sound knowledge of the system or wehave been able to learn from past experiences.Where possible, predictions of both positive andnegative effects of conservation actions shouldbe explicitly incorporated into the priority set-ting processes.

Establishing a protected area in a region canhave positive effects. For example, the liveli-hood of residents might be improved throughthe provision of opportunities for ecotourismventures, which can potentially reduce the de-pendence of local communities on natural re-sources. Successful conservation outcomes in aregion can also be the key that unlocks furtherfunds for conservation. In order to account forsuch leverage, we can modify Equation (1) toinclude an increased budget to represent addi-tional funds, and we can modify Equation (2) toinclude the additional benefits that have beengained.

Negative feedbacks can also occur. For ex-ample, the purchase of properties for a newprotected area may drive land prices up due toincreased demand for land for other purposes.This can have perverse effects on local commu-nities, and also make it more expensive to es-tablish additional protected areas (Costello andPolasky 2004). The protection of a particularparcel of land because it is biologically valuableand also threatened by conversion, may simplydisplace the threat to nearby areas. There ex-ist approaches to predicting and incorporating

258 Annals of the New York Academy of Sciences

negative feedbacks in dynamic conservationplanning situations (Armsworth et al. 2006).Whenever a parcel of land is purchased, therelative price of other parcels of land is alteredusing knowledge of the dynamics of the localmarket. Likewise, the likelihood of success canbe updated in a similar manner.

It is important to acknowledge that theeconomic and sociopolitical reaction of sys-tems and people to conservation cannot al-ways be predicted. Unforeseen reactions canoccur, leading to perverse conservation and orsocial outcomes. Hence, rather than being top-down and inflexible, conservation priority set-ting should be modest and flexible to allow foradaptation to or mitigation of perverse out-comes. The financial and ecological impact ofunforeseen opportunities or reactions shouldbe explicitly accounted for as they arise andassessed within the prioritization framework.We can then learn from such evaluations tobetter respond to opportunities and negativefeedbacks in the future.

Evaluate, Learn, and Improve

There has been growing concern among theconservation community surrounding the lim-ited availability of evidence that investmentshave achieved sound conservation outcomes(Ferraro and Pattanayak 2006). This situationreflects a general tendency for only successfulconservation stories to be reported (Redfordand Taber 2000; Knight 2006) and the limitedresources allocated to monitoring and evalua-tion (Field et al. 2007). This is problematic, sincea lack of evaluation and honest reporting meansthat the process of learning and improvementis hindered and past mistakes may be repeated.There is a clear need for better evaluation pro-cedures, a systematic process for learning frompast successes and failures, improved commu-nication to extend this knowledge, and adaptivepriority setting.

Adaptive priority setting is a relatively sim-ple process, requiring the fluid updating ofpriorities and priority-setting approaches in

response to new information as it becomesavailable (Higgins and Duane 2008). Againflexibility is essential, both in the mentalityof decision-makers and the tools deployed.Broadly speaking, there are three approaches toadaptive management: passive adaptive man-agement, where we simply learn by analyzingthe outcomes of past interventions (this is themost common approach); experimental adap-tive management, where we actively manipu-late the system to maximize how fast we learn;and active adaptive management, where we op-timally manage taking into account the possi-bility for learning. The latter is optimal, buttechnically complex (Walters 1986; McCarthyand Possingham 2007; Hauser and Possing-ham 2008). A properly formulated active adap-tive management problem leads to solutionsto conservation problems that properly inte-grate managing, monitoring, and learning, andhence leads to a reduction of uncertainty.

Conclusions

A broad range of priority-setting approachesis available, matching the myriad types ofconservation decisions faced by conservationscientists and practitioners. By distilling conser-vation prioritization problems into the compo-nents of age-old decision theory, all problemsbecome comprehensible and solvable. Whileacknowledging the variety and complexity ofconservation problems, we have presented aframework and summarized general rules thatcan help to guide most conservation prioritiza-tion tasks. First, problem formulation is critical.Without a clear definition of goals, as well asthe identification of actions and their costs andlikely benefits, decisions are unlikely to be cost-effective, and outcomes cannot be evaluated.Second, considering the conservation actionsto be undertaken in a particular location to pro-tect key assets is the most comprehensive formof conservation prioritization. Third, conser-vation decisionmakers need not shy away fromproblems with multiple objectives, or those that

Wilson et al.: Setting Conservation Priorities 259

are prone to uncertainty and risk—rather, thisinformation can be explicitly included in thepriority-setting process to improve conserva-tion decision-making. And finally, an adaptableand flexible approach to priority setting is es-sential to ensure we can effectively respond toopportunities, feedbacks, and improved knowl-edge. The principles and examples we presentshould only be applied with flexibility, humility,and a sound knowledge of the socioecologicalsystem.

Acknowledgments

The authors acknowledge funding from theAustralian Research Council, Conservation In-ternational, The Nature Conservancy, the Ap-plied Environmental Decision Analysis Com-monwealth Environmental Research Facility,the Australian Department of Environment,Water, Heritage and the Arts, and the Com-monwealth Scientific and Industrial ResearchOrganization (CSIRO). We are also grateful formany stimulating conversations and discussionswith collaborators and colleagues from aroundthe world that have provided a diverse array oftheories, planning approaches, data, and appli-cations on which we could base this review.

Conflicts of Interest

The authors declare no conflicts of interest.

References

Abbitt, R.J.F., J.M. Scott, et al. 2000. The geographyof vulnerability: incorporating species geographyand human development patterns into conservationplanning. Biol. Conserv. 96: 169–175.

Abell, R.A., D.M. Olson, et al. 1999. Freshwater Ecoregions

of North America: A Conservation Assessment. Island Press.Washington, DC.

Alagona, P.S. 2004. Biography of a “Feathered Pig”:The California Condor Conservation Controversy.J. Hist. Biol. 37: 557–583.

Ando, A., J. Camm, et al. 1998. Species distributions, landvalues, and efficient conservation. Science 279: 2126–2128.

Armsworth, P.R., G.C. Daily, et al. 2006. Land marketfeedbacks can undermine biodiversity conservation.Proc. Natl. Acad. Sci. 103: 5403–5408.

Arponen, A., R.K. Heikkinen, et al. 2005. The valueof biodiversity in reserve selection: representation,species weighting and benefit functions. Conserv. Biol.

19: 2009–2014.Arponen, A., H. Kondelin, et al. 2007. Area-based refine-

ment for selection of reserve sites with the benefit-function approach. Conserv. Biol. 21: 527–533.

Arthur, J.L., J.D. Camm, et al. 2004. Weighing conserva-tion objectives: maximum expected coverage versusendangered species protection. Ecol. Appl. 14: 1936–1945.

Australian Bureau of Statistics. 2004. Environment Expendi-

ture, Local Government, Australia, 2002–2003 Australian

Government. Australian Bureau of Statistis. Canberra,Australia.

Balmford, A. & K.J. Gaston. 1999. Why biodiversity sur-veys are good value. Nature 398: 204–205.

Balmford, A., K.J. Gaston, et al. 2003. Global variation interrestrial conservation costs, conservation benefits,and unmet conservation needs. Proc. Natl. Acad. Sci.

USA 100: 1046–1050.Balmford, A., K.J. Gaston, et al. 2000. Integrating costs

of conservation into international priority setting.Conserv. Biol. 14: 597–605.

Balmford, A., P. Gravestock, et al. 2004. The worldwidecosts of marine protected areas. Proc. Natl. Acad. Sci.

101: 9694–9697.Barrett, C.B., K. Brandon, et al. 2001. Conserving tropi-

cal biodiversity amid weak institutions. Bioscience 51:497–502.

Beger, M., G.P. Jones, et al. 2003. Conservation of coralreef biodiversity: a comparison of reserve selectionprocedures for corals and fishes. Biol. Conserv. 111:53–62.

Ben-Haim, Y. 2001. Information-Gap Decision Theory: De-

cisions Under Severe Uncertainty. Academic Press. SanDiego.

Bengen, D. & I.M. Dutton. 2004. Interactions: man-groves, fisheries and forestry management inIndonesia. In Fishes and Forestry—Worldwide Watershed

Interactions and Management. T.G. Northcote & G.F.Hartman, Eds.: 632–653. Blackwell Science Ltd.Oxford, UK.

Bode, M., K. Wilson, et al. 2008. Cost-effective globalconservation spending is robust to taxonomic group.Proc. Natl. Acad. Sci. 105: 6498–6501.

Bottrill, M., L.N. Joseph, et al. 2008. Is conservation triagejust smart decision-making? Trends Ecol. Evol. 23:649–654.

Brooks, T.M., R.A. Mittermeier, et al. 2006. Global bio-diversity conservation priorities. Science 313: 58–61.

260 Annals of the New York Academy of Sciences

Burgman, M.A., H.P. Possingham, et al. 2001. A methodfor setting the size of plant conservation target areas.Conserv. Biol. 15: 603–616.

Caldecott, J.O., M.D. Jenkins, et al. 1996. Priorities forconserving global species richness and endemism.Biodivers. Conserv. 5: 699–727.

Carwardine, J., C.J. Klein, et al. Hitting the target andmissing the point: target-based conservation plan-ning in context. Conserv. Lett. In Press.

Carwardine, J., K.A. Wilson, et al. 2008a. Cost-effectivepriorities for global mammal conservation. Proc. Natl.

Acad. Sci. 105: 11446–11450.Carwardine, J., K.A. Wilson, et al. 2008b. Avoiding costly

conservation mistakes: the importance of definingactions and costs in spatial priority setting. PLoS One

3: e2586.Cassman, K.G. & S. Wood. 2005. Cultivated Systems.

In Ecosystems and Human Well-Being: Current State and

Trends, Vol. 1. R. Hassan, R. Scholes & N. Ash, Eds.Island Press. Washington, DC.

Ceballos, G. & P.R. Ehrlich. 2006. Global mammal dis-tributions, biodiversity hotspots, and conservation.Proc. Natl. Acad. Sci. 103: 19374–19379.

Chan, K.M.A., M.R. Shaw, et al. 2006. ConservationPlanning for Ecosystem Services. PLoS Biol. 4: 379.

Chandrashekara, U.M. & S. Sankar. 1998. Ecology andmanagement of sacred groves in Kerala, India. For.

Ecol. Manage. 112: 165–177.Chape, S., J. Harrison, et al. 2005. Measuring the extent

and effectiveness of protected areas as an indicatorfor meeting global biodiversity targets. Philos. Trans.

R. Soc. Lond. B Biol. Sci. 360: 443–455.Church, R. & C. ReVelle. 1974. The maximal coverage

location problem. Pap. Reg. Sci. Assoc. 32: 101–118.Clemen, R.T. 1996. Making Hard Decisions: An Introduction to

Decision Analysis. Duxbury Press. Pacific Grove, CA.Conservation International. 1990. The Rainforest Imperative.

Conservation International. Washington, DC.Costello, C. & S. Polasky. 2004. Dynamic reserve site

selection. Resour. Energy Econ. 26: 157–174.Curran, L.M., S.N. Trigg, et al. 2004. Lowland forest loss

in protected areas in Indonesian Borneo. Science 303:1000–1003.

Davis, F.W., C. Costello, et al. 2006. Efficient conservationin a utility-maximization framework. Ecol. Soc. 11:33 [on-line] URL: http://www.ecologyandsociety.org/vol11/iss1/art33/.

Dinerstein, E., G. Powell, et al. 2000. A Workbook for Con-

ducting Biological Assessments and Developing Biodiversity

Visions for Ecoregion-Based Conservation. Part 1: Terrestrial

Ecoregions. Conservation Science Program, WWF-USA. Washington, DC.

Drechsler, M. 2005. Probabilistic approaches to schedul-ing reserve selection. Biol. Conserv. 122: 253–262.

Dutton, I.M. 2001. Engaging communities as partnersin conservation and development. van Zorge Report on

Indonesia III: 24–32.Egoh, B., M. Rouget, et al. 2007. Integrating ecosystem

services into conservation assessments: a review. Ecol.

Econ. 63: 714–721.Eken, G., L. Bennun, et al. 2004. Key biodiversity areas as

site conservation targets. Bioscience 54: 1110–1118.Elster, J. 2003. Pascal and decision theory. In The Cam-

bridge Companion to Pascal. N. Hammond, Ed. Cam-bridge University Press. Cambridge.

Evans, M.I. 1994. Important Bird Areas: Priority Sites for Con-

servation. Birdlife International. Cambirdge, UK.Faith, D.P. 1994. Phylogenetic diversity: a general frame-

work for the prediction of feature diversity. In Sys-

tematics and Conservation Evaluation. P.L. Forey, C.J.Humphries & R.I. Vane-Wright, Eds.: 251–268.Clarendon Press. Oxford.

Faith, D.P., S. Ferrier, et al. 2004. The ED strategy: howspecies-level surrogates indicate general biodiversitypatterns through an “environmental diversity” per-spective. J. Biogeogr. 31: 1207–1217.

FAO. 2008. Food and Agriculture Organisation of theUnited Nations. Statistics Database (FAOSTAT).URL: http://faostat.fao.org. Date accessed: January2009.

Fernandes, L., M.A. Ridgley, et al. 1999. Multiple criteriaanalysis integrates economic ecological and socialobjectives for coral reef managers. Coral Reefs 18:393–402.

Ferraro, P.J. & S. Pattanayak. 2006. Money for nothing?A call for empirical evaluation of biodiversity con-servation investments. PLoS Biol. 4: 482–488.

Ferrier, S., R.L. Pressey, et al. 2000. A new predictor of theirreplaceability of areas for achieving a conservationgoal, its application to real-world planning, and aresearch agenda for further refinement. Biol. Conserv.

93: 303–325.Field, S.A., P. O’Connor, et al. 2007. Making monitoring

meaningful. Aust. Ecol. 32: 485–491.Game, E.T., M.E. Watts, et al. 2008. Planning for persis-

tence in marine reserves: a question of catastrophicimportance. Ecol. Appl. 18: 670–680.

Gardner, T.A., J. Barlow, et al. 2008. The cost-effectivenessof biodiversity surveys in tropical forests. Ecol. Lett.

11: 139–150.Grantham, H.S., A. Moilanen, et al. 2008. Diminishing

return on investment for biodiversity data in conser-vation planning. Conserv. Lett. 1: 190–198.

Groombridge, B. & M.D. Jenkins (Eds.) 2002. World Atlas

of Biodiversity: Earth’s Living Resources in the 21st Century.University of California Press. Berkeley, CA.

Groves, C.R., D.B. Jensen, et al. 2002. Planning for bio-diversity conservation: putting conservation scienceinto practice. Bioscience 52: 499–512.

Wilson et al.: Setting Conservation Priorities 261

Haight, R.G., S.A. Snyder, et al. 2005. Metropolitan open-space protection with uncertain site availability. Con-

serv. Biol. 19: 327–337.Hajkowicz, S.A., A. Higgins, et al. 2007. Optimisation

and the selection of conservation contracts. Aust. J.

Agric. Resour. Econ. 51: 39–56.Hajkowicz, S.A., G.T. McDonald, et al. 2000. An

evaluation of multiple objective decision supportweighting techniques in natural resource man-agement. J. Environ. Plann. Manage. 43: 505–518.

Halpern, B.S., C.R. Pyke, et al. 2006. Gaps and mis-matches between global conservation priorities andspending. Conserv. Biol. 20: 56–65.

Hartmann, K. & M. Steel. 2006. Maximizing phyloge-netic diversity in biodiversity conservation: Greedysolutions to the Noah’s Ark problem. Syst. Biol. 55:644–651.

Hauser, C.E. & H.P. Possingham. 2008. Experimental orprecautionary? Adaptive management over a rangeof time horizons. J. Appl. Ecol. 45: 72–81.

Hiedanpaa, J. 2002. European-wide conservation versuslocal well-being: the reception of the Natura 2000Reserve Network in Karvia, SW-Finland. Landscape

Urban Plann. 61: 113–123.Higgins, T.L. & T.P. Duane. 2008. Incorporating com-

plex adaptive systems theory into strategic planning:The Sierra Nevada Conservancy. J. Environ. Plann.

Manage. 51: 141–162.Hobbs, R.J. & L.J. Kristjanson. 2003. Triage: How do we

prioritize health care for landscapes? Ecol. Manage.

Restor. 4: 39–45.Hoekstra, J.M., T.M. Boucher, et al. 2005. Confronting

a biome crisis: global disparities of habitat loss andprotection. Ecol. Lett. 8: 23–29.

Howard, P.C., P. Viskanic, et al. 1998. Complementarityand the use of indicator groups for reserve selectionin Uganda. Nature 394: 472–475.

Isaac, N.J., S.T. Turvey, et al. 2007. Mammals on theEDGE: conservation priorities based on threat andphylogeny. PLoS ONE 2: e296.

IUCN (International Union for Conservation of Natureand Natural Resources). 2003. 2003 IUCN Red List of

Threatened Species. IUCN Species Survival Commis-sion. Gland and Cambridge.

James, A., K. Gaston, et al. 2001. Can we afford to con-serve biodiversity? Bioscience 51: 43–52.

Jepson, P., J.K. Jarvie, et al. 2001. The end of Indonesia’slowland forests? Science 292: 859–861.

Joseph, L.N., R.F. Maloney, et al. Improving methods forallocating resources among threatened species: thecase for a new national approach in New Zealand.Pac. Conserv. Biol. In Press-a.

Joseph, L.N., R.F. Maloney, et al. Evaluating costs, bene-fits and probability of success for threatened species

management: a project prioritisation protocol. Con-

serv. Biol. In Press-b.Justus, J. & S. Sarkar. 2002. The principle of complemen-

tarity in the design of reserve networks to conservebiodiversity: a preliminary history. J. Biosci. 27: 421–435.

Klein, C., A. Chan, et al. 2008. Striking a balance betweenbiodiversity conservation and socioeconomic viabil-ity in marine protected area design. Conserv. Biol. 22:691–700.

Knight, A.T. 2006. Failing but learning: writing thewrongs after Redford and Taber. Conserv. Biol. 20:1312–1314.

Knight, A.T. & R.M. Cowling. 2007. Embracing oppor-tunism in the selection of priority conservation areas.Conserv. Biol. 21: 1124–1126.

Knight, F.H. 1921. Risk, Uncertainty, and Profit. HoughtonMifflin. Boston.

Lambeck, R.J. 1997. Focal species: a multi-species um-brella for nature conservation. Conserv. Biol. 11: 849–856.

Landres, P.B., J. Verber, et al. 1988. Ecological uses ofvertebrate indicator species: a critique. Conserv. Biol.

2: 316–328.Lawton, J.H. & R.M. May. 1995. Extinction Rates. Oxford

University Press. Oxford.Lindenmayer, D.B., A.D. Manning, et al. 2002. The focal-

species approach and landscape restoration: a cri-tique. Conserv. Biol. 16: 338–345.

Luck, G.W., T.H. Ricketts, et al. 2004. Alleviating spatialconflict between people and biodiversity. Proc. Natl.

Acad. Sci. 101: 182–186.Mace, G.M., H.P. Possingham, et al. 2007. Prioritising

choices in conservation. In Key Topics in Conservation

Biology. D. MacDonald & K. Service, Eds. OxfordUniversity Press. Oxford.

Margules, C. & S. Sarkar. 2007. Systematic Conservation Plan-

ning. Cambridge University Press. Cambridge.Margules, C.R. & R.L. Pressey. 2000. Systematic conser-

vation planning. Nature 405: 243–253.Marsh, H., A. Dennis, et al. 2007. Optimizing allocation

of management resources for wildlife. Conserv. Biol.

21: 387–399.Martello, S. & P. Toth. 1990. Knapsack Problems: Algo-

rithms and Computer Implementations. John Wiley & Sons.Chichester, UK.

McBride, M., K.A. Wilson, et al. 2007. Incorporatingthe effects of socioeconomic uncertainty into prioritysetting for conservation investment. Conserv. Biol. 21:1463–1474.

McCarthy, M.A. & H.P. Possingham. 2007. Active adap-tive management for conservation. Conserv. Biol. 21:956–963.

Meijaard, E. & D. Sheil. 2007. A logged forest in Borneois better than none at all. Nature 446: 974.

262 Annals of the New York Academy of Sciences

Meijaard, E. & D. Sheil. 2008. The persistence and con-servation of Borneo’s mammals in lowland forestsmanaged for timber: observations, overviews and op-portunities. Ecol. Res. 23: 21–34.

Meijaard, E., D. Sheil, et al. 2006. Wildlife conservationin Bornean timber concessions. Ecol. Soc. 11: 47.

Meir, E., S. Andelman, et al. 2004. Does conservationplanning matter in a dynamic and uncertain world?Ecol. Lett. 7: 615–622.

Mendoza, G.A. & H. Martins. 2006. Multi-criteria deci-sion analysis in natural resource management: a criti-cal review of methods and new modelling paradigms.For. Ecol. Manage. 230: 1–22.

Mittermeier, R.A., C.G. Mittermeier, et al. 2003. Wilder-ness and biodiversity conservation. Proc. Natl. Acad.

Sci. 100: 10309–10313.Mittermeier, R.A., N. Myers, et al. 1998. Biodiversity

hotspots and major tropical wilderness areas: ap-proaches to setting conservation priorities. Conserv.

Biol. 12: 516–520.Mittermeier, R.A., P. Robles Gil, et al. 2004. Hotspots:

Revisited. CEMEX. Mexico. Mexico City.Moffett, A. & S. Sarkar. 2006. Incorporating multiple cri-

teria into the design of conservation area networks:a minireview with recommendations. Divers. Distrib.

12: 125–137.Moore, J., A. Balmford, et al. 2004. Integrating costs into

conservation planning across Africa. Biol. Conserv.

117: 343–350.Murdoch, W., S. Polasky, et al. 2007. Maximising return

on investment in conservation. Biol. Conserv. 139:375–388.

Myers, N., R.A. Mittermeier, et al. 2000. Biodiversityhotspots for conservation priorities. Nature 403: 853–858.

Naidoo, R., A. Balmford, et al. 2008. Global mapping ofecosystem services and conservation priorities. Proc.

Natl. Acad. Sci. 105: 9495–9500.Naidoo, R., A. Balmford, et al. 2006. Integrating eco-

nomic costs into conservation planning. Trends Ecol.

Evol. 21: 681–687.Naidoo, R. & T. Iwamura. 2007. Global-scale mapping

of economic benefits from agricultural lands: impli-cations for conservation priorities. Biol. Conserv. 140:40–49.

Nakagawa, M., H. Miguchi, et al. 2006. The effects ofvarious forest uses on small mammal communitiesin Sarawak, Malaysia. For. Ecol. Manage. 231: 55–62.

Nalle, D.J., C.A. Montgomery, et al. 2004. Modeling jointproduction of wildlife and timber. J. Environ. Econ.

Manage. 48: 997–1017.Nicholson, E. & H.P. Possingham. 2007. Making con-

servation decisions under uncertainty for the persis-tence of multiple species. Ecol. Appl. 17: 251–265.

Noss, R.F. 1990. Indicators for monitoring biodiversity: ahierarchical approach. Conserv. Biol. 4: 355–364.

Noss, R.F. 2004. Conservation targets and informationneeds for regional conservation planning. Nat. Areas

J. 24: 223–231.Odling-Smee, L. 2005. Conservation: dollars and sense.

Nature 437: 614–616.Oliver, I., A.J. Beattie, et al. 1998. Spatial fidelity of plant,

vertebrate, and invertebrate assemblages in multiple-use forest in Eastern Australia. Conserv. Biol. 12: 822–825.

Olson, D., E. Dinerstein, et al. 1998. Freshwater Biodiversity of

Latin America and the Caribbean: A Conservation Assessment.Biodiversity Support Program. Washington, DC.

Olson, D.M. & E. Dinerstein. 1998. The Global 200: arepresentation approach to conserving the Earth’smost biologically valuable ecoregions. Conserv. Biol.

12: 502–515.Olson, D.M. & E. Dinerstein. 2002. The Global 200:

Priority ecoregions for global conservation. Ann. Mo.

Bot. Gard. 89: 199–224.Orme, C.D.L., R.G. Davies, et al. 2005. Global hotspots

of species richness are not congruent with endemismor threat. Nature 436: 1016–1019.

Paine, R.T. 1969. A note on trophic complexity and com-munity stability. Am. Nat. 103: 91–93.

Pence, G., M.A. Botha, et al. 2003. Evaluating combina-tions of on-and off-reserve conservation strategies forthe Agulhas Plain, South Africa: a financial perspec-tive. Biol. Conserv. 112: 253–273.

Pimm, S.L. 2000. Against triage. Science 289: 2289.Polasky, S., J.D. Camm, et al. 2001. Selecting biological

reserves cost effectively: an application to terrestrialvertebrate conservation in Oregon. Land Econ. 77:68–78.

Possingham, H.P., S.J. Andelman, et al. 2002. Limits tothe use of threatened species lists. Trends Ecol. Evol.

17: 503–507.Possingham, H.P., S.J. Andelman, et al. 2001. Making

smart conservation decisions. In Research Priorities for

Conservation Biology. G. Orians & M. Soule, Eds.: 225–244. Island Press. Washington, DC.

Possingham, H.P., K.A. Wilson, et al. 2006. Protected ar-eas: goals, limitations, and design. In Principles of Con-

servation Biology. M.J. Groom, G.K. Meffe & C.R.Carroll, Eds: 509–533. Sinauer Associates Inc. Sun-daland, MA.

Power, M.E., D. Tilman, et al. 1996. Challenges in thequest for a keystone. Bioscience 46: 609–620.

Pressey, R.L. 2002. The first reserve selection algorithm—A retrospective on Jamie Kirkpatrick’s 1983 paper.Prog. Phys. Geogr. 26: 434–441.

Pressey, R.L., M. Cabeza, et al. 2007. Conservation plan-ning in a changing world. Trends Ecol. Evol. 22: 583–592.

Wilson et al.: Setting Conservation Priorities 263

Pressey, R.L. & A.O. Nicholls. 1989. Efficiency in conser-vation planning: scoring versus iterative approaches.Biol. Conserv. 50: 199–218.

Redford, K.H., P. Coppolillo, et al. 2003. Mapping theconservation landscape. Conserv. Biol. 17: 116–131.

Redford, K.H. & A. Taber. 2000. Writing the wrongs:developing a safe-fail culture in conservation. Conserv.

Biol. 14: 1567–1568.Reichelderfer, K. & W.G. Boggess. 1988. Government

decision making and program performance: the caseof the Conservation Reserve Program. Am. J. Agric.

Econ. 70: 1–11.Reyers, B. & A.S. van Jaarsveld. 2000. Assessment tech-

niques for biodiversity surrogates. S. Afr. J. Sci. 96:406–408.

Ricketts, T.H., E. Dinerstein, et al. 2005. Pinpointing andpreventing imminent extinctions. Proc. Natl. Acad. Sci.

102: 18497–18501.Ricketts, T.H., E. Dinerstein, et al. 1999. Terrestrial Ecore-

gions of North America: A Conservation Assessment. IslandPress. Washington, DC.

Roberge, J.M. & P. Angelstam. 2004. Usefulness of theumbrella species concept as a conservation tool. Con-

serv. Biol. 18: 76–85.Rodrigues, A.S. & T.M. Brooks. 2007. Shortcuts for bio-

diversity conservation planning: the effectiveness ofsurrogates. Annu. Rev. Ecol. Evol. Syst. 38: 717–737.

Rodrıguez, J.P., F. Rojas-Suarez, et al. 2004. Setting prior-ities for the conservation of Venezuela’s threatenedbirds. Oryx 38: 373–382.

Rouget, M., D.M. Richardson, et al. 2003. Current pat-terns of habitat transformation and future threatsto biodiversity in terrestrial ecosystems of the CapeFloristic Region, South Africa. Biol. Conserv. 112: 63–85.

Rubinoff, D. 2001. Evaluating the California Gnatcatcheras an umbrella species for conservation of southernCalifornia coastal sage scrub. Conserv. Biol. 15: 1374–1383.

Sanderson, E.W., M. Jaiteh, et al. 2002. The human foot-print and the last of the wild. Bioscience 52: 891–904.

Sarkar, S., R.L. Pressey, et al. 2006. Biodiversity conser-vation planning tools: present status and challengesfor the future. Annu. Rev. Environ. Resour. 31: 123–159.

Shrader-Frechette, K.S. & E.D. McCoy. 1993. Method in

Ecology. Strategies for Conservation. Cambridge Univer-sity Press. Cambridge.

Simberloff, D. 1998. Flagships, umbrellas, and keystones:is single-species management passe in the landscapeera. Biol. Conserv. 83: 247–257.

Smith, P.G.R. & J.B. Theberge. 1987. Evaluating naturalareas using multiple criteria: theory and practice.Environ. Manage. 11: 447–460.

Smith, R.J., R.D.J. Muir, et al. 2003. Governance and theloss of biodiversity. Nature 426: 67–70.

Soule, M.E. & M.A. Sanjayan. 1998. Ecology—Conservation targets: Do they help? Science 279:2060–2061.

Stattersfield, A.J., M.J. Crosby, et al. 1998. Endemic Bird

Areas of the World: Priorities for Biodiversity Conservation.Birdlife International. Cambridge, MA.

Stewart, R.R., T. Noyce, et al. 2003. Opportunity cost ofad hoc marine reserve design decisions: an examplefrom South Australia. Mar. Ecol. Prog. Ser. 253: 25–38.

Stewart, R.R. & H.P. Possingham. 2005. Efficiency, costsand tradeoffs in marine reserve design. Environ. Model.

Assess. 10: 203–213.Stoneham, G., V. Chaudhri, et al. 2003. Auctions for con-

servation contracts: an empirical examination of Vic-toria’s BushTender trial. Aust. J. Agric. Resour. Econ. 47:477–500.

The Nature Conservancy. 2000. Conservation by Design: A

Framework for Mission Success. The Nature Conser-vancy. Arlington, VA.

The Nature Conservancy. 2003. The Five-S Framework for

Site Conservation: A Practitioner’s Handbook for Site Con-

servation Planning and Measuring Conservation Success, 3rded. Vols. I and II. The Nature Conservancy. Arling-ton, VA.

Turner, W.R., K. Brandon, et al. 2007. Global conserva-tion of biodiversity and ecosystem services. Biosicence

57: 868–873.Underwood, E.C., M.R. Shaw, et al. 2008. Protecting bio-

diversity when money matters: Maximizing returnon investment. PLoS One 3: e1515.

van Jaarsveld, A.S., S. Freitag, et al. 1998. Biodiversityassessment and conservation strategies. Science 279:2106–2108.

Van Teeffelen, A.J.A. & A. Moilanen. 2008. Where andhow to manage: optimal selection of conservationactions for multiple species. Biodivers. Inf. 5: 1–13.

Vane-Wright, R.I., C.R. Smith, et al. 1994. Systematicassessment of taxic diversity by summation. In Sys-

tematics and Conservation Evaluation. P.L. Forey, C.J.Humphries & R.I. Vane-Wright, Eds.: 309–326. Ox-ford University Press. Oxford.

Villa, F., L. Tunsei, et al. 2002. Zoning marine protectedareas through spatial multiple-criteria analysis: thecase of the Asinara Island National Marine Reserveof Italy. Conserv. Biol. 16: 515–526.

Walters, C. 1986. Adaptive Management of Renewable Re-

sources. Blackburn Press. Caldwell.Walters, J.R., S.R. Derrickson, et al. 2008. Status of the

California Condor and Efforts to Achieve its Recovery.The American Ornithologists’ Union and Audubon.San Francisco, CA.

WDPA Consortium. 2004. World Database on Protected Ar-

eas. World Conservation Union (IUCN) and UNEP-World Conservation Monitoring Centre (UNEP-WCMC).

264 Annals of the New York Academy of Sciences

Weitzman, M.L. 1998. The Noah’s Ark problem. Econo-

metrica 66: 1279.Wells, K., E.K.V. Kalko, et al. 2007. Effects of rain forest

logging on species richness and assemblage composi-tion of small mammals in Southeast Asia. J. Biogeogr.

34: 1087–1099.Wilson, K.A., M.F. McBride, et al. 2006. Prioritis-

ing global conservation efforts. Nature 440: 337–340.

Wilson, K.A., R.L. Pressey, et al. 2005. Measuring and

incorporating vulnerability into conservation plan-ning. Environ. Manage. 35: 527–543.

Wilson, K.A., E.C. Underwood, et al. 2007. Conservingbiodiversity efficiently: what to do, where and when.PLoS Biol. 5: e223.

Wright, R.G. & D.J. Mattson. 1996. The origin and pur-pose of national parks and protected areas. In National

Parks and Protected Areas: Their Role in Environmental Pro-

tection. R.G. Wright, Ed.: 3–14. Blackwell Science.Cambridge, MA.