Contributed Paper

Current practices in the identification of criticalhabitat for threatened speciesAbbey E. Camaclang,∗ †† Martine Maron,† Tara G. Martin,‡ and Hugh P. Possingham∗§∗∗∗School of Biological Sciences, The University of Queensland, Brisbane, QLD 4072, Australia†School of Geography, Planning, and Environmental Management, The University of Queensland, Brisbane, QLD 4072, Australia‡CSIRO Land and Water, 41 Boggo Road, Dutton Park, QLD 4102, Australia§School of Mathematics and Physics, The University of Queensland, Brisbane, QLD 4072, Australia∗∗Department of Life Sciences, Imperial College London, Silwood Park, Ascot SL5 7PY, Berkshire, England, United Kingdom

Abstract: The term critical habitat is used to describe the subset of habitat that is essential to the survival andrecovery of species. Some countries legally require that critical habitat of listed threatened and endangeredspecies be identified and protected. However, there is little evidence to suggest that the identification of criticalhabitat has had much impact on species recovery. We hypothesized that this may be due at least partly to amismatch between the intent of critical habitat identification, which is to protect sufficient habitat for speciespersistence and recovery, and its practice. We used content analysis to systematically review critical habitatdocuments from the United States, Canada, and Australia. In particular, we identified the major trends in typeof information used to identify critical habitat and in occupancy of habitat identified as critical. Informationabout population viability was used to identify critical habitat for only 1% of the species reviewed, and formost species, designated critical habitat did not include unoccupied habitat. Without reference to populationviability, it is difficult to determine how much of a species’ occupied and unoccupied habitat will be requiredfor persistence. We therefore conclude that the identification of critical habitat remains inconsistent with thegoal of protecting sufficient habitat to support persistence and recovery of the species. Ensuring that criticalhabitat identification aligns more closely with its intent will improve the accuracy of the designations andmay therefore help improve the benefits to species recovery when combined with adequate implementationand enforcement of legal protections.

Keywords: compliance, conservation policy, Endangered Species Act, Environment Protection and BiodiversityConservation Act, habitat protection, recovery plan, Species at Risk Act

Practicas Actuales en la Identificacion del Habitat Crıtico para las Especies Amenazadas

Resumen: El termino habitat crıtico se usa para describir al subconjunto de habitats que son esencialespara la supervivencia y recuperacion de las especies. Algunos paıses requieren legalmente que se identifiquey proteja el habitat crıtico de las especies enlistadas como amenazadas o en peligro. Sin embargo, existepoca evidencia que sugiera que la identificacion del habitat crıtico haya tenido mucho impacto sobrela recuperacion de especies. Planteamos la hipotesis de que esto tal vez se debe, por lo menos en parte, auna discordancia entre el intento de identificar al habitat crıtico, que es proteger el habitat suficiente para lapersistencia y recuperacion de la especie, y su puesta en practica. Usamos el analisis de contenido para revisarsistematicamente documentos de habitat crıtico de los Estados Unidos, Canada y Australia. Identificamos enparticular a las principales tendencias de tipo de informacion usada para identificar habitats crıticos y alas de ocupacion de habitats identificados como crıticos. La informacion sobre la viabilidad de la poblacionse uso para identificar el habitat crıtico de solamente el 1% de las especies revisadas, y para la mayorıa delas especies, el habitat crıtico designado no incluıa habitats desocupados. Sin referencias de viabilidad depoblacion, es difıcil determinar cuanto del habitat ocupado y desocupado de una especie se necesitara parala persistencia. Por eso concluimos que la identificacion del habitat crıtico sigue siendo inconsistente con el

††email abbey.camaclang@uqconnect.edu.auPaper submitted January 6, 2014; revised manuscript accepted August 17, 2014.

482Conservation Biology, Volume 29, No. 2, 482–492C© 2014 Society for Conservation BiologyDOI: 10.1111/cobi.12428

Camaclang et al. 483

objetivo de proteger el habitat suficiente para asegurar la persistencia y recuperacion de la especie. Asegurarque la identificacion de habitats crıticos se asocie mas de cerca con su intento mejorara la certeza de lasdesignaciones y por lo tanto puede ayudar a mejorar los beneficios para la recuperacion de la especie cuandose combinen con la implementacion y aplicacion adecuada de protecciones legales.

Palabras Clave: Acta de la Conservacion de la Biodiversidad y la Proteccion Ambiental, Acta de las Especies enPeligro, Acta de las Especies en Riesgo, cumplimiento, plan de recuperacion, polıtica de conservacion, proteccionde habitat

Introduction

Habitat protection plays a central role in the conserva-tion of biodiversity. In recognition of this, endangeredspecies legislation in a number of countries, such as theUnited States, Australia, and Canada, provides protectionfor the portion of a species’ habitat considered “essentialto the conservation” (Endangered Species Act of 1973[ESA], 16 USC §§ 1531–1544, 2013), “critical to survival”(Environment Protection and Biodiversity ConservationAct 1999 [EPBC] [Cth]), or “necessary for the survivalor recovery” (Species at Risk Act [SARA], SC 2002, c 29)of listed threatened and endangered species. Referred toas critical habitat, these areas are typically protected byprohibiting activities that may result in “adverse modifi-cation” of or that may “significantly damage” or “destroyany part of” the critical habitat.

Before critical habitat can be protected, however, itmust be identified and legally designated. To improvescientific validity and consistency with the underlyingintent of critical habitat designation, biologists recom-mend that critical habitat be defined operationally as thehabitat required to ensure the persistence of a speciesor population (Murphy & Noon 1991; Hall et al. 1997;Rosenfeld & Hatfield 2006). In this sense, critical habitat isthe minimum subset required for persistence (Rosenfeld& Hatfield 2006), and changes in habitat quality out-side critical habitat should have relatively less impacton species persistence. Based on this definition, the pri-mary criteria for identifying critical habitat should bespecies persistence, as evaluated in terms of the accept-able threshold extinction risk, population size, or numberof patches needed to achieve viable populations over aspecified time horizon (Reed et al. 2006; Rosenfeld &Hatfield 2006). These thresholds and time horizons maybe based on predetermined criteria for classifying speciesas at risk of extinction, such as those used for the Inter-national Union for Conservation of Nature’s (IUCN) RedList of Threatened Species (IUCN 2012). In the absenceof spatially explicit assessments of population viabilityover time, estimates of minimum viable population sizes(Shaffer 1981; Traill et al. 2007) may also be used todetermine threshold values. In addition, species-specifictargets for achieving species recovery or delisting, as de-termined during recovery planning (Reed et al. 2006;Rosenfeld & Hatfield 2006), may be used instead as in-

terim goals when the information needed to assess pop-ulation viability is not yet available.

By using contribution to species persistence to iden-tify critical habitat, the quality, size, and spatial config-uration of areas required to achieve and maintain long-term species persistence can be determined (Reed et al.2006; Rosenfeld & Hatfield 2006). For most threatenedand endangered species, these areas include suitable butcurrently unoccupied sites needed for recovery, whererecovery refers to some increase in population size orimprovement in the status of the species necessary toachieve long-term persistence. Habitat availability is animportant factor in determining population size for manyspecies (Fahrig & Merriam 1994; Hanski 1998; Flather &Bevers 2002). However, in metapopulations, some frac-tion of the total habitat remains unoccupied at any giventime due to local extinction and colonization processes(Thomas & Kunin 1999). Unoccupied habitat, therefore,plays an essential role in population persistence. Protec-tion of currently unoccupied habitat ensures that theseareas remain available in the future for recolonization,either through natural processes or through species rein-troductions (Kleiman 1989; Hanski 1998; Rosenfeld &Hatfield 2006). In some cases, inclusion of unoccupiedand degraded habitat for future restoration may also benecessary (Huxel & Hastings 1999), particularly wherehabitat loss and degradation have been the primary causesof a species’ decline.

Evidence for the effectiveness of critical habitat des-ignation to date has been mixed (Schwartz 2008; Gibbs& Currie 2012). Two previous studies revealed a posi-tive correlation between designation of critical habitat inthe United States and the likelihood that the populationtrend of a species is stable or improving (Rachlinski 1997;Taylor et al. 2005). However, neither study accountedfor the influence of government spending on speciesrecovery (Kerkvliet & Langpap 2007). When estimatesof species-specific government spending were included,critical habitat designation under the ESA did not appearto have a significant effect on recovery (e.g., Male &Bean 2005; Kerkvliet & Langpap 2007; Gibbs & Currie2012). The influence of critical habitat identification onspecies recovery in Australia and Canada has not yet beenevaluated.

We hypothesized that the absence of a clear positiveeffect of critical habitat designation on species recovery

Conservation BiologyVolume 29, No. 2, 2015

484 Identifying Critical Habitat

is due at least partly to the mismatch between its intentand actual practice. However, despite extensive debateson the legal and economic implications of critical habitatdesignation (e.g., Darin 2000; Sinden 2004; Owen 2012),its effectiveness on species recovery (e.g., Clark et al.2002; Taylor et al. 2005; Gibbs & Currie 2012), and thefactors influencing the likelihood of designation or useof biological criteria (e.g., Hoekstra et al. 2002; Hodges& Elder 2008; Taylor & Pinkus 2013), we are unawareof previous studies that reviewed the practice of criticalhabitat identification itself. To address this knowledgegap, we reviewed species recovery plans and regulatorydocuments that describe how critical habitats were iden-tified in the United States, which has the longest his-tory of critical habitat identification for threatened andendangered species, and Australia and Canada, both ofwhich have adopted legal provisions for critical habi-tat similar to those of the United States. In particular,we determined the frequency that predictions of speciespersistence were used to identify critical habitats and theextent to which both occupied and unoccupied habitatswere included as part of critical habitat. We aimed todetermine whether the practice of critical habitat identi-fication is consistent with its intent and considered thepotential implications of our findings for the effectivenessof critical habitat identification.

Methods

To examine current practices in critical habitat identi-fication, we conducted a systematic review of officialdocuments that identify and describe critical habitat forindividual species. We used content analysis to extractthe relevant information from the documents on howcritical habitat was identified. Content analysis is usedwidely in the social sciences and humanities (Krippen-dorff 2004) and in health sciences research (Hsieh &Shannon 2005) to analyze the content of text docu-ments for the purpose of making inferences (Krippen-dorff 2004). More recently, content analysis and relatedtechniques have been used in ecology and conservationto review and classify information from texts such asscientific literature, survey questionnaires, or interviewtranscripts (e.g., Bottrill et al. 2011; Fischer & Onyango2012; Martin et al. 2012a). For instance, content analysiswas used to extract information on the objectives, im-plementation, and outcomes of terrestrial conservationprojects in Samoa (Bottrill et al. 2011). Similarly, a study ofthe global distribution and context of terrestrial ecologi-cal studies included a content analysis of site descriptions,provided in the methods sections of reviewed papers, tocategorize field sites into one of 3 different types (Martinet al. 2012a).

The classification of information from documents isachieved through the process of coding, which assigns

a code—a word or a short phrase—that captures theessence or attribute of a particular unit of text (Saldana2009). Clear definitions for each code and a set of instruc-tions or decision rules must be developed to guide theconsistent and repeatable classification of the textual data(White & Marsh 2006). Codes that share similar attributesor address the same issue may then be grouped into largercategories, which may be further organized into broaderthemes or concepts (Saldana 2009).

We used emergent coding (Haney et al. 1998), guidedby foreshadowing or open research questions (White &Marsh 2006), to develop codes that describe and classifyinformation in the reviewed documents. In emergentcoding, codes and categories are established followingsome preliminary examination of the texts (Haney et al.1998). To do this, we selected a random sample of listedspecies from each country. A single coder reviewed thecritical habitat documents for these species, with partic-ular focus on developing codes to classify the type ofinformation critical habitat identification was based on,the occupancy of sites identified as critical habitat, andthe criterion used to select the subset of habitat identifiedas critical. The codes were refined by conducting anotherreview of the documents, this time classifying texts withthe codes developed and revising the definitions of eachcode as needed to improve the consistency or reliabilityof classification. We repeated this process with the samesample set until we were reasonably confident of theconsistency of the resulting classification. The final list ofcodes and their definitions used in the content analysisare available in Supporting Information.

Once the coding scheme was developed, we selecteda larger sample of species for the formal content analy-sis than was used for the preliminary review. For eachcountry, 15 species from each of the major taxonomicgroups (mammals, birds, reptiles and amphibians, fishes,invertebrates, and plants) were randomly selected. Forgroups with <15 species for which critical habitat hadbeen identified, all species in the group were includedin the review. We limited the selection to species withpublicly available critical habitat documents publishedbetween 2003 and 2012 to reduce potential biases relatedto changes in regulatory practice in the United States andimprovements in data collection and analyses capabili-ties. This period also corresponds to when informationon species’ critical habitats began to appear in recoveryplans in Canada and Australia. Critical habitat documentsconsisted of proposed and final rules for critical habi-tat designation published in the U.S. Federal Register(USFWS 2013a), Canadian recovery strategies and actionplans with information on critical habitats (Governmentof Canada 2012), and Australian recovery plans identify-ing the habitats critical to species survival (Departmentof the Environment 2013). In total, critical habitat doc-uments for 218 species—90 from the United States, 78from Australia, and 50 from Canada—were included in

Conservation BiologyVolume 29, No. 2, 2015

Camaclang et al. 485

the review. A complete list of the documents reviewedis available in Supporting Information.

We conducted the formal content analysis by review-ing the documents and assigning codes to classify dif-ferent categories of information about critical habitatsbased on our coding scheme (Supporting Information).In addition to reviewing the critical habitat narrative, wealso recorded information about the taxonomic group(taxa), geographic range (extent), distribution (distribu-tion), movement (dispersal), specificity of habitat re-quirements (specificity), primary threats (threat), andconservation status (status) of the species, as well as thecountry of designation (country) and the existence ofprior court rulings or settlements on critical habitat desig-nation (litigation), to determine whether these attributesaffected how critical habitats were identified (Support-ing Information). We selected these potential covariatesbased on previous studies that reviewed the influence oftaxonomic group, species distributions, degree of threat,recovery potential, and court-ordered status on the likeli-hood of critical habitat designation (Hoekstra et al. 2002)or on the types of criteria used in designation (Hodges &Elder 2008).

We used ensembles of classification trees to evaluatethe relative influence of potential covariates on the typesof information used to identify critical habitats, whichwere re-classified into 2 categories for this analysis: oc-currence data, which includes species occurrence andhabitat feature information, and model-based data, con-sisting of habitat quality, spatial structure, minimum vi-able population size, and spatially explicit populationviability information (Table 1). We also determined therelative influence of covariates on the inclusion of un-occupied habitat as part of the species’ critical habitat.Ensemble methods, such as bagging (Breiman 1996) orrandom forests (Cutler et al. 2007), provide an alterna-tive to some traditional parametric approaches becausethey are not based on assumptions about the distributionof predictor or response variables and can be used tomodel more complex interactions with a greater num-ber of predictor variables (Cutler et al. 2007). In theseapproaches, classification or regression trees are used toexplain the variation in the response variable throughrecursive partitioning, that is, by repeatedly splitting thedata into increasingly homogeneous groups based ondifferent categories or values of the predictor variables(De’ath & Fabricius 2000). Ensemble methods then com-bine many trees to improve the accuracy and stabilityof classification (Breiman 2001) and provide a measureof the relative importance of predictor variables in accu-rately predicting the observed result (Cutler et al. 2007).

We used a conditional permutation scheme, whichreduces selection bias for correlated predictors (Stroblet al. 2008), to aggregate multiple classification treesand evaluate variable importance. The conditional

importance of each variable was estimated in R version2.15.3 (R Core Team 2013) with the cforest function avail-able in the “party” package (Hothorn et al. 2006; Stroblet al. 2008). To parameterize the function, we set ntree =2000 (number of conditional inference trees in the forest)and mtry = 3 (number of randomly preselected splittingvariables), which corresponds with

√p, where p is the

number of predictors included in the model. Once in-fluential variables were identified, we used contingencytables and mosaic plots to explore the relationship be-tween each influential variable and how critical habitatwas identified.

To assess the reliability of content analysis, 2 additionalcoders used the coding scheme we developed (Support-ing Information) to perform the content analysis inde-pendently on 50 of the 218 species. The independentdata sets generated by the 2 additional coders, in combi-nation with the original data set for the same 50 species,were used to evaluate intercoder reliability of eachcategory with Krippendorff’s agreement coefficient α

(Krippendorff 2004). We examined the sensitivity of ourresults to differences in how coders classified informationfrom critical habitat documents by comparing the codingfrequencies obtained by each coder. We also evaluatedthe relative influence of potential covariates separatelyfor each data set to determine how uncertainty in thecontent analysis may have influenced the identificationof important variables. One of the data sets did not havesufficient frequencies in one or more of the categoriesassessed for this analysis, and therefore, only 2 data setswere analyzed.

Results

The contribution of habitat to species persistence, asevaluated with either minimum viable population sizesor spatially explicit population viability, was used toidentify critical habitat for 3 of 218 species (Table 1). Incontrast, species occurrence and habitat features wereused to identify critical habitat for a majority of species,while additional information on habitat quality or spatialstructure was used for around 8% of species (Table 1).Taxonomic group influenced this finding, based onestimates of the conditional variable importance (Fig. 1).Contingency tables and mosaic plots revealed that data onknown occurrences of species or habitat features wereused relatively more often for invertebrates, fishes, andherpetofauna than for plants, birds, and mammals (Fig. 2).Other variables that influenced the information used forcritical habitat identification included country and pre-vious litigation (Fig. 1). Occurrence-based identificationwas used more frequently in Australia and Canada (Fig. 2),whereas in the United States, it was used relatively more

Conservation BiologyVolume 29, No. 2, 2015

486 Identifying Critical Habitat

Table 1. Results of the content analysis of critical habitat documents and the intercoder reliability test.

Intercoder reliability test (n = 50)

number of species

coder coder coderNumber of species (% of total) 1 2 3 Krippendorff’s α

Information typeoccurrence data 0.5

species occurrence 136 (62.4) 30 15 17habitat features 61 (28.0) 11 25 28

model-based datahabitat quality 7 (3.2) 2 4 3spatial structure 11 (5.0) 5 3 0minimum viable population size 2 (0.9) 1 3 1spatially explicit population viability 1 (0.5) 1 0 1

total 218 50 50 50Habitat occupancy

occupied habitat 0.48all occupied 108 (49.5) 23 18 33selected occupied 103 (47.2) 25 30 15none occupied 7 (3.2) 2 2 2total 218 50 50 50

unoccupied habitat 0.38selected unoccupied 62 (28.4) 11 16 9none unoccupied 156 (71.6) 39 34 41total 218 50 50 50

previously occupied habitat 0.27previously occupied 25 (40.3) 2 7 8no previously occupiedhabitat 37 (59.7) 9 9 1total 62 11 16 9

potential habitat 0.41potential habitat 42 (67.7) 10 11 6no potential habitat 20 (32.3) 1 5 3total 62 11 16 9

Selection criteriaoccupied 0.17

disproportionatelyimportant 31 (30.1) 11 15 5consistent use or stable population 32 (31.1) 4 5 0habitat quality 26 (25.2) 6 5 1recovery goals 7 (6.8) 3 3 4ownership or management status 7 (6.8) 1 2 5total 103 25 30 15

previously occupied 1a

recolonization 12 (48.0) 0 2 2reintroduction 7 (28.0) 1 2 2undetected presence 6 (24.0) 1 3 4total 25 2 7 8

potential habitat 0.72a

connectivity 9 (21.4) 3 4 2expansion 21 (50.0) 3 5 1introduction 6 (14.3) 2 1 2undetected presence 6 (14.3) 2 1 1total 42 10 11 6

aHigh values of α agreement may be due to insufficient sample sizes for these categories. Coders identified from 2 to 8 species as having criticalhabitat that included previously occupied habitat and from 6 to 11 species with unoccupied potential habitat.

often when there has been no previous court ruling orsettlement regarding critical habitat designation (Fig. 2).

Critical habitat identification included unoccupiedhabitat for about 28% of the species we reviewed

(Table 1). The inclusion of unoccupied habitat in criticalhabitat identification appeared to be influenced primar-ily by the dispersal ability of the species (Fig. 1). Crit-ical habitat for species with limited mobility was more

Conservation BiologyVolume 29, No. 2, 2015

Camaclang et al. 487

extent

distribution

threat

specificity

status

dispersal

country

litigation

taxa

0.0000 0.0010 0.0020

Information type usedC

ovar

iate

s

(Mean decrease in accuracy)Conditional variable importance score

threat

status

specificity

extent

distribution

litigation

country

taxa

dispersal

0.000 0.010 0.020

Unoccupied habitat included

(Mean decrease in accuracy)Conditional variable importance score

(a) (b)Figure 1. Relative influence ofcovariates on (a) the type ofinformation used to identifycritical habitat and (b) theinclusion of unoccupied habitatas part of critical habitatdesignation, both estimated withthe variable importance scorefrom conditional inference trees.Higher scores indicate greaterinfluence of the predictorvariable. The vertical dashed lineindicates the absolute value ofthe lowest score, which is used asa threshold value for evaluatingvariable significance.

likely to include unoccupied habitat (Fig. 3). Taxonomicgroup and the country of designation also influenced theinclusion of unoccupied sites as critical habitat (Fig. 1).Plants were relatively more likely to have unoccupiedcritical habitat identified than invertebrates (even oncemovement abilities were accounted for), whereas wide-ranging birds and mammals were more likely than othervertebrate taxa to have critical habitats consisting only ofknown occupied sites (Fig. 3). Species in Canada werealso less likely to have unoccupied habitats recognized aspart of designated critical habitat (Fig. 3).

In general, α-agreement values ranged from 0.17 to0.50 (Table 1), indicating greater than chance agree-ment. Reliability values for the content analysis werelower than the recommended minimum threshold valueof 0.667 (Krippendorff 2004), reflecting the difficultiesexperienced by the coders in identifying and interpret-ing critical habitat information. In particular, it was notalways clearly stated whether any unoccupied habitat hadbeen included in critical habitat and what decision ruleswere used to select areas for inclusion. However, for therelevant categories—the information type used in criticalhabitat identification and the inclusion of unoccupiedhabitats—we considered, similar results were obtainedwhen data generated by each coder were analyzed sepa-rately. For instance, content analysis results from each ofthe 3 coders all indicated that information on occurrenceand habitat features were used more often to identifycritical habitat than other data types (Table 1). Similarly,all 3 coders reported that unoccupied habitats were morelikely to be excluded than included as part of criticalhabitat (Table 1). Despite variation in the conditionalimportance scores, the set of covariates ranked as in-fluential was also similar across 2 different data sets(Fig. 4). These results would suggest that the generaltrends identified are somewhat robust to low intercoderreliability.

Discussion

Limited Role of Species Persistence in Critical HabitatIdentification

Using species persistence as the criterion to identify crit-ical habitat requires information about the amount andquality of habitats, the relationship between habitat andpopulation viability, the target population size or distri-bution needed for long-term population viability, and theamount and location of habitat needed to achieve thedesired targets (Pulliam & Danielson 1991; Rosenfeld &Hatfield 2006). However, these data may not be avail-able for most threatened and endangered species (Neelet al. 2012), resulting in the use of less rigorous criteria(Reed et al. 2006) and therefore a limited role for speciespersistence in critical habitat identification.

Our results indicate that data limitation may be themain reason for inadequate identification of critical habi-tat. For instance, critical habitat for most of the speciesreviewed was identified primarily on the basis of knownlocations of populations or particular habitat features.Relatively few species’ critical habitats were identifiedusing habitat suitability models, knowledge of spatialstructure, or estimates of population viability, all of whichrequire additional data on population size, popula-tion processes, and habitat requirements (Akcakaya &Sjogren-Gulve 2000; Guisan et al. 2013). The influence oftaxonomic group on the types of data used also reflectedthe differences in the amount and quality of informa-tion available for each group. Critical habitat for plants,birds, and mammals, which have been relatively wellstudied, were more likely to be identified using additionalinformation on habitat spatial structure, habitat quality, orpopulation viability. Previous studies have proposed sim-ilar explanations for the taxonomic bias in the frequencyof designation (Hoekstra et al. 2002) and the applica-tion of biological criteria to critical habitat identification

Conservation BiologyVolume 29, No. 2, 2015

488 Identifying Critical Habitat

(a) Information used

Taxo

nom

ic g

roup

Pla

ntIn

vF

ish

Her

pM

amm

Bird

Occurrence Model

(b) Information used

Cou

ntry

Aus

tral

iaC

anad

aU

nite

d S

tate

s

Occurrence Model

(c) Information used

Litig

atio

n (f

or U

.S. s

peci

es o

nly)

Yes

No

Occurrence Model

Ο

Ο

Figure 2. Relative proportion of species whose criticalhabitat is identified based on occurrence data (speciesoccurrence and habitat feature information) andmodel-based data (habitat quality, spatial structure,minimum viable population size, and spatiallyexplicit population viability information) relative to(a) taxonomic group, (b) country of designation, and(c) existence (yes or no) of litigation in the UnitedStates over critical habitat. The area of each barrepresents the proportion of observations in eachvariable combination; zero observations arerepresented by a line and circle.

(Hodges & Elder 2008), although these differences mayalso be due in part to the higher public profile of thesespecies or real biological differences among taxa.

The desire to minimize potential conflicts, particularlyin the form of legal challenges, may have also had aninfluence on critical habitat identification. We found thatcritical habitat identification for species for which therehad been litigation over critical habitat was more likelyto be based on the most information-heavy and rigorousapproaches to identify habitat for inclusion. This find-ing corresponds with those of Hodges and Elder (2008),who found that a greater number of biological crite-ria were considered in critical habitat identification forcourt-ordered designations. We believe this correlationreflects the need for greater clarity and scientific rigor inidentifying critical habitat when designation is likely tobe challenged in court (Hoekstra et al. 2002; Hodges &Elder 2008). For instance, the only species in the reviewfor which critical habitat was identified with explicit ref-erence to population viability was the Northern SpottedOwl (Strix occidentalis caurina), which has been thesubject of considerable controversy in the United Statessince its listing in 1990 (USFWS 2012).

Designation of Unoccupied Habitat

Without reference to population viability or other interimrecovery goals, it is difficult to determine how muchhabitat is critical for long-term persistence, including theproportion of unoccupied habitat that might be requiredfor recovery. It is, therefore, perhaps unsurprising thatcritical habitat designation for most species reviewed didnot include unoccupied habitat. Excluding unoccupiedhabitat from designated critical habitat is likely to hin-der species recovery when recovery depends on speciesreintroduction into, or colonization of, currently unoccu-pied areas. For instance, recovery of black-footed ferrets(Mustela nigripes) in the United States, from 18 wildindividuals in the 1980s to at least 362 breeding adultsin 2012, was achieved through the successful reintroduc-tion of captive bred individuals into unoccupied habitatwith sufficient prey density (USFWS 2013b). The avail-ability of suitable reintroduction sites for black-footedferrets is considered a limiting factor of recovery (USFWS2013b). Down listing of this species is thought to befeasible with successful reintroduction to 6 sites overthe next 10 years (USFWS 2013b). Delisting will requiresuccessful reintroduction to 20 additional sites over thenext 20 years (USFWS 2013b).

Given that mobile species may be more likely to usecurrently unoccupied habitat over the short term, it issurprising that unoccupied habitat was less frequentlyincluded in critical habitat for these species. We be-lieve this trend is also a consequence of limited dataand the desire to minimize potential conflicts. For wide-ranging species, information on population dynamics and

Conservation BiologyVolume 29, No. 2, 2015

Camaclang et al. 489

(a) Unoccupied habitats included

Dis

pers

alS

essi

le/N

eglig

ible

Lim

ited

Wid

e R

angi

ng No Yes(c) Taxonomic group

Dis

pers

al

Uno

ccup

ied

habi

tats

incl

uded

Ses

sile

/Neg

ligib

le

Yes

No

Lim

ited

Yes

No

Wid

e R

angi

ng Bird Mamm HerpFish I P

Yes

No

(b) Unoccupied habitats included

Taxo

nom

ic g

roup

Pla

ntIn

vF

ish

Her

pM

amm

Bird

No Yes(d) Unoccupied habitats included

Cou

ntry

Aus

tral

iaC

anad

aU

nite

d S

tate

s

No Yes

Figure 3. Relative proportion of species for which unoccupied habitat was included (yes or no) in critical habitatdesignation grouped according to (a) species’ dispersal ability, (b) taxonomic group, (c) dispersal ability andtaxonomic group combined, and (d) country of designation. The area of each bar represents the proportion ofobservations in each variable combination; zero observations are represented by a line and circle.

habitat requirements may be more difficult to determine,and justifying the need to protect unoccupied habitat ischallenging. In addition, critical habitats of wide-rangingspecies are more likely to encompass larger areas andmultiple jurisdictions, resulting in greater potential forconflict with landowners and other stakeholders whomay be affected by designation.

Effectiveness of Critical Habitat Identification

Over the past 10 years, the identification of critical habi-tat for threatened species has been inconsistent withits underlying intent—to protect habitat essential forpersistence. In particular, targets for species persistence,such as threshold extinction risks, minimum viable popu-lation sizes, or delisting and recovery goals, were used toidentify critical habitat for a small proportion of speciesin our review. In addition, unoccupied habitat was

included as part of critical habitat for less than one-third of the species we considered. Hence, whether theidentified critical habitat is sufficient to achieve speciespersistence or other recovery goals is unclear. Withoutreference to population viability or other recovery tar-gets, it is difficult to determine how much of a species’habitat will be required. This mismatch between theintent and actual practice poses a challenge to the ef-fectiveness of critical habitat identification: lack of ex-plicit consideration for population viability or recoverytargets, as well as bias against including unoccupied habi-tat, may result in failure to identify and protect enoughhabitat for the recovery and long-term persistence of thespecies.

Bridging the gap between the intent and practice ofcritical habitat identification should help improve itsimpact on species recovery by ensuring that sufficienthabitat is protected in the right locations. Accurately

Conservation BiologyVolume 29, No. 2, 2015

490 Identifying Critical Habitat

dispersal

status

specificity

distribution

extent

threat

taxa

country

litigation

0.00 0.01 0.02 0.03

Coder 1C

ovar

iate

s

specificity

taxa

threat

distribution

extent

dispersal

country

status

litigation

0.000 0.004 0.008 0.012

Coder 2

extent

specificity

status

litigation

country

distribution

dispersal

threat

taxa

0.000 0.004 0.008

Coder 1

(Mean decrease in accuracy)Conditional variable importance score

Cov

aria

tes

threat

litigation

country

distribution

specificity

status

extent

taxa

dispersal

0.000 0.005 0.010

Coder 2

(Mean decrease in accuracy)Conditional variable importance score

(a) Information type used

(b) Unoccupied habitat included

Figure 4. Relative influence of covariates on (a) the type of information used to identify critical habitat and (b)inclusion of unoccupied habitat as part of critical habitat designation for data on 50 species and with 2 differentcontent analysts. Relative influence was estimated with the variable importance score from conditional inferencetrees. Higher scores indicate greater influence of the predictor variable. The vertical dashed line indicates theabsolute value of the lowest score and is used as a threshold value for evaluating variable significance.

identifying critical habitat can also provide greater cer-tainty to landowners and other stakeholders by explicitlymapping areas wherein damage or destruction must beavoided. However, budget considerations and mandatorytimelines in some jurisdictions often place a limit on theamount of time or money that can be spent improv-ing the accuracy of designation (Darin 2000). Calls fortimely protection of habitat from further destruction ordegradation (Possingham et al. 2007) and the needto take advantage of transient conservation opportu-nities (Knight & Cowling 2007; Martin et al. 2012b)also mean that decisions on critical habitat often need

to be made in the absence of perfect information.In short, there is a trade-off between maximizing theaccuracy of designation and minimizing the poten-tial for habitat loss and degradation over time, andthe effectiveness of critical habitat designation onspecies recovery may depend not only on the accuracybut also on the timeliness of designation. Therefore,although critical habitat should initially be identifiedbased on currently available information regardless of thequality of the data, it is also important to ensure that thesedesignations are updated and refined as new informationbecomes available.

Conservation BiologyVolume 29, No. 2, 2015

Camaclang et al. 491

The effectiveness of critical habitat designations, how-ever, depends not just on where prohibitions againstdestruction or damage apply, but also on how strictlythese prohibitions are enforced. Designation of criticalhabitat may not necessarily translate into adequate pro-tection because legal critical habitat prohibitions typi-cally apply only to federal lands in Canada (Mooers et al.2010) or to actions regulated by federal agencies in theUnited States (Gibbs & Currie 2012). In addition, politicalopposition and fear of negative public response to criticalhabitat designation, combined with the perception thatdesignation provides little added benefit to species pro-tection, may have also discouraged government agenciesfrom fully enforcing the legal provisions to designate andprotect critical habitat (e.g., Greenwald et al. 2012; Owen2012; Taylor & Pinkus 2013). Therefore, there is also aneed to evaluate the implementation and enforcementof critical habitat protection before conclusions aboutthe true impact of critical habitat designations on speciesrecovery can be made. Ultimately, improving the effec-tiveness of critical habitat designation as a tool for speciesrecovery will likely require decisive governance to ensurethat both identification and protection of critical habitatare in line with the intent of the law—preventing extinc-tions and promoting species recovery.

Acknowledgments

We thank T. Holmes for advice on performingcontent analyses of government documents, L. VerdeArregoitia and N. Shumway for their assistance inperforming the content analysis, and P. Johanson, N.Auerbach, R. Clemens, Y. Buckley, and 2 anonymousreviewers for their thoughts and comments on an earlierversion of this manuscript. This research was supportedby a Natural Sciences and Engineering Research Councilof Canada Post Graduate Scholarship and a Universityof Queensland International Scholarship for A.E.C., andadditional funding from the Australian Research CouncilCentre of Excellence for Environmental Decisions andthe Australian Government National EnvironmentalResearch Program’s Environmental Decisions Hub.

Supporting Information

The coding scheme used in the content analysis and thelist of potential covariates (Appendix 1) and the list ofspecies and critical habitat documents included in thereview (Appendix 2) are available online. The authorsare solely responsible for the content and functionalityof these materials. Queries (other than absence ofthe material) should be directed to the correspondingauthor.

Literature Cited

Akcakaya HR, Sjogren-Gulve P. 2000. Population viability analyses inconservation planning: an overview. Ecological Bulletins 48:9–21.

Bottrill MC, Hockings M, Possingham HP. 2011. In pursuit ofknowledge: addressing barriers to effective conservation evalu-ation. Ecology and Society 16: http://www.ecologyandsociety.org/vol16/iss2/art14/.

Breiman L. 1996. Bagging predictors. Machine Learning 24:123–140.Breiman L. 2001. Random forests. Machine Learning 45:5–32.Clark JA, Hoekstra JM, Boersma PD, Kareiva P. 2002. Improving U.S.

Endangered Species Act recovery plans: key findings and recom-mendations of the SCB recovery plan project. Conservation Biology16:1510–1519.

Cutler DR, Edwards Jr TC, Beard KH, Cutler A, Hess KT. 2007. Randomforests for classification in ecology. Ecology 88:2783–2792.

Darin TF. 2000. Designating critical habitat under the EndangeredSpecies Act: habitat protection versus agency discretion. HarvardEnvironmental Law Review 24:209–235.

De’ath G, Fabricius KE. 2000. Classification and regression trees: apowerful yet simple technique for ecological data analysis. Ecology81:3178–3192.

Department of the Environment. 2013. Recovery plans adopted un-der the EPBC Act. Species profile and threats database. Depart-ment of the Environment, Canberra. Available from http://www.environment.gov.au/cgi-bin/sprat/public/publicshowallrps.pl (ac-cessed 22 Nov 2013).

Fahrig L, Merriam G. 1994. Conservation of fragmented populations.Conservation Biology 8:50–59.

Fischer TB, Onyango V. 2012. Strategic environmental assessment-related research projects and journal articles: an overview of thepast 20 years. Impact Assessment and Project Appraisal 30:253–263.

Flather CH, Bevers M. 2002. Patchy reaction-diffusion and populationabundance: the relative importance of habitat amount and arrange-ment. The American Naturalist 159:40–56.

Gibbs KE, Currie DJ. 2012. Protecting endangered species: Do themain legislative tools work? PLoS ONE 7:e35730. DOI: 10.1371/jour-nal.pone.0035730.

Government of Canada. 2012. Species at risk public registry. Environ-ment Canada, Ottawa. Available from http://sararegistry.gc.ca/ (ac-cessed 17 Oct 2013).

Greenwald DN, Suckling KF, Pimm SL. 2012. Critical habitat and the roleof peer review in government decisions. BioScience 62:686–690.

Guisan A, et al. 2013. Predicting species distributions for conservationdecisions. Ecology Letters 16:1424–1435.

Hall LS, Krausman PR, Morrison ML. 1997. The habitat concept and aplea for standard terminology. Wildlife Society Bulletin 25:173–182.

Haney W, Russell M, Gulek C, Fierros E. 1998. Drawing on education:using student drawings to promote middle school improvement.Schools in the Middle 7:38–43.

Hanski I. 1998. Metapopulation dynamics. Nature 396:41–49.Hodges KE, Elder J. 2008. Critical habitat designation under the US

Endangered Species Act: How are biological criteria used? BiologicalConservation 141:2662–2668.

Hoekstra JM, Fagan WF, Bradley JE. 2002. A critical role for criticalhabitat in the recovery planning process? Not yet. Ecological Appli-cations 12:701–707.

Hothorn T, Buhlmann P, Dudoit S, Molinaro A, Van Der Laan MJ. 2006.Survival ensembles. Biostatistics 7:355–373.

Hsieh H-F, Shannon SE. 2005. Three approaches to qualitative contentanalysis. Qualitative Health Research 15:1277–1288.

Huxel GR, Hastings A. 1999. Habitat loss, fragmentation, and restora-tion. Restoration Ecology 7:309–315.

IUCN (International Union for Conservation of Nature). 2012. IUCN RedList categories and criteria: version 3.1. IUCN, Gland, Switzerland.

Conservation BiologyVolume 29, No. 2, 2015

492 Identifying Critical Habitat

Kerkvliet J, Langpap C. 2007. Learning from endangered and threat-ened species recovery programs: a case study using U.S. EndangeredSpecies Act recovery scores. Ecological Economics 63:499–510.

Kleiman DG. 1989. Reintroduction of captive mammals for conserva-tion. BioScience 39:152–161.

Knight AT, Cowling RM. 2007. Embracing opportunism in the selectionof priority conservation areas. Conservation Biology 21:1124–1126.

Krippendorff K. 2004. Content analysis: an introduction to its method-ology. Sage, Thousand Oaks, California.

Male TD, Bean MJ. 2005. Measuring progress in US endangered speciesconservation. Ecology Letters 8:986–992.

Martin LJ, Blossey B, Ellis E. 2012a. Mapping where ecologists work:biases in the global distribution of terrestrial ecological observations.Frontiers in Ecology and the Environment 10:195–201.

Martin TG, Nally S, Burbidge AA, Arnall S, Garnett ST, Hayward MW,Lumsden LF, Menkhorst P, McDonald-Madden E, Possingham HP.2012b. Acting fast helps avoid extinction. Conservation Letters5:274–280.

Mooers AO, Doak DF, Findlay CS, Green DM, Grouios C, Manne LL,Rashvand A, Rudd MA, Whitton J. 2010. Science, policy, and speciesat risk in Canada. BioScience 60:843–849.

Murphy DD, Noon BD. 1991. Coping with uncertainty in wildlife biol-ogy. The Journal of Wildlife Management 55:773–782.

Neel MC, Leidner AK, Haines A, Goble DD, Scott JM. 2012. By thenumbers: How is recovery defined by the US Endangered SpeciesAct? BioScience 62:646–657.

Owen D. 2012. Critical habitat and critical challenge of regulating smallharms. Florida Law Review 64:141–199.

Possingham HP, Grantham H, Rondinini C. 2007. How can you conservespecies that haven’t been found? Journal of Biogeography 34:758–759.

Pulliam HR, Danielson BJ. 1991. Sources, sinks, and habitat selection:a landscape perspective on population dynamics. The AmericanNaturalist 137:S50–S66.

R Core Team. 2013. R: a language and environment for statistical com-puting. R Foundation for Statistical Computing, Vienna.

Rachlinski JJ. 1997. Noah by the numbers: an empirical evaluationof the Endangered Species Act. Cornell Law Review 82:356–389.

Reed JM, Akcakaya HR, Burgman M, Bender D, Beissinger SR, Scott JM.2006. Critical habitat. Pages 164–177 in Scott JM, Goble DD, DavisFW, editors. The Endangered Species Act at thirty: conserving biodi-

versity in human-dominated landscapes. Island Press, Washington,D.C.

Rosenfeld JS, Hatfield T. 2006. Information needs for assessing criticalhabitat of freshwater fish. Canadian Journal of Fisheries and AquaticSciences 63:683–698.

Saldana J. 2009. The coding manual for qualitative researchers. SagePublications, London.

Schwartz MW. 2008. The performance of the Endangered Species Act.Annual Review of Ecology, Evolution, and Systematics 39:279–299.

Shaffer ML. 1981. Minimum population sizes for species conservation.BioScience 31:131–134.

Sinden A. 2004. The economics of endangered species: why less is morein the economic analysis of critical habitat designations. HarvardEnvironmental Law Review 28:129–214.

Strobl C, Boulesteix A-L, Kneib T, Augustin T, Zeileis A. 2008. Condi-tional variable importance for random forests. BMC BioinformaticsDOI: 10.1186/1471-2105-9-307.

Taylor EB, Pinkus S. 2013. The effects of lead agency, nongovernmentalorganizations, and recovery team membership on the identificationof critical habitat for species at risk: insights from the Canadianexperience. Environmental Reviews 21:93–102.

Taylor MFJ, Suckling KF, Rachlinski JJ. 2005. The effectiveness of theEndangered Species Act: a quantitative analysis. BioScience 55:360–367.

Thomas CD, Kunin WE. 1999. The spatial structure of populations.Journal of Animal Ecology 68:647–657.

Traill LW, Bradshaw CJA, Brook BW. 2007. Minimum viable populationsize: a meta-analysis of 30 years of published estimates. BiologicalConservation 139:159–166.

USFWS (US Fish and Wildlife Service). 2012. Endangered and threat-ened wildlife and plants; designation of revised critical habitat forthe Northern Spotted Owl: final rule. Federal Register 77:71876–72068.

USFWS (US Fish and Wildlife Service). 2013a. Environmental conserva-tion online system. Available from http://ecos.fws.gov/ecos/home.action (accessed October 2013).

USFWS (US Fish and Wildlife Service). 2013b. Recovery plan for theblack-footed ferret (Mustela nigripes). US Fish and Wildlife Service,Denver, Colorado.

White MD, Marsh EE. 2006. Content analysis: a flexible methodology.Library Trends 55:22–45.

Conservation BiologyVolume 29, No. 2, 2015