486

www.frontiersinecology.org © The Ecological Society of America

Write Back

extinction threats: towards a reevalua-tion of IUCN threatened species cate-gories. Conserv Biol 5: 148–57.

Mace GM, Collar NJ, Gaston KJ, et al.2008. Quantification of extinctionrisk: IUCN’s system for classifyingthreatened species. Conserv Biol 22:1424–42.

doi:10.1890/11.WB.025

No safety in numbers Clements et al. (Front Ecol Environ2011; 9[9]: 521–525) proposed a sin-gle metric that describes a “species’ability to forestall extinction”(referred to by the acronym “SAFE”)as a “scientifically defendable rule ofthumb for when complete demo-graphic data are unavailable” to rankthe relative threat status of a species.SAFE is calculated on a logarithmicscale and reflects the differencebetween a species’ current populationsize and 5000, the estimate for a uni-versal minimum viable population(MVP) promoted by Traill et al.(2010). Clements et al. advocatedSAFE as a useful tool for triage toallocate resources in conservation,and as a measure of population via-bility that would be more easilyunderstood by the public than theIUCN Red List categories (Mace etal. 2008). We believe that SAFE isnot a useful metric to guide conserva-tion planning for three main reasons.

First, a universal MVP of 5000 indi-viduals, regardless of taxon or circum-stance, is poorly supported (Flather etal. 2011). Studies promoting thisbenchmark overlooked substantialuncertainty in standardized MVPestimates that span several orders ofmagnitude for the same species, sug-gesting 5000 is likely to be a poor esti-mate for any specific population.Methods used to standardize MVPestimates across disparate studies werenot robust (Flather et al. 2011). MVPestimates depend critically on theenvironmental context of a popula-tion and on the way that contextinteracts with decisions made in thepopulation modeling process.

Second, theory and practice stronglysuggest that metrics other than popula-tion size are equally or more important

in determining a population’s viability(Lande 1993; Caughley 1994; Flatheret al. 2011). Viability of a species is acomposite of many characteristics,such as the mean and variance of itsgrowth rate, the number and connec-tivity of its populations, its range sizeand trends, and its life history, ratherthan simply its distance from 5000.Clements et al. implicitly acknowledgethis by comparing the performance oftheir SAFE metric versus range change(both independent variables) to theIUCN Criteria (the dependent vari-able or “truth”), which is based on acomplex series of factors combined toassess status. The ordinal logisticregressions only accounted for 6% ofthe deviance.

Third, SAFE offers little to informthe conservation of threatenedspecies. Populations can only be con-served if the factors that cause them todecline are identified and thosethreats are ameliorated. Obtaining areliable estimate of population size forcomparison to the unreliable MVPestimate of 5000 suggests that suffi-cient information on species’ ecology,habitat, and current threats is likely toexist to inform conservation. Triagedecisions based on population sizealone are pointless, ignoring circum-stance, trends, taxonomic uniqueness,desirability, and other important fac-tors that affect such decisions.

Population size is one indicator ofpopulation viability, much like apatient’s body temperature is oneindicator of health. However, thereis no single number that represents ahealthy temperature for all people,because time of day and many othercircumstances affect it. Moreover,physicians do not use body tempera-ture alone to determine a livingpatient’s prognosis, make triage deci-sions, or diagnose cause of illness.

The way forward to develop mea-sures that assist conservation plan-ning is not through oversimplifica-tion. Classification systems, such asthose developed by IUCN and oth-ers, are useful for ranking the degreeof threat because they incorporate awide range of information related topopulation viability. They do so

because no single population charac-teristic is sufficient to describe popu-lation viability. Conservation plan-ning advances when it combinescomprehensive measures of popula-tion viability with knowledge of howthese factors relate to threats, anunderstanding of social desires, andestimates of the cost of recovery.Steven R Beissinger1*, Curtis HFlather2, Gregory D Hayward3,and Philip A Stephens4

1University of California, Berkeley,Berkeley, CA *(beis@berkeley.edu); 2USDA Forest Service, Fort Collins,CO; 3USDA Forest Service,Anchorage, AK; 4Durham University,Durham, UK

Caughley G. 1994. Directions in conserva-tion biology. J Anim Ecol 63: 215–44.

Flather CH, Hayward GD, Beissinger SR,and Stephens PA. 2011. Minimumviable populations: is there a “magicnumber” for conservation practition-ers? Trends Ecol Evol 26: 307–16.

Lande R. 1993. Risks of population extinc-tion from demographic and environ-mental stochasticity and random cata-strophes. Am Nat 142: 911–27.

Mace GM, Collar NJ, Gaston KJ, et al.2008. Quantification of extinctionrisk: IUCN’s system for classifyingthreatened species. Conserv Biol 22:1424–42.

Traill LW, Brook BW, Frankham R, andBradshaw CJA. 2010. Pragmatic popula-tion viability targets in a rapidly chang-ing world. Biol Conserv 143: 28–34.

doi:10.1890/11.WB.026

The SAFE index should notbe used for prioritization Clements et al. (Front Ecol Environ2011; 9[9]: 521–525) proposed theSAFE index to measure a “species’ability to forestall extinction”. How-ever, we believe that this index can, atbest, only measure threat – not theability to forestall extinction. We noteand concur with other concernsregarding the index (letters by Akça-kaya et al. and Beissinger et al., thisissue), but focus on the points below forthe sake of brevity.

The SAFE index is simply a mea-sure of how far the population size(N) is from the minimum viable pop-ulation size (MVP). If the MVP were

488

www.frontiersinecology.org © The Ecological Society of America

Write Back

Of the 95 mammal species weassessed for the SAFE index, 63 areIUCN threat-listed. Of these, 51% arenot assessed by the IUCN on popula-tion size thresholds at all, and onlyone assessment is even partially basedon PVA. Indeed, based on our recent(July 2011) examination of CriticallyEndangered, Endangered, or Vulner-able species, not one of 1370 mammalor 1288 bird species relies entirely oncriterion E data, and only 4 mammaland no bird assessments include anyPVA information. Hence, the asser-tion that the SAFE index (a measureof distance from MVP) simply repro-duces the Red List is demonstrablyincorrect. It is debatable to whatextent the Red List categories predictreal extinction risk (O’Grady et al.2004); regardless, they must largelyinvoke reductions in geographic rangeand population size to do so.

(2) SAFE replaces the Red List.Under no circumstances did we assertthat the SAFE index should replacethe Red List, or that conservation-based prioritization should be based“solely on population size”. Weclearly called for SAFE to be used inconjunction with the Red List to pro-vide a more heuristic measure of rela-tive species-extinction threat. Weagree that assessments made on popu-lation size (and their distance toMVP) alone are inadequate toexplain all elements of risk – claimingotherwise would be astonishinglynaïve (Brook et al. 2011). The contri-bution of SAFE to the existing RedList categories is that, in addition toreflecting susceptibility to stochasticextinction processes, it provides acontinuous measure both among andwithin risk categories (somewhat anal-ogous to RAMAS software’s Red Listfuzzy-number categorization method[www.ramas.com/redlist.htm]). Thisis pertinent given the ambiguousnature of categorical terms like“endangered”, “threatened”, and“vulnerable” that are often confusedby lay persons and used interchange-ably or inconsistently in national-level legislation. In a triage context,the choice to invest in conservingparticular species can be informed, at

least partially, by MVP (Traill et al.2010) and SAFE by indicating howurgently a species requires attention.

(3) SAFE simplifies to population size(N). We incorporated a logarithmictransformation in SAFE to ease inter-pretability for our “distance fromextinction and to MVP” conceptacross many species, and for standard-ization purposes. For example, takehypothetical species A and B – com-prising 200 and 2 000 000 individuals,respectively – and assume a thresholdMVP target of 5000. Even for special-ists, explaining the relative risk as“species A is 4800 individuals awayfrom the threshold target”, and“species B is 1 995 000 individualsabove the threshold” becomes a con-fusing mix of largely irrelevant num-bers and qualifiers. We maintain thatit is far easier to infer whether speciesA is in trouble based on a negativeSAFE index (–1.40, in this case), andthat species B is at far less risk basedon its positive SAFE value (2.60). Aswe originally stated in our paper, thethreshold MVP value need not neces-sarily be 5000; if one has sufficientdata to estimate, for instance, ataxon-specific MVP, then differentdenominator values could be used fordifferent taxa (Traill et al. 2010;Brook et al. 2011). This process wouldact to normalize comparisons ofSAFE-based extinction risks amonggroups (taxa or otherwise) withintrinsically different MVP sizes.Commonly used biodiversity even-ness metrics such as Shannon’s Indexalso use logarithms to make large andsmall sample sizes comparable.

(4) MVP size is not generalizable.Several authors took exception to ourconcept of a generalizable MVP size foruse as a target threshold, based mainlyon arguments raised in a recent cri-tique (Flather et al. 2011). We haveaddressed these concerns elsewhere(Brook et al. 2011), but summarize ourprincipal defense here. AlthoughMVP does vary among species, the keyemergent result is that thousands, andnot hundreds, of individuals areneeded to minimize the risk of stochas-tic extinction − this is the essence ofthe MVP “rule of thumb” (Traill et al.

2010). PVAs are unavailable to esti-mate MVPs for most species, so gener-alizations are required in mostinstances. The alternative – to arguethat the problem is too intractable anduncertain and that all species areunique − leads nowhere in terms ofpractical conservation management.

In conclusion, we are surprised thata heuristic concept designed toenhance conservation decision mak-ing has evoked such spirited criti-cisms from the progenitors of the RedList (Akçakaya et al.) and other con-servation decision-theory specialists(Beissinger et al. and McCarthy etal.). Putting aside arguments aboutuncertainty and relative merit, thereal test of the SAFE concept’s utilitywill be determined by whether it cancontribute usefully to on-the-groundconservation decisions.Corey JA Bradshaw1,2*, GopalasamyReuben Clements3,4, William FLaurance3, and Barry W Brook1

1The Environment Institute and Schoolof Earth and Environmental Sciences,The University of Adelaide, Adelaide,Australia *(corey.bradshaw@adelaide.edu.au); 2South Australian Researchand Development Institute, HenleyBeach, Australia; 3Centre for TropicalEnvironmental and SustainabilityScience and School of Marine andTropical Biology, James CookUniversity, Cairns, Australia; 4Centrefor Malaysian Indigenous Studies,Universiti Malaya, Kuala Lumpur,Malaysia

Brook BW, Bradshaw CJA, Traill LW, andFrankham R. 2011. Minimum viablepopulations size: not magic, but neces-sary. Trends Ecol Evol; doi:10.1016/j.tree.2011.09.006.

Flather CH, Hayward GD, Beissinger SR,and Stephens PA. 2011. Minimumviable populations: is there a “magicnumber” for conservation practition-ers? Trends Ecol Evol 26: 307–16.

O’Grady JJ, Burgman MA, Keith DA, et al.2004. Correlations among extinctionrisks assessed by different systems ofthreatened species categorization.Conserv Biol 18: 1624–35.

Traill LW, Brook BW, Frankham R, andBradshaw CJA. 2010. Pragmatic popu-lation viability targets in a rapidlychanging world. Biol Conserv 143:28–34.

doi:10.1890/11.WB.028