Targeting Global Protected Area Expansion for Imperiled Biodiversity — TEI-tagged view

Oscar VenterpersName, Richard A. FullerpersName, Daniel B. SeganpersName, Josie CarwardinepersName, Thomas BrookspersName, Stuart H. M. ButchartpersName, Moreno Di MarcopersName, Takuya IwamurapersName, Liana JosephpersName, Damien O'GradypersName, Hugh P. PossinghampersName, Carlo RondininipersName, Robert J. SmithpersName, Michelle VenterpersName, James E. M. WatsonpersName (2014date). “Targeting Global Protected Area Expansion for Imperiled Biodiversity.” PLoS Biologytitle·j.

Object of study — where global protected-area expansion should occur to meet area and threatened-species targets cost-efficiently

Research question — how can countries expand terrestrial protected areasterm while balancing agricultural opportunity costterm, ecological representation, and coverage of threatened vertebratesterm

Contribution — the first optimization-based global scenarios jointly targeting Aichi protected-area coverage and threatened-species adequacy at minimum agricultural opportunity costterm

Full text — tagged

The reference list is not part of the text we hold for this paper. Our copy was extracted from the publisher's page and the bibliography did not survive that extraction, so the article below ends before its references. Nothing has been removed by us.

Targeting Global Protected Areaterm Expansion for Imperiled Biodiversityterm Abstract Meeting international targets for expanding protected areasterm could simultaneously contribute to species conservationterm, but only if the distribution of threatened speciesterm informs the future establishment of protected areasterm.

Introduction In 2010 the 193 parties to the Convention of Biological Diversity (CBD) adopted a new strategic plan and set of targets to tackle the continuing decline in biodiversity [1] , [2] . A key element of this plan is Aichi target 11term, which includes a commitment to expand the global coverage of terrestrial protected areasterm from the current 13% to 17% by 2020 [1] . This could drive the most rapid expansion of the global protected area networkterm in history [3] , but corresponding biodiversity benefits are far from guaranteed. This is because protected areasterm are often preferentially established in locations that are remote or have little agricultural valueterm [4] , failing to protect the imperiled biodiversityterm found on more valuable land.

Recognizing the failures of past protected areaterm expansion, the current CBD text directs that protected areasterm should target places of “importance for biodiversity” that are “ecologically representativeterm” [1] . However, these locations can be expensive to protect. For instance, the cost of expanding protected areasterm to cover all “important bird areas” (IBAs) has been estimated at US$58 billion annually (although these sums are still small compared to government budgets) [5] . Moreover, the majority of terrestrial regions have been identified as important for biodiversity by one or more global prioritization schemes [6] , which provides myriad alternatives for meeting protected areaterm targets in locations that are cheap. Given this, where should new protected areasterm be located to deliver on the Aichi biodiversity targets? One option could be based on Aichi target 12term, which aims to “prevent the extinction of all known threatened speciesterm and improve and sustain their conservation status.” In situ conservation of viable populations in natural ecosystems has long been recognized as the fundamental requirement for the maintenance of biodiversity [7] . Hence measuring “biodiversity importanceterm” in terms of protected area coverageterm·var of threatened speciesterm would help countries to simultaneously meet these two CBD targets.

Using new data from the World Database on Protected Areasterm [3] and distribution maps for 4,118 globally threatenedterm birds [8] , mammals [9] , [10] , and amphibians [10] , [11] , as well as ecoregions [12] , we first perform a gap analysisterm to determine the representation of these species in the current global protected area networkterm. We then use a systematic conservation planning frameworkterm [13] to build scenarios for cost-efficiently expanding the global protected area networkterm to contribute to meeting the protected areaterm and threatened speciesterm Aichi targets. Recent works have investigated strategies for achieving Aichi Target 11term by protecting IBAs [5] , [14] or meeting the Global Strategy for Plant Conservation [15] . Our study is the first, to our knowledge, to use an optimization approach to develop scenarios for meeting the Aichi targets in a cost-efficient mannerterm. Incorporating cost efficiency allows the identification of options for meeting Aichi target 11term that contribute optimally to target 12 while minimizing conflict with agricultural production.

Methods All spatial overlays were performed at a spatial resolution of 500 mterm and then aggregated into 30 km×30 km pixelsterm to identify candidate land for protection. By processing data at the finer resolution, we are able to account for protected areasterm at the subpixel level, thereby minimizing omission of small-sized protected areasterm. This resolution of ∼⅓ degree (at the Equator) falls in the midrange between scales of ½ degree [16] and of ⅛ degree [17] typically used in such analyses.

Protected Areasterm To determine the extent of current protected areasterm, we extracted data on International Union for Conservation of Nature (IUCN) category I–VI protected areasterm from the 2012 World Database on Protected Areasterm [3] , excluding all proposed protected areasterm and those lacking “national” designation. For terrestrial protected areasterm with a known areal extent but lacking polygonal representation, we created a circular buffer of the appropriate area around its centroid. To prevent overestimation of the areal coverage of protected areasterm caused by overlapping designations, we merged buffered points and polygons into a single layer. Our final protected area layerterm contained 135,062 protected areasterm covering a total of 17,026,214 km 2 , or 12.9% of the Earth's non-Antarctic land surface ( Figure 1A ).

Key data inputs and output map from the systematic conservation planning frameworkterm. (A) Protected areasterm mapped using polygons and buffered points for nationally designated protected areasterm [3] . (B) The number of native and extant globally threatenedterm terrestrial and freshwater birds [8] , mammals [10] , and amphibians [10] per grid square. (C) The average annual agricultural opportunity costterm of protecting each 30 km grid square in 2012 $US [17] . (D) The distribution of priorities for establishing new protected areasterm to meet the national-level 17% targets under Aichi target 11term at minimal cost and ignoring ecological representation (red), for covering threatened speciesterm (green), and locations selected under both scenarios (yellow). The sizes of the circles in the Venn diagrams are proportional to the area required in each of the three categories.

Distribution of Biodiversity We used distribution maps for birds [8] , mammals [10] , and amphibians [10] . We focused on these taxa as they are the only major terrestrial taxonomic groups that have been comprehensively assessed for their distribution and extinction risk [10] . We excluded marine species and areas, noting that there are specific coverage targets for protecting the marine realm. For all three taxonomic groups, we focused on those species that are listed by the IUCN Red List as Critically Endangeredterm, Endangeredterm, or Vulnerableterm, hereafter referred to as “threatenedterm,” resulting in 4,118 species in total (birds = 1,135, mammals = 1,107, amphibians = 1,876; Figure 1B ). We focus only on threatened speciesterm as these are by definition the most likely species to go extinct, and therefore are most important for slowing biodiversity loss and contributing to CBD Aichi target 12term. We excluded all portions of species ranges where the species was identified as extinct, introduced, or of uncertain origin. In addition to these data, we used data on the distribution of ecoregions as defined by the World Wildlife Fund [12] .

Protected Areaterm Opportunity Cost To account for the spatial variation in the cost of protected areaterm expansion, we used a dataset on agricultural opportunity costterm [18] , converted to 2012 US$ and with no data values filled using regularized spline interpolation with tension ( Figure 1C ). The dataset provides the estimated gross agricultural rentsterm for terrestrial areas mapped at approximately the 5 km resolution. We use these data as our surrogate for the opportunity costs of establishing new protected areasterm, as agricultural expansionterm is the greatest single cause of habitat lossterm, as well as the one most commonly associated with habitat lossterm driven by multiple factors [19] , [20] . Agricultural opportunity coststerm also reflect the reduction in food security and tax revenue that national governments face when implementing protected areasterm. We applied a fixed cost of US$100 per km 2 to reflect the transaction costs of acquiring new protected areasterm [21] , although we recognize there is likely to be considerable spatial variation in these costs. We did not attempt to estimate the ongoing management coststerm·var of protected areasterm following establishment, as this metric needs to account for a number of difficult-to-measure social and socioeconomic factors [22] , but a recent analysis estimated that these equate to ∼14% of the agricultural opportunity coststerm of protection [5] .

Gap Analysisterm We assessed the occurrence of threatened vertebratesterm within protected areasterm using a representation targetterm and an adequacy targetterm. The representation targetterm was achieved if any portion of the species' distribution overlapped with the protected area networkterm. To set adequacy targetsterm we followed the method of Rodrigues et al. [23] to scale the target to the species' overall geographic range sizeterm. Complete (i.e., 100%) coverage by protected areasterm was required for species with a geographic range of <1,000 km 2 . For wide-ranging speciesterm (>250,000 km 2 ), the target was reduced to 10% coverage, and where geographic range sizeterm was intermediate between these extremes, the target was log-linearly interpolated.

Scenarios for Protected Areaterm Expansion To explore future scenarios for the growth of the global protected area networkterm we used the systematic conservation planning software Marxanterm [24] . Marxanterm uses a simulated annealing algorithmterm to select multiple alternative sets of areas that meet prespecified conservation targetsterm·var (described in the following section) while trying to minimize overall cost. All spatial data on the distribution of conservation featuresterm and conservation coststerm were summarized into a “planning unitterm” layer consisting of 30 km×30 km square pixels comprising the world's non-Antarctic terrestrial areas. We intersected this planning unitterm layer with the protected areasterm and agricultural opportunity layers and the geographic distribution of each of the 4,118 threatened speciesterm and ecoregions at a 500 m resolution. This allowed us to determine the agricultural opportunity costterm of the unprotected portion of each planning unitterm and the protected and unprotected extent of each biodiversity feature within each planning unitterm.

To explore the costs and benefits of alternate scenarios for achieving 17% protection of terrestrial areas, we developed four separate spatial scenarios using contrasting conservation targetsterm·var. We accounted for the existing protected area networkterm's contribution to the targets in each scenario, and then added additional protected areasterm to ensure all targets are met. In each scenario, the aim is to minimize the costs of meeting the conservation targetsterm·var. However, to avoid the global protected areaterm target being met only through increased protection in low-cost countries, which would reduce the total cost of the target, in all scenarios we maintain the constraint that each country must meet its national protected area targetterm. Moreover, it is at the national level that the target is being interpreted and implemented. For each scenario, we used Marxanterm to perform 10 runs of 1 billion iterations each, each of which represents an alternate near optimal reserve network for meeting the relevant conservation targetsterm·var at the lowest overall cost. From these 10 runs, we select and report on the results from the lowest cost solution.

National targets In the first scenario, we set the conservation targetterm·var as each country meeting its protected areaterm target at the lowest agricultural opportunity costterm. In this scenario, we set all countries' terrestrial protected areaterm target to 17%, except for the 73 countries that have indicated in CBD workshops that they proposed alternative targets [25] , in which case we used these targets. As countries have tended in the past to meet their targets by favoring high, far, and otherwise agriculturally low-value areas [4] , we view this as our business-as-usual protected area expansion scenarioterm. We also determine the conservation benefits of protected areaterm target levels above the current Aichi 17% targets by setting national levels up to 30% of each country.

Ecoregional target In this scenario, we maintain the national-level 17% targets from scenario a but add the additional constraint that countries meet their target in a way that ensures that each of the 821 terrestrial ecoregionsterm receive at least 17% protection. We include this scenario as Target 11 calls for areas protected to be “ecologically representativeterm” [1] .

Threatened speciesterm target In this scenario, we maintain the national-level 17% targets from scenario a but add the additional constraint that all threatened speciesterm must be covered to the level of their adequacy targetsterm [23] .

Threatened speciesterm preference In this scenario, we construct an efficiency frontierterm between the cost of meeting the 17% target as in a and attaining threatened speciesterm conservation targetsterm·var as in c. The tradeoff curveterm is established by iteratively increasing the value given to meeting species adequacy targetsterm·var, from no value to a value equal to that given to the 17% target itself. The 17% target is always met at the national level across the tradeoff frontier.

Commission Errorsterm in Range Mapsterm The IUCN [10] and Birdlife International and NatureServe [8] range mapsterm used in this study comprise polygons showing distribution of 4,118 globally threatenedterm birds, mammals, and amphibians. These maps may be subject to commission errorsterm [26] – [29] , where the species is mapped as present in locations where it is in fact not present. As they affect range-based species conservationterm targets and lead to an overestimation of occurrenceterm in existing or prioritized areas, commission errorsterm could influence our study's main conclusions. We performed two analyses to determine the sensitivity of our primary results to commission errorsterm ( Text S1 ). First we created 100 range mapsterm for each of the 4,118 species of birds, mammals, and amphibians that simulated commission errorterm rates [25] by deleting 50% of the range of narrow-ranged species (range<1,000 km 2 ), by deleting 25% of the range of wide-ranging speciesterm (range>250,000 km 2 ), and by linearly extrapolating the deletion rate for species of intermediate ranges. Second, we identified the “Extent of Suitable Habitatterm” (ESH) using high-resolution species distribution models for 1,063 mammals [30] . The ESH maps were used to identify locations in the original maps for mammals that are likely to be commission errorsterm. We then reran our analyses using (a) the maps with simulated commission errorsterm and (b) the ESH maps, to quantify the effects of the simulated and mapped commission errorsterm on our estimated biodiversity value of meeting the 17% protected areaterm target, and the shape of the efficiency frontierterm between cost and threatened vertebrateterm coverage.

Results We find that 17% of threatened vertebratesterm are not found in a single protected areaterm and 85% are not covered to the level of our adequacy targetsterm ( Figure S1A ). A decade ago, 20% of globally threatenedterm terrestrial birds, mammals, and amphibians were not found in a single protected areaterm and 89% were inadequately protected [15] . Our analysis using updated datasets indicates that the global protected area networkterm has made little progress since then toward securing a future for the world's threatenedterm biodiversity.

We discover that if countries choose to expand their protected areasterm in a manner that minimizes agricultural opportunity costterm, meeting their national-level targets for 17% coverage would entail a once-off transaction cost of US$0.9 billion and an annual agricultural opportunity cost of $4.9 billionterm ( Table 1 ). As this option aligns with the previous pattern of protected areaterm establishment, we view it as a likely business-as-usual scenario for meeting the terrestrial coverage aspect of Aichi target 11term. We find that this would result in only 852 (21%) threatened vertebratesterm reaching targets for adequate coverage ( Figure S1B ), an increase of only 249 species over existing protection ( Table 1 ) and arguably a failure to meet Aichi target 12term. Moreover, even if highly ambitious areal targets were to drive further growth of the global protected area networkterm beyond 2020, the costs of expansion would rise steeply without providing cost-effective coverage for threatened speciesterm ( Figure 2 ).

The number of globally threatened vertebratesterm that reach our adequacy targetsterm (black), and the agricultural opportunity costterm of establishing new protected areasterm (red), as the proportion of global land areas protected increases above 17%.

Costs and benefits of the current protected area networkterm and for future protection scenarios that (a) meet country-level targets for protected area coverageterm·var; (b) meet these targets while also achieving 17% protection of each terrestrial ecoregionterm; (c) meet the targets from scenario a and protect a scaled fraction of the geographic ranges of threatenedterm terrestrial birds, mammals, and amphibians; and (d) achieve the country-level targets for protected area coverageterm·var while also achieving five times the level of biodiversity protection relative to scenario a.

An alternative is to ensure a representative sample of major vegetation communities is protected, as this would protect a broader range of habitats and could lead to improved conservation outcomes. Target 11 calls for ecologically representativeterm protected area coverageterm·var. We find that if countries meet their 17% coverage targets in a way that distributes protection across ecoregions equally, the opportunity cost of establishing the additional protected areasterm would be 4.5 times higher than the business-as-usual scenario ($24.8 billion annually; Table 1 ), but that coverage of threatened speciesterm would increase only marginally ( Figure S1C ). Moreover, the majority of species that reach their adequacy targetsterm are those with a geographic range sizeterm ≥250,000 km 2 ( Figure S1C ), as their wide distribution renders them more easily captured when distributing protected areasterm equitably across ecoregions. The species most likely to be left unprotected are narrowly distributed speciesterm, which often are those in greatest need of protection [31] , [32] .

These results indicate that protected areaterm expansion targeting either the cheapest landterm or representation of ecoregions is not an efficient approach for covering threatened speciesterm. Alternatively, we find that locating protected areasterm to ensure they meet targets for adequate coverage of all 4,118 threatened speciesterm would cost about $42.5 billion annually ( Table 1 ), which is about 7.5 times more than the cheapest option for meeting the 17% target. This difference in costterm is driven by low concordanceterm between areas that are cheap to protect and those that capture the distributions of threatened speciesterm ( Figure 1D ). Land selected for threatened speciesterm tends to align with tropical forest hotspots ( Figure 1B ), such as the tropical Andes and eastern Madagascar, whereas the cheapest landterm to protect is remote and often in more arid zones ( Figure 1D ). This lack of overlap helps explain why the existing protected area networkterm, which has favored low-cost areas in each country [4] , represents threatened speciesterm rather poorly.

How can countries reconcile the attraction of low-cost conservation with the benefits of protecting places that contribute to threatened speciesterm conservation? By varying the importance placed on meeting targets for adequate coverage of threatened speciesterm, we discover a nonlinear tradeoffterm between the cost of establishing additional protected areasterm and the proportion of threatened vertebratesterm covered by these areas ( Figure 3 ). The shape of the curve illustrates that large gains in the number of species potentially protectedterm could be achieved for relatively small increases in costterm. For instance, increasing by 5-fold the number of species protected relative to the low-cost, business-us-usual scenario would increase opportunity costs to only $7.4 billion annually (1.5 times as much; Table 1 ).

Efficiency frontierterm between the cost of establishing additional protected areasterm to achieve 17% coverage and the number of species covered. The y -axis presents the proportion of each species adequacy targetterm·var that is met within protected areasterm, summed across all species, and is not directly comparable to that of the other figures, which only count species whose protected area coverageterm·var meets or exceeds their target.

We find that our primary results are robust to randomly simulated commission errorsterm in the range mapsterm. Although the number of species meeting range-based coverage targets generally decreases once commission errorsterm are simulated ( Text S1 ), this drop averages only 5% across the tradeoff curveterm ( Figure S2 ). Moreover, both a visual interpretation and a quantitative measure of the shape of the tradeoff curveterm reveals that the original and commission errorterm updated curves are similarly nonlinear. Moreover, using high-resolution expert-based habitat suitability modelsterm for 1,063 threatenedterm mammals, we again find that commission errorsterm are unlikely to alter our primary findings ( Figure S3 ).

Discussion A small minority (15%) of threatened vertebratesterm are adequately covered by existing protected areasterm. However, the adoption of the Aichi targets marks an historic opportunity for achieving conservation of the world's biodiversity. If countries are to meet the protected areaterm Aichi target, at least 5.8 million km 2 of new protected areasterm will need to be created by 2020. Although this is a significant opportunity for biodiversity conservation, we have shown that protected areaterm expansion that targets low-cost areas in each country and ignores threatened speciesterm is unlikely to protect such species incidentally. This remains the case even if protected areasterm are further expanded to cover 30% of land areas, or if they are located to cover a representative sample of Earth's terrestrial ecoregionsterm. On the other hand, we find that if protected areasterm are directed in a cost-efficient mannerterm to protect threatened vertebratesterm, these species could be protected for an estimated agricultural opportunity costterm of about $42.5 billion annually. We also find that there is a nonlinear relationship between cost and species protection, indicating that options exist for increasing threatened speciesterm protection above the business-as-usual level at little additional cost.

Our estimate of the cost of reaching adequacy targetsterm for all threatenedterm birds, mammals, and amphibians is lower than the $58 billion annually estimated for protecting the world's IBAs [5] , though each option comprises a similar land area. There are three primary reasons for this. First, the estimated costs of protecting IBAs include management coststerm·var, which are estimated at ∼$7 billion annually [5] . Second, IBAs are identified for their contribution to global bird conservation, without consideration of the cost of protecting these areas, whereas we used an optimization approach to identify low-cost options for meeting conservation targetsterm·var [33] , [34] . Third, IBAs are identified based on the presence of both threatenedterm and nonthreatened species (e.g., congregatory species), while we focused on threatened speciesterm alone.

Our analyses are subject to a number of caveats. First, we considered relative cost based on gross agricultural rentsterm, not management coststerm·var or the opportunity costs for other land uses [33] , nor the practicalities of establishing reserves among these competing land uses. Second, overlay of coarse scale maps of species distributions onto fine-scale protected areaterm maps generates commission errorsterm [26] , [35] , though these are unlikely to qualitatively change our results. Still, as commission errorsterm mean that species distributions overlap less than these coarse-scale maps suggest, our estimate of the area needed to protect all threatened speciesterm is a minimum [30] . Locations identified here should therefore be considered as broad indications of where specific areas for protection might be located, and our estimates of cost and the area requiring protection will be minima. Third, although we recognize that our analyses have limited taxonomic breadth, no other taxonomic groups (e.g., plants) have undergone comprehensive assessment of both extinction risk and distribution at a sufficiently fine scale for a comparable analysis [10] . Yet good indications exist from the literature that protected areasterm identified for broad taxonomic groups cover the majority of species in other, nontarget groups [36] , [37] . Finally, our species-specific targets for protection do not account for minimum viable protected areasterm or connectivityterm and do not guarantee the long-term survival of all species. Moreover, many species are threatenedterm by processes other than habitat lossterm and therefore require additional conservation actionsterm both inside and outside protected areasterm [38] .

For the global protected area networkterm to fulfill its potential role as the cornerstone of biodiversity conservationterm [39] , and for governments to meet their commitments on protected areasterm and species extinctions, the distribution of threatened speciesterm must inform future protected areaterm establishment. Preventing the further loss of all threatened speciesterm is a lofty goal and will require substantial efforts. But expanding protected areasterm requires managing tradeoffs among societal objectivesterm [40] , and here we have shown that considerable increases in protected area coverageterm·var of species could be achieved at modest additional cost. Exploiting the nonlinearity of this tradeoff will require directly linking the Aichi targets on protected areasterm and threatened speciesterm (as well as other targets, including target 5 on slowing habitat lossterm), thereby formalizing the interdependence of these key commitmentsterm.

Supporting Information The total extant geographic range sizeterm, in logarithmic scale, and the percent of that range in protected areasterm for 4,118 threatened vertebratesterm, with the red line detailing the range-based conservation targetsterm·var used in the analyses. “a” shows the protection afforded by the current protected areasterm, “b” shows the protection from the current network plus new protected areasterm necessary to meet national-level 17% targets, and “c” shows the protection from the current network plus new protected areasterm to meet national-level 17% targets in a way that ensures terrestrial ecoregionsterm are protected to the level of 17%. Numbers in the graphs give the number of threatened speciesterm that have their adequacy targetterm fully met in each scenario.

(EPS) Click here for additional data file.

Efficiency frontierterm between the cost of establishing additional protected areasterm to achieve 17% coverage and the number of species potentially covered for the original range mapsterm (black circles) and the randomly reduced species range mapsterm (red stars). The y -axis presents the proportion of each species adequacy targetterm·var that is met within protected areasterm, summed across all species. The red stars show the average results from 100 iterations of randomly deleting a portion of each species range; standard deviations for the 100 runs average ±0.82% across the tradeoff frontier and are therefore too small to graph.

(EPS) Click here for additional data file.

Efficiency frontierterm between the cost of establishing additional protected areasterm to achieve 17% coverage and the number of mammal species potentially covered for original range mapsterm (black circles) and the ESH maps (red stars). The y -axis presents the proportion of each species adequacy targetterm·var that is met within protected areasterm, summed across all species.

(EPS) Click here for additional data file.

Analyses of sensitivity to range mapterm commission errorsterm.

(DOCX) Click here for additional data file.