Diversity and Distributions, (Diversity Distrib.) (2011) 1–9

BIODIVERSITY Integrating spatially explicit RESEARCH projections into extinction risk assessments: a reassessment of Amazonian avifauna incorporating projected deforestation Jeremy P. Bird1*, Graeme M. Buchanan2, Alexander C. Lees3, Rob P. Clay4,5, Pedro F. Develey4,5, Itala Ye´pez4,5 and Stuart H. M. Butchart1

1BirdLife International, Wellbrook Court, ABSTRACT Girton Road, Cambridge CB3 0NA, UK, Aim We aimed to complete the first systematic assessment of extinction risk 2Royal Society for the Protection of , 2 Lochside View, Edinburgh Park, Edinburgh based on projected population declines derived from spatially explicit habitat EH12 9DH, UK, 3Departamento de Zoologia, projections for any taxonomic group at a regional scale, to use the outputs to Museu Paraense Emı´lio Goeldi, Caixa Postal ascertain the efficacy of an existing protected area network in covering species of 399, CEP 66040-170, Bele´m – Para´, , conservation concern, and identify gaps therein. 4 BirdLife International Americas Secretariat, Location This study focused on Amazonia; an area of exceptional biodiversity, A Journal of Conservation Biogeography Juan de Dios Martı´nez Mera N35-76 y Av., currently experiencing the highest absolute rate of forest loss globally but where 5 Portugal, Quito, Ecuador, BirdLife the proportion of species assessed as ‘threatened’ on the International Union for International Americas Secretariat, Casilla the Conservation of Nature (IUCN) Red List in the region is below global 17-17-717, Quito, Ecuador averages.

Methods For all forest-dependent Amazonian species (814), we revised extinction risk estimates by combining data from a spatially explicit deforestation model with generation length estimates. By overlaying distribution maps for these revised threatened species, we identified crisis areas (areas of projected deforestation supporting the highest numbers of threatened species), refugia (areas projected to retain forest supporting the highest numbers of threatened species) and areas of high irreplaceability: short- and long-term priorities for new protected areas (PAs).

Results The number of species qualifying as threatened rose substantially from 24 (3%) to 64–92 (8–11%). Areas of particular concern are the crisis and highly irreplaceable areas within the ‘arc of deforestation’ in the southern Brazilian Amazon states of Rondoˆnia, Mato Grosso and Para´.

Main conclusions Through a novel application of the IUCN Red List criteria, we present a spatially accurate rendering of the extinction risks of Amazonian birds. Important areas in the Amazon are not secure. We identify priorities for expansion of the PAs network and key locations where protection should be enforced. We recommend a collaborative approach employing our methods to

and Distributions repeat this process for other taxonomic groups. *Correspondence: Jeremy P. Bird, BirdLife Keywords International Pacific Secretariat, 10 MacGregor Road, Suva, Fiji. Amazon, deforestation, IUCN Red List, population trends, priority-setting, E-mail: jez@birdlifepacific.org.fj threatened species.

but is threatened by the highest absolute rate of global forest INTRODUCTION loss and degradation (Hansen et al. 2008; Nepstad et al., 2009). 2 Diversity Amazonia’s 6.2 million km of tropical forest hosts perhaps the To date, c.18% of the region’s tropical forest has been cleared, greatest biological diversity on Earth (Myers & Myers, 1992), with average annual losses in the last decade of 1.8 million

DOI: 10.1111/j.1472-4642.2011.00843.x ª 2011 Blackwell Publishing Ltd http://wileyonlinelibrary.com/journal/ddi 1 J. P. Bird et al. hectares per year (INPE 2011). Development of infrastructure requires the identification of individual species and sites under is recognized as a contributing driver of this forest loss. greatest threat from deforestation: past, present or future Extrapolations interpreting planned infrastructure develop- (Feeley & Silman, 2009). ments across the Amazon basin predict that up to 40% of Here we address this by estimating projected rates of forest cover will be lost during the period 2002–2050 (Soares- population decline based on modelled deforestation under Filho et al., 2006); or up to 42% of the Brazilian Amazon by governance [GOV] and business as usual [BAU] scenarios 2020 (Laurance et al., 2001). In addition to the 32 ± 8 Pg of sensu Soares-Filho et al. (2006) for all Amazonian species in carbon expected to be released into the atmosphere, two-thirds one taxonomic group: birds (class Aves; nomenclature follows of the forest cover of six major watersheds and 12 ecoregions is BirdLife International 2010 throughout). We applied these predicted to be lost, resulting in a 40% decline of forest cover estimates to the ‘A4’ criterion of the IUCN Red List categories within the ranges of 382 mammals (Soares-Filho et al., 2006). and criteria (IUCN 2001) to revise extinction risk categories. Between 5% and 18% of endemic mammals in the Brazilian The ‘A4’ criterion requires ‘an observed, estimated, inferred, Amazon could be threatened with extinction (Grelle, 2005), projected or suspected population reduction (up to a maximum of while 5–9% of plants may be ‘committed to extinction’ (Feeley 100 years) where the time period must include both the past and & Silman, 2009). the future, and where the causes of reduction may not have ceased Widespread, but targeted forest protection through coordi- or may not be understood or may not be reversible’. The IUCN nated land use and conservation planning is required to avoid Red List guidelines (IUCN 2008) stipulate thresholds for these outcomes. The Brazilian government currently addresses declines which, if passed within 10 years or three generations forest conservation through the Sistema de Protec¸a˜oda (whichever is longer), qualify species for classification as Amazoˆnia (SIPAM), the Amazon Region Protected Areas Critically Endangered (decline ‡ 80%), Endangered (‡ 50%) program (ARPA) initiatives (Presideˆncia da Repu´ blica, Casa or Vulnerable (‡ 30%). This approach was not previously Civil 2004) and the 1965 Brazilian forest code that requires possible as generation lengths had not been calculated for all landowners to preserve forest on 80% of their property in the birds, but we use for the first time a new global dataset of avian Amazon (Brasil 1965). Concurrently, conservation organiza- generation lengths to generate assessments of extinction risk tions are working to increase the numbers of protected areas that account for the considerable impact that generation times (PAs) (Bates & Demos, 2001) drawing on area- and species- have on them. We then examined how well the revised set of based priority-setting approaches (Rodrigues et al., 2006) and threatened species is covered by existing PAs and the network rigorous analyses of regional, national, site and species-level of Key Biodiversity Areas (Eken et al., 2004; Langhammer threats (TNC 2000; Salafsky et al., 2008; Jarvis et al., 2010). et al., 2007; Devenish et al., 2009) identified for birds [the These approaches rely on accurate and recent assessments of Important Bird Areas (IBAs)]. We identified highly irreplace- species ranges and conservation status. able areas and refugia in the current and future forest cover of Despite recognition of the region’s importance for biodi- the Amazon, and compared these areas with the distribution of versity, the threats it faces and the responsive investment in PAs and IBAs to consider (1) their capacity to buffer against conservation from governments and conservation Non Gov- future forest loss in the region and (2) priority sites for future ernmental Organizations alike, the Amazon does not feature expansion and management of the PA network in the face of prominently in major area-based priority-setting mechanisms projected land use change. (Bates & Demos, 2001; Jarvis et al., 2010), nor in species-based approaches. The numbers of threatened species in different METHODS taxonomic groups fall well below global averages on the International Union for the Conservation of Nature (IUCN) Calculating forest loss within species’ ranges Red List (e.g. Grelle et al., 1999; Feeley & Silman, 2009; BirdLife International 2010; IUCN 2010) – the most highly We used digital distribution maps (Ridgely et al., 2003; regarded and widely adopted system for measuring extinction BirdLife International 2010), and altitudinal data from BirdLife risk (Mace et al., 2008). This may be because IUCN Red List International (2010), to calculate the percentage of forest cover assessments have not yet taken account of recently available within species’ geographic and altitudinal ranges that falls information on projected regional patterns of future defores- within the PanAmazon [‘Amazonia’; defined as the Amazon tation. Only one study (Vale et al., 2008) has attempted to river watershed, the Legal Amazon in Brazil and the Guiana interpret the impacts of predicted deforestation on Red List region (Soares-Filho et al., 2006)]. Forest cover was extracted assessments of extinction risk. This identified eight species from GLC2000 (classes 1 – 8; Bartholome & Belward, 2005), within the Brazilian Amazon that may hypothetically qualify as while altitude came from 30 arc second SRTM data (USGS, threatened in the future, once their ranges contract below a 2004). There were 814 resident bird species with ‘high’ or threshold extent of occurrence (20,000 km2; see IUCN 2001) ‘medium’ dependence on forest (BirdLife International 2010) on the basis of predicted deforestation rates. Unfortunately, for which Amazonia contained the majority of remaining this method does not inform changes to the current Red List, habitat (i.e. > 50% – see Supporting Information Data S1). and so cannot be used proactively in conservation planning. Recognizing the influence of seasonally inundated va´rzea Targeting of resources by conservation practitioners instead versus unflooded terra firme forest on species’ distributions

2 Diversity and Distributions, 1–9, ª 2011 Blackwell Publishing Ltd Extinction risk of Amazonian avifauna

(Remsen & Parker, 1983; Aleixo, 2002), we used Remsen & List category revisions requires expert review on a species by Parker (1983) and Stotz et al. (1996) to identify va´rzea forest species basis of projected population changes from this endemics. We used Hess et al. (2003) to further refine range analysis, plus generation length, altitudinal range, geographic maps to represent the extent of va´rzea forest within these range and other sources of uncertainty). species’ ranges. The area of forest in 2002 within each of the 814 focal species’ altitudinal range was used to define a baseline Identifying priority areas for conservation extent of suitable habitat (ESH). We used individual yearly predictions from ‘SimAmazonia 1’ (Soares-Filho et al., 2006) We identified priority areas for conservation in three ways. (1) to revise ESH estimates for each species after three generations Crisis areas – maps of 2002 forest cover within the ranges of all [following the IUCN Red List guidelines (2008); generation revised threatened species were overlaid to illustrate current length estimates from BirdLife International’s World Bird concentrations of such species. From these maps the top 10% Database were used] under BAU and GOV scenarios (Soares- of pixels that support the highest number of revised threatened Filho et al., 2006). All spatial data manipulations were species in 2002 forest cover, but where forest cover is predicted undertaken in imagine 8.5 (Erdas 2001) or arcmap 9.1 (Esri to be lost over the next three generations, were identified as 2006). Analyses were conducted at a 1-km resolution and used ‘crisis areas’. Such areas are priorities for urgent action to try to an equal area projection. prevent projected habitat loss. (2) Refugia – the 10% of pixels Changes in the area of forest cover were converted into in Amazonia that support the greatest number of revised percentage declines, and equivalent population declines over threatened species where forest cover is predicted to persist three generations were inferred (see Data S1 for discussion of after three generations. Such areas are priorities for long-term sources of uncertainty). To account for uncertainty in gener- protection and safeguard. (3) Highly irreplaceable areas – the ation length estimates, we calculated declines over minimum value of each 1 km2 pixel to each species was calculated as 1/ and maximum trend periods (on the basis of minimum and [total ESH], and these values were summed for all species maximum generation length estimates from BirdLife Interna- occurring in each pixel to assess aggregate irreplaceability. tional World Bird Database; or where only one figure was Irreplaceability was calculated for (1) currently threatened available by modifying this by ±25% – one standard deviation species under 2002 forest cover to identify areas that should in the minimum and maximum estimates around the mean already be recognized as highly irreplaceable; (2) revised generation length). For ‘medium’ dependence forest general- threatened species under 2002 forest cover to identify ists, we assumed population declines were 10% lower than % additional recently/currently forested and highly irreplaceable ESH declines (as such species can tolerate secondary to areas and (3) revised threatened species under forest cover a degree), while for species in families known to be susceptible remaining in three generations time to identify areas with the to forest fragmentation, edge effects and hunting pressure, we greatest potential for long-term conservation aimed at max- assumed population declines were 10% higher than ESH imizing the proportion of species’ populations persisting. We declines (see Data S1 for further details). compared the distribution of crisis areas, refugia and highly irreplaceable areas with the distribution of IBAs and PAs. Identifying threatened species RESULTS We based projected population declines on forest loss accord- ing to BAU and GOV scenarios and adjusted decline estimates The generation length for Amazonian forest-dependent birds incorporating uncertainty in generation lengths and non- ranges from 2.5 to 18.5 years with a mean of 4.9 years. Under proportional population responses (see Data S1). These were the IUCN Red List criterion A4, this requires population compared with the thresholds for criterion A4 of the IUCN trends over 10–56 (mean 14.8) years to be assessed (depending Red List (2001). Revised categories were assigned where the on the species). We found that the area of habitat available to rate of decline warranted uplisting the species to a higher Amazonian forest-dependent birds is predicted to decrease category of extinction risk. Revised categories were compared over the next three generations by 15.7% (SE ± 3.3%) and with the current Red List category (to ensure that no species 17.9% (SE ± 3.4%) on average under GOV and BAU scenar- were mistakenly recommended for downlisting when they ios, respectively (see Fig. S1; Appendix S1). Considering the already qualify at higher categories of threat on the basis of degree of forest dependence for each species and their other Red List criteria) and a final revised optimistic and susceptibility to edge effects (see Data S1), our projected pessimistic Red List category was assigned to each species. population declines based on these rates of habitat loss qualify Optimistic categories assume that mitigation measures pro- 64 species (7.8%) of Amazonian birds as threatened under a posed under the GOV scenario are implemented while GOV scenario and 92 species (11.3%) under BAU (Table 1), pessimistic assessments assume BAU. ‘Currently threatened compared with just 24 species (2.9%) currently listed as species’ refers to threatened species on the 2010 IUCN Red List threatened on the 2010 IUCN Red List (BirdLife International while ‘revised threatened species’ refers to all species appar- 2010). The number of species ‘of conservation concern’ ently qualifying as threatened under a BAU scenario (‘appar- (threatened plus near threatened species) increases from 60 ently’ is used here as a qualifier because implementation of Red species (7.4%) currently to 117–172 species (14.3–21.1%)

Diversity and Distributions, 1–9, ª 2011 Blackwell Publishing Ltd 3 J. P. Bird et al.

Table 1 Number of Amazon forest-dependent species that qualify for different International Union for the Conservation of Nature (IUCN) categories: (1) on the IUCN Red List in 2010, under a business as usual (BAU) scenario, and under a governance (GOV) scenario; (2) when uncertainty in species’ generation lengths is incorporated and (3) when uncertainty in species’ responses to deforestation is incorporated.

(2) Incorporating uncertainty in (3) Incorporating uncertainty in (1) IUCN Red List generation length population responses to fragmentation

Revised category Revised category Revised category 2010 Red Revised Revised Revised category (minimum (corrected for (corrected for species List category category (maximum generation species resilient vulnerable to Category (BAU) (GOV) generation length) length) to fragmentation) fragmentation)

Least concern 754 642 697 723 592 713 543 Near threatened 36 80 53 46 100 46 118 Vulnerable 13 66 44 25 90 35 118 Endangered 11 22 17 19 27 18 31 Critically endangered 0 4 3 1 5 2 4 Threatened 24 92 64 45 122 55 153 % threatened 2.9 11.3 7.8 5.5 15.0 6.7 18.8 Of cons. concern* 60 172 117 91 222 101 271 % of cons. concern 7.4 21.1 14.3 11.2 27.2 12.4 33.2

*Of cons. concern combines threatened species categories and Near Threatened species to identify all species of conservation concern. under GOV and BAU scenarios; with 112 species uplisted by threatened species in the Andean foothills and western Guyana one or more Red List categories (Table 1). Of particular are expected to remain largely intact because forest loss is concern are the species that appear to qualify for uplisting to predicted to be minimal in these areas, but those along the Critically Endangered or Endangered that have short genera- , in Amazonas and Para´ states and around the tion times, such as Hoary-throated Spinetail (Synallaxis kollari) Atlantic coast correspond closely with crisis areas and are and Varzea ( varzeae), as these are the species predicted to be lost over the next three generations (Figs S2g & projected to lose considerable proportions of suitable habitat S3c,d). within their ranges most rapidly. A high proportion (54.9%) of the most important areas for Incorporating uncertainty into our estimates of generation revised threatened species are currently protected (Fig. 1a), lengths gives a range of 45–122 species (5.5–15.0%) potentially with some 24.9% covered by the IBA network (see Fig. S3a). qualifying as threatened (Table 1), while incorporating species’ Projected refugia (the 10% of Amazonia that supports the relative resilience or vulnerability to forest fragmentation, edge greatest number of revised threatened species where forest effects and hunting suggested that 55–153 (6.7–18.8%) species cover is predicted to persist after three generations) are may qualify as threatened (Table 1). particularly well captured by the existing PA networks with Current and revised threatened species are not distributed 64.5% of the areas identified as refugia falling within PAs evenly across Amazonia (Fig. S2a,b). The former occur (Fig. 1b), although only 18.2% fall within IBAs (see Fig. S3b). predominantly in western Amazonia particularly in the Indigenous reserves such as the Munduruku, Menkragnoti and foothills of the Andes, while the latter are concentrated in Trincheira Bacaja in Para´ state are theoretically already southern Amazonia, east of the Madeira river on the Brazilian protecting particularly important refugia. As might be Shield and particularly in the highly fragmented Bele´m Centre expected, crisis areas (the top 10% of Amazonia that supports of Endemism (after Cracraft, 1985). These core areas become the highest number of revised threatened species in 2002 forest much reduced in size and highly fragmented following BAU cover, but where forest cover is predicted to be lost over the forest loss (Fig. S2c); refugia are projected to remain in parts of next three generations) are largely unprotected; however, central-southern Amazonia, particularly in Para´ and eastern 19.2% of pixels lie within PA boundaries (and 19.2% of pixels Amazonas states (see Fig. 1b). Crisis areas at greatest risk of lie within IBA boundaries – see Figs 1c & S3c). The areas that losing revised threatened species occur in the central-southern are within, for example, Triunfo do Xingu and Serra do Pardo watersheds of the rivers Tapajo´s and Xingu (sensu Soares-Filho correspond with crisis areas despite being designated PAs. Type et al., 2006; Fig. 1c, and see Fig. S2d). The areas of highest ii highly irreplaceable areas (the top 10% of Amazonia with the irreplaceability for currently threatened species lie in the north greatest irreplaceability index values – see Methods ‘Identifying and west of Amazonia (Fig. S2e), but those for revised priority areas for conservation’) are not well represented threatened species lie in coastal Amazonia, along parts of the within the existing PA and IBA networks, with only 22.7% and Amazon river, Para´ and Amazonas states, western Guyana and 29.9% of pixels currently falling within the respective networks where the Amazon meets the Andean foothills in Peru (see Figs 1d & S3d). There is little overlap (only 0.9% of pixels (Fig. S2f). Highly irreplaceable areas of forest for currently shared) between refugia and type ii highly irreplaceable areas

4 Diversity and Distributions, 1–9, ª 2011 Blackwell Publishing Ltd Extinction risk of Amazonian avifauna

(a) (b)

(c) (d)

Figure 1 Priority areas for conservation in the Amazon basin with Protected Areas (PAs) overlaid: (a) distribution of the top 10% of cells that support the highest number of revised threatened species in 2002 forest cover; (b) distribution of refugia, or the top 10% of cells that support the highest number of revised threatened species in future forest cover; (c) distribution of crisis areas, or the top 10% of cells that support the highest number of revised threatened species that are predicted to be deforested; (d) distribution of the top 10% of cells with the highest irreplaceability for revised threatened species in 2002 forest cover. Areas outside existing PAs are shown in black while those already occurring inside PAs are shown in dark grey.

for revised threatened species (see Fig. 1b,d), but there is tion of responsible practices within the cattle and soy sectors, considerable overlap (21.4% of pixels shared) between type ii and enforcement and expansion of a public and private highly irreplaceable areas and crisis areas (see Fig. 1c,d). network of PAs. The effectiveness of PAs at conserving biodiversity has been questioned, but they evidently can serve to reduce habitat loss and degradation (Ferraro & Pattanayak, DISCUSSION 2006; Andam et al., 2008; Nelson & Chomitz, 2009). To Incorporating projected deforestation into avian threat assess- maximize the benefits of PAs in the region, we advocate ments increases the number of species qualifying as threatened urgently extending protection to priority sites that cover crisis on the IUCN Red List considerably and provides a more areas where important forest will imminently be lost. This accurate reflection of the extinction risk facing species in the analysis highlights gaps in existing site-based priorities and the region. Our revised estimates of extinction risk to Amazonian current reach of PAs. Integration of these priorities into birds brings them closer in line with global averages, despite national and regional development agendas is a recommended the Amazon currently retaining vast areas of primary forest step towards protecting Amazonia (BirdLife International, and supporting many widespread, homogenously distributed 2009). species. The additional revised threatened species identified The crisis areas that we identify correspond with the ‘arc of (and the proposed elevated categories of extinction risk for deforestation’ spanning the states of Rondoˆnia, Para´ and Mato currently threatened species) should be considered priorities Grosso. This broad ribbon of southern Amazonia has the for incorporation into the IUCN Red List following review dubious distinction of being the most active land use frontier through the existing Red List evaluation processes (see http:// in the world, but coverage of strictly PAs is low (Morton et al., www.birdlife.org/globally-threatened-bird-forums/ for further 2006); a concern for its highly fragmentation-sensitive avifau- details). nal community (Lees & Peres, 2006). In spite of arguments that The conservation of threatened species in Amazonia will these areas have been undervalued by global-scale area-based require measures identified under the GOV scenario outlined prioritization approaches (Bates & Demos, 2001), they were in Nepstad et al. (2009) and Soares-Filho et al. (2006), identified as an important priority over the next 2–5 years for nominally: support of low-deforestation livelihoods, promo- conservation investment (Jarvis et al., 2010), even though

Diversity and Distributions, 1–9, ª 2011 Blackwell Publishing Ltd 5 J. P. Bird et al.

Jarvis et al. (2010) explicitly excluded from their model threats setting tool, the IUCN Red List, and the effective protection of like new infrastructure that operate over longer timescales. key sites that support revised threatened species could avert Conservation on privately owned land, of which 80% must be imminent crises. set-aside as forest reserves (Brasil 1965), has long held promise, Effective conservation planning must marry spatially explicit but remains a major doubt considering proposed changes to assessments of threat and biodiversity. Opportunities are the Brazilian Forestry code (Metzger et al., 2010). In May 2011, emerging to finance PA management in Amazonia with the Brazil’s lower parliamentary house approved a bill to revise the joint aims of reducing carbon emissions and maximizing Forestry Code and this legislation is currently being debated in biodiversity conservation (Nepstad et al., 2009; Venter et al., Brazil’s Senate. If passed, the bill would reduce the percentage 2009). This will require a better understanding of the of land a private landowner is legally required to maintain as distribution of both biodiversity and the threats that it faces forest (including, critically, a reduction in the width of forest (Feeley & Silman, 2009; Jarvis et al., 2010). There is now scope buffers alongside perennial steams) and includes an amnesty to synthesize existing analyses for a range of taxonomic groups for landowners who deforested before July 2008 (who would in Amazonia, and incorporate emerging threats like the subsequently be absolved of the need to reforest illegally expansion of biofuel crops and climate change impacts into cleared land). A spike in deforestation rates in March–April forest cover scenarios, to assess the future efficacy of the 2011 (593 km2 cleared, 81% in the crisis region of Mato existing PA network for all biodiversity. Grosso) represents a 473% increase on the same time period in Our approach could be developed to support optimized 2010 and was likely sparked by speculation associated with the land use and conservation planning. First to utilize species- new bill (INPE 2011). The existing network of PAs needs specific data on forest dependence so that likely responses to strengthening in specific areas (illustrated in Fig. 1c) to help deforestation, degradation and fragmentation can be modelled avert pending crises where significant proportions of the ranges on a case-by-case basis; at present, some species with low of revised threatened species are predicted to be lost in the near sensitivity to deforestation may be included in the analyses (see future. The refugia we identify already benefit from some legal Data S1). Secondly for those species that are uplisted on the protection (see Fig. 1b), but unprotected sites should be Red List, regular review of their status in future will be treated as long-term priorities for protected status. important to determine whether the projected rates of habitat Modelling exercises are only as good as the assumptions (and hence population) loss eventuate. In wilderness areas, upon which they are based. Nepstad et al. (2009) argued that where biodiversity monitoring is absent, the most effective BAU predictions (sensu Soares-Filho et al., 2006) may have approach to monitoring is likely to be the use of surrogate been overly pessimistic given a decline in deforestation within spatially explicit data (including remote sensing) to inform the region to 36% of historical levels between 2005 and 2009 trends in land cover change (Buchanan et al., 2008; Hansen (although the recent deforestation spike mentioned earlier et al., 2008). Thirdly, lists of threatened species will require renews concern), and Heywood & Stuart (1992) point out that regular reappraisal whenever better distributional data deforestation may not automatically lead to a dramatic loss of becomes available and to include newly discovered species biodiversity. While some areas have been clear felled for cattle and new species-level taxa that arise through taxonomic ranching, others originally logged or cleared by small holders inflation (Peterson, 1998; Isaac et al., 2004; Padial & De la are regenerating and may buffer some species responses to Riva, 2006; see Data S1). This latter driver of the description of deforestation and degradation (Stork, 2010). Nevertheless, no new species has been brought about by the now widespread use model to date has incorporated the synergisms associated with of molecular and vocal characters in assessing species limits in the interconnected phenomena of logging, climate change, neotropical birds which has revealed the presence of sub- altered fire frequency and resultant potential savannization – stantial cryptic ornithodiversity within polytypic parent species all of which may have profound implications for species between Amazonian interfluvia. It has the potential to distributions and persistence – (Thiollay, 1992; Laurance et al., substantially inflate the number of threatened taxa and 2000; Barlow & Peres, 2004; Hansen et al., 2008; Senna et al., intensify the patterns described here. Finally, future analyses 2009; Jarvis et al., 2010) with models of future infrastructure should seek to incorporate projected patterns in threats other development and biofuel/agricultural expansion. In the face of than deforestation. For example, forests in Roraima are such uncertainty, IUCN recommends a precautionary predicted to remain under the BAU scenario (Fig. S2g), but approach to evaluating extinction risk through the Red List the region is seriously threatened by a recent increase in fire (IUCN 2008). We suggest that even if some of the revised frequency (Jarvis et al., 2010) which is not incorporated into threatened species are found, following case-by-case reviews the SimAmazonia model (see Methods). (or a greater knowledge of species life histories), not to merit Historically, the vast majority of avian extinctions have uplisting, the broad spatial patterns of threatened species occurred on islands driven by the twin threats of over distribution and irreplaceability described here will remain exploitation and invasive alien species (Butchart et al., 2010). valid. As such, our results can be used now to inform The extinction rate on islands is now slowing, presumably conservation activities in the region. Broadly, they show that because many susceptible species have already gone extinct, but the current extinction risk facing species in Amazonia has been by contrast the extinction rate on continents is increasing. The underestimated on the most prominent species-based priority- widespread conversion of natural habitats represents a major

6 Diversity and Distributions, 1–9, ª 2011 Blackwell Publishing Ltd Extinction risk of Amazonian avifauna threat to the long-term diversity of continental bird species Bartholome, E. & Belward, A. (2005) GLC2000: a new ap- (Butchart et al., 2010). This study provides a compelling proach to global land cover mapping from Earth observation illustration of this shifting pattern quantifying the impact of data. International Journal of Remote Sensing, 26, 1959–1977. deforestation within one ‘region’ which constitutes a mosaic of Bates, J.M. & Demos, T.C. (2001) Do we need to devalue parapatric areas of endemism – behaving simultaneously as Amazonia and other large tropical forests? Diversity and both an island and a continental system. Distributions, 7, 249–255. Projected extinctions often do not occur for a long time after BirdLife International (2010). The IUCN Red List for birds. the original threatening process. It can be difficult to predict Available at: http://www.birdlife.org (accessed 20 September the length of the lag time between the extinction debt being 2010). accrued now and when it will be paid (Tilman et al., 1994; BirdLife International (2009) Important bird areas in the Brooks et al., 1999), but if accurate assessments of extinction Amazon basin. BirdLife International Americas Secretariat, risk are available this lag time allows conservation practitioners Quito, Ecuador. to target actions aimed at repaying the debt before extinctions Brasil (1965) Brasil, Ministe´rio da Agricultura, Instituto Bra- occur. It has been shown that there is a good fit between the sileiro de Desenvolvimento Florestal lei no. 4.771 (1965) number of birds that would be predicted to go extinct by the Brasilia. species-area relationship (given the scale of deforestation) and Brooks, T., Tobias, J. & Balmford, A. (1999) Deforestation and the number of birds classified as threatened on the IUCN Red bird extinctions in the Atlantic forest. Conservation, List with ‘a high risk of extinction in the wild in the medium- 2, 211–222. term future’, for example in the Atlantic Forest of South Buchanan, G.M., Butchart, S.H.M., Dutson, G., Pilgrim, J.D., America (Brooks et al., 1999). When revised threatened species Steininger, M.K., Bishop, K.D. & Mayaux, P. (2008) Using are taken into consideration this pattern is upheld by our data, remote sensing to inform conservation status assessment: highlighting the significance of deforestation in elevating estimates of recent deforestation rates on New Britain and overall extinction risk in Amazonia (see Data S1 and Table S1). the impacts upon endemic birds. Biological Conservation, Undocumented ‘Centinelan’ extinctions (sensu Wilson, 1992) 141, 56–66. notwithstanding, no Amazonian birds have yet been driven to Butchart, S.H.M., Collar, N.J., Stattersfield, A.J. & Bennun, extinction by humans. To ensure that none are, and that other L.A. (2010) Conservation of the world’s birds: the view from taxonomic groups fare equally well, further investment is 2010. Handbook of the birds of the world, Vol. 15. Weavers to needed to protect areas of remaining habitat, focusing on crisis new world warblers (ed. by J. del Hoyo, A. Elliott and D.A. areas most urgently, followed by areas of high irreplaceability Christie), pp. 13–68. Lynx Edicions, Barcelona, Spain. and refugia. Cracraft, J. (1985) Historical Biogeography and Patterns of Differentiation within the South American Avifauna: Areas ACKNOWLEDGEMENTS of Endemism. Neotropical Ornithology (ed. by P.A. Buckley, M.S. Foster, E.S. Morton, R.S. Ridgely and E.G. Buckley), pp. For helpful comments during preparation of this manuscript 49–84, Ornithological Monographs, No. 36. American we would like to thank T.B., and for the kind provision of Ornithologists’ Union, New Mexico. contributing data we thank L.P. and J.B. The authors’ time for Devenish, C., Dı´az Ferna´ndez, D.F., Clay, R.P., Davidson, I. & completing this analysis was funded from the core programs of Ye´pez Zabala, I. (eds) (2009) Important bird areas Americas – BirdLife International and RSPB. We are extremely grateful for priority sites for biodiversity conservation. BirdLife Inter- the input of two anonymous referees and Nigel Stork, whose national (BirdLife Conservation Series No. 16), Quito, comments greatly helped to improve this manuscript. Ecuador. Eken, G., Bennun, L., Brooks, T.M., Darwall, W., Fishpool, REFERENCES L.D.C., Foster, M., Knox, D., Langhammer, P., Matiku, P., Radford, E., Salaman, P., Sechrest, W., Smith, M.L., Spector, Aleixo, A. (2002) Molecular systematics and the role of the S. & Tordoff, A. (2004) Key biodiversity areas as site con- ‘‘Va´rzea’’-‘‘Terra-firme’’ ecotone in the diversification of servation targets. BioScience, 54, 1110–1118. Xiphorhynchus woodcreepers (Aves: Dendrocolaptidae). The Erdas (2001) ERDAS IMAGINE v8.5. ERDAS Inc, Atlanta, GA, Auk, 119, 621–640. USA. Andam, K.S., Ferraro, P.J., Pfaff, A., Sanchez-Azofeifa, G.A. & Esri (2006) ArcMap 9.1. ESRI, Redlands, CA, USA. Robalino, J.A. (2008) Measuring the effectiveness of protected Feeley, K.J. & Silman, M.R. (2009) Extinction risks of Ama- area networks in reducing deforestation. Proceedings of the zonian plant species. Proceedings of the National Academy of National Academy of Sciences of the USA, 105, 16089–16094. Sciences of the USA, 106, 12382–12387. Barlow, J. & Peres, C.A. (2004) Ecological responses to El Ferraro, P.J. & Pattanayak, S.K. (2006) Money for Nothing? A Nin˜o-induced surface fires in central Amazonia: manage- call for empirical evaluation of biodiversity conservation ment implications for flammable tropical forests. Philo- investments. PLoS Biology, 4, e105. Doi:10.1371/journal. sophical Transactions of the Royal Society B: Biological pbio.0040105. Sciences, 359, 367–380.

Diversity and Distributions, 1–9, ª 2011 Blackwell Publishing Ltd 7 J. P. Bird et al.

Grelle, C.E.V. (2005) Predicting extinction of mammals in the (2001) The future of the Brazilian Amazon. Science, 291, Brazilian Amazon. Oryx, 39, 347–350. 438–439. Grelle, C.E.V., Fonseca, G.A.B., Fonseca, M.T. & Costa, L.P. Lees, A.C. & Peres, C.A. (2006) Rapid avifaunal collapse along (1999) The question of scale in threat analysis; a case study the Amazonian deforestation frontier. Biological Conserva- with Brazilian mammals. Animal Conservation, 2, 149–152. tion, 133, 198–211. Hansen, M.C., Stehman, S.V., Potapov, P.V., Loveland, T.R., Mace, G.M., Collar, N.J., Gaston, K.J., Hilton-Taylor, C., Townshend, G.R.G., DeFries, R.S., Pittman, K.W., Arun- Akc¸akaya, H.R., Leader-Williams, N., Milner-Gulland, E.J. & arwati, B., Stolle, F., Steiniger, M.K., Carroll, M. & DiMiceli, Stuart, S. (2008) Quantification of extinction risk: IUCN’s C. (2008) Humid tropical forest clearing from 2000 to 2005 system for classifying threatened species. Conservation Biol- quantified by using multitemporal and multiresolution ogy, 22, 1424–1442. remotely sensed data. Proceedings of the National Academy of Metzger, J.P.M., Lewinsohn, T.M., Joly, C.A., Verdade, L.M., Sciences of the USA, 105, 9439–9444. Martinelli, L.A. & Rodrigues, R.R. (2010) Brazilian law: full Hess, L.L., Melack, J., Novo, E.M.L.M., Barbosa, C.C.F. & speed in reverse? Science, 329, 276. Gastil, M. (2003) Dual-season mapping of wetland inunda- Morton, D.C., DeFries, R.S., Shimabukuro, Y.E., Anderson, tion and vegetation for the central Amazon basin. Remote L.O., Arai, E., Espirito-Santo, F.D.B., Freitas, R. & Morisette, Sensing of Environment, 87, 404–428. J. (2006) Cropland expansion changes deforestation dynamics Heywood, V.H. & Stuart, S.N. (1992) Species extinctions in in the southern Brazilian Amazon. Proceedings of the National tropical forests. Tropical deforestation and species extinction Academy of Sciences of the USA, 103, 14637–14641. (ed. by T.C. Whitmore and J.A. Sayer), pp. 91–117, Chap- Myers, N. & Myers, N.J. (1992) The primary source: tropical man & Hall, London. forests and our future – updated for the 1990s. WW Norton INPE (2011) Monitoramento da Floresta Amazoˆnica Brasileira and Company, New York. por Sate´lite: Projeto PRODES (in Portuguese). Instituto Nelson, A. & Chomitz, K.M. (2009) Protected area effectiveness Nacional de Pesquisas Espaciais, Sa˜o Jose´ dos Campos, Sa˜o in reducing tropical deforestation: a global analysis of the Paulo. Available at: http://www.obt.inpe.br/prodes/index. impact of protection status. Evaluation Brief 7. The World html (accessed 15 June 2011). Bank, Washington, DC, pp. 31. Isaac, N.J.B., Mallet, J. & Mace, G.M. (2004) Taxonomic Nepstad, D., Soares-Filho, B.S., Merry, F., Lima, A., Moutinho, inflation: its influence on macroecology and conservation. P., Carter, J., Bowman, M., Cattaneo, A., Rodrigues, H., Trends in Ecology and Evolution, 19, 464–469. Schwartzman, S., McGrath, D.G., Stickler, M., Lubowski, R., IUCN (2001) IUCN Red List categories and criteria: version 3.1. Piris-Cabezas, P., Rivero, S., Alencar, A., Almeida, O. & IUCN Species Survival Commission. International Union for Stella, O. (2009) The end of deforestation in the Brazilian the Conservation of Nature, Gland, Switzerland and Cam- Amazon. Science, 326, 1350–1351. bridge, UK. Padial, J.M. & De la Riva, I. (2006) Taxonomic inflation and IUCN (2008) Guidelines for using the IUCN Red List cate- the stability of species lists: the perils of ostrich’s behaviour. gories and criteria. Version 7.0. Available at: http://www. Systematic Biology, 55, 859–867. iucnredlist.org/documents/redlist_guidelines_ v1223290226. Peterson, A.T. (1998) New species and new species limits in pdf (accessed 10 November 2010). birds. The Auk, 115, 555–558. IUCN (2010) Summary statistics for globally threatened Presideˆncia da Repu´ blica, Casa Civil (2004) Plano de ac¸a˜o para species. Available at: http://www.iucnredlist.org/static/stats a prevenc¸a˜o e controle do desmatamento na Amazoˆnia (accessed 20 September 2010). Legal. Grupo Permanente de Trabalho Interministerial para Jarvis, A., Touval, J.L., Schmitz, M.C., Sotomayor, L. & a Reduc¸a˜o dos I´ndices de Desmatamento da Amazoˆnia Hyman, G.G. (2010) Assessment of threats to ecosystems in Legal. Decreto de 3 de Julho de 2003. Presideˆncia da South America. Journal for Nature Conservation, 18, 180– Repu´ blica, Casa Civil, Brası´lia. Available at: http://www. 188. presidencia.gov.br/casacivil/desmat.pdf (accessed 10 November Langhammer, P.F., Bakarr, M.I., Bennun, L.A., Brooks, T.M., 2010). Clay, R.P., Darwall, W., De Silva, N., Edgar, G.J., Eken, G., Remsen, J.V. Jr & Parker, T.A. III (1983) Contribution of river Fishpool, L.D.C., da Fonseca, G.A.B., Foster, M.N., Knox, created habitats to bird species richness in Amazonia. Bio- D.H., Matiku, P., Radford, E.A., Rodrigues, A.S.L., Salaman, tropica, 15, 223–231. P., Sechrest, W. & Tordoff, A.W. (2007) Identification and Ridgely, R.S., Allnutt, T.F., Brooks, T., McNicol, D.K., Mehl- gap analysis of key biodiversity areas: targets for comprehensive man, D.W., Young, B.E. & Zook, J.R. (2003) Digital distri- protected area systems. International Union for the Conser- bution maps of the birds of the western hemisphere, version 1.0. vation of Nature, Gland, Switzerland. NatureServe, Arlington, VA, USA. Laurance, W.F., Vasconcelos, H.L. & Lovejoy, T.E. (2000) Rodrigues, A.S.L., Pilgrim, J.D., Lamoreux, J.F., Hoffmann, M. Forest loss and fragmentation in the Amazon: implications & Brooks, T.M. (2006) The value of the IUCN Red List for for wildlife conservation. Oryx, 34, 39–45. conservation. Trends in Ecology and Evolution, 21, 71–76. Laurance, W.F., Cochrane, M.A., Bergen, S., Fearnside, P.M., Salafsky, N., Salzer, D., Stattersfield, A.J., Hilton-Taylor, C., Delamoˆnica, P., Barber, C., D’Angelo, S. & Fernandes, T. Neugarten, R., Butchart, S.H.M., Collen, B., Cox, N., Master,

8 Diversity and Distributions, 1–9, ª 2011 Blackwell Publishing Ltd Extinction risk of Amazonian avifauna

L.L., O’connor, S. & Wilkie, D. (2008) A standard lexicon for Data S1 Additional background to materials and methods. biodiversity conservation: unified classifications of threats Appendix S1 Individual percentage reductions in forest cover and actions. Conservation Biology, 22, 897–911. within species ranges. Senna, M.C.A., Costa, M.H. & Pires, G.F. (2009) Vegetation- atmosphere-soil nutrient feedbacks in the Amazon for Figure S1 Number of species predicted to experience different different deforestation scenarios. Journal of Geophysical percentile losses of forest. Research, 114, D04104, pp. 9. DOI:10.1029/2008JD010401. Figure S2 Density overlays of threatened species distributions Soares-Filho, B.S., Nepstad, D.C., Curran, L.M., Cerqueira, under different scenarios. G.C., Garcia, R.A., Ramos, C.A., Voll, E., McDonald, A., Lefebvre, P. & Schlesinger, P. (2006) Modelling conservation Figure S3 Priority areas for conservation and Important Bird in the Amazon basin. Nature, 440, 520–523. Areas. Stork, N.E. (2010) Re-assessing current extinction rates. Bio- Table S1 Summary of predicted changes to the International diversity Conservation, 19, 357–371. Union for the Conservation of Nature Red List. Stotz, D.F., Fitzpatrick, J.W., Parker, T.A. III & Muskovits, D.K. (1996) Neotropical birds: ecology and conservation. The As a service to our authors and readers, this journal provides University of Chicago Press, Chicago. supporting information supplied by the authors. Such mate- Thiollay, J.M. (1992) Influence of selective logging on bird rials are peer reviewed and may be reorganized for online species diversity in a Guianan rain forest. Conservation delivery, but are not copy edited or typeset. Technical support Biology, 6, 47–63. issues arising from supporting information (other than missing Tilman, D., May, R.M., Lehman, C.L. & Nowak, M.A. (1994) files) should be addressed to the authors. Habitat destruction and the extinction debt. Nature, 371, 65–66. BIOSKETCH TNC (2000) The five-S framework for site conservation: a Practitioner’s handbook for site conservation planning and The authors work in conservation and academia with individ- measuring conservation success, Vol. I, 2nd edn. The Nature uals focused globally or on South America. The authors are Conservancy, Arlington, VA. directly involved in assessing avian extinction risk as BirdLife USGS (United States Geological Survey) (2004) Shuttle radar International is the International Union for the Conservation topography mission, 30 arc second scene SRTM. Global Land of Nature Red List authority for birds. They are also involved Cover Facility, University of Maryland, College Park, MD, in the spatial prioritization of sites for avian conservation in USA. South America through BirdLife International’s Important Vale, M.M., Cohn-Haft, M., Bergen, S. & Pimm, S.L. (2008) Bird Areas programme. Effects of future infrastructure development on threat status Author contributions: J.B., G.B. and S.B. conceived the ideas; and occurrence of Amazonian birds. Conservation Biology, J.B., G.B. and S.B. collected and analysed data, J.B., G.B., A.L. 22, 1006–1015. and S.B. led the writing and R.C., P.D. and I.Y. contributed to Venter, O., Laurance, W.F., Iwamura, T., Wilson, K.A., Fuller, the writing. R.A. & Possingham, H.P. (2009) Harnessing carbon payments to protect biodiversity. Science, 326, 1368. Wilson, E.O. (1992) The diversity of life. Harvard University Editor: Mark Burgman Press, Cambridge, MA, USA.

SUPPORTING INFORMATION

Additional Supporting Informatinon may be found in the online version of this article:

Diversity and Distributions, 1–9, ª 2011 Blackwell Publishing Ltd 9