Survey
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
Biodiversity and Conservation (2005) 14:2835–2853 DOI 10.1007/s10531-004-0219-9 Ó Springer 2005 How well will Brazil’s system of Atlantic forest reserves maintain viable bird populations? STUART J. MARSDEN1,*, MARK WHIFFIN2, MAURO GALETTI3 and ALAN H. FIELDING4 1 Applied Ecology Group, Department of Environmental and Geographical Sciences, Manchester Metropolitan University, Chester St. Manchester M1 5GD, UK; 253 Shipman Road, Market Weighton, East Yorkshire YO43 3RA, UK; 3Laboratório de Biologia da Conservação, Departamento de Ecologia, Universidade Estudual Paulista (UNESP), C.P. 199, 13506-900Rio Claro, SP - Brazil; 4 Behavioural and Environmental Biology Group, Department of Biological Sciences, Manchester Metropolitan University, Chester Street,Manchester M1 5GD, UK; *Author for correspondence (e-mail: [email protected]; phone: +(0)161-247-6215; fax: +(0)161-247-6318) Received 23 September 2003; accepted in revised form 13 April 2004 Key words: Atlantic forest, Birds, Brazil, Distance sampling, Extinction, Population viability, Protected areas Abstract. It is crucial for biodiversity conservation that protected areas are large and effective enough to support viable populations of their original species. We used a point count distance sampling method to estimate population sizes of a range of bird species in three Atlantic forest protected areas of size 5600, 22,500, and 46,050 ha. Population sizes were generally related to reserve area, although in the mid-sized reserve, there were many rare species reflecting a high degree of habitat heterogeneity. The proportions of forest species having estimated populations >500 ranged from 55% of 210 species in the largest reserve to just 25% of 140 species in the smallest reserve. All forest species in the largest reserves had expected populations >100, but in the small reserve, 28% (38 species) had populations <100 individuals. Atlantic forest endemics were no more or less likely to have small populations than widespread species. There are 79 reserves (>1000 ha) in the Atlantic forest lowlands. However, all but three reserves in the north of the region (Espı́rito Santo and states north) are smaller than 10,000 ha, and we predict serious levels of local extinction from these reserves. Habitat heterogeneity within reserves may promote species richness within them, but it may also be important in determining species loss over time by suppressing populations of individual species. We suggest that most reserves in the region are so small that homogeneity in the habitat/altitude within them is beneficial for maintenance of their (comparatively small) original species compliment. A lack of protection in the north, continued detrimental human activity inside reserves, and our poor knowledge of how well the reserve system protects individual taxa, are crucial considerations in biodiversity management in the region. Introduction There is a consensus that protecting areas as reserves is the best way to maintain wildlife populations in species-rich countries (Bruner et al. 2001). Consequently there is an extensive literature on the minimum size of protected areas needed within a region to contain all its important species 2836 (reviewed by Rodrigues and Gaston 2001). The aim of these complimentarity-based methods is generally to select minimum-sized protected area networks that include every endemic species within at least one area (e.g., Peterson et al. 2000). What has been far less extensively studied is the likely persistence of populations within protected areas, especially if reserves are chosen to maximise species richness, rather than large populations of individual taxa. How much of its original compliment of species a protected area will fail to sustain long-term is a crucial question in protected area design and maintenance. Degree of isolation (Saunders et al. 1991), suitability of the matrix for species survival and movement (e.g., Sekercioglu et al. 2002), and habitat alteration or direct exploitation inside the reserve (Stouffer and Bierregaard 1995) can all influence species persistence. However, the underlying factor influencing the degree to which fragments lose species over what could be very long periods of ‘relaxation’ (Brooks et al. 1999) is size of the protected area (e.g., Gurd et al. 2001) and ultimately, the sizes of populations that it supports (e.g., Pimm et al. 1988). There has been much debate as to the minimum sizes of populations that are viable in the long-term, and the factors that might affect these values (e.g., Lynch and Lande 1998). Estimates of minimum viable populations (MVPs) can range from tens or a few hundred individuals (Howells and Edwards-Jones 1997; Wielgus 2002) to tens of thousands (Tscharntke 1992), with enormous within-taxon variability (Harcourt 2002). Clearly, MVPs are variable, dependent on many factors, not least the timescale over which extinction risk is assessed. In this study, we acknowledge the fact that we do not know the true values of MVPs for the study species, and use a range of MVPs from 100 to 3000, within which the real values are generally thought to fall (e.g., Franklin 1980; Gilpin and Soulé 1986; Lynch and Lande 1998). The danger of significant loss of species from the entire eco-region is particularly acute in hotspots such as the endemic-rich Atlantic forest (Fonseca 1985; Brooks et al. 1999). Only around 7.6% of the original forest remains (Morellato and Haddad 2000), and in some states much less (e.g., Brown and Brown 1992). What is left is generally small and isolated protected areas (Ranta et al. 1998; Chiarello 1999), many of which are impacted by local people (e.g., Redford 1989), and lying within a matrix dominated by agriculture, disallowing persistence or passage by much of the region’s wildlife (Terborgh and Weske 1969; Stouffer and Bierregaard 1995; Marsden et al. 2000). Here we estimate population sizes for a range of bird species in three lowland Atlantic forest reserves of size 5600, 22,500 and 46,050 ha (Figure 1). We identify the proportions of endemics amongst those species with small populations to see whether such taxa are in particular danger of local extinction. Finally, we estimate likely bird population sizes in other Atlantic Forest protected areas to examine the degree to which the current reserve system might protect birds from extinction. 2837 Figure 1. Map of southeastern Brazil showing the locations of Sooretama/Linhares, Ilha do Cardoso and Ilha Grande. States are labeled as follows: PA, Paranà; SP, São Paulo; RJ, Rio de Janeiro; ES, Espı́rito Santo; MG, Minas Gerais; BH Bahia. Methods Study sites Birds were censused in three reserves/reserve complexes (Figure 1): 1. The Sooretama/Linhares complex of east-central Espı́rito Santo state, comprises the 24,250 ha Sooretama Biological Reserve (19 030 S 40 000 W), which lies adjacent to the 21,800 ha Linhares Forestry Reserve, owned by Compania Vale do Rio Doce (Wege and Long 1995; Chiarello 1999). The complex represents a Key Area for Threatened Birds in the Neotropics (Wege and Long 1995). Both reserves lie close to sea level. Besides a small marsh and clearings for administration buildings, the reserves are wholly forested, being composed mostly of ‘terra firme’ forest up to 40 m in height, with some ‘mussununga’ forest containing much smaller trees on sandier soils (Simão et al. 1997). There is a history of logging within the reserves (according to park guards, some illegal logging persists in Sooretama), and some areas are secondary growth. Most of the area around the reserve is pasture, Eucalyptus plantations, arable farmland and coffee plantation, although some small private preserves remain. Hunting is a common practice in the area and may affect abundance of some gamebirds (Chiarello 2000). Bird surveys were conducted in both reserves between 11 August and 16 October 1998 (transect positions and density estimates can be found in Marsden et al. (2001)). 2. Ilha Grande State Park (23 090 S 44 100 W) covers 5600 ha of Ilha Grande (19,000 ha) in the Baia da Ilha Grande, 2 km from the mainland of Rio de 2838 Janeiro State. The dominant habitat is lowland Atlantic forest, although the island is dominated by two peaks, the higher of which reaches 1033 m (Alves 2001). All bird surveys were conducted between June and August 1999 within the state park itself. The island has a long history of human occupation and original forest habitat is highly fragmented (Alves 2001). Further details of study sites and bird density estimates can be found in Marsden et al. (2004). 3. Ilha do Cardoso (25 040 S 47 550 W), in southernmost São Paulo state, is wholly protected within the Ilha do Cardoso State Park covering an area of 22,500 ha. The island rises to 800 m, with major habitats comprising sandplain forest (restinga), lowland evergreen forest through to humid montane forest. Bird surveys were done during July 1999 along the network of narrow trails cut especially for researchers, and were restricted to forest and restinga below 200 m. Further details of study sites and bird density estimates can be found in Marsden et al. (2004). Bird surveys In 1998, a period of 7 weeks was spent practising bird identification and distance estimation prior to the census in Sooretama/Linhares. This followed an extensive period of practice in bird identification from literature and especially from bird recordings. In the 1999 surveys of Ilha Grande and Ilha do Cardoso, data were collected by the same team, and a period of 3 weeks was spent re-training in bird identification and distance estimation prior to the survey. All bird surveying was done by a pair of observers, with MW always the principal recorder. A variable circular plot method (e.g., Reynolds et al. 1980; Marsden et al. 2001) was used to produce density estimates (individuals per km2 ) for birds in the two reserves. Unidentified birds (usually ones which were heard only) were noted. The distance sampling method used allows for unidentified bird contacts, so long as all birds encountered very close to the recorder, are identified (Marsden 1998). In each reserve, transects were set up along existing paths of width <3 m. Census stations were marked out along these transects at intervals of 200 paces (around 200 m). Ideally, each census station should have been positioned randomly, but this was not possible for the given the timescale of the fieldwork. Bird censuses were carried out between 07:00 and 11:30 and only in the absence of rain or heavy mist. This census period was seen as appropriate, because during a pilot study, bird activity tended to be high during the whole morning with a noticeable fall off in activity after 11:30 which lasted much of the afternoon. Birds were counted at each census station for a period of 10 min. The count period commenced immediately on arrival at the station. Recording effort was concentrated within 30 m of the plot’s central point, although for some species such as parrots, birds seen, or more usually heard, at distances up to 50 m away, were also recorded. The distance from the recorder to each bird 2839 encountered was estimated to the nearest metre. The number of individuals in each encounter was recorded, if ascertainable. When approaching a census station, if any birds were disturbed (flushed) from the plot, these were recorded as being present during the census period. Species such as parrots, toucans, pigeons if encountered in flight at census stations were noted, but records were omitted from the density calculations because aerial birds violate an assumption of the census method used (Marsden 1999). For contacts where birds were heard only, the mean group size for visual contacts with that species was substituted for the missing group size values. Data analysis Density and population estimation Data from VCPM were analysed using the DISTANCE 3.5 program (http:// www.ruwpa.st-and.ac.uk/distance). We produced density estimates only for species which were recorded ten or more times. In many species, we pooled bird records across sites to allow more robust modelling of detection functions (e.g., Marsden et al. 2001). Bird groups were entered as clusters and in ungrouped format. All key functions and associated series expansions were considered, and choice of model for detection function governed by Akaike’s information criterion as detailed in Buckland et al. (2001). Data in some instances were right-hand truncated to remove outliers, but because search effort was usually restricted to within 30 m of the recorder, truncation was sometimes unnecessary and at most 10% of the furthest records were excluded. Actual values for truncation and subsequent grouping of records into distance bands was done following visual inspection of detection histograms and checking of density estimate precision under different analysis conditions. Density estimates were used in conjunction with data on reserve sizes to calculate, where possible, total population estimates for each species in each reserve (total population = population density area). Standard errors for population estimates were based on those computed for density estimates by DISTANCE. They may be underestimated, although the standard error that the program attaches to density estimates does have a variance component calculated to account for survey effort replicate design (Buckland et al. 2001). Population sizes across the three reserves Many species were recorded at census stations, but on too few occasions to produce a density estimate. However, it is still important to estimate their likely population size within the reserves as these rarely recorded species are those most likely to have small populations. We examined the relationship between the number of records accumulated and the density estimate derived for species 2840 which were recorded more often. To do this, we used regression analysis with each species’ density estimate as the dependent variable and the number of records accumulated for the species as the independent variable. We then used the regression equation to predict the population density for species for which we had at least one bird record, but for which we could not calculate a density estimate. We considered a number of possible regression curve shapes and retained the equation that explained the largest proportion of variability. We then plotted a cumulative population size curve (most to least abundant in rank order) for all species for which we had a predicted population size. In order to predict population sizes for species that were unrecorded in our survey, but listed as present within the reserve, we fitted an asymptotic curve (cumulative population size ¼ b1 x=ðx þ b2 Þ using species rank as the independent variable. The coefficients (b1 and b2 ) were estimated iteratively using Solver from Excel (Office 2000). Population sizes for unrecorded species were estimated by extrapolating beyond the last recorded species. Although it is always dangerous to extrapolate beyond the limits of the independent variable, it is unlikely that unrecorded species have anything other than tiny populations and the errors are likely to be small relative to the MVPs used in subsequent analyses. To ascertain whether species endemic to the Atlantic forest were more or less likely to have small populations than widespread species, we used chi-squared tests, with Yates’ correction, to examine the proportions of endemic versus nonendemic species that were actually recorded in the reserves. Because there is no definitive list of bird species recorded on Ilha Grande (see Table 1), we did not undertake an analysis for that reserve. Each species known from the reserves was coded according to whether it is endemic to the Atlantic forest or not according to the list in Brooks et al. (1999). To test whether Atlantic forest endemics predominated amongst species likely to have low population sizes, we grouped all bird species known from a reserve into the following: endemic and recorded at points, endemic and not recorded at points, non-endemic and recorded at points, and non-endemic and unrecorded. In effect, a species from Sooretama/Linhares that was actually recorded at one census station or more has an estimated population equal to or greater that 715 individuals, while a species with at least one record from Cardoso should have a population in excess of 473 individuals. Therefore, we tested for differences in proportions of endemic and non-endemic species having populations greater than 714 in Sooretama and 473 on Cardoso. Consideration of other Atlantic forest reserves Available data on the sizes of protected areas within the lowland Atlantic forest were collated from Fernandes (1997). Only protected areas greater than 1000 ha, and having most of their land area within the Atlantic forest lowlands [Endemic Bird Area 075 as defined by Stattersfield et al. (1998)], were included. Highland protected areas were included only if they contain substantial areas of forest below 500 m a.s.l. Protected areas were split by region: the north 2841 Table 1. Summary statistics for the three reserves, their avifaunas and the surveys. Area (ha) Total avifauna spp. No. surveyable forest spp.d No. endemic forest spp.d No. points/counts Point counts per ha No. surveyable species recorded at points (%) No. species for which population estimation possible Sooretama/ Linhares Ilha do Cardoso Ilha Grande 46,050 >286a 221a 48 273 (546) 1:84 112 (51%) 40 22,500 418b 226 73 105 (210) 1:107 71 (31%) 27 5600 200c 140c 76 (152) 1:37 62 (44%) 23 a Data from Parker and Goerck (1997). Data from Martuscelli unpubl. data. c Figures are estimates as no complete list of birds recorded is available. d Excluding waterbirds, waders, rails, birds of prey, kingfishers, swifts, hirundines, owls, nightjars, potoos, mockingbirds, house sparrow and grassland birds such as Anthus, Volatina, Sicalis and Emberizoides. Also excluded on Cardoso are species occupying only the higher altitudes (>500 m) of the island. b which includes the states of Alagoas, Sergipe, Bahia, and Espı́rito Santo, and the south including the states of Rio de Janeiro, Săo Paulo, Paraná, Santa Catarina, and Rio Grande do Sul. Data on the proportion of bird species in surveyed reserves with population estimates less than given thresholds are used to estimate similar proportions in other protected forest areas. A species can be considered in serious danger of extinction if its population size falls below its MVP. As few data are available on MVPs, the numbers and percentages of species under threat were estimated for each reserve over a range of realistic MVPs. Results In total, 546 counts were made at 273 points at Sooretama/Linhares (Table 1). This figure equates to 1 point count per 84 ha. There were 210 counts at 105 points in Ilha do Cardoso State Park (1 per 107 ha) and 152 counts at 76 points in Ilha Grande State Park (1 per 37 ha). There were sufficient bird records to produce density estimates, and hence population estimates, for 40 species from Sooretama/Linhares, 27 species from Cardoso and 23 species from Ilha Grande (Table 1). Population sizes in the reserves In all three reserves, the best fit for population estimates of species against number of records accumulated was a power curve where lnðyÞ ¼ lnðb0 Þþ 2842 b1 lnðtÞ. Regression equations for all three reserves were highly significant (Fmin ¼ 15:4, df ¼ 38, p < 0:001). Figure 2 shows predicted populations of all bird species actually recorded in the reserves based on the regression equations of population estimate against the number of records accumulated for each species. Species are ranked by decreasing predicted population size. The fit of the cumulative population size curve to the data was good for all three sites (Figure 3) allowing estimates of population size for all species in each reserve (Figure 4). The bird species list for Cardoso, the mid-sized reserve, is actually larger than that for the largest reserve, Sooretama. In turn, the nth commonest species in Cardoso is far less numerous than the corresponding species in Sooretama (Figure 4). Figure 5 shows the actual numbers of species, and the percentages of the species in the avifauna of each reserve that have estimated populations lower than a series of realistic MVPs. Numbers of ‘threatened’ species are highly dependent on MVP size, and are especially sensitive to changes at low values of MVP. Importantly, at an MVP of 100 individuals, Cardoso has relatively few species under threat, because while it has many rare species, they still have populations greater than 100. If, however, the MVP rises to just 200, then it has a relatively large proportion of species under threat (because many species have populations estimated between 100 and 200). This, we believe, is a product of the reserve’s habitat heterogeneity, and the scale at which it manifests itself in terms of bird populations. There was no difference in the proportions of endemics and non-endemics with ‘small’ populations in either Sooretama/Linhares (v21 ¼ 0:03, p ¼ 0:88) or Cardoso (v21 ¼ 1:96, p ¼ 0:16). If anything, in the latter reserve, there was a greater proportion of endemics with large populations than expected. Estimates of population size across the protected area network Figure 6 shows the proportions of the avifauna which may be at risk under a range of MVP scenarios (100–3000) for reserves of different size. MVP has a very clear influence on our ideas as to the percentage of bird species safe in reserves. Reserves around 200,000 ha and over appear to be large enough to protect full, or almost full avifaunas, even at medium-to-high MVP values. Conversely, reserves between 1000 and 10,000 ha are expected to lose very high proportions of species: even with an MVP of 150, a reserve of 10,000 ha has around 45%, and a reserve of 1000, almost 50 % of its species having populations below the MVP level. Figure 7 shows the size of each protected area (>1000 ha) in the north and south of the Atlantic forest region. Large reserves are rare: only eight reserves in the south, and not one reserve in the north of the region is larger than 50,000 ha (Sooretama/Linhares is the largest reserve in the north). Twenty-six of 63 (41%) reserves in the south, but only three of 16 (19%) in the north, are larger than 10,000 ha. 2843 Figure 2. Ranked population sizes in the three reserves. 2844 Figure 3. Population accumulation curves showing power regression population estimates (symbols) and fitted curves extrapolated to last known species using an asymptotic function. Order from highest to lowest: Sooretama/Linhares, Ilha do Cardoso and Ilha Grande. Figure 4. Estimated population sizes from the fitted asymptotic function curve. Order from highest to lowest: Sooretama/Linhares, Ilha do Cardoso and Ilha Grande. Discussion How reliable are the population figures? Our dataset from Sooretama/Linhares is one of the largest for birds using point counts in the neotropics, and yet only around half of Sooretama/Linhares’s forest species were recorded during more than 500 point counts, and density 2845 Figure 5. The numbers of species (a) and the percentages of species (b) in the three protected areas (of size 5600, 22,500 and 46,050 ha) with populations greater than a series of MVPs. estimation was possible in only 10–15% of species. One problem is that, even if most birds are detected within a radius of 20 m at each census point, then the area of census at Sooretama was only around 0.7 km2 . With such a small area of census then there is a high likelihood of missing rare species from the census. Most bird species are undoubtedly rare, in Sooretama as in tropical forests generally (e.g., Thiollay 1994), so even with large datasets, indications of abundance for rare species will be either highly imprecise, or lacking altogether. While this is a major problem for conservation biologists conducting autecological studies of such taxa, we were interested in the species abundance curve for the whole community, where estimates of the actual populations of individual taxa were less important. The principal recorder conducted every point in the survey and was well trained in bird identification and distance estimation. However, while the common birds were no doubt identified well by sight and sound, some of the rarer species (of more than 200 species) are more problematic. The problem is 2846 Figure 6. The percentages of species within the avifaunas of reserves with expected populations lower than a series of MVPs against reserve area. Figure 7. Ranked size distribution of protected areas >1000 ha in the northern and southern Atlantic forest regions. Data from Fernandes (1997). not so much with individuals seen close to the recorder, but largely with birds heard but not observed. Distance sampling, in theory at least, allows for undetected or unidentified encounters so long as all birds at, and very close to the recorder, are detected (e.g., Buckland et al. 2001). More problematic is that if in one species, most birds are identified within say 20 m, and in another, within just 5–10 m, then the ratio of encounter rates to resultant density estimates will be different in the two (the first species will have a higher encounter rate for a given density estimate). This is a potential problem when detection functions are combined across habitats or species, when data are pooled across 2847 periods when recorders are becoming better at detecting and identifying birds, or as in this study, where a relationship between encounter rate and density estimates is being identified (Marsden 1998). No study has as yet addressed this problem so in this study we can only caution that population estimates for some of the rarer species, themselves predicted from the relationship between encounter rate and population density in the commoner species, may be underestimated due to identification problems. Two issues concerning placement of sampling points warrant discussion. First, multiplying density estimates by the amount of available habitat is a reasonable method to estimate total populations so long as the sample taken is representative of the area as a whole. Bird surveys in Sooretama/Linhares covered the reserves quite well as many paths cross the reserves (Marsden et al. 2001), although the sample was not placed randomly. In the other areas, bird surveying was more restricted, and this may have caused more serious problems of bias. For example, on Cardoso, we could survey birds only at lower altitudes, but we know that several mid-altitude species are recorded from the island’s peak, where no surveying was done. The same island is also rather different to Sooretama/Linhares in that it contains rather distinct forest types (e.g., restinga) and holds several altitudinal migrants (e.g., Galetti 2001, pers. obs.). The second issue is that bird censusing was usually done from existing paths and this may, to some extent, have biased our estimates of abundance for some bird species. For example, any edge habitat bordering paths open to increased sunlight may be preferentially chosen by hummingbirds, tanagers and other bird species while at the same time may be avoided by other shade-loving, forest interior species. While we do not pretend to know the full effect of such bias on population estimates for individual species, we suggest that the effect is manageable when we identify the trends in abundance across the whole bird community. Inaccuracy in density estimates for individual species may not be such a problem, unless there is systematic underestimation or overestimation across the whole assemblage due to problems with the census method. There have been no rigorous tests of the accuracy of VCPM (where actual density is known) in the census of tropical birds, and data from temperate habitats are few and the results usually ambiguous (Buckland et al. 2001). The most valid tests of point count density estimation against known bird densities remain the two studies by DeSante (1981, 1986). DeSantes (1981) found, for eight bird species in a scrub habitat, that point count density estimates were underestimated by an average of 18% (2–70% in individual species). In a later study of subalpine forest birds, estimates were far less precise, and could be overestimates as well as underestimates (DeSante 1986). Although falling short of actual tests, various studies have attempted to identify factors influencing point count census accuracy, particularly with a view to tailoring field methods and analysis procedures to avoid violations of key method assumptions (e.g., Marsden 1998; Marsden 1999; Tobias and Seddon 2848 2002). Taken together, we suggest that our estimates, or rather the species abundance curve derived from these estimates, may have possible errors of up to 33%. If this was the case, then our calculations of the proportions of species at risk in reserves under the various MVP scenarios would reflect these errors. Implications for species extinction/conservation The main problem with the approach taken in this study is that we do not know the MVP of any of the species. However, there seems little doubt that populations of many species in most reserves are small – over 70% of bird species in reserves less than 10,000 ha in area are expected to have populations less than 500 individuals. Using a species-area related approach, Brooks et al. (1999) reasoned that while there have been no documented extinctions amongst Atlantic forest birds, around 40% of endemic taxa may succumb to extinction in the coming years due to past habitat loss. While there are problems with the approach taken by Brooks et al. (Kinzig and Harte 2000), which may overestimate the magnitude of the extinction ‘debt’ owed, our study identifies a mechanism by which a significant number of local extinctions are expected, primarily as a result of habitat loss/fragmentation. The degree to which these predicted local extinctions meld to produce global extinctions will depend on precise patterns of distribution amongst Atlantic forest endemics and the degree of protection offered to them within the current reserve network. We know far too little about the above to predict future extinction patterns in the Atlantic forest accurately, and work on these issues across the region as a whole is a real priority (see later). The species on the brink of extinction or at least those with very small and isolated populations are, in some cases, well known. Cracids such as Red-billed Curassow Crax blumenbachii, known only from Sooretama and a few other patches, and Jacutinga Pipile jacutinga (Galetti et al. 1997) are prime candidates for global extinction in the medium term. Even more serious is the plight of several critically endangered species, including Alagoas Foliage-gleaner Phylidor novaesi, and Alagoas Antwren Myrmotherula snowi, occurring only in tiny patches in the north (Stattersfield et al. 1998). The message from our study is that there are many other taxa heading for local extinction in small reserves across the region. For some endemic species, this will spell global extinction. In others, local losses may fall short of becoming global extinctions, but we believe that many will end up being restricted to just one or two protected areas. Such species may include almost all the Alagoas endemics, several ‘Espı́rito Santoan/Bahian’ species such as Red-browed Amazon Amazona rhodocorytha and Banded Cotinga Cotinga maculata, which may, in time, contract to the Sooretama/Linhares reserve, and other, currently ‘widespread’ species, including several southern coastal birds. The minimum area requirement to avoid extinction of populations is clearly influenced by population density, and there is greatest concern for the viability, 2849 within reserves, of populations of the higher vertebrates such as birds and large mammals (Schonewald-Cox 1983; Belovsky 1987). Minimum-sized areas to avoid extinction of mammal species from North American reserves was calculated to be upwards of 2700 km2 (Gurd et al. 2001) and up to 18,000 km2 for grizzly bears Ursus arctos (Wielgus 2002). No study to date has examined population sizes for tropical forest bird communities across whole reserves, although areas of 5000–25,000 ha may be needed to hold viable populations of the temperate gamebird Tetrao urogallus (Marshall and Edwards-Jones 1998; Grimm and Storch 2000). Our study suggests that while the region’s largest reserves should retain their full species compliments, reserves smaller than 10,000 ha are in serious danger of losing considerable components of their avifaunas in the long term. Just as some species may need greater areas than others, some reserves may tend to have a higher proportion of very rare species than we might expect for their size. In this study, Cardoso has very high bird species richness, but many species are extremely uncommon or localised on the island. In fact, a high proportion of species are known from just one or a tiny number of records (e.g., Harpia harpyja, Triclaria malachitacea, Platyrinchus leucoryphus). Despite its relatively large size, we suggest that populations of many birds are extremely small, as some are associated with upland areas (e.g., Touit melanonota), or restricted to certain habitats such as restinga scrub (e.g., Phylloscartes kronei) or bamboo (Sporophila frontalis) (Sick 1993). In this way, reserves with particularly large faunal lists for their size (after controlling for how well the area has been studied), may not support large populations of individual species, as high species richness may be a reflection of habitat diversity, while population sizes of species may more reflect the quantity of preferred habitat. The habitat heterogeneity of reserves may be crucial in determining rates of local extinction across the Atlantic forest reserve network. For a given size, we might expect homogeneous reserves to preserve a greater proportion of their current avifaunas than will heterogeneous ones. We suggest that many of the Atlantic forest reserves are so small, in terms of the populations of rarer species that they can support, that homogeneity of habitat, with its accompanying low total species list, is a desirable thing for long-term conservation of their present avifaunas. Perhaps heterogeneity is only desirable in the very largest reserves. We did not actually measure habitat heterogeneity across the reserves, and neither do we pretend that such measures, if we had them, would be valid for all species at all scales (e.g. Ricklefs and Lovette 1999; Tews et al. 2004). However, altitudinal range within the reserves is a useful correlate of habitat heterogeneity and species richness, and we can say with certainty that the altitudinal range represented within the Cardoso protected area was much greater than that in either Sooretama or Ilha Grande. It may be important, and somewhat fortuitous, that Sooretama and Monte Pascoal, the biggest reserves in the north, are comparatively homogeneous (they are low-lying so have little altitudinal zonation and lack really 2850 distinct habitat zonation). In contrast, the large reserves in the Serro do Mar of the south have broad altitudinal ranges with marked altitudinal zonation of endemic birds (Goerck 1999). An important consideration is whether species survive in reasonable numbers in the habitats and areas surrounding protected areas or not (e.g., Marsden et al. 2001; Sekercioglu et al. 2002). For species intolerant of the matrix, maintenance of populations inside reserves or in undesignated forest patches on private land is the only conservation option available. A few highly mobile species, such as some parrots, may be able to move between some Atlantic forest protected areas, especially those around the Serro do Mar or where the matrix retains some forest blocks. This is important as it may allow linkage in the form of metapopulation structure (Gilpin and Hanski 1991), which may assist in both genetic mixing and in the rescue of extinct subpopulations. Realistically, however, the great majority of protected areas in the Atlantic forest, and especially those in the north of the region, must be seen by conservationists as isolates, as far as almost every taxon is concerned. In terms of minimising extinction risk in the future, the biggest reserves (and those with large areas of a given habitat) are the only areas which are likely to maintain their species richness over time. We believe that smaller reserves will likely go through extended periods of faunal relaxation at the end of which the smallest are predicted to be highly impoverished. Protection of the stronghold reserves such as Serra do Mar, Iguaçu, Sooretama/Linhares and Monte Pascoal is critical if species richness, especially amongst endemics, is to be maintained in the long term. Protection will be particularly important in the north of the region where there has been so little provision of large reserves (e.g., Silva and Tabarelli 2000). The situation is nowhere worse than in the Atlantic slope of Alagoas and Pernambuco which has several endemics not found in the south of the region, and where remains just 878 km2 of forest in patches averaging just 1.5 km2 (Stattersfield et al. 1998). The next step is to compare the provision of protection for individual endemic species across the whole network of reserves, to find out (1) whether all endemic species do occur in reasonable numbers within at least one stronghold protected area, (2) to identify gaps in protected area provision for individual taxa or clusters of taxa that could be remediated by additional protected area provision, and (3) to examine the role for extending habitat management efforts to areas adjacent to protected areas to boost populations of key species they contain. To make this step, many more data on population sizes are needed from across important sites in the region, along with an analysis of the spatial arrangement of protected areas, and the suitability of the matrix to properly assess how isolated bird populations are in individual reserves. The issues raised in this study are, of course, not specific to the Atlantic forest, and assessments of the viability of wildlife populations in reserve systems from other regions is needed if we are to know how well systems will conserve their full compliment of species long-term. 2851 Acknowledgements For various help in the field, we thank Lisa Sadgrove, Paulo Guimarães Jr and Eliana Cazetta. Permission to conduct research in Sooretama and Linhares reserves, logistical help and accommodation was kindly provided by Sr Gilberto Gerhardt of IBAMA- Espı́rito Santo, and Sr Renato de Jesus, of Linhares CVRD. Our project in Sooretama and Linhares was funded by Blackpool Zoo. Fieldwork on Ilha do Cardoso and Ilha Grande was funded by Royal Geographical Society, Gilchrist Educational Trust, Percy Sladen Memorial Fund, and the Department of Biological Sciences, Manchester Metropolitan University. For permission to conduct research on Cardoso, logistical help and accommodation, we thank Marcos Campolim. Mauro Galetti receives a CNPq fellowship (Proc. 300025/97-1). Martin Jones kindly improved the manuscript. References Alves M.A. 2001. Estudos de ecologia de aves na Ilha Grande, Rio de Janeiro. In: Albuquerque J., Cândido J.R. Jr., Straube F. and Roos A.L. (eds), Ornitologia e Conservação: da ciência ás estratégias. Editora Unisul, Santa Catarina, Brazil, pp. 61–68. Belovsky G.E. 1987. Extinction models and mammalian persistence. In: Soulé M.E. (ed.), Viable Populations for Conservation. Cambridge University Press, Cambridge, pp. 35–57. Brooks T., Tobias J. and Balmford A. 1999. Deforestation and bird extinctions in the Atlantic forest. Animal Conserv. 2: 211–222. Brown K.S. Jr. and Brown C.G. 1992. Habitat alteration and species loss in Brazil. In: Whitmore T.C. and Sayer J.A. (eds), Tropical Deforestation and Species Extinction. Chapman and Hall, London, pp. 119–142. Bruner A.G., Gullison R.E., Rice R.E. and da Fonseca G.A.B. 2001. Effectiveness of parks in protecting tropical biodiversity. Science 291: 125–128. Buckland S.T., Anderson D.R., Burnham K.P., Laake J.L., Borchers D.L. and Thomas L. 2001. Introduction to Distance Sampling: Estimating Abundance of Biological Populations. Oxford University Press, Oxford, UK. Chiarello A.G. 1999. Effects of fragmentation of the Atlantic forest on mammal communities in south-eastern Brazil. Biol. Conserv. 89: 71–82. Chiarello A. 2000. Influência da caça ilegal sobre mamı́feros e aves das matas de tabuleiros do norte do estado do Espirito Santo. Boletim do Museu de Biologia Mello Leitão 11/12: 229–247. DeSante D.F. 1981. A field test of the variable circular-plot census technique in a California coastal scrub breeding bird community. In: Ralph C.J. and Scott J.M. (eds), Estimating Numbers of Terrestrial Birds. Studies in Avian Biology No. 6. Cooper Ornithological Society, pp. 177–185. DeSante D.F. 1986. A field test of the variable circular-plot censusing method in a Sierran subalpine forest habitat. Condor 88: 129–142. Fernandes M.L.B. 1997. Unidades de Conservação do domı́nio da Mata Atlântica. Documentos do Instituto Socio-ambiental (ISA) 4: 19–54. da Fonseca G.A.B. 1985. The vanishing Brazilian Atlantic forest. Biol. Conserv. 34: 17–34. Franklin I.R. 1980. Evolutionary change in small populations. In: Soulé M.E.and Wilcox B.A. (eds), Conservation Biology: An Evolutionary-Ecological Perspective. Sinauer, Sunderland, pp. 135–150. Galetti M. 2001. Seasonal movements and diet of the Plumbeous Pigeon (Columba plumbea) in a Brazilian Atlantic forest. Melopsittacus 4: 39–43. 2852 Galetti M., Martuscelli P., Olmos F. and Aleixo A. 1997. Ecology and conservation of the Jacutinga Pipile jacutinga in the Atlantic Forest of Brazil. Biol. Conserv. 82: 31–39. Gilpin M.E. and Hanski I. 1991. Metapopulation Dynamics: Empirical and Theoretical Investigations. Academic Press, San Diego, CA. Gilpin M.E. and Soulé M.E. 1986. Minimum viable populations: processes of species extinction. In: Soulé M.E. (ed.), Conservation Biology: The Science of Scarcity and Diversity. Sinauer Associates, Sunderland, MA, pp. 19–34. Goerck J.M. 1999. Distribution of birds along an elevational gradient in the Atlantic forest of Brazil: implications for the conservation of endemic and endangered species. Bird Conserv. Int. 9: 235–253. Grimm V. and Storch I. 2000. Minimum viable population size of capercaillie Tetrao urogallus: results from a stochastic model. Wildlife Biol. 6: 219–225. Gurd D.B., Nudds T.D. and Rivard D.H. 2001. Conservation of mammals in eastern North American wildlife reserves: how small is too small? Conserv. Biol. 15: 1355–1363. Harcourt A.H. 2002. Empirical estimates of minimum viable population sizes for primates: tens to tens of thousands? Animal Conserv. 5: 237–244. Howells O. and Edwards-Jones G. 1997. A feasibility study of reintroducing wild boar Sus scrofa to Scotland: are existing woodlands large enough to support minimum viable populations? Biol. Conserv. 81: 77–89. Kinzig A.P. and Harte J. 2000. Implications of endemics–area relationships for estimates of species extinctions. Ecology 81: 3305–3311. Lynch M. and Lande R. 1998. The critical effective size for a genetically secure population. Animal Conserv. 1: 70–72. Marsden S.J. 1998. Counting single species. In: Bibby C.J., Jones M.J. and Marsden S.J. (eds), Expedition Field Techniques: Bird Surveys. Birdlife International/Royal Geographical Society, London, pp. 53–75. Marsden S.J. 1999. Estimation of parrot and hornbill densities using a point count distance sampling method. Ibis 141: 377–390. Marsden S.J., Whiffin M. and Galetti M. 2001. Bird diversity and abundance in forest fragments and Eucalyptus plantations surrounding a Brazilian Atlantic forest reserve. Biodiv. Conserv. 10: 737–751. Marsden S.J., Whiffin M., Sadgrove L. and Guimarães P. Jr. 2000. Parrot populations and habitat use in and around two lowland Atlantic forest reserves, Brazil. Biol. Conserv. 96: 209–217. Marsden S.J., Whiffin M., Sadgrove L. and Guimarães P.R. Jr. 2004. Bird community composition and species abundance on two inshore islands in the Atlantic forest region of Brazil. Ararajuba 11: 29–36. Marshall K. and Edwards-Jones G. 1998. Reintroducing capercaillie (Tetrao urogallus) into southern Scotland: identification of minimum viable populations at potential release sites. Biodiv. Conserv. 7: 275–296. Morellato L.P.C. and Haddad C.F.B. 2000. Introduction: the Brazilian Atlantic forest. Biotropica 32: 786–792. Parker III T.A. and Goerck J.M. 1997. The importance of national parks and biological reserves to bird conservation in the Atlantic forest region of Brazil. Ornithol. Monographs 48: 527–541. Peterson A.T., Ball L.G. and Brady K.W. 2000. Distribution of the birds of the Philippines: biogeography and conservation priorities. Bird Conserv. Int. 10: 149–167. Pimm S.L., Gittleman J.L., McCracken G.F. and Gilpin M. 1988. On the risk of extinction. Am. Naturalist 132: 757–785. Ranta P., Blom T., Niemela J., Joensuu E. and Siitonen M. 1998. The fragmented Atlantic rain forest of Brazil: size, shape and distribution of forest fragments. Biodiv. Conserv. 7: 385–403. Redford K.H. 1989. Monte Pascoal – indigenous rights and conservation in conflict. Oryx 23: 33– 36. Reynolds R.T., Scott J.M. and Nussbaum R.A. 1980. A variable circular plot method for estimating bird numbers. Condor 82: 309–313. 2853 Ricklefs R.E. and Lovette I.J. 1999. The roles of island area per se and habitat diversity in the species–area relationships of four Lesser Antillean faunal groups. J. Animal Ecol. 68: 1142–1160. Rodrigues A.S.L. and Gaston K.J. 2001. Maximising phylogenetic diversity in the selection of networks of conservation areas. Biol. Conserv. 105: 103–111. Saunders D.A., Hobbs R.J. and Margules C.R. 1991. Biological consequences of ecosystem fragmentation: a review. Conserv. Biol. 5: 18–32. Schonewald-Cox C.M. 1983. Conclusions: guidelines to management: a beginning attempt. In: Schonewald-Cox C.M., Chambers S.M., MacBryde B. and Thomas L. (eds), Genetics and Conservation: A Reference for Managing Wild Animal and Plant Populations. Benjamin/ Cummings, Menlo Park, CA, pp. 414–445. Sekercioglu C.H., Ehlrich P.R., Daily G.C., Aygen D., Goehring D.and Sandi R.F. 2002. Disappearance of insectivorous birds from tropical forest fragments. Proc. Natl. Acad. USA 99: 263–267. Sick H. 1993. Birds in Brazil: A Natural History. Princeton University Press, Princeton, NJ. da Silva J.M.C. and Tabarelli M. 2000. Tree species impoverishment and the future flora of the Atlantic forest of northeast Brazil. Nature 404: 72–74. Simão I., Maësdos Santos F.A. and Aurelio Pizo M. 1997. Vertical stratification and diet of psittacids in a tropical lowland forest of Brazil. Ararajuba 5: 169–174. Stattersfield A.J., Crosby M.J., Long A.J. and Wege D.C. 1998. Endemic Bird Areas of the World: Priorities for Biodiversity Conservation. BirdLife Conservation Series No. 7. BirdLife International, Cambridge, UK. Stouffer P.C. and Bierregaard R.O. Jr. 1995. Use of Amazonian forest fragments by understory insectivorous birds. Ecology 76: 2429–2445. Terborgh J. and Weske J.S. 1969. Colonisation of secondary habitats by Peruvian birds. Ecology 50: 765–782. Tews J., Brose U., Grimm V., Tielborger K., Wichmann M.C., Schwager M. and Jeltsch F. 2004. Animal species diversity driven by habitat heterogeneity/diversity: the importance of keystone structures. J. Biogeography 31: 79–92. Thiollay J.M. 1994. Structure, density and rarity in an Amazonian rainforest bird community. J. Tropical Ecol. 10: 449–481. Tobias J.A. and Seddon N. 2002. Estimating population size in the subdesert mesite (Monias bensci): new methods and implications for conservation. Biol. Conserv. 108: 199–212. Tscharntke T. 1992. Fragmentation of Phragmites habitats, minimum viable population-size, habitat suitability, and local extinction of moths, midges, flies, aphids, and birds. Conserv. Biol. 6: 530–536. Wege D.C. and Long A.J. 1995. Key Areas for Threatened Birds in the Neotropics. BirdLife Conservation Series, No. 5. BirdLife International, Cambridge, UK. Wielgus R.B. 2002. Minimum viable population and reserve sizes for naturally regulated grizzly bears in British Columbia. Biol. Conserv. 106: 381–388. .