Quantifying the biodiversity value of tropical primary,
secondary, and plantation forests
J. Barlow*†‡, T. A. Gardner*, I. S. Araujo†, T. C. A´vila-Pires†, A. B. Bonaldo†, J. E. Costa†, M. C. Esposito†, L. V. Ferreira†,
J. Hawes*, M. I. M. Hernandez§, M. S. Hoogmoed†, R. N. Leite¶, N. F. Lo-Man-Hung†, J. R. Malcolm?, M. B. Martins†,
L. A. M. Mestre**, R. Miranda-Santos†, A. L. Nunes-Gutjahr†, W. L. Overal†, L. Parry*, S. L. Peters††, M. A. Ribeiro-Junior†,
M. N. F. da Silva§, C. da Silva Motta§, and C. A. Peres*
*Centre for Ecology, Evolution and Conservation, School of Environmental Sciences, University of East Anglia, Norwich NR4 7TJ, United Kingdom;†Museu
Paraense Emı ´lio Goeldi (MPEG), Avenida Perimetral 1901, Bairro Terra Firme, 66077-530 Bele ´m, Para ´, Brazil;§Universidade Federal da Paraiba, Cidade
Universita ´ria-Campus I, CEP 58059-900 Joa ˜o Pessoa, Paraiba, Brazil;¶Insituto Nacional de Pesquisas da Amazonia, Avenida Andre ´ Arau ´jo 2936, Petro ´polis,
CEP 69083-000 Manaus, Amazonas, Brazil;?Faculty of Forestry, University of Toronto, Toronto, ON, Canada M5S 3B3; **Laborato ´rio de Ornitologia,
CEM Universidade Federal do Parana ´, Avenida Beira-mar, s/n., CEP 83.255-000 Pontal do Sul, Parana ´, Brazil; and††University of Western Ontario,
London, ON, Canada N6A 4B8
Edited by Rodolfo Dirzo, Stanford University, Stanford, CA, and approved October 5, 2007 (received for review April 11, 2007)
Biodiversity loss from deforestation may be partly offset by the
expansion of secondary forests and plantation forestry in the
tropics. However, our current knowledge of the value of these
habitats for biodiversity conservation is limited to very few taxa,
and many studies are severely confounded by methodological
shortcomings. We examined the conservation value of tropical
primary, secondary, and plantation forests for 15 taxonomic
edge effects. Different taxa varied markedly in their response to
of species restricted to primary forest (varying from 5% to 57%),
yet almost all between-forest comparisons showed marked differ-
ences in community structure and composition. Cross-taxon con-
abundance or species richness data, but much stronger when using
metrics based upon community similarity. Our results show that,
whereas the biodiversity indicator group concept may hold some
validity for several taxa that are frequently sampled (such as birds
and fruit-feeding butterflies), it fails for those exhibiting highly
idiosyncratic responses to tropical land-use change (including
highly vagile species groups such as bats and orchid bees), high-
lighting the problems associated with quantifying the biodiversity
value of anthropogenic habitats. Finally, although we show that
areas of native regeneration and exotic tree plantations can
provide complementary conservation services, we also provide
clear empirical evidence demonstrating the irreplaceable value of
biodiversity indicators ? congruence ? conservation ? tropical forests ?
areas are a vital tool in preventing these losses, their coverage is
currently limited (2) and their integrity is often threatened in
areas undergoing widespread deforestation (3). The potential
limitations of protected areas have led to a growing interest in
the conservation value of the wider anthropogenic landscape
(4–6), an approach that is enshrined in the Ecosystem Principles
of the Convention on Biological Diversity (7). Secondary forests
and tree plantations are of particular importance for biodiversity
conservation as their coverage is rapidly expanding in the tropics
(1, 8) and they may help to retain more forest species than
alternative and more intensive agricultural land uses (6). Fur-
thermore, it has been argued that both planted and naturally
regenerating forests could provide important collateral benefits
in terms of ecosystem goods and services (9), and both habitats
feature prominently in ongoing carbon sequestration initia-
igh rates of deforestation have led to an unprecedented loss
of biodiversity in the humid tropics (1). Although protected
be easily predicted because of three widespread problems in
quantifying biodiversity. First, many biodiversity studies are
either poorly replicated or conducted in very small plots that are
vulnerable to edge effects from adjacent primary forest (11, 12)
and may systematically over-estimate the value of secondary and
plantation forests for forest biodiversity (11, 13, 14). Second, our
understanding of the value of these habitats for different taxa is
incomplete because of a research bias toward certain groups of
vertebrates and pristine habitats (15), which is exacerbated by
studies reporting low levels of congruence between taxa along
gradients of forest degradation (16–18). Finally, there has been
a bias toward examining changes in species richness rather than
community similarity and species composition (19), particularly
Here, we provide a comprehensive assessment of the biodi-
versity conservation value of tropical primary and secondary
forests and tree plantations, using a variety of focal taxa and
response metrics to quantify biodiversity, and a large-scale
replicated experimental design that minimized edge and frag-
mentation effects. The Jari forestry project, established in the
1970s in the north-eastern Brazilian Amazon, provided a land-
scape containing large blocks of 14- to 19-yr-old secondary
forests and 4- to 6-yr-old Eucalyptus plantations, embedded in a
largely undisturbed primary forest matrix [see supporting infor-
mation (SI) Fig. 5] . We selected five study sites in each forest
type (average nearest-neighbor distances between sites in pri-
mary, secondary and Eucalyptus forests were 30 km, 9 km, and
11 km, respectively) that were both spatially independent (SI Fig.
5 and SI Table 1) and large enough to minimize edge effects
(mean size of secondary forest and Eucalyptus blocks were 26
km2and 17 km2, and mean distances of our sample transects
from primary forests were 1.3 km and 1.1 km, respectively). We
sampled 15 different taxa based on the availability of standard-
ized sampling methodologies and taxonomic expertise, and
Author contributions: J.B., T.A.G., T.C.A´.-P., M.C.E., L.V.F., J.H., M.I.M.H., R.N.L., J.R.M.,
M.B.M., L.A.M.M., A.L.N.-G., W.L.O., L.P., S.L.P., M.A.R.-J., M.N.F.d.S., and C.A.P. designed
research; J.B., T.A.G., I.S.A., T.C.A´.-P., A.B.B., J.E.C., M.C.E., L.V.F., J.H., M.I.M.H., M.S.H.,
R.N.L., N.F.L.-M.-H., J.R.M., M.B.M., L.A.M.M., R.M.-S., A.L.N.-G., W.L.O., L.P., S.L.P., M.A.R.-
J., M.N.F.d.S., C.d.S.M., and C.A.P. performed research; J.B. and T.A.G. analyzed data; J.B.,
T.A.G., and C.A.P. wrote the paper.
The authors declare no conflict of interest.
This article is a PNAS Direct Submission.
‡To whom correspondence should be addressed at the present address: Lancaster Environ-
ment Centre, Lancaster University, Lancaster LA1 4YQ, United Kingdom. E-mail:
This article contains supporting information online at www.pnas.org/cgi/content/full/
© 2007 by The National Academy of Sciences of the USA
November 20, 2007 ?
vol. 104 ?
no. 47 ?
assessed the similarity of their responses to habitat conversion
using five different response metrics [overall abundance, ob-
served richness, estimated total richness, community composi-
tion (based on species presence-absence data), and community
structure (based on species composition and relative abundance
data; see Materials and Methods)]. Taxa were grouped by sam-
pling method rather than phylogenetic differences, reflecting the
focus of the existing literature on indicator taxa (16–18), the
unresolved taxonomy of some groups, and the practicalities of
sampling biodiversity in tropical forests.
Results and Discussion
Our examination of 15 different taxa provided the opportunity
to make novel insights into how we interpret and understand the
conservation value of anthropogenic habitats. First, we exam-
ined patterns of biodiversity in the three habitats across all taxa
by comparing patterns of observed species richness, the percent-
age of species unique to each habitat, and community turnover
between habitats. We then explored how the choice of response
metric could effect conclusions regarding the validity of the
indicator taxa concept, and finally examined whether taxa re-
spond in similar ways to land-use change, and identified which
taxa are outliers.
In terms of species richness, five distinct response types were
evident from our samples of each of the 15 taxa within each of
the three forest types (Fig. 1). Only four taxa (trees and lianas,
birds, fruit-feeding butterflies, and leaf-litter amphibians) fol-
lowed the expected gradient with the most and least species-rich
assemblages occurring in primary forest and plantations respec-
tively. In contrast, five taxa showed idiosyncratic responses, with
arachnids, lizards, dung beetles, and bats). Finally, the five
remaining taxa had similar levels of species richness in all
habitats, although one taxon (moths) was least species-rich in
primary forest (although this group also had the lowest sample
representation; see SI Fig. 6).
The percentage of species unique to each habitat was also
highly variable across taxa (Fig. 2). Almost 60% of tree and liana
genera were only ever recorded in primary forest (and the
coarser taxonomic resolution for this taxon probably greatly
underestimates the number of unique species), whereas ?5% of
Eucalyptus plantations (dotted-dash line). (Letters a–e) Five response types, grouping taxa according to our analytical criteria (see Materials and Methods) that
showed the following: significant differences between samples from all habitat types (letter a), no clear significant difference between samples from secondary
forest and Eucalyptus (letter b), no clear significant difference between samples from primary and secondary forest (letter c), no clear difference between any
habitat (letter d), and primary forest appearing as less species-rich than other forest habitats (letter e).
www.pnas.org?cgi?doi?10.1073?pnas.0703333104Barlow et al.
species of orchid bees were unique to primary forest, and both
secondary forests and plantations held almost all of the species
of orchid bee also found in primary forest (Fig. 2). Averaging
(25%) than secondary forests or plantations (8 and 11%, re-
spectively), and secondary forests held more primary forest
species (59%) than plantations (47%). However, there were also
exceptions, and plantations held a higher percentage of primary
forest grasshoppers than secondary forests, and the same per-
centage of bats (Fig. 2). These estimates of the uniqueness of
primary habitat are likely to be highly conservative as they are
influenced by the presence of vagile and transient species
recorded infrequently outside primary forest and by the often
disproportionate difficulties of sampling the full range of diver-
sity found within primary habitats.
between primary forest and other habitats based on community
structure were significant at P ? 0.05, and seven taxa showed
highly significant (P ? 0.01) differences between all pairs of
habitat types (SI Table 2; see SI Fig. 7 for taxon-specific
ordination plots). Only three taxa (orchid bees, fruit-flies, and
small mammals) showed nonsignificant interhabitat compari-
sons at P ? 0.05 between secondary forests and Eucalyptus
plantations (SI Table 2). Comparisons based on community
composition (presence-absence data) were qualitatively similar
to the abundance-weighted comparisons, although there were
fewer significant results in the primary-secondary and second-
ary-plantation comparisons (SI Table 2).
In summary, the results of between-habitat comparisons were
highly variable and highly dependent upon the choice of focal
taxon and the value metric that was selected for analysis. To
further understand this variability, we compared levels of con-
gruency between taxa across our 15 sample sites using five
different response metrics. Community structure showed the
highest number of significant correlations across all taxa (79/105
possible pairwise correlations were significant at P ? 0.05),
whereas very few correlations of species richness (observed or
estimated) or abundance were significant (Fig. 3). Our results
help explain why previous attempts to find taxa that are effective
at indicating the responses of other taxa along a gradient of
structural forest modification have been unsuccessful (refs. 16
and 17; see also refs. 20 and 21), as these empirical tests of the
indicator species concept have based their conclusions on uni-
variate response metrics such as species richness, which retain
little biological information, and are highly sensitive to sampling
effort (22) and the invasion of degraded areas by disturbance-
tolerant species that are often of least conservation concern (23,
Because the highest levels of congruence were found by using
which taxa shared the most similar responses to land-use change,
and which were relative outliers. Mean correlation coefficients
for each taxon show that fruit-feeding butterflies were the best
56% of the variance in community responses), closely followed
by large mammals, lizards, and birds (SI Fig. 6). These patterns
are demonstrated by a two-stage ordination plot (25) based on
the multitaxon response correlation matrix (SI Table 3). Inter-
pretation of this plot is intuitively straightforward, as taxa that
have similar community responses group toward the center of
the plot, whereas taxa that exhibit apparently idiosyncratic
also recorded in secondary forest and plantations (B). Primary, secondary, and plantation forests are represented by gray bars, black circles, and white circles,
The percentage of species unique to primary, secondary, and plantation forests (A) and the percentage of species recorded in primary forest that were
sites based on community structure (CS), community composition (CC), ob-
served species richness (OR), estimated species richness (ER), and abun-
Barlow et al. PNAS ?
November 20, 2007 ?
vol. 104 ?
no. 47 ?
responses appear as outliers (Fig. 4). The distinct responses of
these outlying taxa might be attributed to their high vagility [e.g.,
orchid bees and bats can travel very long distances to locate
ephemeral resources (26, 27), or the difficulty in achieving
adequate sample representation at the site level (small mam-
mals; see SI Fig. 6]. Contrary to previous suggestions (16), our
analysis based on community structure rather than species
richness shows that well studied taxa (such as trees and woody
lianas, birds, large mammals, fruit-feeding butterflies, lizards,
dung beetles, and epigeic arachnids) do seem to share broadly
similar community responses to land-use change (Fig. 4; and see
ref. 19), and the average matrix correlation coefficient between
this subset of taxa was 0.70. Furthermore, the generality of this
result is supported by four independent lines of evidence (see SI
Text), and these selected taxa also clearly distinguish between
different forest types (SI Table 2), providing support for their
potential value as indicators of structural habitat integrity (28,
29). Nevertheless, these data also reinforce the need for focused
research on other taxa that seem to have more unique and less
predictable response patterns, as well as those that are partic-
ularly difficult to sample (30).
Conservation science can help inform political decision-
making regarding land-use strategies by providing clear answers
to questions regarding the biodiversity value of different habi-
tats. This study demonstrates that surprisingly high numbers of
primary forest species can be found in areas of native regener-
ation and exotic tree plantations with an understorey of native
shrubs (Fig. 2), and suggests these habitats can provide impor-
tant conservation services that are complementary to strictly
protected areas (4–6). However, the intact nature of the forest
matrix surrounding the plantations and secondary forests mean
our results should be seen as a best-case scenario, and we also
provide some of the clearest empirical evidence available re-
garding the unique importance of undisturbed primary forests
(Fig. 2), highlighting the importance of retaining comprehensive
reserve networks as part of a wider landscape management
strategy. Finally, we provide insights regarding the complexities
involved in answering simple questions about the biodiversity
conservation value of degraded habitats and caution against
drawing firm conclusions from studies that focus on a limited
number of taxa or inappropriate response metrics.
Materials and Methods
Study Area. The study was conducted within the 17,000-km2Jari
landholding located in the State of Para ´ in north-eastern Bra-
zilian Amazonia (00°27?00??–01°30?00?? S, 51°40?00??–53°20?00??
W). Fifteen transects were established in April 2004, with five
replicate sites in each of three forest types (primary, secondary,
and plantation forests). Although 130,000 ha of Amazonian
primary forests were converted to fast-growing tree monocul-
tures, 80% of the primary forest was left virtually intact. The Jari
landholding currently contains 53,000 ha of Eucalyptus urogran-
dis plantations and a similar amount of regenerating native
vegetation in areas where Eucalyptus or Gmelina stands were
once planted and harvested but subsequently abandoned (SI Fig.
After a detailed landscape structure analysis using the com-
pany GIS database and a recent Landsat image, 15 transects
were preselected and then established across the Jari landscape,
comprising five replicate sites in each habitat type (primary,
secondary, and plantation forests; SI Fig. 5). Transects in
primary forests were selected to reflect regional-scale differ-
ences in soil type and topography, thereby capturing the highest
amount of baseline regional diversity represented before forest
clearance. Primary forest plots had experienced minimal levels
of anthropogenic disturbance, although a few stems of commer-
cially valuable timber (Manilkara sp. and Dinizia spp.) were
felled 30–40 years ago for latex or wood. Floristically, primary
forest tree plots were dominated by the Burseraceae, Sapota-
ceae, Lecythidaceae, and Leguminosae (Caesalpinioideae).
and bulldozed between 1970 and 1980. We selected even-aged
second-growth sites of similar age (14–19 yr) that contained a
high basal area of palms, Inga spp., and other pioneers. Transects
in Eucalyptus stands were located in 4- to 6-yr-old plantations
(stands are harvested at ages 5–7 yr), which were commercially
managed, and the native vegetation in the understorey is peri-
odically cleared or suppressed by using a herbicidal treatment
(Glyphosate and Isoxaflutole). However, all areas were charac-
terized by dense understorey vegetation, consisting of annuals
(including many Rubiaceae and Piperaceae), lianas (e.g., Davilla
sp. Dilleniaceae), and small trees such as Vismia spp. (Clusi-
aceae), Jacaranda copaia (Bignoniaceae), Cecropia sp-
p.(Cecropiaceae), Mabea taquari, and Aparisthmium cordatum
(Euphorbiaceae). Our results are not explained by the spatial
layout of the sampling transects, and all Martel-type RELATE
test correlations between community structure and the geo-
graphic distance between sites were insignificant (SI Table 1).
Sampling Methodologies. Detailed sampling methodologies are
available from the authors on request. Where collections were
necessary, specimens were first sorted to morphospecies, and
voucher specimens were later identified by an expert taxonomist.
Unless stated otherwise, all within-site trapping stations were
100 m apart along a 1-km transect. All sampling was carried out
between May 2004 and June 2005. Samples were repeated during
different seasons wherever possible, and all sampling was ran-
domly alternated across habitat types. In the following, numbers
in parentheses relate to the number of individuals and species
Leaf-litter amphibians (1,739; 23). Each forest type was sampled by
using a combination of pitfall traps (35- and 62-liter buckets in
four-trap arrays, a total of 3,150 and 630 array-nights, respec-
tively), funnel traps (6,300 trap-nights), and transect walks (153
h). All sites were sampled in both 2004 and 2005.
Bats (4,125; 54). Each site was sampled for seven nights by using
ten 12-m mistnets that were set at 100-m intervals in the
understorey, and two 12-m mistnets set in the subcanopy.
Mistnets were moved after two consecutive sampling nights, and
an interval of at least 4 weeks was maintained before resampling
any site. Sampling was conducted during the wet and dry seasons
responses of any given taxon to land-use change across the 15 sites is repre-
sented by their relative proximity on the plot. See Fig. 2 for taxa codes.
Multidimensional scaling ordination of the correlation coefficients
www.pnas.org?cgi?doi?10.1073?pnas.0703333104Barlow et al.
Birds (6,865; 255). Point counts (ten 10-min counts spaced 200 m
apart) and mist netting (twenty-four 12-m mistnets strung along
the transect) were used at all sites. Each primary forest site was
sampled for 3 days. Each secondary forest and plantation site
was sampled for 2 days.
Large mammals [1,227 (direct and indirect observations); 30]. Monthly
diurnal transect walks (0630 hours to 1030 hours) were made
along 2- to 5-km transects at each site for 12 months.
Lizards (1,937; 30). See leaf-litter amphibians for methods, with
additional use of glue traps (3,324 trap-nights).
Small mammals (219; 32). Individuals were captured from the same
pitfall traps as the herpetofauna. A total of 160 baited (Sherman
and Tomahawk) live-traps were deployed for a 10-day period at
Epigeic arachnids [Arachnida; orders Amblypygi, Araneae, Opiliones, Scor-
piones, and Uropygi (3,176; 116; adults only)]. Individuals were
trapped by using the same pitfall traps as the herpetofauna
(samples collected in 2005 only).
Scavenger flies [Diptera; Calliphoridae, Mesembrinellidae, and Sarcoph-
agidae (5,365; 30)]. Each forest type was sampled in the dry (2004)
and wet season (2005) by using suspended traps baited with a piece
of 24-hour-old cow lung. Five traps were used at each site, sampled
across two nights in each year (300 trap-nights in total).
Dung beetles [Coleoptera; Scarabaeinae (9,203; 85)]. Pitfall traps were
baited with human dung. Trapping protocol was as for scavenger
Fruit-feeding butterflies [Lepidoptera; Nymphalidae (10,987; 128)]. Cy-
lindrical traps were baited with fermented banana. Eight traps
were placed in each site, each 100 m apart. Eight additional
canopy traps were suspended in primary forest. Sampling was
conducted for 20 days in each site, spaced over four seasonal
Fruit flies [Diptera; Drosophilinae (5,085; 38; males only)]. Ten traps
baited with fermented banana were suspended for 2 days in each
site during the wet season in 2004.
Grasshoppers [Orthoptera; Acridiidae (932; 44; adults only)]. Sweep-
netting was conducted once along each 1-km transect and within a
10 ? 10-m plot at each site during the wet season of 2005. We also
collected all grasshoppers attracted to light traps (see Moths).
Moths [Lepidoptera; Arctiidae, Saturniidae and Sphingidae (1,848, 335)].
We conducted two nights of light-trapping in the wet season of
2005 at each site (with 30 days between repeated surveys) using
sheets suspended next to two mercury vapor lamps and one UV
lamp, held 100 m apart. Sheets were checked hourly from 1800
hours to 0600 hours.
Orchid bees [Hymenoptera; Apinae, Euglossini (2,363, 22)]. Eight traps
baited with the attractant Methyl Salicylate were suspended for
2 days at each site, repeated over two seasonal replicates in both
2004 and 2005 (480 trap-days in total).
Trees and woody lianas [8,077; 219 (genera)]. All trees ?10 cm and
lianas ?5 cm in diameter at breast height (DBH) were measured
and identified to genus within 1-ha plots at all native (primary
and secondary) forest sites.
Overall Survey Effort. We spent a total of 18,200 person-hours
(excluding all travel time) sampling and identifying specimens
and recorded ?2,000 vertebrate, invertebrate, and plant species
from ?60,000 registration events. Habitat level sample repre-
sentation was ?75% in 34 of 44 cases (see SI Fig. 6).
Data Analysis. Patterns of species richness between different
forest types were compared by using a sample-based rarefaction
procedure within the program EstimateS (v.7) (31), where
individuals are set as samples and the curves are then calculated
by using the Mao Tao estimator (31). This approach allowed
direct comparison of results between groups that differ in
patterns of abundance and were sampled by using very different
techniques. Results were qualitatively indistinguishable from
patterns shown using sample-based curves. The significance of
0.05) was evaluated by visually comparing rarefaction curves and
their associated 95% confidence intervals. If the total observed
richness (for a given taxon) of a more species-poor habitat fell
outside the 95% confidence interval of a more species-rich
habitat, then we inferred that the former sample contained
significantly fewer species than the richer community. Estimated
richness was calculated by using the abundance-based coverage
estimator (ACE) (31). Ordination analyses were implemented in
Primer v. 5 (25) using nonmetric multidimensional scaling
(MDS) and the Bray–Curtis similarity index. There are many
different options for standardizing data, and we decided on two
frequently used alternatives. First, we used square-root trans-
formed and site-standardized the data, which retains some
abundance information but reduces the contribution of the most
by ref. 25). Second, we performed the most severe possible
transformation of abundance data by converting each species-
site abundance matrix into presence/absence data (community
composition) (SI Table 4). Analyses of Similarity (ANOSIM) (a
nonparametric permutation test) were used to test for significant
differences in community structure between disturbance treat-
ments. Congruence between taxa across the 15 sites was com-
pared by using five different response metrics. We used non-
parametric Spearman’s correlations to evaluate congruence for
observed species richness, estimated species richness, and abun-
dance. Abundance data were corrected by sample effort where
this differed across sites. Nonparametric RELATE tests were
used to evaluate congruence and correlate distance matrices
based on community structure (defined as site-standardized and
square-root transformed data) and community composition
(presence-absence data). The correlation coefficients from the
RELATE tests based on community structure formed the basis
of the MDS ordination of response similarity (Fig. 3).
We thank Grupo Orsa and the staff of Orsa Florestal and Jari Celulose in
Jari for logistical support and permission to work in their landholding.
Manoel Reis Cordeiro, Mike Hopkins, and Fernando Vaz-de-Mello as-
sisted with identification work. Pat Hardcastle helped define project objec-
tives. Comments from two anonymous reviewers helped improve the
manuscript. We thank the Brazilian Ministe ´rio de Cie ˆncias e Tecnologia
(MCT-CNPq) and Ministe ´rio do Meio Ambiente (MMA-IBAMA license
numbers 043/2004-COMON, 0079/2004-CGFAU/LIC, 127/2005, and 048/
2005) for permission for this research. The project was funded by the U.K.
Government Darwin Initiative, the National Environment Research Coun-
cil (U.K.), the National Geographic Society, the Conservation Food and
Health Foundation, and Conservation International. This is publication
number 14 of the Land-Use Change and Amazonian Biodiversity project.
1. Mace G, Masundire H, Baillie J (2005) in Ecosystems and Human Well Being,
Current State and Trends: Millenium Ecosystem Assessment, eds Hassan R,
Scholes R, Ash N (Island Press, New York), Vol 1.
2. Rodrigues ASL, Andelman SJ, Bakarr MI, Boitani L, Brooks TM, Cowling
RM, Fishpool LDC, da Fonseca GAB, Gaston KJ, Hoffmann M, et al. (2004)
3. Pedlowski MA, Matricardi EAT, Skole D, Cameron SR, Chomentowski W,
Fernandes C, Lisboa A (2005) Environ Conserv 32:149–155.
4. Daily GC (2001) Nature 411:245.
5. Vandermeer J, Perfecto I (2007) Conserv Biol 21:274–277.
6. Lindenmayer DB, Franklin JF (2002) Conserving Biodiversity: A Comprehensive
Multiscaled Approach (Island Press, Washington, DC).
7. Sayer J, Maginis S (2005) Forests in Landscapes: Ecosystem Approaches to
Sustainability (Earthscan, London).
8. Wright SJ (2005) Trends Ecol Evol 20:553–560.
9. Myers JP (1997) in Nature’s Services: Societal Dependence on Natural Ecosys-
tems, ed Daily GC (Island Press, Washington, DC), pp 215–237.
10. Yu CM (2004) Sequ ¨estro Florestal de Carbono no brasil: Dimenso ˜es Polı ´ticas,
Socioecono ˆmicas e Ecolo ´gicas (Annablume, Sao Paulo).
11. Gardner TA, Barlow J, Parry LTW, Peres CA (2007) Biotropica 39:25–30.
Barlow et al. PNAS ?
November 20, 2007 ?
vol. 104 ?
no. 47 ?
12. Gardner TA, Barlow J, Peres CA (2007) Biol Conserv 138:166–179. Download full-text
13. Barlow J, Mestre LAM, Gardner TA, Peres CA (2007) Biol Conserv 136:212–231.
14. Barlow J, Overal WL, Araujo IS, Gardner TA, Peres CA (2007) J Appl Ecol
15. Fazey I, Fischer J, Lindenmayer DB (2005) Biol Conserv 124:63–73.
16. Lawton JH, Bignell DE, Bolton B, Bloemers GF, Eggleton P, Hammond PM,
Hodda M, Holt RD, Larsen TB, Mawdsley NA, et al. (1998) Nature 391:72–76.
17. Schulze CH, Waltert M, Kessler PJA, Pitopang R, Shahabuddin, Veddeler D,
Muhlenberg M, Gradstein SR, Leuschner C, Steffan-Dewenter I, Tscharntke
T (2004) Ecol Appl 14:1321–1333.
18. Wolters V, Bengtsson J, Zaitsev AS (2006) Ecology 87:1886–1895.
19. Su JC, Debinski DM, Jakubauskas ME, Kindscher K (2004) Conserv Biol
20. Prendergast JR, Eversham BC (1997) Ecography 20:210–216.
21. Caro TM, O’Doherty G (1999) Conservation Biology 13:805–814.
22. Gotelli NJ, Colwell RK (2001) Ecol Lett 4:379–391.
23. Brown KS, Jr (1997) Journal of Insect Conservation 1, 25–42.
24. Pearman PB (1997) Conserv Biol 11:1211–1225.
25. Clarke KR, Warwick RM (2001) Change in Marine Communities: An Approach
to Statistical Analysis and Intepretation (Primer-E Ltd, Plymouth).
26. Janzen DH (1971) Science 71:203–205.
27. Morrison DW (1978) J Mammal 59:622–624.
28. Niemi GJ, McDonald ME (2004) Annu Rev Ecol Evol Systematics 35:89–111.
29. Gardner TA, Barlow J, Araujo IS, A ´vila-Pires TCS, Bonaldo AB, Costa JE,
Esposito MC, Ferreira LV, Hawes J, Hernandez MIM, et al., Ecol Lett, in press.
30. Pawar S (2003) Bioscience 53:861–864.
31. Colwell RK (2004) EstimateS: Statistical Estimation of Species Richness and
Shared Species from Samples (http://purl.oclc.org/estimates), Version 7.
www.pnas.org?cgi?doi?10.1073?pnas.0703333104 Barlow et al.