, 169 (2008);
et al.Yadvinder Malhi,
Climate Change, Deforestation, and the Fate of the
www.sciencemag.org (this information is current as of January 11, 2008 ):
The following resources related to this article are available online at
version of this article at:
including high-resolution figures, can be found in the online
Updated information and services,
can be found at:
Supporting Online Material
, 9 of which can be accessed for free:
cites 43 articles
This article appears in the following
in whole or in part can be found at:
permission to reproduce
of this article or about obtaining
Information about obtaining
registered trademark of AAAS.
is aScience 2008 by the American Association for the Advancement of Science; all rights reserved. The title
Copyright American Association for the Advancement of Science, 1200 New York Avenue NW, Washington, DC 20005.
(print ISSN 0036-8075; online ISSN 1095-9203) is published weekly, except the last week in December, by theScience
on January 11, 2008
Climate Change, Deforestation, and
the Fate of the Amazon
Yadvinder Malhi,1* J. Timmons Roberts,1,2Richard A. Betts,3Timothy J. Killeen,4
Wenhong Li,5Carlos A. Nobre6
The forest biome of Amazonia is one of Earth’s greatest biological treasures and a major
component of the Earth system. This century, it faces the dual threats of deforestation and stress
from climate change. Here, we summarize some of the latest findings and thinking on these
threats, explore the consequences for the forest ecosystem and its human residents, and outline
options for the future of Amazonia. We also discuss the implications of new proposals to finance
preservation of Amazonian forests.
in Brazil. They host perhaps a quarter of the
world’s terrestrial species (3) and account for
about 15% of global terrestrial photosynthesis
(4). Evaporation and condensation over Amazonia
are engines of the global atmospheric cir-
culation, having downstream effects on precipi-
tation across South America and further afield
across the Northern Hemisphere (5, 6). Amazo-
nian forests have been an important and con-
tinuous part of Earth system functioning since
the Cretaceous (7).
By 2001, about 837,000 km2of Amazonian
forests had been cleared (2), with 1990s gross
rates of ~25,000 km2year–1(8). Clearance is
concentrated in the “arc of deforestation” on the
southern and eastern margins, driven primarily
by expansion of cattle and soybean production,
and along the Andean piedmont. Amazonia lies
inside nine nations, but 80% of deforestation has
been in Brazil (2) and 70% of that is provoked
by cattle ranching. From 1988 to 2006, deforest-
ation rates in Brazilian Amazonia averaged 18,100
km2year–1, recently reaching 27,400 km2year–1
in 2004. Brazilian deforestation rates had more
of a combination of falling prices for soy, increased
strength of the Brazilian currency, and active
Brazilian government intervention (9). Roughly
6% of deforested land has remained in cropland,
62% in pastures, and 32% in regrowing vegeta-
tion (10). The overall direct footprint of human
activity in Amazonia is much greater than defor-
he forests of Amazonia (1) covered about
5.4 million km2in 2001, approximately
87% of their original extent (2), with 62%
estation alone and includes logging, hunting, and
fire leakage (see supporting online text).
Global Drivers of Amazonian
In recent decades, the rate of warming in Ama-
zonia (11) has been about 0.25°C decade–1.
Under midrange greenhouse-gas emission sce-
narios, temperatures are projected to rise 3.3°C
(range 1.8 to 5.1°C) this century, slightly more
in the interior in the dry season (12), or by up to
8°C if substantial forest dieback affects regional
biophysical properties (13). At the end of the
last glacial period, Amazonia warmed (14) at
only ~0.1°C century–1.
Changes in precipitation, particularly in the
dry season, are probably the most critical deter-
minant of the climatic fate of the Amazon. There
has been a drying trend in northern Amazonia
since the mid-1970s and no consistent multi-
decadal trend in the south (15), but some global
climate models (GCMs) project significant
Amazonian drying over the 21st century. Pacific
sea surface temperature (SST) variation, dominated
by the El Niño–Southern Oscillation (ENSO), is
particularly important for wet-season rainfall:
El Niño events (warm eastern Pacific) suppress
convection in northern and eastern Amazonia.
However, dry-season rainfall is strongly influ-
enced by the tropical Atlantic north-south SST
gradient; intensification of the gradient (warm-
ing of northern SSTs relative to the south) shifts
the Intertropical Convergence Zone northwards
(interannual time scales) and strengthens the
Hadley Cell circulation (longer time scales), en-
hancing the duration and intensity of the dry
season in much of southern and eastern Amazonia
(16), as occurred in 2005. Interannual variability in
the Atlantic gradient is influenced by remote forc-
ing such as ENSO and the North Atlantic Os-
cillation, as well as by variations in evaporation
induced by strengthening/weakening of the local
trade winds (17). On longer time scales, the Atlan-
tic SST gradient may be strengthened by changes
in the north Atlantic, such as changes of the ther-
mohaline circulation driven by subpolar melting
(18), or a warmer north Atlantic associated with
warmer northern hemisphere continents.
Forest Influences on Regional and
Amazonian forests have a substantial influence
on regional and global climates. Hence, their
removal by deforestation can itself be a driver of
climate change and a positive feedback on ex-
ternally forced climate change. They store 120 ±
30 Pg C in biomass carbon (19), of which 0.5
Pg C year–1(0.3 to 1.1) were released through
deforestation in the 1990s (10). Similar or greater
amounts may be held in soil carbon, but these
are less vulnerable to loss after deforestation
(20). In addition, forest plot studies suggest that
intact forests are a carbon sink (~0.6 Pg C year–1)
(21), particularly in more fertile western Ama-
zonia. The existence of this sink is debated (22)
but is strongly supported by a recent reevalua-
tion of global sources and sinks of atmospheric
carbon dioxide (23). It may be driven by en-
hanced productivity associated with CO2fertil-
ization, changes in light regime, or other factors
not yet identified (24).
The extraction of soil water by tree roots up
to 10 m deep, and its return to the atmosphere (a
“transpiration service”), is perhaps the most im-
portant regional ecosystem service. Basin-wide,
25 to 50% of rainfall is recycled from forests
(25), but this effect is particularly important in
regions where most precipitation is derived from
local convection (see below). Moderate and lo-
calized deforestation may locally enhance con-
vection and rainfall, but large-scale forest loss
tends to reduce rainfall (26), the magnitude of
reduction being dependent on how regional cir-
culation of atmospheric moisture is affected. Some
model studies suggest that the regional forest-
climate system may have two stable states: Re-
moval of 30 to 40% of the forest could push much
of Amazonia into a permanently drier climate re-
gime (27). Dry season rainfall, the most critical
for determining vegetation patterns, is more often
driven by locally generated convection and may
be more strongly affected by deforestation.
Loss of forest also results in (i) decreased
the cloud effect (28), (iii) changes in the aerosol
ocean” atmosphere to a smoky and dusty con-
tinental atmosphere that can modify rainfall pat-
terns (29), and (iv) changes in surface roughness
(and hence wind speeds) and the large-scale con-
vergence of atmospheric moisture that generates
Risks of Amazon Forest Loss Due to
Global Climate Change
Risks of a drying climate. The climate models
employed in the 2007 Intergovernmental Panel
on Climate Change (IPCC) Fourth Assessment
Report (12) show no consistent trend in annual,
1Environmental Change Institute, Oxford University Centre
for the Environment, South Parks Road, Oxford OX1 3QY,
23187, USA.3Met Office Hadley Centre, Exeter EX1 3PB,
UK.4Conservation International, Washington, DC 20036,
USA.5School of Earth and Atmospheric Sciences, Georgia
Institute of Technology, Atlanta, GA 30332–0340, USA.
6Instituto Nacional de Pesquisas Espaciais, São Jose dos
Campos, SP, Brazil.
*To whom correspondence should be addressed. E-mail:
2College of William and Mary, Williamsburg, VA
VOL 31911 JANUARY 2008
on January 11, 2008
Amazon-wide rainfall over the 21st century, but
a tendency to less dry season rain in the east
and more rain in the west and in the wet season
(Fig. 1). Taking the ensemble of 23 IPCC mod-
els as a crude metric of probabilities, some inten-
sification of dry seasons is about 80% probable
in the southeast Amazon and Guyanas, 70% in
the east, 60% in the center, and 30% in the west
(Fig. 1A). The probabilities of more substantial
decline are slightly lower: 70% in the southeast,
60% in the Guyanas, 50% in the east, 40% in the
center, and 20% in the west. The probabilities of
severe decline in dry season rainfall are 50% in
the southeast, 30% in the Guyanas and east, and
10% in the center and west. This metric is not
ideal, as models may share systematic biases and
vary in their ability to represent current Amazo-
nian climates; further, most underestimate current
Amazonian rainfall and most do not incorporate
the climatic feedbacks from forest loss. Therefore,
a more careful evaluation of model ability to
capture key elements of Amazonian climate
variability is needed. When the effects of rising
temperatures on evapotranspiration are included,
almost all models indicate increasing seasonal
water deficit in eastern Amazonia (30). This
drying is exacerbated by ecosystem feedbacks
such as forest die-back and reduced transpiration
in remaining forests (13).
The zone of highest drought risk (south-
east and east) is also the zone of most active
deforestation (Fig. 2). Deforestation-driven
changes in precipitation may be strongest in
the eastern dry corridor, 700 km inland from
the coast, where geographical positioning re-
sults in ocean-generated squall lines passing
through the region at night and being unable
to trigger much rainfall (31), leaving a greater
fraction of precipitation generated locally. This
area includes important agricultural and ranch-
ing frontiers that are experiencing high levels of
In contrast, the northwestern Amazon is least
likely to experience major drought. The high
precipitation in this region is controlled by mois-
ture convergence forced by the Andes, although
it may be vulnerable to reduced precipitation re-
cycling upwind in eastern Amazonia. This re-
gion hosts the highest biodiversity and has been
least affected by historical climate variability and
The Andean flank of the Amazon has excep-
tional rates of biodiversity, adjoins the most bio-
diverse regions of lowland Amazonia, and also
hosts a number of orographic wet spots in other-
wise dry areas (32). As conditions warmed at the
end of the last ice age, the Andes acted as refugia
for many “lowland” (now exclusively Andean)
tree species that were ill-adapted to warming
temperatures (33). Andean ecosystems have
their own form of vulnerability, however: The
cloud forests between 1500 and 3000 m eleva-
tion are susceptible to drying as cloud levels rise
in the face of warming temperatures (34), and
higher elevation restricted endemics would be
Resilience of Amazonian forest ecosystems.
Understanding of Amazonian forest processes
has greatly advanced through the recent Large-
Scale Biosphere-Atmosphere program in Ama-
zonia (LBA) (35). There is mounting evidence
from artificial drought experiments (36), flux
towers, and satellite remote sensing of forest
greenness (37) that intact Amazonian forests
are more resilient (although not invulnerable)
to climatic drying than is currently represented
in vegetation-climate models. First, dry season
water supply is greatly enhanced by root sys-
tems accessing deep soil water and redistribut-
ing it into the surface soil through the process of
hydraulic lift, enabling the whole forest eco-
system to maintain high transpiration and photo-
synthesis rates (38). Second, plant acclimation
to higher temperatures may limit detrimental
effects below 45°C [when proteins begin to de-
nature (39)], although selective advantage by
favored species may cause changes in commu-
nity composition, as occurred at the last glacial-
interglacial transition (33). Finally, rising CO2
may improve plant water use efficiencies and
offset the negative transpiration effects of rising
temperatures. Southern Amazonia was consid-
erably drier as recently as the early-mid Holo-
cene, yet the region seems to have remained
largely forested (33).
The Interaction Between Human
Pressures and Forest Resilience
The speed and magnitude of current human pres-
sures on forests are affecting forest resilience.
Forests close to edges are vulnerable to elevated
dessication, tree mortality (40), and fire impacts.
Rain forests may become seasonally flammable
in dry years, but without anthropogenic ignition
sources fire is a rare occurrence. Hence, fire has
been a weak evolutionary selective force, and
as a result many tree species lack adaptations
Fig. 1. A metric of the probability of enhanced drought in Amazonia: the proportion of 23 climate
models that show a decline in rainfall between 1980 to 1999 and 2080 to 2099 under midrange
(A1B) global greenhouse gas emissions scenarios. (A) Any decline (rainfall decline > 0%); (B)
substantial decline (rainfall decline > 20%); (C) severe decline (rainfall decline > 50%). Dry
season rainfall is particularly important. Left column: December-January-February (dry season in
north); right column: June-July-August (dry season in central and southern Amazonia).
11 JANUARY 2008VOL 319
on January 11, 2008
that allow them to survive even low-intensity
Fire use for land management is nearly
ubiquitous in rural Amazonia. About 28% of
the Brazilian Amazon faces incipient fire pres-
sure, being within 10 km of a fire source (42).
Logging and forest fragmentation also increase
the flammability of forests by providing substan-
tial combustion material, opening up the canopy
and drying the understory and litter layer and
greatly increasing the amount of dry fire-prone
forest edge. This synergism between fragmenta-
tion and fire is becoming increasingly important,
with 20,000 to 50,000 km2of new forest edge
being created annually in Brazilian Amazonia
alone (43). Once burnt, a forest becomes more
vulnerable to further burns (44), loses many pri-
mary forest species, and decreases sharply in
biomass (41). A tipping point may be reached
when grasses can establish in the forest under-
story, providing a renewable source of fuel for
In scenarios of increased drying, it is pos-
sible to see this logging, fragmentation, dessi-
cation, and repeated burning as a likely fate for
many of Amazonia’s forests. The 2005 drought
provides evidence of this in southwest Amazonia:
Remote forests remained fairly unaffected, but
there was substantial penetration of fires from
agricultural areas into surrounding, temporarily
flammable forests (45).
Despite the very recent slowdown in defor-
estation rates, there is potential for extensive
deforestation in Amazonia, as more roads (both
official and unplanned) are built through its core
and connect across to Pacific ports and as inter-
national demand for tropical timber, soybeans,
and free-range beef continues to grow, particu-
larly from rapidly expanding Asian economies
(2, 46, 47). Existing pressures might be exac-
erbated by accelerating worldwide demand for
biofuels. Current plans for infrastructure expan-
sion and integration could reduce forest cover
from 5.4 million km2(2001, 87% of original
area) to 3.2 million km2(53%) by 2050 (2) (Fig.
2A). This exceeds the likely threshold for rain-
fall maintenance and would emit 32 ± 8 Pg of
carbon. Deforestation will be more concentrated
in the south and east, with >50% forest loss, and
along the Andean piedmont, isolating the warm-
ing lowlands from potential biotic refuges in the
cooler mountains (46). In this scenario, the north-
western Amazon is protected by its remoteness
and wetness, but longer term, this region is also
vulnerable to hydrocarbon exploration and oil-
palm plantations that are suitable for wet cli-
mates and acidic soils and have already replaced
many of Asia’s tropical rainforests (46). Drying
of Amazonia, whether caused by local or global
drivers, could greatly expand the area suitable
for soy, cattle, and sugarcane, accelerating forest
Planning for Climate Change
The probability of substantially enhanced drought
(Fig. 1B) under mid-range greenhouse gas emis-
sions scenarios ranges from >60% in the south-
east to <20% in the west. The severity of this
potential threat merits planning for development,
conservation, and adaptation in all regions. Even
if the drought does not come, a well-conceived
and implemented plan will have built resilience
into the Amazon social ecological system.
It is almost inevitable that substantial further
conversion of forest into agricultural and pasture
lands will occur as part of the economic devel-
opment of Amazonian countries (2, 46). The
danger is that degradation of ecosystem services
could push some subregions into a permanent-
ly drier climate regime and greatly weaken the
resilience of the entire region to possible large-
scale drought driven by SST changes. Hence,
the challenge is to manage the economic devel-
opment of Amazonia so that it occurs where ap-
propriate and sustainable, in a way that maintains
the inherent climatic resilience that the intact
forest provides. Simultaneously, this
would preserve the region’s carbon
store and sink and its exceptional
biodiversity, contributing both toward
mitigating global warming and assist-
ing that biodiversity to adapt to cli-
Key aspects of such a plan for
Amazonia could include
(i) Keeping the total extent of
deforestation safely below possible
climatic threshold values (about 30
to 40% cleared) in a matrix that in-
cludes large protected areas with
limited fragmentation and managed
landscapes that maintain sufficient
forest cover and landscape connectiv-
ity to preserve species migration cor-
ridors and forest transpiration services.
(ii) Controlling fire use through
both education and regulation, prob-
ably for net economic benefit.
(iii) Maintaining broad species
migration corridors in ecotonal areas
that are most likely to show early sig-
nals of climate impacts, such as those
between forest and savanna, between lowlands
and the Brazilian and Guyana shield uplands,
between the Andean piedmont and montane
forest, and between montane forest and highland
(iv) Conserving river corridors to act as hu-
mid refugia and migration corridors for terres-
trial ecosystems and as sedimentation buffers
and refugia for aquatic systems. Many of the
southern tributaries of the Amazon river run
from dry fringes to the wet core and could assist
the migration of wet-adapted species.
(v) Keeping the core northwest Amazon
largely intact as a biological refuge that hosts
the highest biodiversity and is the least vulner-
able to climatic drying.
Is such a plan feasible? With the expansion of
protected areas and effective legal enforcement
of private land use, the projections of loss of 47%
of original forest area by 2050 could be reduced to
28% loss (2), avoiding ~17 PgC emissions (Fig.
2B). Recent developments suggest that such good
governance is achievable. Details of the role that
can be played by protected areas, indigenous peo-
ples, smallholders, agroindustries, and governments
are discussed in the supporting online text.
Financing a Climate-Resilience
Plan for Amazonia
A plan for keeping Amazonia from ecological
and climatic decline faces several challenges:
Fig. 2. The potential overlap between deforestation and climate change. Potential loss in forest cover (brown) by
2050 under (A) business as usual and (B) increased governance scenarios [derived from (2)], superimposed on the
probability of substantial drought, which is defined as a >20% reduction in dry-season rainfall by the late 21st
century, as shown in Figure 1B. The dry season is defined as from December to February (south of the equator) and
from June to August (north of the equator). Precipitation scenarios are from mid-range (A1B) global greenhouse gas
emissions scenarios, from the 21 climate models employed in IPCC Fourth Assessment Report [extracted and
modified from (15)].
VOL 319 11 JANUARY 2008
on January 11, 2008
the drive of globalizing market forces, insufficient Download full-text
financial resources, provision of open access to
information, limited technical and governance
capacity, and ineffective enforcement of rule of
law. In particular, new financial incentives are
needed to act as a countervailing force to the
economic pressures for deforestation.
Such incentives are now a serious possibility
through the international markets in carbon
spawned by the Kyoto Protocol, such as the
European Union’s Emissions Trading System.
The recently agreed-upon “Bali Roadmap” for
extension of the Kyoto Protocol beyond 2012
includes plans for rainforest nations to be paid
for reducing emissions from deforestation and
degradation (REDD), either through international
carbon markets or a voluntary fund (48–50).
Tropical forest carbon credits have particular
value within a climate mitigation strategy because
they bring additional direct climatic services
[cloud formation and precipitation, local cooling
by evapotranspiration (28)], as well as other eco-
system services such as biodiversity conservation,
watershed protection, and pollination.
These plans have the potential to shift the
balance of underlying economic market forces
that currently favor deforestation (45) by raising
billions of dollars for the ecosystem services
provided by rainforest regions but will require
exceptional planning, execution, and long-term
follow-through. Such resources could support
the expansion of capacity in forest monitoring
(e.g., freely available satellite-based monitoring,
as already achieved by Brazil) and improved
governance and rule-of-law in frontier regions,
but in particular would need to ensure that they
bring benefits and incentives (e.g., improved social
services like health and education) to the individ-
uals and groups making decisions about Amazon
land use on a daily basis, be they indigenous peo-
ples, rural subsistence dwellers, smallholder mi-
grants, or large private landholders.
The interaction between global climate change
forests vulnerable to large-scale degradation.
Ironically, it is also this linkage between the
global (carbon sequestering) ecosystem service,
for which the world may be more willing to pay,
and regional (transpiration) services that maintain
the region’s climate that provides an opportunity
to sustain the climatic resilience of Amazonia
while contributing toward its conservation and
The next few years represent a unique op-
portunity, perhaps the last, to maintain the resil-
ience, biodiversity, and ecosystem services of
Amazonia in the face of a medium threat of
significant drying and a high threat of significant
deforestation. The best climate, ecological, eco-
nomic, and social science will be needed to de-
velop, implement, and monitor effective policy
responses for securing the region’s future. The
other key requirement is political will at the local,
national, and international levels.
References and Notes
1. Here defined as the Legal Amazon in Brazil and the Amazon
river watershed and Guyanas region outside of Brazil.
2. B. S. Soares-Filho et al., Nature 440, 520 (2006).
3. R. Dirzo, P. H. Raven, Annu. Rev. Env. Res. 28, 137 (2003).
4. C. B. Field, M. J. Behrenfeld, J. T. Randerson, P. Falkowski,
Science 281, 237 (1998).
5. N.Gedney,P.J. Valdes,Geophys. Res. Lett. 27,3053(2000).
6. D. Werth, R. Avissar, J. Geophys. Res. Atmos. 107, 8087
7. M. Maslin, Y. Malhi, O. Phillips, S. Cowling, Trans. Inst.
Br. Geogr. 30, 477 (2005).
8. F. Achard et al., Science 297, 999 (2002).
9. D. Nepstad et al., Conserv. Biol. 20, 65 (2006).
10. N. Ramankutty et al., Glob. Change Biol. 13, 51 (2007).
11. Y. Malhi, J. Wright, Philos. Trans. R. Soc. Lond. B Biol.
Sci. 359, 311 (2004).
12. J. H. Christensen et al., in Climate Change 2007: The
Physical Science Basis. Contribution of Working Group I
to the Fourth Assessment Report of the Intergovernmental
Panel on Climate Change, S. D. Solomon et al., Eds.
(Cambridge Univ. Press, Cambridge and New York,
2007), chap. 11.
13. R. A. Betts et al., Theor. Appl. Clim. 78, 157 (2004).
14. M. B. Bush, M. R. Silman, D. H. Urrego, Science 303, 827
15. J. A. Marengo, Theor. Appl. Clim. 78, 79 (2004).
16. W. H. Li, R. Fu, R. E. Dickinson, J. Geophys. Res. 111,
D02111, 10.1029/2005JD006355 (2006).
17. P. Chang, R. Saravanan, L. Ji, Geophys. Res. Lett. 30,
18. B. W. Dong, R. T. Sutton, Geophys. Res. Lett. 29, 1728
19. Y. Malhi et al., Glob. Change Biol. 12, 1107 (2006).
20. K. W. Holmes et al., Global Biogeochem. Cycles 20,
21. T. R. Baker et al., Philos. Trans. R. Soc. Lond. B Biol. Sci.
359, 353 (2004).
22. S. J. Wright, Trends Ecol. Evol. 20, 553 (2005).
23. B. B. Stephens et al., Science 316, 1732 (2007).
24. S. L. Lewis et al., Philos. Trans. R. Soc. Lond. B Biol. Sci.
359, 421 (2004).
25. E. A. B. Eltahir, R. L. Bras, Q. J. R. Meteorol. Soc. 120,
26. S. B. Roy, R. Avissar, J. Geophys. Res. 107, 8037 (2002).
27. M. D. Oyama, C. A. Nobre, Geophys. Res. Lett. 30, 2199
28. G. Bala et al., Proc. Natl. Acad. Sci. U.S.A. 104, 6550
29. M. O. Andreae et al., Science 303, 1337 (2004).
30. L. F. Salazar, C. A. Nobre, M. D. Oyama, Geophys. Res.
Lett. 34, L09708 (2007).
31. M. Garstang, H. L. Massie Jr., J. Halverson, S. Greco,
J. Scala, Mon. Weather Rev. 122, 608 (1994).
32. T. J. Killeen, M. Douglas, T. Consiglio, P. M. Jorgensen,
J. Mejia, J. Biogeogr. 34, 1357 (2007).
33. F. E. Mayle, D. J. Beerling, W. D. Gosling, M. B. Bush,
Philos. Trans. R. Soc. Lond. B Biol. Sci. 359, 499 (2004).
34. P. Foster, Earth Sci. Rev. 55, 73 (2001).
35. LBA Web site: http://lba.cptec.inpe.br/lba/site.
36. R. A. Fisher et al., Glob. Change Biol. 13, 2361 (2007).
37. A. R. Huete et al., Geophys. Res. Lett. 33, L06405
38. R. S. Oliveira, T. E. Dawson, S. S. O. Burgess, D. C. Nepstad,
Oecologia 145, 354 (2005).
39. T. D. Sharkey, S. M. Schrader, in Physiology and
Molecular Ecology of Stress Tolerance in Plants (Kluwer,
Netherlands, 2004), pp. 101–130.
40. W. F. Laurance et al., Science 278, 1117 (1997).
41. J. Barlow, C. A. Peres, Philos. Trans. R. Soc. Lond. B Biol. Sci.
359, 367 (2004).
42. P. Barreto et al., Human Pressure on the Brazilian
Amazon Forests (World Resources Institute, Washington,
DC, USA, and Imazon, Belém, Brazil, 2006).
43. M. A. Cochrane, W. F. Laurance, J. Trop. Ecol. 18, 311
44. M. A. Cochrane et al., Science 284, 1832 (1999).
45. L. E. O. C. Aragão et al., Geophys. Res. Lett. 34, L07701
46. T. J. Killeen, Advances in Applied Biodiversity Science No. 7
(Center for Applied Biodiversity Science, Conservation
International, Washington, DC, 2007).
47. A. A. C. Alencar, L. A. Solorzano, D. C. Nepstad,
Ecol. Appl. 14, S139 (2004).
48. M. Santilli et al., Clim. Change 71, 267 (2005).
49. R. E. Gullison et al., Science 316, 985 (2007).
50. P. Moutinho, S. Schwartzman, Tropical Deforestation and
Climate Change (IPAM, Belem, Brazil, 2005).
51. This paper is based largely on the conference “Climate
Change and the Fate of the Amazon,” held at Oriel
College, University of Oxford, 20 to 22 March 2007, and
funded by the James Martin 21st Century School, the
Environmental Change Institute, and the Centre for
Brazilian Studies, University of Oxford. Conference
presentations are available at www.eci.ox.ac.uk/news/
events/070320presentations.php. We thank all
participants at the conference, in particular E. Boyd,
M. Gloor, P. Harris, J. Lloyd, J. Marengo, D. Nepstad,
O. Phillips, and B. Soares, for their comments on this
manuscript. Y.M. is supported by the Jackson Foundation,
J.T.R. by the James Martin 21st Century School and the
College of William and Mary, and R.A.B. by the joint
Defra and MoD Integrated Climate Programme. We
thank D. Maniatis and P. Zelazowski for assistance in
manuscript preparation. We acknowledge the Coupled
Model Intercomparison Project and the international
modeling groups for providing their data for analysis.
Supporting Online Material
References and Notes
Published online 22 November 2007;
Include this information when citing this paper.
11 JANUARY 2008VOL 319
on January 11, 2008