info:doi/10.1038/nature04389 2006-03-23T00:00:00Z 2006-03-23T00:00:00Z Britaldo Silveira Soares-Filho Daniel Curtis Nepstad Lisa M. Curran Gustavo Coutinho Cerqueira Ricardo Alexandrino Garcia Claudia Azevedo Ramos Eliane Voll Alice McDonald Paul Lefebvre Peter Schlesinger © 2006 Nature Publishing Group Nature 440 7083 520 523 0028-0836 1476-4679 Expansion of the cattle and soy industries in the Amazon basin has increased deforestation rates and will soon push all-weather highways into the region's core . In the face of this growing pressure, a comprehensive conservation strategy for the Amazon basin should protect its watersheds, the full range of species and ecosystem diversity, and the stability of regional climates. Here we report that protected areas in the Amazon basin—the central feature of prevailing conservation approaches —are an important but insufficient component of this strategy, based on policy-sensitive simulations of future deforestation. By 2050, current trends in agricultural expansion will eliminate a total of 40% of Amazon forests, including at least two-thirds of the forest cover of six major watersheds and 12 ecoregions, releasing 32 ± 8 Pg of carbon to the atmosphere. One-quarter of the 382 mammalian species examined will lose more than 40% of the forest within their Amazon ranges. Although an expanded and enforced network of protected areas could avoid as much as one-third of this projected forest loss, conservation on private lands is also essential. Expanding market pressures for sound land management and prevention of forest clearing on lands unsuitable for agriculture are critical ingredients of a strategy for comprehensive conservation . Expansion cattle soy industries Amazon basin has increased deforestation rates will soon push all-weather highways into region's core . face growing pressure, comprehensive conservation strategy Amazon basin should protect its watersheds, full range species ecosystem diversity, stability regional climates. Here report protected areas Amazon basin— central feature prevailing conservation approaches — important but insufficient component strategy, based policy-sensitive simulations future deforestation. 2050, current trends agricultural expansion will eliminate total 40% Amazon forests, including at least two-thirds forest cover six major watersheds 12 ecoregions, releasing 32 ± 8 Pg carbon atmosphere. One-quarter 382 mammalian species examined will lose more than 40% forest within their Amazon ranges. Although expanded enforced network protected areas could avoid much one-third projected forest loss, conservation private lands also essential. Expanding market pressures sound land management prevention forest clearing lands unsuitable agriculture critical ingredients strategy comprehensive conservation . (?-mix:\s+\W+|\W+\s+|\s+|\/) Amazon conservation forest will strategy comprehensive basin lands protected deforestation watersheds species areas 40 32 Expanding least increased one-third basin—the expansion soy Expansion critical unsuitable also Although carbon rates push including simulations essential avoid more agricultural pressure than Pg ecoregions six range at climates future based projected forests prevention cover market industries policy-sensitive should private loss releasing report full ingredients highways two-thirds 2050 eliminate component land could mammalian region's stability clearing their core total insufficient important has growing prevailing much its current all-weather enforced lose diversity central 8 into examined agriculture One-quarter trends face sound pressures atmosphere 12 major ranges but ecosystem expanded within 382 protect feature management network regional cattle soon Here approaches (?-mix:\.\s+(?=[A-Z])) Although an expanded and enforced network of protected areas could avoid as much as one-third of this projected forest loss, conservation on private lands is also essential By 2050, current trends in agricultural expansion will eliminate a total of 40% of Amazon forests, including at least two-thirds of the forest cover of six major watersheds and 12 ecoregions, releasing 32 ± 8 Pg of carbon to the atmosphere Expanding market pressures for sound land management and prevention of forest clearing on lands unsuitable for agriculture are critical ingredients of a strategy for comprehensive conservation . Expansion of the cattle and soy industries in the Amazon basin has increased deforestation rates and will soon push all-weather highways into the region's core Here we report that protected areas in the Amazon basin—the central feature of prevailing conservation approaches —are an important but insufficient component of this strategy, based on policy-sensitive simulations of future deforestation In the face of this growing pressure, a comprehensive conservation strategy for the Amazon basin should protect its watersheds, the full range of species and ecosystem diversity, and the stability of regional climates One-quarter of the 382 mammalian species examined will lose more than 40% of the forest within their Amazon ranges (?-mix:\s+\W+|\W+\s+|\s+|\/) deforestation model Supplementary spatial rates roads Amazon km Table paved road land proximity paving which deforested across highways subregion PA maps 180 S2 subregions represented distance simulated 1 each basin historical SimAmazonia Methods relationships incorporates socioeconomic highway existing call PAs secondary over annually areas automata allocates updated variables biophysical annual network empirical thereby all customized 2 Landsat individualized evolving description example 432 into Information Thematic patterns previously S1 these scarce proposed basis attained derived Portuguese inflection use counties studies employing resolution simulation output raster entire forests expansion simulação when case 12 endogenous Fig trends S5 start accelerates Amazônia component cellular scenarios Mapper zoning towns schedule Spatial provided up stratified protected rivers Increasing regional constructor its planned parameters 2002 ceilings outside becoming da point effectiveness driver 47 material simulates under detailed features within estimated scene Brazilian latter 1997 distribution validated defined relationship major data user-defined produces future satellite-based design forecast Proximity networks empirically unpublished B.S.S.-F (from effect variable integrity grids calibrated (?-mix:\.\s+(?=[A-Z])) A detailed description of the model design is provided in Supplementary Information . Increasing proximity to paved highways accelerates deforestation within a subregion up to an inflection point when forests outside protected areas start becoming scarce Methods The model, which we call ‘SimAmazonia 1’ (from the Portuguese simulação da Amazônia ) produces annual maps of simulated future deforestation under user-defined scenarios of highway paving, PA networks, PA effectiveness, deforestation rates and deforested land ceilings Proximity to paved highways is the major driver of deforestation rates in the model and this relationship was defined empirically from data on deforestation and paved roads for 432 counties of the Brazilian Amazon ( Supplementary Fig S5 ) Spatial integrity across subregions is attained by employing spatial variables (for example, distance to previously deforested land and distance to all roads) that are updated annually over the entire basin The spatial distribution of deforestation across the Amazon is simulated with a cellular automata model, with parameters customized for each subregion, that allocates deforestation on the basis of its empirical relationships with proximity to roads, rivers and towns, land use zoning, and biophysical features represented in raster grids of 1 km 2 resolution The spatial simulation model was calibrated and validated for 12 regional case studies (B.S.S.-F., unpublished material), each represented by a Landsat Thematic Mapper scene (180 km × 180 km) This latter variable is output of the road constructor model , a component that simulates the expansion of the secondary road network and thereby incorporates the effect of endogenous roads on the evolving spatial patterns of deforestation We stratified the Amazon basin into 47 socioeconomic subregions for which individualized deforestation rates are forecast; these rates were estimated from historical trends (from 1997 to 2002) derived from satellite-based deforestation maps ( Supplementary Table S2 ), the planned road paving schedule ( Supplementary Table S1 ) and existing and proposed PAs ( Supplementary Table S2 ) (?-mix:\s+\W+|\W+\s+|\s+|\/) Amazon forest deforestation scenario PAs will carbon governance Fig more Supplementary emissions than their reserves would highway paving legislation through conservation expansion highways ecoregions new forests watersheds land but frontier current network potential environmental BAU scenarios species may lose do paved PA Brazilian other region future lands 2050 which under cover markets if soy not regional they private assumes properties cattle Pg example million worldwide loss outside could regions mammalian These 2 vegetation region's km model clearing within include planned atmosphere avoid international range southwestern growing low Xingu highest 40 continue Interoceanica where protection currently 32 85 into BR-163 climate major area effects Table management trends sound vulnerable remain global large equivalent eastern measures extensive 70 increase decrease Paving 1a intact developed coastal Acre least projected At Maranhão all system highly created 12 most compared Manaus–Porto ranchers although Grosso human half subject Velho enforcement Conservation requiring consider two-thirds protected two decades northwest productive mammals production experiments Parque Maraj parks also along licensing ranching have market imperilled closed-canopy pressures only higher alone found conservative impoverishment provided landholders fire indigenous additional extreme beef trees few savanna types 7 must zoning Brazil farmers 4 incentives planning much Cuiabá–Santarém rates riparian expanded encompass systems estimate both active recent expanding between concentrated concentration years result increases endemism implemented Mato limit S7 projections multiplication avoidance high Some substantial entered richness Wetland Par landscapes sustain workshops threat reinforce restricts effectively accruing stimulate strategies perceive simulated taxa approaches Carbon retained warming no occupation regime dry enforced flow Brasil–Cuzco derived lowered Gurupi less processes rate state Guianas influence Tapajós number such succeeds many centres necessary developing soundly 41 Peru logging native regulations Compliance released uninhabited estimated River Moreover road run substitution river change become 8 babaçu 1.5 anthropogenic initially priorities emerging fall one until assuming 93 mid-century simply intermediate mammal food below 17 provoke Bolivia contrasted 60–100 forested constructed Rondônia However credits gap eight affect 5.3 maintain Florencia analysis conserve superimposing 15 S11–S13 access 97 flooding contain comprise countries collapse strategy investments gas requirement settlement Eight unabated permits costs latent did use reduce others transfer Hence Santa satellite-based financial either core lowering reduction Beyond part savannas portion benefits Land-use tributaries favour who 100 protect Brazil's S14 compensation restricting contained coalesces demonstrated enforcing Eighteen non-flying extrapolated crop soybeans 4.5 estimates inaccessible drives time 2003 Atlantic Given 2004 restrictions Paramaribo existing compliance Protected 2005 sufficient representation All BR-364 rainfall has Nor rising participatory inhibit our proxies its ranges—is exceed modified blocks Watershed 1b 1c 2008 zone 20 human-induced larger Indigena inside achievements individual 1e Both maintenance observation forest-dependent Kyoto mandatory politically 1g undergo ran magnitude Assis reserve declines Thirty-five heat scheduled avoided headwaters biomass achieved unlikely ecologically Terrestrial 76 extractive 28 funding abundant lucrative trajectories property designating Figs instead described occur unpublished exception formation Felix emitted 78 development Pucallpa disaggregated 119 occurring willing further across S5–S7 Nacional certification might eightfold based mammals—those Region baseline Under beyond at Our owing exercises areas profitability Regional east-central successful swamp effective successes traverses command-and-control retain extinction Amazon's advances opportunity annual eastward economic protocol pastures plausibility Tables boundaries remoteness states Andes scarce implementation business-as-usual phase empirically underway climbing timber single target depends ecosystems far São Amazonian simulating b several likely determined permitted greenhouse Current four primate 420,000 d extend agro-ecological entire Program manage f ‘passively some 50 peak era ha made convention after h indicator first coupled focused over 3.2 86 agricultural required basin just efforts nor 53 reduces conserving accelerated whose largely sale standards southeastern diversity plausible public refinement already ambitious captured pay generally potentially total critical large-scale feasible Santarém impose impact humanity possible inhabited Areas how summarized biological period tropics rainforest woodlands Amap disruption scrub without lax Southwestern ecological make each close preserved considers proposed policy-sensitive during levels comply 2010 reduced B.S.S.-F little commodities about expected S1 assessments running maps law preceding Cruz serve moving reducing institutions Recent indicate S3 Uncertainty because Information dependent replace accounted S9 simulations Within elsewhere relaxes quadrant cause combined Colombia original three density S8 (?-mix:\.\s+(?=[A-Z])) All conservation measures combined (but without PA expansion) accounted for 86% of the deforestation that is avoided in the ‘governance’ scenario An expanded conservation strategy must encompass lands that fall outside PAs if it is to avoid the collapse of regional rainforest ecosystems that is already occurring elsewhere in the tropics . An expanded network of PAs that is effectively implemented and enforced provided half of this reduction in deforestation At one extreme is the ‘business-as-usual’ scenario (BAU), which assumes that: recent deforestation trends will continue; highways currently scheduled for paving will be paved; compliance with legislation requiring forest reserves on private land will remain low; and new PAs will not be created At the other extreme, the ‘governance’ scenario assumes that Brazilian environmental legislation is implemented across the Amazon basin through the refinement and multiplication of current experiments in frontier governance Beyond the market incentives for ecologically sound land management that may be captured through the environmental certification of beef, soybeans and timber , part of this funding might be provided through the sale of carbon credits derived from the avoidance of 17 ± 4 Pg of carbon emissions within a modified climate change convention Both regional and global climate systems are also coupled to the Amazon forest through latent heat transfer By 2050, 32 ± 8 Pg of carbon is emitted under the BAU scenario, equivalent to four years of current annual emissions worldwide, contrasted with 15 ± 4 Pg under the governance scenario Carbon emissions expected for each scenario were estimated by assuming that 85% of the carbon contained in forest trees is released to the atmosphere after deforestation and by superimposing deforestation simulations on a range of forest biomass maps , extrapolated to the entire Amazon Compliance with Brazil's environmental legislation requiring the protection of riparian vegetation and forest reserves on private land could increase as cattle ranchers and soy farmers perceive sound land management to be a requirement to access lucrative international markets Conservation achievements of this magnitude are unlikely to result from command-and-control implementation of environmental legislation alone, but instead become more likely as international markets impose higher environmental standards on beef, soy and other food commodities Conservation strategies for the Amazon region have focused on protected areas (PAs), both inhabited (‘extractive reserves’, public forests and indigenous lands) and uninhabited (parks and biological reserves) Current legislation permits landholders to exceed the limit on forest clearing on their property if they retain other properties on which forest reserves comprise more than this limit Eight of the 12 major watersheds of the Amazon will lose more than half of their forest cover under the 2050 BAU scenario Eighteen of the Amazon's 32 major forested ecoregions will lose more than 40% of their forest cover by 2050 and 12 will lose more than 70% ( Fig. 1e , f , and Supplementary Table S9 ) For example, Amazon trees contain 119 ± 28 Pg of carbon , equivalent to 1.5 decades of current worldwide anthropogenic carbon emissions to the atmosphere Given the concentration of projected deforestation in the eastern Amazon, some watersheds, ecoregions and mammalian species are far more vulnerable to disruption than others Hence, although the highest levels of mammal endemism and species richness are found in the southwestern Amazon close to the Andes ( Supplementary Fig However, our analysis considers only existing or planned highways In all scenarios, future deforestation is concentrated in the eastern Amazon, where the density of paved highways will continue to be highest for several decades, and along the BR-364 highway from Rondônia into Acre in the Southwestern Amazon ( Fig. 1a , b ) Land-use zoning systems under development by Amazon states could reinforce a system of deforestation licensing that restricts the clearing of lands with low productive potential as it relaxes restrictions on lands with high productive potential Moreover, we did not include the loss of Amazon savannas Nor will PAs be sufficient to sustain Amazon climate; more than 70% of the forest cover of Amazon landscapes may be necessary to maintain the forest-dependent rainfall regime Our projections of future deforestation would increase if additional highways were constructed into currently inaccessible regions PAs inhibit both deforestation and fire but they are less effective in conserving watersheds, whose headwaters generally extend beyond reserve boundaries Paving of the Interoceanica highway would cause the least impact, but would provoke extensive deforestation in the southwestern Brazilian state of Acre Paving of the Manaus–Porto Velho highway, which traverses a region with few PAs and little human settlement, would stimulate more deforestation than the paving of either the Cuiabá–Santarém or the Interoceanica highways alone ( Supplementary Figs S11–S13 and Supplementary Tables S5–S7 ) Recent advances in enforcing the region's ambitious deforestation legislation and regional planning exercises underway along the major highways planned for paving are just two of the large-scale conservation efforts that could be the target of investments made for lowering the region's carbon emissions Regional workshops and gap assessments have developed conservation priorities and determined the representation of vegetation types and centres of endemism within current and proposed PAs S1 ) S14 ), the larger mammalian extinction threat may occur in the eastern Amazon, where projected deforestation rates are much higher S7 ) S7 ) Some Atlantic coastal watersheds (Pará, Maranhão, Marajó and Amapá) and southeastern tributaries of the Amazon River (Tapajós and Xingu) will all lose at least two-thirds of their forest cover ( Fig. 1c , d , and Supplementary Table S8 ) and may undergo substantial increases in peak river flow and flooding as pastures replace forests Terrestrial non-flying mammals serve as conservative indicator taxa or proxies to estimate how future deforestation will affect individual species The Amazon PA network may protect a large portion of the region's mammalian diversity, but will not avoid the impoverishment of critical watersheds and ecoregions The Amazon has entered a new era as the growing profitability of cattle ranching and soy production increases deforestation rates and drives the expansion of the highway network into the region's core ( Supplementary Fig The BAU scenario assumes that as much as 40% of the forests inside of PAs are subject to deforestation (B.S.S.-F., unpublished observation), climbing to 85% outside The conservation measures simulated in the ‘governance’ scenario would reduce the number of imperilled watersheds, ecoregions and mammalian species by about two-thirds and would avoid carbon emissions to the atmosphere equivalent to two years of global human-induced emissions The ecological effects of this new phase of accelerated deforestation are potentially large The economic opportunity costs of reducing deforestation in the Amazon could be lowered by restricting forest clearing on lands with low potential for crop production or cattle ranching, which are abundant in the region The governance scenario assumes that the planned expansion of the PA network in the Brazilian Amazon , from 32% to 41% of the total forest area, succeeds and 100% of the forests in PAs are preserved intact, with only 50% of the forests outside of PAs subject to deforestation (compared with 20% currently permitted by Brazilian regulations) The highest concentration of ‘imperilled’ mammals—those species that lose more than 40% of the forest in their Amazon ranges—is in the east-central Amazon ( Fig. 1g , h ) The intermediate scenarios that we ran in the model indicate that simply expanding the PA network, but with lax enforcement, reduces new deforestation 7% below the BAU scenario baseline The model was run under eight scenarios that encompass a plausible range of future trajectories of deforestation The most vulnerable ecoregions are found between savanna woodlands and closed-canopy forest where human occupation is concentrated and PAs are scarce The paving of the BR-163 highway would result in an extensive region of deforestation as the agricultural frontier moving northwest from São Felix do Xingu coalesces with the eastward expansion from the BR-163 The plausibility of a scenario of expanding frontier governance is demonstrated by growing pressures on Amazon cattle ranchers and soy farmers from international markets and financial institutions to comply with environmental legislation and manage their land soundly , recent successes in designating new PAs in regions of active frontier expansion (for example, 7 million ha of new reserves were created in active frontier regions in 2004 and 2005) and the regional, participatory planning processes that are preceding the paving of new highways between Santarém and Mato Grosso , and in the southwestern Amazon The potential decrease in Amazon carbon emissions within the governance scenario is more than eightfold the worldwide decrease in greenhouse gas emissions that will be achieved during the first compensation period of the Kyoto protocol The scenarios are conservative because they do not consider forest impoverishment through logging and fire , nor do they consider the potential for forest substitution by savanna scrub through global warming These ecoregions are highly dependent on the successful protection and possible expansion of a few indigenous reserves (for example, Parque Indigena do Xingu ) and parks (for example, Parque Nacional do Gurupi ) These experiments include enforcement of mandatory forest reserves on private properties through a satellite-based licensing system , agro-ecological zoning of land use , and the expansion of the PA network (Amazon Region Protected Areas Program ) These include the Maranhão babaçu forest (97% loss) and the 420,000 km 2 Mato Grosso dry forest (76% loss) These scenarios are summarized in Supplementary Table S3 Thirty-five primate species lose 60–100% of their Amazonian range Uncertainty estimates are described in Supplementary Information Under the governance scenario, 4.5 million km 2 of forest would remain in 2050 ( Fig. 1b and Supplementary Fig Watershed protection depends on the maintenance of riparian zone vegetation, required by law on private properties in Brazil We compared the potential influence of PAs and other conservation approaches on future trends in Amazon watersheds, vegetation types (ecoregions), mammals and carbon emissions by developing an empirically based, policy-sensitive model of Amazon deforestation We disaggregated the effects of highway paving by running the model with no further paving and with three additional scenarios in which a single highway is paved: the BR-163 (Cuiabá–Santarém) and the Interoceanica (Assis Brasil–Cuzco) paved in 2008 and the Manaus–Porto Velho highway paved in 2010 We estimate that the closed-canopy forest formation of the Amazon will be reduced from its current area of 5.3 million km 2 (2003, 85% of the original area) to 3.2 million km 2 (53%) by 2050 if current trends continue unabated (BAU) ( Fig. 1a and Supplementary Fig Wetland forests are also highly vulnerable, such as the Paramaribo swamp forest (93% forest loss) and the Marajó forest (78%) With many of the benefits of Amazon conservation accruing to humanity worldwide, developed countries must be willing to pay to make frontier governance politically feasible With the exception of Santa Cruz (Bolivia), Pucallpa (Peru), the Florencia region of Colombia, and coastal regions of the Guianas, large blocks of forest outside Brazil and most of the northwest quadrant of the Brazilian Amazon may remain largely intact until the mid-century, ‘passively’ protected by their remoteness Within the governance scenario, the deforestation rate, although rising initially owing to road paving, declines over time, simulating the effects of growing market pressures in favour of sound land management, emerging markets for carbon retained in native forests and other incentives for landholders who conserve forest on their properties Model results for the extreme-case scenarios in the year 2050. a , b , Forest cover for BAU ( a ) and governance ( b ) scenarios. c , d , Percentage forest loss by major watershed for BAU ( c ) and governance ( d ) scenarios in 2050. e , f , Percentage forest loss by ecoregion for BAU ( e ) and governance ( f ) scenarios in 2050. g , h , Numbers of imperilled mammals for BAU ( g ) and governance ( h ) scenarios in 2050 ( n = 105). Ecoregions: avdf, Apure/Villavicencio dry forests; mbf, Maranhão Babaçu forests; mdf, Maranon dry forests; mgtdf, Mato Grosso tropical dry forests; mvf, Marajo Varzea forests; nr, Northeastern restingas; pm, Pará mangroves; pmmf, Purus/Madeira moist forests; psf, Paramaribo swamp forests; tammf, Tocantins–Araguaia/Maranhão moist forests; txmf, Tapajos/Xingu moist forests; xtamf, Xingu/Tocantins–Araguaia moist forests. info:doi/10.1126/science.1067053 info:doi/10.1046/j.1523-1739.1995.09010034.x info:doi/10.1126/science.1072807 info:doi/10.1046/j.1523-1739.2000.99329.x info:doi/10.1046/j.1365-2486.2001.00426.x info:doi/10.1029/2001JD000717 info:doi/10.1029/2001JD000335 info:doi/10.1046/j.1523-1739.1995.9051134.x info:doi/10.1007/s10584-005-8074-6 info:doi/10.1038/19066 info:doi/10.1029/2003GL018600 info:doi/10.1016/S0022-1694(03)00267-1 info:doi/10.1038/35002501 info:doi/10.1038/371065a0 info:doi/10.1038/35002062 info:doi/10.1126/science.1091714 info:doi/10.1111/j.1529-8817.2003.00769.x 12