March 2002 Vol 52 No 3 BioScience 225

Articles

Tropical savannas cover between 15 and 246 x

106 square kilometers (km2) of the land surface in SouthAmerica Africa and Asia Most people know these savannasbecause of their unique assemblages of abundant and exquisitewildlife however they have only recently begun to receive thekind of attention from a conservation viewpoint that hasbeen given to tropical rain forests (Myers et al 2000) Thelargest richest and possibly most threatened tropical sa-vanna in the world is the Cerrado a large region that occu-pies the center of South America In an effort to identify theworldrsquos most important biodiversity hotspots Myers andcolleagues (2000) ranked the Cerrado among the 25 most im-portant terrestrial hotspots It is the only region on their listdominated by tropical savannas The biodiversity of the Cer-rado is impressive in an area of 186 million km2 10000 plantspecies 161 mammal species 837 bird species 120 reptilespecies and 150 amphibian species have been recorded (Myers et al 2000) However the present-day situation of theCerradorsquos biodiversity is tragic because by some estimates only20 of the region remains undisturbed and only 12 ispreserved in protected areas (Mittermeier et al 2000)

Although the Cerrado has been recognized as an impor-tant South American area of endemism for several groups oforganisms (Muumlller 1973 Rizzini 1979 Cracraft 1985 Haffer1985) its biogeography remains poorly known Here we pre-sent an overview of the main biogeographic patterns foundfor birds in the Cerrado comparing these patterns with thosefound in other groups of organisms and we examine how thisinformation can be brought to bear on efforts to conserve theregionrsquos biotic diversity

The Cerrado Geographic settingTropical savannas are an important component of the ter-restrial vegetation in South America In some regions suchas the Llanos of northern South America and Cerrado sa-vannas dominate the landscape whereas forests are found only

along the rivers or in small isolated patches In contrast in largeforested regions such as Amazonia and Atlantic Forest sa-vannas occur as more or less isolated patches amidst a con-tinuous cover of forests Cerrado Llanos Roraima (or Gratilde-Sabana) Llanos de Mojos and Pantanal are the largestcontinuous blocks of savanna in South America (Figure 1)

The Cerrado is by far the largest savanna region in SouthAmerica It is also the second largest South American biomeexceeded only by Amazonia It includes much of centralBrazil and parts of northeastern Paraguay and eastern Bolivia(AbrsquoSaber 1977) As a result the Cerrado occupies a centralposition in relation to other large South American biomes

Joseacute Maria Cardoso da Silva (e-mail jsilvaconservationorgbr) is

director for Amazonia Conservation International do Brasil Av

Nazareacute 541310 66035-170 Beleacutem Paraacute Brazil His research is

focused on biogeography systematics ecology and conservation of

Neotropical birds John M Bates is an associate curator of birds at

the Field Museum of Natural History Department of Zoology 1400

S Lake Shore Drive Chicago Illinois 60605-2496 His research in-

terests are in genetic diversity of tropical birds and the implications

of this diversity for evolution and conservation copy 2002 American In-

stitute of Biological Sciences

Biogeographic Patterns andConservation in the SouthAmerican Cerrado A TropicalSavanna HotspotJOSEacute MARIA CARDOSO DA SILVA AND JOHN M BATES

THE CERRADO WHICH INCLUDES BOTH

FOREST AND SAVANNA HABITATS IS THE

SECOND LARGEST SOUTH AMERICAN

BIOME AND AMONG THE MOST

THREATENED ON THE CONTINENT

226 BioScience March 2002 Vol 52 No 3

Articles

(Figure 1) with extensiveborders with the two largestforest biomes (Amazonia andAtlantic Forest) as well as withthe two largest dry regions(Caatinga and Chaco)

The Cerrado region has atropical wet and dry climatewith intermediate rainfall be-tween the wetter regions tothe northwest and southeastand the drier areas to thenortheast Average annualrainfall varies between 125and 200 cm and average an-nual temperature between20deg and 26deg C (Nimer 1979AbrsquoSaber 1983) The dry pe-riod lasts from 3 to 5 consec-utive months (generally be-tween May and September orOctober) during the winterof the Southern Hemisphere(Nimer 1979 Sarmiento1983)

Most of the Cerrado is on plateaus of crystalline or sedi-mentary blocks whose continuity is broken by an extensivenetwork of peripheral or interplateau depressions (Brasiland Alvarenga 1989) This geomorphological variation ex-plains much of the distribution of plants in the Cerrado(Cole 1986) Savannas are by far the dominant vegetation cov-ering around 72 of the region but patches of dry forests andcomplex belts composed of both dry forests and savannas givethe region a mosaic-like aspect (Figure 2) We estimate that24 of the region represents this mosaic that we call savannandashforest transition with a remaining 4 being dry forest

The tops of the plateaus are flat gently rolling surfaces atelevations ranging from 500 to 1700 meters (m) They are cov-ered primarily by cerrado a semideciduous to evergreen sa-vanna-like vegetation growing on the deep well-drainedand nutrient-poor soils (Eiten 1972 1990 Furley and Ratter1988) Throughout cerrado vegetation varies in physiog-nomy and composition (Furley and Ratter 1988 Ribeiro andWalter 1998) Five main structural types of cerrado are rec-ognized by botanists (Eiten 1972)

1 cerradatildeo a dense forest type (8ndash15 m tall) that often hasa completely closed canopy

2 cerrado sensu stricto a woodland (5ndash8 m tall) withclosed scrub and more scattered trees than in cerradatildeo(Figure 3c)

3 campo cerrado an open scrubland (3ndash6 m tall) with fewtrees (Figure 3a)

Figure 1 Distribution of major lowland tropical biomes inSouth America The Cerrado occupies a central position in rela-tion to other lowland tropical biomes on the continent Othermajor savanna regions are the Llanos Roraima the Llanos deMojos and the Pantanal Note the presence of several savannapatches within Amazonia and Atlantic Forest

Figure 2 Major vegetation types in the Cerrado The savannandashforest transition is a mosaiccomposed of patches of forest and savannas

4 campo sujo grassland (2ndash3 m tall) with scattered shrubs

5 campo limpo grassland with few or no shrubs or tallerwoody plants (Figure 3b)

Tree species common in the cerrado vegetation are Annonacrassiflora Astronium fraxinifolium Byrsonima coccolobifoliaB verbascifolia Hancornia speciosa Kielmeyera coriaceaQualea grandifolia and Q parviflora among others (Ratterand Dargie 1992) Besides cerrado other types of distinct sa-vanna vegetation called campos rupestres (Figure 3a) andmiscellaneous lithosolic campos occur on rocky outcropsThese habitats are limited to small patches on the plateaus and

have a highly endemic flora (Giullietti and Pirani 1988 Eiten1990) Narrow fringes of gallery forest often border small riversand streams of the region Gallery forests on the plateausgrow on narrow belts of soils rich in organic matter (Eiten1990) They are evergreen have trees that are on average20ndash30 m tall and possess a dense understory of low stature(Ribeiro and Walter 1998) Associated with gallery forestsare stands of a species of palm Mauritia flexuosa locallyknown as veredas

The peripheral depressions are generally flat surfaces of lit-tle relief (100ndash500 m in elevation) that are ocasionally inter-rupted by the presence of steep-sided hills The landscape pat-

March 2002 Vol 52 No 3 BioScience 227

Articles

Figure 3 Examples of several of the structural habitat typesthat grade into one another across the Cerrado Bird species endemic to the cerrado region and typical of these habitats aregiven in parentheses (Table 1) (a) Campo rupestre (blue finchPorphyrospiza caerulescens) campo sujo (collared crescent-chest Melanopareia torquata) and campo cerrado (black-throated saltator Saltator atricollis and curl-crested jayCyanocorax cristatellus) atop the Serraniacutea de Huanchaca (b)Campo limpo (campo miner Geobates poecilopterus) with asmall finger of gallery forest (helmeted manakin Antilophiagaleata) in the distance (c) Campo cerrado (chapada fly-catcher Suiriri islerorum) (d) Cerrado sensu stricto (coal-crested finch Charitospiza eucosma) All photographs taken at Noel Kempff Mercado National Park Bolivia

ba

d

c

tern on these depressions is much more heterogeneous thanthat of the plateaus in that it includes more habitats (river-ine forests tropical dry forests all types of cerrado andmarshlands) distributed in a mosaic fashion Riverine forests(called matas ciliares Ribeiro and Walter 1998) differ fromgallery forests because they are found only along the large riversin the Cerrado They are not wide (up to 100 m from each sideof the river) and can be semideciduous to evergreen Trees are20ndash30 m tall Common species are Anadenanthera sppApeiba tibourbou Aspidosperma spp Tabebuia spp andTrema micrantha among others (Ribeiro and Walter 1998)Tropical dry forests (matas secas) are particularly associatedwith peripheral depressions and in some areas (eg in theParanatilde River valley) are the dominant vegetation type Theyare deciduous or semideciduous and grow on patches ofmoderately productive soils derived from basic rocks such aslimestone (Ratter et al 1978 Prado and Gibbs 1993) Canopyheight averages 20ndash45 m (Ribeiro and Walter 1998) Treespecies common in dry forests include Astronium urundeuvaPiptadenia macrocarpa Chorisia speciosa Tabebuia spp Ca-vanillesia arborea and Cedrella fissilis (Ratter et al 1978)

Biogeographic patterns

Forest organisms in a tropical savanna regionThe most basic question about the biota of a region is howit is distributed in major vegetation types In the Cerrado onemight expect that most of the biota would inhabit the savannasthat dominate the region However studies of mammalsbirds and plants do not support this hypothesis Redford andFonseca (1986) found that most (563) of the nonvolantmammal species in the Cerrado inhabit forests SimilarlySilva (1995b) classified the 759 birds species known or as-sumed to breed in the Cerrado as independent (occurring ex-clusively in forests) semidependent (occurring in both forestsand open vegetation) or dependent (occurring exclusively inopen vegetation) The majority of species (393 or 518) aredependent on the various forest habitats 208 (158) aresemidependent and 274 (208) are independent speciesThus approximately 826 of Cerrado birds require foreststo some degree Mendonccedila and colleagues (1998) listed 6671plant taxa (species and varieties) for the Cerrado of these 38occur only in forests

In general these data support the hypothesis that galleryand dry forests which occupy less than 20 of the Cerradoare necessary for a large portion of regional biodiversityHow did this pattern evolve Silva (1995b) suggested that biotic diversity of the Cerrado region increased through in-terchange with adjacent regions during the Quaternary climaticndashvegetational fluctuations Thus Atlantic and Ama-zonian forest species expanded their ranges into the Cerradoduring the wet periods following the expansion of the riverinendashgallery forest network (Silva 1995b) whereas Caatingaand Chaco elements colonized the Cerrado during the Qua-ternary dry periods as part of the coalescence of dry forestsin peripheral depressions (AbrsquoSaber 1983 Pennington et al

2000) These species probably originated in adjacent regionsand maintained viable populations within the Cerrado regionbecause their forest habitats did not vanish entirely during thepaleoclimatic period that followed their range expansion Ina fashion similar to what is seen today large rivers and streamsmust have provided suitable environmental conditions tomaintain wide galleryndashriverine forests during the dry periods(AbrsquoSaber 1983) whereas large patches of soils derived fromlimestone were important for maintaining dry forests duringthe wet periods (Prado and Gibbs 1993 Pennington et al2000)

Endemic species Origin and evolution The per-centage of species endemic to the Cerrado varies across dif-ferent groups of organisms vascular plants (44) amphib-ians (30) reptiles (20) mammals (118) and birds(14) (Silva 1995a Myers et al 2000) However because largeareas of the Cerrado remain unexplored (Silva 1995c) thesenumbers will likely change when additional biological in-ventories are conducted At present we still know little abouthow these endemic species are distributed across the Cerradoor how they evolved A key question for conservation purposesis whether one can identify subareas of endemism withinthe large Cerrado region

To determine if there are subareas of endemism within theCerrado Silva (1997) mapped and created overlays of all en-demic bird speciesrsquo ranges smaller than 60000 km2 Only 10of the 30 endemic species met this requirement (Table 1) Theranges of these species delimit three main areas

Espinhaccedilo Plateau Four species that inhabit camposrupestres are Augastes scutatus (Trochilidae) Asthenesluizae (Furnariidae) Polystictus superciliaris (Tyran-nidae) and Embernagra longicauda (Emberizidae)

Araguaia River Valley Three species that inhabit galleryforests are Synallaxis simoni (Furnariidae) Cercomacraferdinandi (Thamnophilidae) and Paroaria baeri(Emberizidae)

Paranatilde River Valley Two species found only in tropicaldry forests are Pyrrhura pfrimeri (Psittacidae) andKnipolegus franciscanus (Tyrannidae) (Figure 4)

The subareas of endemism identified for birds are at leastpartially supported by studies of other groups of organismsThe Espinhaccedilo Plateau is well known for its unique biotacomposed of hundreds of endemic species of plants (Giulli-eti and Pirani 1988 Harley 1988) and some endemic speciesof other vertebrates such as lizards and amphibians (Haddadet al 1988 Rodrigues 1988) The Araguaia and Paranatilde rivervalleys are still poorly known for nonavian groups but theriverine forests and dry forests likely harbor endemic speciesof other organisms also For example a very distinctive speciesof rodent of the genus Kerodon was described recently fromParanatilde dry forests (Moojen et al 1997) Although not knownto possess endemic birds several other areas possess endemictaxa from other groups For instance Erwin and Pogue (1988)

228 BioScience March 2002 Vol 52 No 3

Articles

indicated that the Mato Grosso Plateau is an important areaof endemism for beetles of the genus Agra Thus bird distri-butions indicate three subareas of endemism however thisnumber is almost certainly an underestimate for other groupssuch as some insects vascular plants and amphibians Froma taxonomic perspective these groups are still too poorlyknown to be used for assessing patterns of endemism withinthe Cerrado

Investigation of the timing and mode of the evolution ofthe endemic species of the Cerrado is in its infancy butnonetheless raises controversy Some authors (eg Rizzini1979) have suggested that the original vegetation in the Cer-rado was forest rather than savanna In this view present-daysavannas succeeded forests by natural processes such as fireand climatic changes At the opposite extreme Cole (1986)suggests that within the Cerrado the savanna vegetation isolder than the forests based on geomorphological informa-

tion If one assumes that avian diversification within the Cer-rado followed at least roughly the same trend as evolutionof the landscape in this region then one would predict basedon Rizzinirsquos hypothesis that older species are associatedmainly with forest vegetation and younger species are asso-ciated mainly with savannas Under Colersquos hypothesis the pre-diction would be inverted To evaluate these two hypotheseswe classified each endemic species into one of two evolutionaryage categories paleoendemics species that are either taxo-nomically distinct (ie well-differentiated monotypic gen-era) or considered to be the sister taxon of a large radiationin which one or more members also occur in the Cerrado re-gion and neoendemics species with sister taxa in adjacentSouth American regions Based on recent molecular work onSouth American birds (Bates et al 1999) we suggest thatneoendemics are younger than the PliocenendashPleistocene tran-sition (about 18 million years ago) whereas paleoendemicsoriginated before this period This is only a tentative classi-fication that must be evaluated in future molecular studiesIn addition we classified each avian endemic as either a for-est or nonforest species All 30 endemic species were analyzed(Table 1) We determined 12 species to be paleoendemicsand 18 to be neoendemics The majority (11 of 12) of the pa-leoendemic species are nonforest species whereas most of theneoendemics (12 of 18) are forest birds These results supportColersquos rather than Rizzinirsquos hypothesis

An additional aspect of these data concerns recent sug-gestions that ecotones or transition zones between two ormore distinctive types of vegetation provide important op-portunities for speciation (Smith et al 1997 Schlithuizen2000) For the Cerrado avifauna this does not appear to bewell supported with respect to the ecotone between savannaand forest Although 158 bird species inhabit both forest and

March 2002 Vol 52 No 3 BioScience 229

Articles

Table 1 Endemic bird species of the Cerrado regiona

Species Ageb Habitatc

Nothura minor P NTaoniscus nanusd P NPenelope ochrogaster N FColumbina cyanopis P NPyrrhura pfrimeri N FAmazona xanthops P NCaprimulgus candicans P NAugastes scutatus N NGeobates poecilopterusd P NSynallaxis simoni N FAsthenes luizae N NPhilydor dimidiatus N FAutomolus rectirostris N FHerpsilochmus longirostris N FCercomacra ferdinandi N FMelanopareia torquata N NScytalopus novacapitalis N FPhyllomyias reiseri N FSuiriri islerorum N FPolystictus superciliaris P NKnipolegus fraciscanus N FAntilophia galeata N FPoospiza cinerea N NEmbernagra longicauda N NCharitospiza eucosmad P NParoaria baeri N FSaltator atricollis P NPorphyrospiza caerulescensd P NBasileuterus leucophrys N FCyanocorax cristatellus P N

a This data set has been modified from Silva (1997) as Geobates

poecilopterus (a monotypic genus) has been moved from neoendem-

ic to paleoendemic and Antilophia galeata has been changed to

neoendemic because of the recent discovery of Antilophia bokerman-

ni (Coelho and Silva 1998) Suiriri islerorum is a newly described

endemic species (Zimmer et al 2001)

b P paleoendemic N neoendemic

c N nonforest F forest

d Monotypic genus

Figure 4 Three subareas of endemism are identified for birdswithin the Cerrado Espinhaccedilo Plateau (with four endemicspecies) Araguaia River Valley (with three endemic species)and Paranatilde River Valley (with two endemic species)

savanna no single case of intraspecific differentiation acrossthis transition has been reported so far In addition few gen-era have species in both forest and savanna In cases in whichavian genera do have species in both habitats those speciesare generally considered distantly related

Connections among Neotropical savannas SouthAmerican tropical savanna regions can be grouped into twomajor groups that are currently separated by the Amazon val-ley The northern block is formed by Llanos Roraima ParuMonte Alegre Amapaacute and Marajoacute whereas the second blockis formed by Cerrado Pantanal Llanos de Mojos and patchesof savannas located close to the transition between Amazo-nia and the Cerrado (Figure 5) Because South American sa-vannas exhibit biotic similarity to one another (Sarmiento1983) a consensus opinion is that these regions were con-nected in the recent past and the biotic disjunctions ob-served today are the result of vicariance rather than long-dis-tance dispersal across large expanses of Amazonian forest(Haffer 1967 Sarmiento 1983) Under this scenario SouthAmerican savannas are believed to have expanded and re-tracted their ranges during Quaternary climatic cycles Dur-ing the cold dry periods savannas expanded in Amazoniawhile humid forests retracted to peripheral ecological refugesDuring moist and warm periods humid forests spread againwhile savannas retracted to areas approximating their present-day ranges (Mayle et al 2000) These Quaternary cyclic cli-maticndashvegetational changes have been postulated by some bio-geographers to be the most important factor driving thespeciation process in both forest and savanna organisms intropical South America (Whitmore and Prance 1987 Haffer2001)

Three main corridors connecting northern and southernsavanna regions have been proposed (Haffer 1967 1974Webb 1991) (a) Andean connecting the southern block ofsavannas directly with Llanos and Roraima through the An-dean slopes (b) central Amazonian connecting the southernblock of savannas directly with some patches of savannaslocated north of the Amazon such as Monte Alegre andParu roughly following a belt of low precipitation acrosscentral Amazonia and (c) coastal connecting the southernand northern blocks through savanna patches located closeto the Atlantic coast such as Marajoacute and Amapaacute (Figure 5)Silva (1995a) examined the general distribution patterns ofsavanna-associated birds in the Cerrado region to determinehow well these patterns fit with these three proposed corri-dors It was assumed that if these corridors did exist currentpatterns of distribution of savanna-adapted birds might showevidence of their legacy Thus we predicted spatial congru-ence should exist between speciesrsquo ranges and the position ofthese past biotic corridors as mapped by their proponents (Fig-ure 5)

Present-day ranges of savanna birds support two of the threehypothesized biogeographic corridors (Figure 6) The Andeanconnection is supported by the distribution of five species andlinks the Cerrado region with the Llanos and Gratilde-Sabana The

Atlantic coast savanna corridor is consistent with the distri-butions of 33 species All these species occur in Amapaacute orMarajoacute with some of them extending their ranges in other sa-vannas of the northern block (Pinto Henriques and Oren1997 Silva et al 1997) No single species was recorded onlyin the savannas that follow the central Amazonian savanna cor-ridor (Silva 1995a) These two pieces of information suggestthat a savanna corridor following a belt of low precipitationacross the Amazon basin may not have existed and the mostrecent biotic connections between Amazonian savannas andthe Cerrado region were mostly through the savannas locatedalong the Atlantic coast (Silva 1995a Silva et al 1997) Aacutevila-Pires (1995) analyzed the distribution patterns of all lizardspecies within Amazonia and also found little support for acentral Amazonian savanna corridor Thus even though sa-vanna biotas have expanded and contracted their ranges dur-ing the Quaternary climatic cycles (Mayle et al 2000) thereis little evidence to suggest that these changes resulted in abroad savanna corridor across central Amazonia Penningtonand colleagues (2000) suggested that several lines of evidencepoint to tropical dry forests rather than savannas as replac-ing rain forests in Amazonia during the coolest periods of theQuaternary

Biogeography and conservationUntil very recently conservation of biological resources in theCerrado had received little international attention To date

230 BioScience March 2002 Vol 52 No 3

Articles

Figure 5 Major savanna corridors that have been hy-pothesized to historically connect northern and southernblocks of South American tropical savannas (a) Andeancorridor (b) central Amazonian corridor and (c) coastalcorridor

less than 2 of the total area of the Cerrado is protected inreserves (Mittermeier et al 2000) but there are important dif-ferences in how countries are approaching Cerrado conser-vation The Bolivian government created Noel Kempff Na-tional Park in 1988 thereby protecting the countryrsquos largestcerrado (Killeen and Schulenberg 1998) These roughly 42000ha of unaltered cerrado near the western edge of the biomeis one of the worldrsquos largest continuous protected parcels ofthis habitat (Figure 7) Because of the parkrsquos remoteness andBoliviarsquos comparatively low human population density (es-pecially in the countryrsquos eastern lowlands) the most sub-stantial human threats to this area come from burning asso-ciated with ranching operations in neighboring Brazil Boliviaalso possesses several smaller and more isolated plateaus withcerrado habitat these remain unprotected and largely unex-plored (Parker and Rocha 1991)

In Paraguay at the southern limit of the biome Cerradois protected by two national parks Serraniacuteas de San Luiacutes andCerro Coraacute (Clay et al 1998 Robbins et al 1999) Avian sur-vey work has been conducted at both sites however the re-gions are still not considered to be thoroughly surveyed andboth Clay (1998) and Robbins and their colleagues (1999) em-phasize that the rapid expansion of agricultural operationsthreaten all unprotected savanna regions in the country

In Brazil which possesses the vast majority of Cerrado thesituation is drastically worse There are few large reserves(more than 25000 hectares [ha]) and they are not distributedevenly across the biome (Figure 7) Consequently an im-portant part of the Cerradorsquos environmental diversity hasnot been incorporated in a network of protected areas Largeareas of Brazilian Cerrado and dry forests have been convertedto soybean and rice plantations Most of this large-scale habi-tat modification has not followed the most basic principlesof environmental conservation and problems of erosion anddeterioration of important streams and rivers are increasingInternational support for this alteration has come from ma-jor development agencies and some senior scientists whohave encouraged increased land use in the Cerrado to reducehuman pressure on Amazonian forests

The Brazilian governmentrsquos appreciation of the uniquenessof the Cerradorsquos biodiversity has developed slowly The firstmajor initiative occurred in 1997 when the Brazilian Ministryof Environment promoted a workshop which included over200 scientific experts on the region This group met to definepriority areas for conservation in the Brazilian Cerrado It fol-lowed the same methodology applied earlier for the AtlanticForest and Amazonia (Cavalcanti 1999) Several criteria forassessing biodiversity value were used such as number ofendemic species species richness presence of rare or en-dangered species or both and sites of unique communitiesor key areas for migratory species The group identified a to-tal of 87 priority areas The urgency for conservation actionin a priority area was determined by cross-referencing bio-diversity data with the human-pressure and land cover-change data (Cavalcanti 1999) As of August 1999 the Brazil-ian government has indicated that it will follow most of the

recommendations of this workshop and that new protectedareas and ecological corridors will be established In factsome important actions were taken during 2001 but twodeserve special comments The first was the expansion ofthe National Park of Chapada dos Veadeiros an important re-serve that includes pristine cerrados and patches of camposrupestres with several endemic plant species from 60000 to235000 ha The second was the creation of the Ecological Sta-tion of Serra Geral do Tocantins with 716306 ha close to theState Park of Jalapatildeo with 158885 ha Together these two ar-eas form the worldrsquos largest continuous area of protectedCerrado (Figure 7)

How can the biogeographic information discussed here beused to help protect and preserve the Cerradorsquos biodiversityFirst we know now that the Cerradorsquos biota is not homoge-neous so additional distribution data for different groups oforganisms must be collected and organized in a retrievable wayto help conservationists determine whether there are addi-tional unidentified subareas of endemism Second eventhough the three subareas of endemism identified here as har-boring restricted-range endemic species of birds are partiallycovered by one or more reserves (for comparison refer to Fig-ures 4 and 7 Silva 1997) most reserves need to be more for-mally implemented and connected through biodiversity cor-ridors to insure their preservation over the long term Analysesof other nonavian taxa are certain to uncover additional ar-eas that are not currently protected by reserves Third despiteoccupying a relatively small part of the total landscape

March 2002 Vol 52 No 3 BioScience 231

Articles

Figure 6 Examples of ranges of bird species supportingthe hypothesis (a b) of the Andean savanna corridor and(c d) of the coastal savanna corridor

ba

c d

galleryndashriverine forest cerradatildeo anddry forest have been shown to be keyhabitats for a substantial portion ofthe Cerradorsquos biodiversity Thus theirconversion into pastures and agricul-tural fields should be halted and theirconnectivity maintained Fourth newprojects in areas currently covered bycerrado or other types of open vegeta-tion should be temporarily banned un-til their impact on the flora and faunacan be rigorously assessed and conser-vation strategies for those areas de-signed Fifth it is critical that agricul-tural technology be developed andimplemented to help landowners in-crease productivity of lands alreadyunder cultivation (Macedo 1994 Silva1998) This would reduce pressure onlands still covered by natural vegetationFinally a special conservation actionplan must be developed to guaranteethe conservation of at least part of thelargest savannas in Amazonia and At-lantic Forest as they are important lab-oratories for ecological and evolution-ary studies in tropical South AmericaIt is urgent now because efforts are underway to establish soy-bean plantations in these enclaves

The Cerradorsquos conservation is a challenge from both sci-entific and political viewpoints Because important financialand political interests are involved in the destruction of theCerrado scientists and conservationists must be creative ingathering and synthesizing the best and most complete in-formation possible to generate a viable strategy to insureconservation of the worldrsquos richest tropical savanna

AcknowledgmentsWe thank Jon Fjeldsaring David C Oren Roberto B CavalcantiFernando C Novaes Bent O Poulsen Mary E PetersenTownsend Peterson Carsten Rahbek James A Ratter NeivaGuedes and Shannon J Hackett for discussions about Cer-rado biogeography and comments on the manuscript LuisBarbosa produced the figures Research support came fromConselho Nacional de Desenvolvimento Cientiacutefico e Tec-noloacutegico (grant no 30246488-3) National Geographic So-ciety (grant no 4964-93) Danish Natural Science ResearchCouncil (grant J no 11-0390) Museu Paraense Emiacutelio Goeldiand Universidade de Brasiacutelia

References citedAbrsquoSaber AN 1977 Os Domiacutenios morfoclimaacuteticos da Ameacuterica do Sul

Primeira Aproximaccedilatildeo Geomorfologia 52 1ndash21mdashmdashmdash 1983 O domiacutenio dos cerrados Introduccedilatildeo ao conhecimento Re-

vista do Servidor Puacuteblico 111 41ndash55Aacutevila-Pires TC 1995 Lizards of Brazilian Amazonia (Reptilia Squamata)

Zoologische Verhandelingen Leiden 299 1ndash706

Bates JM Hackett SJ Goerck J 1999 High levels of mitochondrial DNA dif-ferentiation in two lineages of antbirds (Drymophila and Hypocnemis)The Auk 1161093ndash1106

Brasil AE Alvarenga SM 1989 Relevo Pages 53ndash72 in Duarte AC ed Ge-ografia do Brasil Regiatildeo Centro-Oeste Rio de Janeiro IBGE

Cavalcanti RB ed 1999 Accedilotildees Prioritaacuterias para a conservaccedilatildeo da biodiver-sidade do Cerrado e Pantanal Belo Horizonte Conservation Internationaldo Brasil

Clay RP Cooper DR Mazar Barnett J Burfiled IZ Esquivel EZ Farintildea RKennedy CP Perrens M Pople RG 1998 White-winged NightjarsCaprimulgus candicans and cerrado conservation The key findings of pro-ject Aguaraacute Nu 1997 Cotinga 9 52ndash55

Coelho G Silva W 1998 A new species of Antilophia (Paseriformes Pipri-dae) from Chapada do Araripe Cearaacute Brazil Ararajuba 6 81ndash86

Cole MM 1986 The Savannas Biogeography and Geobotany LondonAcademic Press

Cracraft J 1985 Historical biogeography and patterns of differentiationwithin the South America avifauna Areas of endemism OrnithologicalMonographs 36 49ndash84

Eiten G 1972 The Cerrado vegetation of Brazil Botanical Review 38201ndash341

mdashmdashmdash 1990 Vegetaccedilatildeo Pages 9ndash65 in Pinto MN ed Cerrado Caracteri-zaccedilatildeo Ocupaccedilatildeo e Perspectiva Brasiacutelia Universidade de Brasiacutelia and SE-MATEC

Erwin TL Pogue MG 1988 Agra arboreal beetles of Neotropical forests Bio-geography and the forest refugium hypothesis (Carabidae) Pages 161ndash188in Vanzolini PE Heyer WR eds Proceedings of a Workshop on Neotrop-ical Distribution Patterns Rio de Janeiro Academia Brasileira de Ciecircn-cias

Furley PA Ratter JA 1988 Soil resources and plant communities of the cen-tral Brazilian cerrado and their development Journal of Biogeography15 97ndash108

Giullietti AM Pirani JR 1988 Patterns of geographic distribution of someplant species from the Espinhaccedilo Range Minas Gerais and Bahia Brazil

232 BioScience March 2002 Vol 52 No 3

Articles

Figure 7 Parks and ecological stations in the Cerrado region Only those reserveslarger than 25000 ha are shown in Brazil PE state park PN national park EEecological station

226 BioScience March 2002 Vol 52 No 3

Articles

(Figure 1) with extensiveborders with the two largestforest biomes (Amazonia andAtlantic Forest) as well as withthe two largest dry regions(Caatinga and Chaco)

The Cerrado region has atropical wet and dry climatewith intermediate rainfall be-tween the wetter regions tothe northwest and southeastand the drier areas to thenortheast Average annualrainfall varies between 125and 200 cm and average an-nual temperature between20deg and 26deg C (Nimer 1979AbrsquoSaber 1983) The dry pe-riod lasts from 3 to 5 consec-utive months (generally be-tween May and September orOctober) during the winterof the Southern Hemisphere(Nimer 1979 Sarmiento1983)

Most of the Cerrado is on plateaus of crystalline or sedi-mentary blocks whose continuity is broken by an extensivenetwork of peripheral or interplateau depressions (Brasiland Alvarenga 1989) This geomorphological variation ex-plains much of the distribution of plants in the Cerrado(Cole 1986) Savannas are by far the dominant vegetation cov-ering around 72 of the region but patches of dry forests andcomplex belts composed of both dry forests and savannas givethe region a mosaic-like aspect (Figure 2) We estimate that24 of the region represents this mosaic that we call savannandashforest transition with a remaining 4 being dry forest

The tops of the plateaus are flat gently rolling surfaces atelevations ranging from 500 to 1700 meters (m) They are cov-ered primarily by cerrado a semideciduous to evergreen sa-vanna-like vegetation growing on the deep well-drainedand nutrient-poor soils (Eiten 1972 1990 Furley and Ratter1988) Throughout cerrado vegetation varies in physiog-nomy and composition (Furley and Ratter 1988 Ribeiro andWalter 1998) Five main structural types of cerrado are rec-ognized by botanists (Eiten 1972)

1 cerradatildeo a dense forest type (8ndash15 m tall) that often hasa completely closed canopy

2 cerrado sensu stricto a woodland (5ndash8 m tall) withclosed scrub and more scattered trees than in cerradatildeo(Figure 3c)

3 campo cerrado an open scrubland (3ndash6 m tall) with fewtrees (Figure 3a)

Figure 1 Distribution of major lowland tropical biomes inSouth America The Cerrado occupies a central position in rela-tion to other lowland tropical biomes on the continent Othermajor savanna regions are the Llanos Roraima the Llanos deMojos and the Pantanal Note the presence of several savannapatches within Amazonia and Atlantic Forest

Figure 2 Major vegetation types in the Cerrado The savannandashforest transition is a mosaiccomposed of patches of forest and savannas

4 campo sujo grassland (2ndash3 m tall) with scattered shrubs

5 campo limpo grassland with few or no shrubs or tallerwoody plants (Figure 3b)

Tree species common in the cerrado vegetation are Annonacrassiflora Astronium fraxinifolium Byrsonima coccolobifoliaB verbascifolia Hancornia speciosa Kielmeyera coriaceaQualea grandifolia and Q parviflora among others (Ratterand Dargie 1992) Besides cerrado other types of distinct sa-vanna vegetation called campos rupestres (Figure 3a) andmiscellaneous lithosolic campos occur on rocky outcropsThese habitats are limited to small patches on the plateaus and

have a highly endemic flora (Giullietti and Pirani 1988 Eiten1990) Narrow fringes of gallery forest often border small riversand streams of the region Gallery forests on the plateausgrow on narrow belts of soils rich in organic matter (Eiten1990) They are evergreen have trees that are on average20ndash30 m tall and possess a dense understory of low stature(Ribeiro and Walter 1998) Associated with gallery forestsare stands of a species of palm Mauritia flexuosa locallyknown as veredas

The peripheral depressions are generally flat surfaces of lit-tle relief (100ndash500 m in elevation) that are ocasionally inter-rupted by the presence of steep-sided hills The landscape pat-

March 2002 Vol 52 No 3 BioScience 227

Articles

Figure 3 Examples of several of the structural habitat typesthat grade into one another across the Cerrado Bird species endemic to the cerrado region and typical of these habitats aregiven in parentheses (Table 1) (a) Campo rupestre (blue finchPorphyrospiza caerulescens) campo sujo (collared crescent-chest Melanopareia torquata) and campo cerrado (black-throated saltator Saltator atricollis and curl-crested jayCyanocorax cristatellus) atop the Serraniacutea de Huanchaca (b)Campo limpo (campo miner Geobates poecilopterus) with asmall finger of gallery forest (helmeted manakin Antilophiagaleata) in the distance (c) Campo cerrado (chapada fly-catcher Suiriri islerorum) (d) Cerrado sensu stricto (coal-crested finch Charitospiza eucosma) All photographs taken at Noel Kempff Mercado National Park Bolivia

ba

d

c

tern on these depressions is much more heterogeneous thanthat of the plateaus in that it includes more habitats (river-ine forests tropical dry forests all types of cerrado andmarshlands) distributed in a mosaic fashion Riverine forests(called matas ciliares Ribeiro and Walter 1998) differ fromgallery forests because they are found only along the large riversin the Cerrado They are not wide (up to 100 m from each sideof the river) and can be semideciduous to evergreen Trees are20ndash30 m tall Common species are Anadenanthera sppApeiba tibourbou Aspidosperma spp Tabebuia spp andTrema micrantha among others (Ribeiro and Walter 1998)Tropical dry forests (matas secas) are particularly associatedwith peripheral depressions and in some areas (eg in theParanatilde River valley) are the dominant vegetation type Theyare deciduous or semideciduous and grow on patches ofmoderately productive soils derived from basic rocks such aslimestone (Ratter et al 1978 Prado and Gibbs 1993) Canopyheight averages 20ndash45 m (Ribeiro and Walter 1998) Treespecies common in dry forests include Astronium urundeuvaPiptadenia macrocarpa Chorisia speciosa Tabebuia spp Ca-vanillesia arborea and Cedrella fissilis (Ratter et al 1978)

Biogeographic patterns

Forest organisms in a tropical savanna regionThe most basic question about the biota of a region is howit is distributed in major vegetation types In the Cerrado onemight expect that most of the biota would inhabit the savannasthat dominate the region However studies of mammalsbirds and plants do not support this hypothesis Redford andFonseca (1986) found that most (563) of the nonvolantmammal species in the Cerrado inhabit forests SimilarlySilva (1995b) classified the 759 birds species known or as-sumed to breed in the Cerrado as independent (occurring ex-clusively in forests) semidependent (occurring in both forestsand open vegetation) or dependent (occurring exclusively inopen vegetation) The majority of species (393 or 518) aredependent on the various forest habitats 208 (158) aresemidependent and 274 (208) are independent speciesThus approximately 826 of Cerrado birds require foreststo some degree Mendonccedila and colleagues (1998) listed 6671plant taxa (species and varieties) for the Cerrado of these 38occur only in forests

In general these data support the hypothesis that galleryand dry forests which occupy less than 20 of the Cerradoare necessary for a large portion of regional biodiversityHow did this pattern evolve Silva (1995b) suggested that biotic diversity of the Cerrado region increased through in-terchange with adjacent regions during the Quaternary climaticndashvegetational fluctuations Thus Atlantic and Ama-zonian forest species expanded their ranges into the Cerradoduring the wet periods following the expansion of the riverinendashgallery forest network (Silva 1995b) whereas Caatingaand Chaco elements colonized the Cerrado during the Qua-ternary dry periods as part of the coalescence of dry forestsin peripheral depressions (AbrsquoSaber 1983 Pennington et al

2000) These species probably originated in adjacent regionsand maintained viable populations within the Cerrado regionbecause their forest habitats did not vanish entirely during thepaleoclimatic period that followed their range expansion Ina fashion similar to what is seen today large rivers and streamsmust have provided suitable environmental conditions tomaintain wide galleryndashriverine forests during the dry periods(AbrsquoSaber 1983) whereas large patches of soils derived fromlimestone were important for maintaining dry forests duringthe wet periods (Prado and Gibbs 1993 Pennington et al2000)

Endemic species Origin and evolution The per-centage of species endemic to the Cerrado varies across dif-ferent groups of organisms vascular plants (44) amphib-ians (30) reptiles (20) mammals (118) and birds(14) (Silva 1995a Myers et al 2000) However because largeareas of the Cerrado remain unexplored (Silva 1995c) thesenumbers will likely change when additional biological in-ventories are conducted At present we still know little abouthow these endemic species are distributed across the Cerradoor how they evolved A key question for conservation purposesis whether one can identify subareas of endemism withinthe large Cerrado region

To determine if there are subareas of endemism within theCerrado Silva (1997) mapped and created overlays of all en-demic bird speciesrsquo ranges smaller than 60000 km2 Only 10of the 30 endemic species met this requirement (Table 1) Theranges of these species delimit three main areas

Espinhaccedilo Plateau Four species that inhabit camposrupestres are Augastes scutatus (Trochilidae) Asthenesluizae (Furnariidae) Polystictus superciliaris (Tyran-nidae) and Embernagra longicauda (Emberizidae)

Araguaia River Valley Three species that inhabit galleryforests are Synallaxis simoni (Furnariidae) Cercomacraferdinandi (Thamnophilidae) and Paroaria baeri(Emberizidae)

Paranatilde River Valley Two species found only in tropicaldry forests are Pyrrhura pfrimeri (Psittacidae) andKnipolegus franciscanus (Tyrannidae) (Figure 4)

The subareas of endemism identified for birds are at leastpartially supported by studies of other groups of organismsThe Espinhaccedilo Plateau is well known for its unique biotacomposed of hundreds of endemic species of plants (Giulli-eti and Pirani 1988 Harley 1988) and some endemic speciesof other vertebrates such as lizards and amphibians (Haddadet al 1988 Rodrigues 1988) The Araguaia and Paranatilde rivervalleys are still poorly known for nonavian groups but theriverine forests and dry forests likely harbor endemic speciesof other organisms also For example a very distinctive speciesof rodent of the genus Kerodon was described recently fromParanatilde dry forests (Moojen et al 1997) Although not knownto possess endemic birds several other areas possess endemictaxa from other groups For instance Erwin and Pogue (1988)

228 BioScience March 2002 Vol 52 No 3

Articles

indicated that the Mato Grosso Plateau is an important areaof endemism for beetles of the genus Agra Thus bird distri-butions indicate three subareas of endemism however thisnumber is almost certainly an underestimate for other groupssuch as some insects vascular plants and amphibians Froma taxonomic perspective these groups are still too poorlyknown to be used for assessing patterns of endemism withinthe Cerrado

Investigation of the timing and mode of the evolution ofthe endemic species of the Cerrado is in its infancy butnonetheless raises controversy Some authors (eg Rizzini1979) have suggested that the original vegetation in the Cer-rado was forest rather than savanna In this view present-daysavannas succeeded forests by natural processes such as fireand climatic changes At the opposite extreme Cole (1986)suggests that within the Cerrado the savanna vegetation isolder than the forests based on geomorphological informa-

tion If one assumes that avian diversification within the Cer-rado followed at least roughly the same trend as evolutionof the landscape in this region then one would predict basedon Rizzinirsquos hypothesis that older species are associatedmainly with forest vegetation and younger species are asso-ciated mainly with savannas Under Colersquos hypothesis the pre-diction would be inverted To evaluate these two hypotheseswe classified each endemic species into one of two evolutionaryage categories paleoendemics species that are either taxo-nomically distinct (ie well-differentiated monotypic gen-era) or considered to be the sister taxon of a large radiationin which one or more members also occur in the Cerrado re-gion and neoendemics species with sister taxa in adjacentSouth American regions Based on recent molecular work onSouth American birds (Bates et al 1999) we suggest thatneoendemics are younger than the PliocenendashPleistocene tran-sition (about 18 million years ago) whereas paleoendemicsoriginated before this period This is only a tentative classi-fication that must be evaluated in future molecular studiesIn addition we classified each avian endemic as either a for-est or nonforest species All 30 endemic species were analyzed(Table 1) We determined 12 species to be paleoendemicsand 18 to be neoendemics The majority (11 of 12) of the pa-leoendemic species are nonforest species whereas most of theneoendemics (12 of 18) are forest birds These results supportColersquos rather than Rizzinirsquos hypothesis

An additional aspect of these data concerns recent sug-gestions that ecotones or transition zones between two ormore distinctive types of vegetation provide important op-portunities for speciation (Smith et al 1997 Schlithuizen2000) For the Cerrado avifauna this does not appear to bewell supported with respect to the ecotone between savannaand forest Although 158 bird species inhabit both forest and

March 2002 Vol 52 No 3 BioScience 229

Articles

Table 1 Endemic bird species of the Cerrado regiona

Species Ageb Habitatc

Nothura minor P NTaoniscus nanusd P NPenelope ochrogaster N FColumbina cyanopis P NPyrrhura pfrimeri N FAmazona xanthops P NCaprimulgus candicans P NAugastes scutatus N NGeobates poecilopterusd P NSynallaxis simoni N FAsthenes luizae N NPhilydor dimidiatus N FAutomolus rectirostris N FHerpsilochmus longirostris N FCercomacra ferdinandi N FMelanopareia torquata N NScytalopus novacapitalis N FPhyllomyias reiseri N FSuiriri islerorum N FPolystictus superciliaris P NKnipolegus fraciscanus N FAntilophia galeata N FPoospiza cinerea N NEmbernagra longicauda N NCharitospiza eucosmad P NParoaria baeri N FSaltator atricollis P NPorphyrospiza caerulescensd P NBasileuterus leucophrys N FCyanocorax cristatellus P N

a This data set has been modified from Silva (1997) as Geobates

poecilopterus (a monotypic genus) has been moved from neoendem-

ic to paleoendemic and Antilophia galeata has been changed to

neoendemic because of the recent discovery of Antilophia bokerman-

ni (Coelho and Silva 1998) Suiriri islerorum is a newly described

endemic species (Zimmer et al 2001)

b P paleoendemic N neoendemic

c N nonforest F forest

d Monotypic genus

Figure 4 Three subareas of endemism are identified for birdswithin the Cerrado Espinhaccedilo Plateau (with four endemicspecies) Araguaia River Valley (with three endemic species)and Paranatilde River Valley (with two endemic species)

savanna no single case of intraspecific differentiation acrossthis transition has been reported so far In addition few gen-era have species in both forest and savanna In cases in whichavian genera do have species in both habitats those speciesare generally considered distantly related

Connections among Neotropical savannas SouthAmerican tropical savanna regions can be grouped into twomajor groups that are currently separated by the Amazon val-ley The northern block is formed by Llanos Roraima ParuMonte Alegre Amapaacute and Marajoacute whereas the second blockis formed by Cerrado Pantanal Llanos de Mojos and patchesof savannas located close to the transition between Amazo-nia and the Cerrado (Figure 5) Because South American sa-vannas exhibit biotic similarity to one another (Sarmiento1983) a consensus opinion is that these regions were con-nected in the recent past and the biotic disjunctions ob-served today are the result of vicariance rather than long-dis-tance dispersal across large expanses of Amazonian forest(Haffer 1967 Sarmiento 1983) Under this scenario SouthAmerican savannas are believed to have expanded and re-tracted their ranges during Quaternary climatic cycles Dur-ing the cold dry periods savannas expanded in Amazoniawhile humid forests retracted to peripheral ecological refugesDuring moist and warm periods humid forests spread againwhile savannas retracted to areas approximating their present-day ranges (Mayle et al 2000) These Quaternary cyclic cli-maticndashvegetational changes have been postulated by some bio-geographers to be the most important factor driving thespeciation process in both forest and savanna organisms intropical South America (Whitmore and Prance 1987 Haffer2001)

Three main corridors connecting northern and southernsavanna regions have been proposed (Haffer 1967 1974Webb 1991) (a) Andean connecting the southern block ofsavannas directly with Llanos and Roraima through the An-dean slopes (b) central Amazonian connecting the southernblock of savannas directly with some patches of savannaslocated north of the Amazon such as Monte Alegre andParu roughly following a belt of low precipitation acrosscentral Amazonia and (c) coastal connecting the southernand northern blocks through savanna patches located closeto the Atlantic coast such as Marajoacute and Amapaacute (Figure 5)Silva (1995a) examined the general distribution patterns ofsavanna-associated birds in the Cerrado region to determinehow well these patterns fit with these three proposed corri-dors It was assumed that if these corridors did exist currentpatterns of distribution of savanna-adapted birds might showevidence of their legacy Thus we predicted spatial congru-ence should exist between speciesrsquo ranges and the position ofthese past biotic corridors as mapped by their proponents (Fig-ure 5)

Present-day ranges of savanna birds support two of the threehypothesized biogeographic corridors (Figure 6) The Andeanconnection is supported by the distribution of five species andlinks the Cerrado region with the Llanos and Gratilde-Sabana The

Atlantic coast savanna corridor is consistent with the distri-butions of 33 species All these species occur in Amapaacute orMarajoacute with some of them extending their ranges in other sa-vannas of the northern block (Pinto Henriques and Oren1997 Silva et al 1997) No single species was recorded onlyin the savannas that follow the central Amazonian savanna cor-ridor (Silva 1995a) These two pieces of information suggestthat a savanna corridor following a belt of low precipitationacross the Amazon basin may not have existed and the mostrecent biotic connections between Amazonian savannas andthe Cerrado region were mostly through the savannas locatedalong the Atlantic coast (Silva 1995a Silva et al 1997) Aacutevila-Pires (1995) analyzed the distribution patterns of all lizardspecies within Amazonia and also found little support for acentral Amazonian savanna corridor Thus even though sa-vanna biotas have expanded and contracted their ranges dur-ing the Quaternary climatic cycles (Mayle et al 2000) thereis little evidence to suggest that these changes resulted in abroad savanna corridor across central Amazonia Penningtonand colleagues (2000) suggested that several lines of evidencepoint to tropical dry forests rather than savannas as replac-ing rain forests in Amazonia during the coolest periods of theQuaternary

Biogeography and conservationUntil very recently conservation of biological resources in theCerrado had received little international attention To date

230 BioScience March 2002 Vol 52 No 3

Articles

Figure 5 Major savanna corridors that have been hy-pothesized to historically connect northern and southernblocks of South American tropical savannas (a) Andeancorridor (b) central Amazonian corridor and (c) coastalcorridor

less than 2 of the total area of the Cerrado is protected inreserves (Mittermeier et al 2000) but there are important dif-ferences in how countries are approaching Cerrado conser-vation The Bolivian government created Noel Kempff Na-tional Park in 1988 thereby protecting the countryrsquos largestcerrado (Killeen and Schulenberg 1998) These roughly 42000ha of unaltered cerrado near the western edge of the biomeis one of the worldrsquos largest continuous protected parcels ofthis habitat (Figure 7) Because of the parkrsquos remoteness andBoliviarsquos comparatively low human population density (es-pecially in the countryrsquos eastern lowlands) the most sub-stantial human threats to this area come from burning asso-ciated with ranching operations in neighboring Brazil Boliviaalso possesses several smaller and more isolated plateaus withcerrado habitat these remain unprotected and largely unex-plored (Parker and Rocha 1991)

In Paraguay at the southern limit of the biome Cerradois protected by two national parks Serraniacuteas de San Luiacutes andCerro Coraacute (Clay et al 1998 Robbins et al 1999) Avian sur-vey work has been conducted at both sites however the re-gions are still not considered to be thoroughly surveyed andboth Clay (1998) and Robbins and their colleagues (1999) em-phasize that the rapid expansion of agricultural operationsthreaten all unprotected savanna regions in the country

In Brazil which possesses the vast majority of Cerrado thesituation is drastically worse There are few large reserves(more than 25000 hectares [ha]) and they are not distributedevenly across the biome (Figure 7) Consequently an im-portant part of the Cerradorsquos environmental diversity hasnot been incorporated in a network of protected areas Largeareas of Brazilian Cerrado and dry forests have been convertedto soybean and rice plantations Most of this large-scale habi-tat modification has not followed the most basic principlesof environmental conservation and problems of erosion anddeterioration of important streams and rivers are increasingInternational support for this alteration has come from ma-jor development agencies and some senior scientists whohave encouraged increased land use in the Cerrado to reducehuman pressure on Amazonian forests

The Brazilian governmentrsquos appreciation of the uniquenessof the Cerradorsquos biodiversity has developed slowly The firstmajor initiative occurred in 1997 when the Brazilian Ministryof Environment promoted a workshop which included over200 scientific experts on the region This group met to definepriority areas for conservation in the Brazilian Cerrado It fol-lowed the same methodology applied earlier for the AtlanticForest and Amazonia (Cavalcanti 1999) Several criteria forassessing biodiversity value were used such as number ofendemic species species richness presence of rare or en-dangered species or both and sites of unique communitiesor key areas for migratory species The group identified a to-tal of 87 priority areas The urgency for conservation actionin a priority area was determined by cross-referencing bio-diversity data with the human-pressure and land cover-change data (Cavalcanti 1999) As of August 1999 the Brazil-ian government has indicated that it will follow most of the

recommendations of this workshop and that new protectedareas and ecological corridors will be established In factsome important actions were taken during 2001 but twodeserve special comments The first was the expansion ofthe National Park of Chapada dos Veadeiros an important re-serve that includes pristine cerrados and patches of camposrupestres with several endemic plant species from 60000 to235000 ha The second was the creation of the Ecological Sta-tion of Serra Geral do Tocantins with 716306 ha close to theState Park of Jalapatildeo with 158885 ha Together these two ar-eas form the worldrsquos largest continuous area of protectedCerrado (Figure 7)

How can the biogeographic information discussed here beused to help protect and preserve the Cerradorsquos biodiversityFirst we know now that the Cerradorsquos biota is not homoge-neous so additional distribution data for different groups oforganisms must be collected and organized in a retrievable wayto help conservationists determine whether there are addi-tional unidentified subareas of endemism Second eventhough the three subareas of endemism identified here as har-boring restricted-range endemic species of birds are partiallycovered by one or more reserves (for comparison refer to Fig-ures 4 and 7 Silva 1997) most reserves need to be more for-mally implemented and connected through biodiversity cor-ridors to insure their preservation over the long term Analysesof other nonavian taxa are certain to uncover additional ar-eas that are not currently protected by reserves Third despiteoccupying a relatively small part of the total landscape

March 2002 Vol 52 No 3 BioScience 231

Articles

Figure 6 Examples of ranges of bird species supportingthe hypothesis (a b) of the Andean savanna corridor and(c d) of the coastal savanna corridor

ba

c d

galleryndashriverine forest cerradatildeo anddry forest have been shown to be keyhabitats for a substantial portion ofthe Cerradorsquos biodiversity Thus theirconversion into pastures and agricul-tural fields should be halted and theirconnectivity maintained Fourth newprojects in areas currently covered bycerrado or other types of open vegeta-tion should be temporarily banned un-til their impact on the flora and faunacan be rigorously assessed and conser-vation strategies for those areas de-signed Fifth it is critical that agricul-tural technology be developed andimplemented to help landowners in-crease productivity of lands alreadyunder cultivation (Macedo 1994 Silva1998) This would reduce pressure onlands still covered by natural vegetationFinally a special conservation actionplan must be developed to guaranteethe conservation of at least part of thelargest savannas in Amazonia and At-lantic Forest as they are important lab-oratories for ecological and evolution-ary studies in tropical South AmericaIt is urgent now because efforts are underway to establish soy-bean plantations in these enclaves

The Cerradorsquos conservation is a challenge from both sci-entific and political viewpoints Because important financialand political interests are involved in the destruction of theCerrado scientists and conservationists must be creative ingathering and synthesizing the best and most complete in-formation possible to generate a viable strategy to insureconservation of the worldrsquos richest tropical savanna

AcknowledgmentsWe thank Jon Fjeldsaring David C Oren Roberto B CavalcantiFernando C Novaes Bent O Poulsen Mary E PetersenTownsend Peterson Carsten Rahbek James A Ratter NeivaGuedes and Shannon J Hackett for discussions about Cer-rado biogeography and comments on the manuscript LuisBarbosa produced the figures Research support came fromConselho Nacional de Desenvolvimento Cientiacutefico e Tec-noloacutegico (grant no 30246488-3) National Geographic So-ciety (grant no 4964-93) Danish Natural Science ResearchCouncil (grant J no 11-0390) Museu Paraense Emiacutelio Goeldiand Universidade de Brasiacutelia

References citedAbrsquoSaber AN 1977 Os Domiacutenios morfoclimaacuteticos da Ameacuterica do Sul

Primeira Aproximaccedilatildeo Geomorfologia 52 1ndash21mdashmdashmdash 1983 O domiacutenio dos cerrados Introduccedilatildeo ao conhecimento Re-

vista do Servidor Puacuteblico 111 41ndash55Aacutevila-Pires TC 1995 Lizards of Brazilian Amazonia (Reptilia Squamata)

Zoologische Verhandelingen Leiden 299 1ndash706

Bates JM Hackett SJ Goerck J 1999 High levels of mitochondrial DNA dif-ferentiation in two lineages of antbirds (Drymophila and Hypocnemis)The Auk 1161093ndash1106

Brasil AE Alvarenga SM 1989 Relevo Pages 53ndash72 in Duarte AC ed Ge-ografia do Brasil Regiatildeo Centro-Oeste Rio de Janeiro IBGE

Cavalcanti RB ed 1999 Accedilotildees Prioritaacuterias para a conservaccedilatildeo da biodiver-sidade do Cerrado e Pantanal Belo Horizonte Conservation Internationaldo Brasil

Clay RP Cooper DR Mazar Barnett J Burfiled IZ Esquivel EZ Farintildea RKennedy CP Perrens M Pople RG 1998 White-winged NightjarsCaprimulgus candicans and cerrado conservation The key findings of pro-ject Aguaraacute Nu 1997 Cotinga 9 52ndash55

Coelho G Silva W 1998 A new species of Antilophia (Paseriformes Pipri-dae) from Chapada do Araripe Cearaacute Brazil Ararajuba 6 81ndash86

Cole MM 1986 The Savannas Biogeography and Geobotany LondonAcademic Press

Cracraft J 1985 Historical biogeography and patterns of differentiationwithin the South America avifauna Areas of endemism OrnithologicalMonographs 36 49ndash84

Eiten G 1972 The Cerrado vegetation of Brazil Botanical Review 38201ndash341

mdashmdashmdash 1990 Vegetaccedilatildeo Pages 9ndash65 in Pinto MN ed Cerrado Caracteri-zaccedilatildeo Ocupaccedilatildeo e Perspectiva Brasiacutelia Universidade de Brasiacutelia and SE-MATEC

Erwin TL Pogue MG 1988 Agra arboreal beetles of Neotropical forests Bio-geography and the forest refugium hypothesis (Carabidae) Pages 161ndash188in Vanzolini PE Heyer WR eds Proceedings of a Workshop on Neotrop-ical Distribution Patterns Rio de Janeiro Academia Brasileira de Ciecircn-cias

Furley PA Ratter JA 1988 Soil resources and plant communities of the cen-tral Brazilian cerrado and their development Journal of Biogeography15 97ndash108

Giullietti AM Pirani JR 1988 Patterns of geographic distribution of someplant species from the Espinhaccedilo Range Minas Gerais and Bahia Brazil

232 BioScience March 2002 Vol 52 No 3

Articles

Figure 7 Parks and ecological stations in the Cerrado region Only those reserveslarger than 25000 ha are shown in Brazil PE state park PN national park EEecological station

4 campo sujo grassland (2ndash3 m tall) with scattered shrubs

5 campo limpo grassland with few or no shrubs or tallerwoody plants (Figure 3b)

Tree species common in the cerrado vegetation are Annonacrassiflora Astronium fraxinifolium Byrsonima coccolobifoliaB verbascifolia Hancornia speciosa Kielmeyera coriaceaQualea grandifolia and Q parviflora among others (Ratterand Dargie 1992) Besides cerrado other types of distinct sa-vanna vegetation called campos rupestres (Figure 3a) andmiscellaneous lithosolic campos occur on rocky outcropsThese habitats are limited to small patches on the plateaus and

have a highly endemic flora (Giullietti and Pirani 1988 Eiten1990) Narrow fringes of gallery forest often border small riversand streams of the region Gallery forests on the plateausgrow on narrow belts of soils rich in organic matter (Eiten1990) They are evergreen have trees that are on average20ndash30 m tall and possess a dense understory of low stature(Ribeiro and Walter 1998) Associated with gallery forestsare stands of a species of palm Mauritia flexuosa locallyknown as veredas

The peripheral depressions are generally flat surfaces of lit-tle relief (100ndash500 m in elevation) that are ocasionally inter-rupted by the presence of steep-sided hills The landscape pat-

March 2002 Vol 52 No 3 BioScience 227

Articles

Figure 3 Examples of several of the structural habitat typesthat grade into one another across the Cerrado Bird species endemic to the cerrado region and typical of these habitats aregiven in parentheses (Table 1) (a) Campo rupestre (blue finchPorphyrospiza caerulescens) campo sujo (collared crescent-chest Melanopareia torquata) and campo cerrado (black-throated saltator Saltator atricollis and curl-crested jayCyanocorax cristatellus) atop the Serraniacutea de Huanchaca (b)Campo limpo (campo miner Geobates poecilopterus) with asmall finger of gallery forest (helmeted manakin Antilophiagaleata) in the distance (c) Campo cerrado (chapada fly-catcher Suiriri islerorum) (d) Cerrado sensu stricto (coal-crested finch Charitospiza eucosma) All photographs taken at Noel Kempff Mercado National Park Bolivia

ba

d

c

tern on these depressions is much more heterogeneous thanthat of the plateaus in that it includes more habitats (river-ine forests tropical dry forests all types of cerrado andmarshlands) distributed in a mosaic fashion Riverine forests(called matas ciliares Ribeiro and Walter 1998) differ fromgallery forests because they are found only along the large riversin the Cerrado They are not wide (up to 100 m from each sideof the river) and can be semideciduous to evergreen Trees are20ndash30 m tall Common species are Anadenanthera sppApeiba tibourbou Aspidosperma spp Tabebuia spp andTrema micrantha among others (Ribeiro and Walter 1998)Tropical dry forests (matas secas) are particularly associatedwith peripheral depressions and in some areas (eg in theParanatilde River valley) are the dominant vegetation type Theyare deciduous or semideciduous and grow on patches ofmoderately productive soils derived from basic rocks such aslimestone (Ratter et al 1978 Prado and Gibbs 1993) Canopyheight averages 20ndash45 m (Ribeiro and Walter 1998) Treespecies common in dry forests include Astronium urundeuvaPiptadenia macrocarpa Chorisia speciosa Tabebuia spp Ca-vanillesia arborea and Cedrella fissilis (Ratter et al 1978)

Biogeographic patterns

Forest organisms in a tropical savanna regionThe most basic question about the biota of a region is howit is distributed in major vegetation types In the Cerrado onemight expect that most of the biota would inhabit the savannasthat dominate the region However studies of mammalsbirds and plants do not support this hypothesis Redford andFonseca (1986) found that most (563) of the nonvolantmammal species in the Cerrado inhabit forests SimilarlySilva (1995b) classified the 759 birds species known or as-sumed to breed in the Cerrado as independent (occurring ex-clusively in forests) semidependent (occurring in both forestsand open vegetation) or dependent (occurring exclusively inopen vegetation) The majority of species (393 or 518) aredependent on the various forest habitats 208 (158) aresemidependent and 274 (208) are independent speciesThus approximately 826 of Cerrado birds require foreststo some degree Mendonccedila and colleagues (1998) listed 6671plant taxa (species and varieties) for the Cerrado of these 38occur only in forests

In general these data support the hypothesis that galleryand dry forests which occupy less than 20 of the Cerradoare necessary for a large portion of regional biodiversityHow did this pattern evolve Silva (1995b) suggested that biotic diversity of the Cerrado region increased through in-terchange with adjacent regions during the Quaternary climaticndashvegetational fluctuations Thus Atlantic and Ama-zonian forest species expanded their ranges into the Cerradoduring the wet periods following the expansion of the riverinendashgallery forest network (Silva 1995b) whereas Caatingaand Chaco elements colonized the Cerrado during the Qua-ternary dry periods as part of the coalescence of dry forestsin peripheral depressions (AbrsquoSaber 1983 Pennington et al

2000) These species probably originated in adjacent regionsand maintained viable populations within the Cerrado regionbecause their forest habitats did not vanish entirely during thepaleoclimatic period that followed their range expansion Ina fashion similar to what is seen today large rivers and streamsmust have provided suitable environmental conditions tomaintain wide galleryndashriverine forests during the dry periods(AbrsquoSaber 1983) whereas large patches of soils derived fromlimestone were important for maintaining dry forests duringthe wet periods (Prado and Gibbs 1993 Pennington et al2000)

Endemic species Origin and evolution The per-centage of species endemic to the Cerrado varies across dif-ferent groups of organisms vascular plants (44) amphib-ians (30) reptiles (20) mammals (118) and birds(14) (Silva 1995a Myers et al 2000) However because largeareas of the Cerrado remain unexplored (Silva 1995c) thesenumbers will likely change when additional biological in-ventories are conducted At present we still know little abouthow these endemic species are distributed across the Cerradoor how they evolved A key question for conservation purposesis whether one can identify subareas of endemism withinthe large Cerrado region

To determine if there are subareas of endemism within theCerrado Silva (1997) mapped and created overlays of all en-demic bird speciesrsquo ranges smaller than 60000 km2 Only 10of the 30 endemic species met this requirement (Table 1) Theranges of these species delimit three main areas

Espinhaccedilo Plateau Four species that inhabit camposrupestres are Augastes scutatus (Trochilidae) Asthenesluizae (Furnariidae) Polystictus superciliaris (Tyran-nidae) and Embernagra longicauda (Emberizidae)

Araguaia River Valley Three species that inhabit galleryforests are Synallaxis simoni (Furnariidae) Cercomacraferdinandi (Thamnophilidae) and Paroaria baeri(Emberizidae)

Paranatilde River Valley Two species found only in tropicaldry forests are Pyrrhura pfrimeri (Psittacidae) andKnipolegus franciscanus (Tyrannidae) (Figure 4)

The subareas of endemism identified for birds are at leastpartially supported by studies of other groups of organismsThe Espinhaccedilo Plateau is well known for its unique biotacomposed of hundreds of endemic species of plants (Giulli-eti and Pirani 1988 Harley 1988) and some endemic speciesof other vertebrates such as lizards and amphibians (Haddadet al 1988 Rodrigues 1988) The Araguaia and Paranatilde rivervalleys are still poorly known for nonavian groups but theriverine forests and dry forests likely harbor endemic speciesof other organisms also For example a very distinctive speciesof rodent of the genus Kerodon was described recently fromParanatilde dry forests (Moojen et al 1997) Although not knownto possess endemic birds several other areas possess endemictaxa from other groups For instance Erwin and Pogue (1988)

228 BioScience March 2002 Vol 52 No 3

Articles

indicated that the Mato Grosso Plateau is an important areaof endemism for beetles of the genus Agra Thus bird distri-butions indicate three subareas of endemism however thisnumber is almost certainly an underestimate for other groupssuch as some insects vascular plants and amphibians Froma taxonomic perspective these groups are still too poorlyknown to be used for assessing patterns of endemism withinthe Cerrado

Investigation of the timing and mode of the evolution ofthe endemic species of the Cerrado is in its infancy butnonetheless raises controversy Some authors (eg Rizzini1979) have suggested that the original vegetation in the Cer-rado was forest rather than savanna In this view present-daysavannas succeeded forests by natural processes such as fireand climatic changes At the opposite extreme Cole (1986)suggests that within the Cerrado the savanna vegetation isolder than the forests based on geomorphological informa-

tion If one assumes that avian diversification within the Cer-rado followed at least roughly the same trend as evolutionof the landscape in this region then one would predict basedon Rizzinirsquos hypothesis that older species are associatedmainly with forest vegetation and younger species are asso-ciated mainly with savannas Under Colersquos hypothesis the pre-diction would be inverted To evaluate these two hypotheseswe classified each endemic species into one of two evolutionaryage categories paleoendemics species that are either taxo-nomically distinct (ie well-differentiated monotypic gen-era) or considered to be the sister taxon of a large radiationin which one or more members also occur in the Cerrado re-gion and neoendemics species with sister taxa in adjacentSouth American regions Based on recent molecular work onSouth American birds (Bates et al 1999) we suggest thatneoendemics are younger than the PliocenendashPleistocene tran-sition (about 18 million years ago) whereas paleoendemicsoriginated before this period This is only a tentative classi-fication that must be evaluated in future molecular studiesIn addition we classified each avian endemic as either a for-est or nonforest species All 30 endemic species were analyzed(Table 1) We determined 12 species to be paleoendemicsand 18 to be neoendemics The majority (11 of 12) of the pa-leoendemic species are nonforest species whereas most of theneoendemics (12 of 18) are forest birds These results supportColersquos rather than Rizzinirsquos hypothesis

An additional aspect of these data concerns recent sug-gestions that ecotones or transition zones between two ormore distinctive types of vegetation provide important op-portunities for speciation (Smith et al 1997 Schlithuizen2000) For the Cerrado avifauna this does not appear to bewell supported with respect to the ecotone between savannaand forest Although 158 bird species inhabit both forest and

March 2002 Vol 52 No 3 BioScience 229

Articles

Table 1 Endemic bird species of the Cerrado regiona

Species Ageb Habitatc

Nothura minor P NTaoniscus nanusd P NPenelope ochrogaster N FColumbina cyanopis P NPyrrhura pfrimeri N FAmazona xanthops P NCaprimulgus candicans P NAugastes scutatus N NGeobates poecilopterusd P NSynallaxis simoni N FAsthenes luizae N NPhilydor dimidiatus N FAutomolus rectirostris N FHerpsilochmus longirostris N FCercomacra ferdinandi N FMelanopareia torquata N NScytalopus novacapitalis N FPhyllomyias reiseri N FSuiriri islerorum N FPolystictus superciliaris P NKnipolegus fraciscanus N FAntilophia galeata N FPoospiza cinerea N NEmbernagra longicauda N NCharitospiza eucosmad P NParoaria baeri N FSaltator atricollis P NPorphyrospiza caerulescensd P NBasileuterus leucophrys N FCyanocorax cristatellus P N

a This data set has been modified from Silva (1997) as Geobates

poecilopterus (a monotypic genus) has been moved from neoendem-

ic to paleoendemic and Antilophia galeata has been changed to

neoendemic because of the recent discovery of Antilophia bokerman-

ni (Coelho and Silva 1998) Suiriri islerorum is a newly described

endemic species (Zimmer et al 2001)

b P paleoendemic N neoendemic

c N nonforest F forest

d Monotypic genus

Figure 4 Three subareas of endemism are identified for birdswithin the Cerrado Espinhaccedilo Plateau (with four endemicspecies) Araguaia River Valley (with three endemic species)and Paranatilde River Valley (with two endemic species)

savanna no single case of intraspecific differentiation acrossthis transition has been reported so far In addition few gen-era have species in both forest and savanna In cases in whichavian genera do have species in both habitats those speciesare generally considered distantly related

Connections among Neotropical savannas SouthAmerican tropical savanna regions can be grouped into twomajor groups that are currently separated by the Amazon val-ley The northern block is formed by Llanos Roraima ParuMonte Alegre Amapaacute and Marajoacute whereas the second blockis formed by Cerrado Pantanal Llanos de Mojos and patchesof savannas located close to the transition between Amazo-nia and the Cerrado (Figure 5) Because South American sa-vannas exhibit biotic similarity to one another (Sarmiento1983) a consensus opinion is that these regions were con-nected in the recent past and the biotic disjunctions ob-served today are the result of vicariance rather than long-dis-tance dispersal across large expanses of Amazonian forest(Haffer 1967 Sarmiento 1983) Under this scenario SouthAmerican savannas are believed to have expanded and re-tracted their ranges during Quaternary climatic cycles Dur-ing the cold dry periods savannas expanded in Amazoniawhile humid forests retracted to peripheral ecological refugesDuring moist and warm periods humid forests spread againwhile savannas retracted to areas approximating their present-day ranges (Mayle et al 2000) These Quaternary cyclic cli-maticndashvegetational changes have been postulated by some bio-geographers to be the most important factor driving thespeciation process in both forest and savanna organisms intropical South America (Whitmore and Prance 1987 Haffer2001)

Three main corridors connecting northern and southernsavanna regions have been proposed (Haffer 1967 1974Webb 1991) (a) Andean connecting the southern block ofsavannas directly with Llanos and Roraima through the An-dean slopes (b) central Amazonian connecting the southernblock of savannas directly with some patches of savannaslocated north of the Amazon such as Monte Alegre andParu roughly following a belt of low precipitation acrosscentral Amazonia and (c) coastal connecting the southernand northern blocks through savanna patches located closeto the Atlantic coast such as Marajoacute and Amapaacute (Figure 5)Silva (1995a) examined the general distribution patterns ofsavanna-associated birds in the Cerrado region to determinehow well these patterns fit with these three proposed corri-dors It was assumed that if these corridors did exist currentpatterns of distribution of savanna-adapted birds might showevidence of their legacy Thus we predicted spatial congru-ence should exist between speciesrsquo ranges and the position ofthese past biotic corridors as mapped by their proponents (Fig-ure 5)

Present-day ranges of savanna birds support two of the threehypothesized biogeographic corridors (Figure 6) The Andeanconnection is supported by the distribution of five species andlinks the Cerrado region with the Llanos and Gratilde-Sabana The

Atlantic coast savanna corridor is consistent with the distri-butions of 33 species All these species occur in Amapaacute orMarajoacute with some of them extending their ranges in other sa-vannas of the northern block (Pinto Henriques and Oren1997 Silva et al 1997) No single species was recorded onlyin the savannas that follow the central Amazonian savanna cor-ridor (Silva 1995a) These two pieces of information suggestthat a savanna corridor following a belt of low precipitationacross the Amazon basin may not have existed and the mostrecent biotic connections between Amazonian savannas andthe Cerrado region were mostly through the savannas locatedalong the Atlantic coast (Silva 1995a Silva et al 1997) Aacutevila-Pires (1995) analyzed the distribution patterns of all lizardspecies within Amazonia and also found little support for acentral Amazonian savanna corridor Thus even though sa-vanna biotas have expanded and contracted their ranges dur-ing the Quaternary climatic cycles (Mayle et al 2000) thereis little evidence to suggest that these changes resulted in abroad savanna corridor across central Amazonia Penningtonand colleagues (2000) suggested that several lines of evidencepoint to tropical dry forests rather than savannas as replac-ing rain forests in Amazonia during the coolest periods of theQuaternary

Biogeography and conservationUntil very recently conservation of biological resources in theCerrado had received little international attention To date

230 BioScience March 2002 Vol 52 No 3

Articles

Figure 5 Major savanna corridors that have been hy-pothesized to historically connect northern and southernblocks of South American tropical savannas (a) Andeancorridor (b) central Amazonian corridor and (c) coastalcorridor

less than 2 of the total area of the Cerrado is protected inreserves (Mittermeier et al 2000) but there are important dif-ferences in how countries are approaching Cerrado conser-vation The Bolivian government created Noel Kempff Na-tional Park in 1988 thereby protecting the countryrsquos largestcerrado (Killeen and Schulenberg 1998) These roughly 42000ha of unaltered cerrado near the western edge of the biomeis one of the worldrsquos largest continuous protected parcels ofthis habitat (Figure 7) Because of the parkrsquos remoteness andBoliviarsquos comparatively low human population density (es-pecially in the countryrsquos eastern lowlands) the most sub-stantial human threats to this area come from burning asso-ciated with ranching operations in neighboring Brazil Boliviaalso possesses several smaller and more isolated plateaus withcerrado habitat these remain unprotected and largely unex-plored (Parker and Rocha 1991)

In Paraguay at the southern limit of the biome Cerradois protected by two national parks Serraniacuteas de San Luiacutes andCerro Coraacute (Clay et al 1998 Robbins et al 1999) Avian sur-vey work has been conducted at both sites however the re-gions are still not considered to be thoroughly surveyed andboth Clay (1998) and Robbins and their colleagues (1999) em-phasize that the rapid expansion of agricultural operationsthreaten all unprotected savanna regions in the country

In Brazil which possesses the vast majority of Cerrado thesituation is drastically worse There are few large reserves(more than 25000 hectares [ha]) and they are not distributedevenly across the biome (Figure 7) Consequently an im-portant part of the Cerradorsquos environmental diversity hasnot been incorporated in a network of protected areas Largeareas of Brazilian Cerrado and dry forests have been convertedto soybean and rice plantations Most of this large-scale habi-tat modification has not followed the most basic principlesof environmental conservation and problems of erosion anddeterioration of important streams and rivers are increasingInternational support for this alteration has come from ma-jor development agencies and some senior scientists whohave encouraged increased land use in the Cerrado to reducehuman pressure on Amazonian forests

The Brazilian governmentrsquos appreciation of the uniquenessof the Cerradorsquos biodiversity has developed slowly The firstmajor initiative occurred in 1997 when the Brazilian Ministryof Environment promoted a workshop which included over200 scientific experts on the region This group met to definepriority areas for conservation in the Brazilian Cerrado It fol-lowed the same methodology applied earlier for the AtlanticForest and Amazonia (Cavalcanti 1999) Several criteria forassessing biodiversity value were used such as number ofendemic species species richness presence of rare or en-dangered species or both and sites of unique communitiesor key areas for migratory species The group identified a to-tal of 87 priority areas The urgency for conservation actionin a priority area was determined by cross-referencing bio-diversity data with the human-pressure and land cover-change data (Cavalcanti 1999) As of August 1999 the Brazil-ian government has indicated that it will follow most of the

recommendations of this workshop and that new protectedareas and ecological corridors will be established In factsome important actions were taken during 2001 but twodeserve special comments The first was the expansion ofthe National Park of Chapada dos Veadeiros an important re-serve that includes pristine cerrados and patches of camposrupestres with several endemic plant species from 60000 to235000 ha The second was the creation of the Ecological Sta-tion of Serra Geral do Tocantins with 716306 ha close to theState Park of Jalapatildeo with 158885 ha Together these two ar-eas form the worldrsquos largest continuous area of protectedCerrado (Figure 7)

How can the biogeographic information discussed here beused to help protect and preserve the Cerradorsquos biodiversityFirst we know now that the Cerradorsquos biota is not homoge-neous so additional distribution data for different groups oforganisms must be collected and organized in a retrievable wayto help conservationists determine whether there are addi-tional unidentified subareas of endemism Second eventhough the three subareas of endemism identified here as har-boring restricted-range endemic species of birds are partiallycovered by one or more reserves (for comparison refer to Fig-ures 4 and 7 Silva 1997) most reserves need to be more for-mally implemented and connected through biodiversity cor-ridors to insure their preservation over the long term Analysesof other nonavian taxa are certain to uncover additional ar-eas that are not currently protected by reserves Third despiteoccupying a relatively small part of the total landscape

March 2002 Vol 52 No 3 BioScience 231

Articles

Figure 6 Examples of ranges of bird species supportingthe hypothesis (a b) of the Andean savanna corridor and(c d) of the coastal savanna corridor

ba

c d

galleryndashriverine forest cerradatildeo anddry forest have been shown to be keyhabitats for a substantial portion ofthe Cerradorsquos biodiversity Thus theirconversion into pastures and agricul-tural fields should be halted and theirconnectivity maintained Fourth newprojects in areas currently covered bycerrado or other types of open vegeta-tion should be temporarily banned un-til their impact on the flora and faunacan be rigorously assessed and conser-vation strategies for those areas de-signed Fifth it is critical that agricul-tural technology be developed andimplemented to help landowners in-crease productivity of lands alreadyunder cultivation (Macedo 1994 Silva1998) This would reduce pressure onlands still covered by natural vegetationFinally a special conservation actionplan must be developed to guaranteethe conservation of at least part of thelargest savannas in Amazonia and At-lantic Forest as they are important lab-oratories for ecological and evolution-ary studies in tropical South AmericaIt is urgent now because efforts are underway to establish soy-bean plantations in these enclaves

The Cerradorsquos conservation is a challenge from both sci-entific and political viewpoints Because important financialand political interests are involved in the destruction of theCerrado scientists and conservationists must be creative ingathering and synthesizing the best and most complete in-formation possible to generate a viable strategy to insureconservation of the worldrsquos richest tropical savanna

AcknowledgmentsWe thank Jon Fjeldsaring David C Oren Roberto B CavalcantiFernando C Novaes Bent O Poulsen Mary E PetersenTownsend Peterson Carsten Rahbek James A Ratter NeivaGuedes and Shannon J Hackett for discussions about Cer-rado biogeography and comments on the manuscript LuisBarbosa produced the figures Research support came fromConselho Nacional de Desenvolvimento Cientiacutefico e Tec-noloacutegico (grant no 30246488-3) National Geographic So-ciety (grant no 4964-93) Danish Natural Science ResearchCouncil (grant J no 11-0390) Museu Paraense Emiacutelio Goeldiand Universidade de Brasiacutelia

References citedAbrsquoSaber AN 1977 Os Domiacutenios morfoclimaacuteticos da Ameacuterica do Sul

Primeira Aproximaccedilatildeo Geomorfologia 52 1ndash21mdashmdashmdash 1983 O domiacutenio dos cerrados Introduccedilatildeo ao conhecimento Re-

vista do Servidor Puacuteblico 111 41ndash55Aacutevila-Pires TC 1995 Lizards of Brazilian Amazonia (Reptilia Squamata)

Zoologische Verhandelingen Leiden 299 1ndash706

Bates JM Hackett SJ Goerck J 1999 High levels of mitochondrial DNA dif-ferentiation in two lineages of antbirds (Drymophila and Hypocnemis)The Auk 1161093ndash1106

Brasil AE Alvarenga SM 1989 Relevo Pages 53ndash72 in Duarte AC ed Ge-ografia do Brasil Regiatildeo Centro-Oeste Rio de Janeiro IBGE

Cavalcanti RB ed 1999 Accedilotildees Prioritaacuterias para a conservaccedilatildeo da biodiver-sidade do Cerrado e Pantanal Belo Horizonte Conservation Internationaldo Brasil

Clay RP Cooper DR Mazar Barnett J Burfiled IZ Esquivel EZ Farintildea RKennedy CP Perrens M Pople RG 1998 White-winged NightjarsCaprimulgus candicans and cerrado conservation The key findings of pro-ject Aguaraacute Nu 1997 Cotinga 9 52ndash55

Coelho G Silva W 1998 A new species of Antilophia (Paseriformes Pipri-dae) from Chapada do Araripe Cearaacute Brazil Ararajuba 6 81ndash86

Cole MM 1986 The Savannas Biogeography and Geobotany LondonAcademic Press

Cracraft J 1985 Historical biogeography and patterns of differentiationwithin the South America avifauna Areas of endemism OrnithologicalMonographs 36 49ndash84

Eiten G 1972 The Cerrado vegetation of Brazil Botanical Review 38201ndash341

mdashmdashmdash 1990 Vegetaccedilatildeo Pages 9ndash65 in Pinto MN ed Cerrado Caracteri-zaccedilatildeo Ocupaccedilatildeo e Perspectiva Brasiacutelia Universidade de Brasiacutelia and SE-MATEC

Erwin TL Pogue MG 1988 Agra arboreal beetles of Neotropical forests Bio-geography and the forest refugium hypothesis (Carabidae) Pages 161ndash188in Vanzolini PE Heyer WR eds Proceedings of a Workshop on Neotrop-ical Distribution Patterns Rio de Janeiro Academia Brasileira de Ciecircn-cias

Furley PA Ratter JA 1988 Soil resources and plant communities of the cen-tral Brazilian cerrado and their development Journal of Biogeography15 97ndash108

Giullietti AM Pirani JR 1988 Patterns of geographic distribution of someplant species from the Espinhaccedilo Range Minas Gerais and Bahia Brazil

232 BioScience March 2002 Vol 52 No 3

Articles

Figure 7 Parks and ecological stations in the Cerrado region Only those reserveslarger than 25000 ha are shown in Brazil PE state park PN national park EEecological station

tern on these depressions is much more heterogeneous thanthat of the plateaus in that it includes more habitats (river-ine forests tropical dry forests all types of cerrado andmarshlands) distributed in a mosaic fashion Riverine forests(called matas ciliares Ribeiro and Walter 1998) differ fromgallery forests because they are found only along the large riversin the Cerrado They are not wide (up to 100 m from each sideof the river) and can be semideciduous to evergreen Trees are20ndash30 m tall Common species are Anadenanthera sppApeiba tibourbou Aspidosperma spp Tabebuia spp andTrema micrantha among others (Ribeiro and Walter 1998)Tropical dry forests (matas secas) are particularly associatedwith peripheral depressions and in some areas (eg in theParanatilde River valley) are the dominant vegetation type Theyare deciduous or semideciduous and grow on patches ofmoderately productive soils derived from basic rocks such aslimestone (Ratter et al 1978 Prado and Gibbs 1993) Canopyheight averages 20ndash45 m (Ribeiro and Walter 1998) Treespecies common in dry forests include Astronium urundeuvaPiptadenia macrocarpa Chorisia speciosa Tabebuia spp Ca-vanillesia arborea and Cedrella fissilis (Ratter et al 1978)

Biogeographic patterns

Forest organisms in a tropical savanna regionThe most basic question about the biota of a region is howit is distributed in major vegetation types In the Cerrado onemight expect that most of the biota would inhabit the savannasthat dominate the region However studies of mammalsbirds and plants do not support this hypothesis Redford andFonseca (1986) found that most (563) of the nonvolantmammal species in the Cerrado inhabit forests SimilarlySilva (1995b) classified the 759 birds species known or as-sumed to breed in the Cerrado as independent (occurring ex-clusively in forests) semidependent (occurring in both forestsand open vegetation) or dependent (occurring exclusively inopen vegetation) The majority of species (393 or 518) aredependent on the various forest habitats 208 (158) aresemidependent and 274 (208) are independent speciesThus approximately 826 of Cerrado birds require foreststo some degree Mendonccedila and colleagues (1998) listed 6671plant taxa (species and varieties) for the Cerrado of these 38occur only in forests

In general these data support the hypothesis that galleryand dry forests which occupy less than 20 of the Cerradoare necessary for a large portion of regional biodiversityHow did this pattern evolve Silva (1995b) suggested that biotic diversity of the Cerrado region increased through in-terchange with adjacent regions during the Quaternary climaticndashvegetational fluctuations Thus Atlantic and Ama-zonian forest species expanded their ranges into the Cerradoduring the wet periods following the expansion of the riverinendashgallery forest network (Silva 1995b) whereas Caatingaand Chaco elements colonized the Cerrado during the Qua-ternary dry periods as part of the coalescence of dry forestsin peripheral depressions (AbrsquoSaber 1983 Pennington et al

2000) These species probably originated in adjacent regionsand maintained viable populations within the Cerrado regionbecause their forest habitats did not vanish entirely during thepaleoclimatic period that followed their range expansion Ina fashion similar to what is seen today large rivers and streamsmust have provided suitable environmental conditions tomaintain wide galleryndashriverine forests during the dry periods(AbrsquoSaber 1983) whereas large patches of soils derived fromlimestone were important for maintaining dry forests duringthe wet periods (Prado and Gibbs 1993 Pennington et al2000)

Endemic species Origin and evolution The per-centage of species endemic to the Cerrado varies across dif-ferent groups of organisms vascular plants (44) amphib-ians (30) reptiles (20) mammals (118) and birds(14) (Silva 1995a Myers et al 2000) However because largeareas of the Cerrado remain unexplored (Silva 1995c) thesenumbers will likely change when additional biological in-ventories are conducted At present we still know little abouthow these endemic species are distributed across the Cerradoor how they evolved A key question for conservation purposesis whether one can identify subareas of endemism withinthe large Cerrado region

To determine if there are subareas of endemism within theCerrado Silva (1997) mapped and created overlays of all en-demic bird speciesrsquo ranges smaller than 60000 km2 Only 10of the 30 endemic species met this requirement (Table 1) Theranges of these species delimit three main areas

Espinhaccedilo Plateau Four species that inhabit camposrupestres are Augastes scutatus (Trochilidae) Asthenesluizae (Furnariidae) Polystictus superciliaris (Tyran-nidae) and Embernagra longicauda (Emberizidae)

Araguaia River Valley Three species that inhabit galleryforests are Synallaxis simoni (Furnariidae) Cercomacraferdinandi (Thamnophilidae) and Paroaria baeri(Emberizidae)

Paranatilde River Valley Two species found only in tropicaldry forests are Pyrrhura pfrimeri (Psittacidae) andKnipolegus franciscanus (Tyrannidae) (Figure 4)

The subareas of endemism identified for birds are at leastpartially supported by studies of other groups of organismsThe Espinhaccedilo Plateau is well known for its unique biotacomposed of hundreds of endemic species of plants (Giulli-eti and Pirani 1988 Harley 1988) and some endemic speciesof other vertebrates such as lizards and amphibians (Haddadet al 1988 Rodrigues 1988) The Araguaia and Paranatilde rivervalleys are still poorly known for nonavian groups but theriverine forests and dry forests likely harbor endemic speciesof other organisms also For example a very distinctive speciesof rodent of the genus Kerodon was described recently fromParanatilde dry forests (Moojen et al 1997) Although not knownto possess endemic birds several other areas possess endemictaxa from other groups For instance Erwin and Pogue (1988)

228 BioScience March 2002 Vol 52 No 3

Articles

indicated that the Mato Grosso Plateau is an important areaof endemism for beetles of the genus Agra Thus bird distri-butions indicate three subareas of endemism however thisnumber is almost certainly an underestimate for other groupssuch as some insects vascular plants and amphibians Froma taxonomic perspective these groups are still too poorlyknown to be used for assessing patterns of endemism withinthe Cerrado

Investigation of the timing and mode of the evolution ofthe endemic species of the Cerrado is in its infancy butnonetheless raises controversy Some authors (eg Rizzini1979) have suggested that the original vegetation in the Cer-rado was forest rather than savanna In this view present-daysavannas succeeded forests by natural processes such as fireand climatic changes At the opposite extreme Cole (1986)suggests that within the Cerrado the savanna vegetation isolder than the forests based on geomorphological informa-

tion If one assumes that avian diversification within the Cer-rado followed at least roughly the same trend as evolutionof the landscape in this region then one would predict basedon Rizzinirsquos hypothesis that older species are associatedmainly with forest vegetation and younger species are asso-ciated mainly with savannas Under Colersquos hypothesis the pre-diction would be inverted To evaluate these two hypotheseswe classified each endemic species into one of two evolutionaryage categories paleoendemics species that are either taxo-nomically distinct (ie well-differentiated monotypic gen-era) or considered to be the sister taxon of a large radiationin which one or more members also occur in the Cerrado re-gion and neoendemics species with sister taxa in adjacentSouth American regions Based on recent molecular work onSouth American birds (Bates et al 1999) we suggest thatneoendemics are younger than the PliocenendashPleistocene tran-sition (about 18 million years ago) whereas paleoendemicsoriginated before this period This is only a tentative classi-fication that must be evaluated in future molecular studiesIn addition we classified each avian endemic as either a for-est or nonforest species All 30 endemic species were analyzed(Table 1) We determined 12 species to be paleoendemicsand 18 to be neoendemics The majority (11 of 12) of the pa-leoendemic species are nonforest species whereas most of theneoendemics (12 of 18) are forest birds These results supportColersquos rather than Rizzinirsquos hypothesis

An additional aspect of these data concerns recent sug-gestions that ecotones or transition zones between two ormore distinctive types of vegetation provide important op-portunities for speciation (Smith et al 1997 Schlithuizen2000) For the Cerrado avifauna this does not appear to bewell supported with respect to the ecotone between savannaand forest Although 158 bird species inhabit both forest and

March 2002 Vol 52 No 3 BioScience 229

Articles

Table 1 Endemic bird species of the Cerrado regiona

Species Ageb Habitatc

Nothura minor P NTaoniscus nanusd P NPenelope ochrogaster N FColumbina cyanopis P NPyrrhura pfrimeri N FAmazona xanthops P NCaprimulgus candicans P NAugastes scutatus N NGeobates poecilopterusd P NSynallaxis simoni N FAsthenes luizae N NPhilydor dimidiatus N FAutomolus rectirostris N FHerpsilochmus longirostris N FCercomacra ferdinandi N FMelanopareia torquata N NScytalopus novacapitalis N FPhyllomyias reiseri N FSuiriri islerorum N FPolystictus superciliaris P NKnipolegus fraciscanus N FAntilophia galeata N FPoospiza cinerea N NEmbernagra longicauda N NCharitospiza eucosmad P NParoaria baeri N FSaltator atricollis P NPorphyrospiza caerulescensd P NBasileuterus leucophrys N FCyanocorax cristatellus P N

a This data set has been modified from Silva (1997) as Geobates

poecilopterus (a monotypic genus) has been moved from neoendem-

ic to paleoendemic and Antilophia galeata has been changed to

neoendemic because of the recent discovery of Antilophia bokerman-

ni (Coelho and Silva 1998) Suiriri islerorum is a newly described

endemic species (Zimmer et al 2001)

b P paleoendemic N neoendemic

c N nonforest F forest

d Monotypic genus

Figure 4 Three subareas of endemism are identified for birdswithin the Cerrado Espinhaccedilo Plateau (with four endemicspecies) Araguaia River Valley (with three endemic species)and Paranatilde River Valley (with two endemic species)

savanna no single case of intraspecific differentiation acrossthis transition has been reported so far In addition few gen-era have species in both forest and savanna In cases in whichavian genera do have species in both habitats those speciesare generally considered distantly related

Connections among Neotropical savannas SouthAmerican tropical savanna regions can be grouped into twomajor groups that are currently separated by the Amazon val-ley The northern block is formed by Llanos Roraima ParuMonte Alegre Amapaacute and Marajoacute whereas the second blockis formed by Cerrado Pantanal Llanos de Mojos and patchesof savannas located close to the transition between Amazo-nia and the Cerrado (Figure 5) Because South American sa-vannas exhibit biotic similarity to one another (Sarmiento1983) a consensus opinion is that these regions were con-nected in the recent past and the biotic disjunctions ob-served today are the result of vicariance rather than long-dis-tance dispersal across large expanses of Amazonian forest(Haffer 1967 Sarmiento 1983) Under this scenario SouthAmerican savannas are believed to have expanded and re-tracted their ranges during Quaternary climatic cycles Dur-ing the cold dry periods savannas expanded in Amazoniawhile humid forests retracted to peripheral ecological refugesDuring moist and warm periods humid forests spread againwhile savannas retracted to areas approximating their present-day ranges (Mayle et al 2000) These Quaternary cyclic cli-maticndashvegetational changes have been postulated by some bio-geographers to be the most important factor driving thespeciation process in both forest and savanna organisms intropical South America (Whitmore and Prance 1987 Haffer2001)

Three main corridors connecting northern and southernsavanna regions have been proposed (Haffer 1967 1974Webb 1991) (a) Andean connecting the southern block ofsavannas directly with Llanos and Roraima through the An-dean slopes (b) central Amazonian connecting the southernblock of savannas directly with some patches of savannaslocated north of the Amazon such as Monte Alegre andParu roughly following a belt of low precipitation acrosscentral Amazonia and (c) coastal connecting the southernand northern blocks through savanna patches located closeto the Atlantic coast such as Marajoacute and Amapaacute (Figure 5)Silva (1995a) examined the general distribution patterns ofsavanna-associated birds in the Cerrado region to determinehow well these patterns fit with these three proposed corri-dors It was assumed that if these corridors did exist currentpatterns of distribution of savanna-adapted birds might showevidence of their legacy Thus we predicted spatial congru-ence should exist between speciesrsquo ranges and the position ofthese past biotic corridors as mapped by their proponents (Fig-ure 5)

Present-day ranges of savanna birds support two of the threehypothesized biogeographic corridors (Figure 6) The Andeanconnection is supported by the distribution of five species andlinks the Cerrado region with the Llanos and Gratilde-Sabana The

Atlantic coast savanna corridor is consistent with the distri-butions of 33 species All these species occur in Amapaacute orMarajoacute with some of them extending their ranges in other sa-vannas of the northern block (Pinto Henriques and Oren1997 Silva et al 1997) No single species was recorded onlyin the savannas that follow the central Amazonian savanna cor-ridor (Silva 1995a) These two pieces of information suggestthat a savanna corridor following a belt of low precipitationacross the Amazon basin may not have existed and the mostrecent biotic connections between Amazonian savannas andthe Cerrado region were mostly through the savannas locatedalong the Atlantic coast (Silva 1995a Silva et al 1997) Aacutevila-Pires (1995) analyzed the distribution patterns of all lizardspecies within Amazonia and also found little support for acentral Amazonian savanna corridor Thus even though sa-vanna biotas have expanded and contracted their ranges dur-ing the Quaternary climatic cycles (Mayle et al 2000) thereis little evidence to suggest that these changes resulted in abroad savanna corridor across central Amazonia Penningtonand colleagues (2000) suggested that several lines of evidencepoint to tropical dry forests rather than savannas as replac-ing rain forests in Amazonia during the coolest periods of theQuaternary

Biogeography and conservationUntil very recently conservation of biological resources in theCerrado had received little international attention To date

230 BioScience March 2002 Vol 52 No 3

Articles

Figure 5 Major savanna corridors that have been hy-pothesized to historically connect northern and southernblocks of South American tropical savannas (a) Andeancorridor (b) central Amazonian corridor and (c) coastalcorridor

less than 2 of the total area of the Cerrado is protected inreserves (Mittermeier et al 2000) but there are important dif-ferences in how countries are approaching Cerrado conser-vation The Bolivian government created Noel Kempff Na-tional Park in 1988 thereby protecting the countryrsquos largestcerrado (Killeen and Schulenberg 1998) These roughly 42000ha of unaltered cerrado near the western edge of the biomeis one of the worldrsquos largest continuous protected parcels ofthis habitat (Figure 7) Because of the parkrsquos remoteness andBoliviarsquos comparatively low human population density (es-pecially in the countryrsquos eastern lowlands) the most sub-stantial human threats to this area come from burning asso-ciated with ranching operations in neighboring Brazil Boliviaalso possesses several smaller and more isolated plateaus withcerrado habitat these remain unprotected and largely unex-plored (Parker and Rocha 1991)

In Paraguay at the southern limit of the biome Cerradois protected by two national parks Serraniacuteas de San Luiacutes andCerro Coraacute (Clay et al 1998 Robbins et al 1999) Avian sur-vey work has been conducted at both sites however the re-gions are still not considered to be thoroughly surveyed andboth Clay (1998) and Robbins and their colleagues (1999) em-phasize that the rapid expansion of agricultural operationsthreaten all unprotected savanna regions in the country

In Brazil which possesses the vast majority of Cerrado thesituation is drastically worse There are few large reserves(more than 25000 hectares [ha]) and they are not distributedevenly across the biome (Figure 7) Consequently an im-portant part of the Cerradorsquos environmental diversity hasnot been incorporated in a network of protected areas Largeareas of Brazilian Cerrado and dry forests have been convertedto soybean and rice plantations Most of this large-scale habi-tat modification has not followed the most basic principlesof environmental conservation and problems of erosion anddeterioration of important streams and rivers are increasingInternational support for this alteration has come from ma-jor development agencies and some senior scientists whohave encouraged increased land use in the Cerrado to reducehuman pressure on Amazonian forests

The Brazilian governmentrsquos appreciation of the uniquenessof the Cerradorsquos biodiversity has developed slowly The firstmajor initiative occurred in 1997 when the Brazilian Ministryof Environment promoted a workshop which included over200 scientific experts on the region This group met to definepriority areas for conservation in the Brazilian Cerrado It fol-lowed the same methodology applied earlier for the AtlanticForest and Amazonia (Cavalcanti 1999) Several criteria forassessing biodiversity value were used such as number ofendemic species species richness presence of rare or en-dangered species or both and sites of unique communitiesor key areas for migratory species The group identified a to-tal of 87 priority areas The urgency for conservation actionin a priority area was determined by cross-referencing bio-diversity data with the human-pressure and land cover-change data (Cavalcanti 1999) As of August 1999 the Brazil-ian government has indicated that it will follow most of the

recommendations of this workshop and that new protectedareas and ecological corridors will be established In factsome important actions were taken during 2001 but twodeserve special comments The first was the expansion ofthe National Park of Chapada dos Veadeiros an important re-serve that includes pristine cerrados and patches of camposrupestres with several endemic plant species from 60000 to235000 ha The second was the creation of the Ecological Sta-tion of Serra Geral do Tocantins with 716306 ha close to theState Park of Jalapatildeo with 158885 ha Together these two ar-eas form the worldrsquos largest continuous area of protectedCerrado (Figure 7)

How can the biogeographic information discussed here beused to help protect and preserve the Cerradorsquos biodiversityFirst we know now that the Cerradorsquos biota is not homoge-neous so additional distribution data for different groups oforganisms must be collected and organized in a retrievable wayto help conservationists determine whether there are addi-tional unidentified subareas of endemism Second eventhough the three subareas of endemism identified here as har-boring restricted-range endemic species of birds are partiallycovered by one or more reserves (for comparison refer to Fig-ures 4 and 7 Silva 1997) most reserves need to be more for-mally implemented and connected through biodiversity cor-ridors to insure their preservation over the long term Analysesof other nonavian taxa are certain to uncover additional ar-eas that are not currently protected by reserves Third despiteoccupying a relatively small part of the total landscape

March 2002 Vol 52 No 3 BioScience 231

Articles

Figure 6 Examples of ranges of bird species supportingthe hypothesis (a b) of the Andean savanna corridor and(c d) of the coastal savanna corridor

ba

c d

galleryndashriverine forest cerradatildeo anddry forest have been shown to be keyhabitats for a substantial portion ofthe Cerradorsquos biodiversity Thus theirconversion into pastures and agricul-tural fields should be halted and theirconnectivity maintained Fourth newprojects in areas currently covered bycerrado or other types of open vegeta-tion should be temporarily banned un-til their impact on the flora and faunacan be rigorously assessed and conser-vation strategies for those areas de-signed Fifth it is critical that agricul-tural technology be developed andimplemented to help landowners in-crease productivity of lands alreadyunder cultivation (Macedo 1994 Silva1998) This would reduce pressure onlands still covered by natural vegetationFinally a special conservation actionplan must be developed to guaranteethe conservation of at least part of thelargest savannas in Amazonia and At-lantic Forest as they are important lab-oratories for ecological and evolution-ary studies in tropical South AmericaIt is urgent now because efforts are underway to establish soy-bean plantations in these enclaves

The Cerradorsquos conservation is a challenge from both sci-entific and political viewpoints Because important financialand political interests are involved in the destruction of theCerrado scientists and conservationists must be creative ingathering and synthesizing the best and most complete in-formation possible to generate a viable strategy to insureconservation of the worldrsquos richest tropical savanna

AcknowledgmentsWe thank Jon Fjeldsaring David C Oren Roberto B CavalcantiFernando C Novaes Bent O Poulsen Mary E PetersenTownsend Peterson Carsten Rahbek James A Ratter NeivaGuedes and Shannon J Hackett for discussions about Cer-rado biogeography and comments on the manuscript LuisBarbosa produced the figures Research support came fromConselho Nacional de Desenvolvimento Cientiacutefico e Tec-noloacutegico (grant no 30246488-3) National Geographic So-ciety (grant no 4964-93) Danish Natural Science ResearchCouncil (grant J no 11-0390) Museu Paraense Emiacutelio Goeldiand Universidade de Brasiacutelia

References citedAbrsquoSaber AN 1977 Os Domiacutenios morfoclimaacuteticos da Ameacuterica do Sul

Primeira Aproximaccedilatildeo Geomorfologia 52 1ndash21mdashmdashmdash 1983 O domiacutenio dos cerrados Introduccedilatildeo ao conhecimento Re-

vista do Servidor Puacuteblico 111 41ndash55Aacutevila-Pires TC 1995 Lizards of Brazilian Amazonia (Reptilia Squamata)

Zoologische Verhandelingen Leiden 299 1ndash706

Bates JM Hackett SJ Goerck J 1999 High levels of mitochondrial DNA dif-ferentiation in two lineages of antbirds (Drymophila and Hypocnemis)The Auk 1161093ndash1106

Brasil AE Alvarenga SM 1989 Relevo Pages 53ndash72 in Duarte AC ed Ge-ografia do Brasil Regiatildeo Centro-Oeste Rio de Janeiro IBGE

Cavalcanti RB ed 1999 Accedilotildees Prioritaacuterias para a conservaccedilatildeo da biodiver-sidade do Cerrado e Pantanal Belo Horizonte Conservation Internationaldo Brasil

Clay RP Cooper DR Mazar Barnett J Burfiled IZ Esquivel EZ Farintildea RKennedy CP Perrens M Pople RG 1998 White-winged NightjarsCaprimulgus candicans and cerrado conservation The key findings of pro-ject Aguaraacute Nu 1997 Cotinga 9 52ndash55

Coelho G Silva W 1998 A new species of Antilophia (Paseriformes Pipri-dae) from Chapada do Araripe Cearaacute Brazil Ararajuba 6 81ndash86

Cole MM 1986 The Savannas Biogeography and Geobotany LondonAcademic Press

Cracraft J 1985 Historical biogeography and patterns of differentiationwithin the South America avifauna Areas of endemism OrnithologicalMonographs 36 49ndash84

Eiten G 1972 The Cerrado vegetation of Brazil Botanical Review 38201ndash341

mdashmdashmdash 1990 Vegetaccedilatildeo Pages 9ndash65 in Pinto MN ed Cerrado Caracteri-zaccedilatildeo Ocupaccedilatildeo e Perspectiva Brasiacutelia Universidade de Brasiacutelia and SE-MATEC

Erwin TL Pogue MG 1988 Agra arboreal beetles of Neotropical forests Bio-geography and the forest refugium hypothesis (Carabidae) Pages 161ndash188in Vanzolini PE Heyer WR eds Proceedings of a Workshop on Neotrop-ical Distribution Patterns Rio de Janeiro Academia Brasileira de Ciecircn-cias

Furley PA Ratter JA 1988 Soil resources and plant communities of the cen-tral Brazilian cerrado and their development Journal of Biogeography15 97ndash108

Giullietti AM Pirani JR 1988 Patterns of geographic distribution of someplant species from the Espinhaccedilo Range Minas Gerais and Bahia Brazil

232 BioScience March 2002 Vol 52 No 3

Articles

Figure 7 Parks and ecological stations in the Cerrado region Only those reserveslarger than 25000 ha are shown in Brazil PE state park PN national park EEecological station

indicated that the Mato Grosso Plateau is an important areaof endemism for beetles of the genus Agra Thus bird distri-butions indicate three subareas of endemism however thisnumber is almost certainly an underestimate for other groupssuch as some insects vascular plants and amphibians Froma taxonomic perspective these groups are still too poorlyknown to be used for assessing patterns of endemism withinthe Cerrado

Investigation of the timing and mode of the evolution ofthe endemic species of the Cerrado is in its infancy butnonetheless raises controversy Some authors (eg Rizzini1979) have suggested that the original vegetation in the Cer-rado was forest rather than savanna In this view present-daysavannas succeeded forests by natural processes such as fireand climatic changes At the opposite extreme Cole (1986)suggests that within the Cerrado the savanna vegetation isolder than the forests based on geomorphological informa-

tion If one assumes that avian diversification within the Cer-rado followed at least roughly the same trend as evolutionof the landscape in this region then one would predict basedon Rizzinirsquos hypothesis that older species are associatedmainly with forest vegetation and younger species are asso-ciated mainly with savannas Under Colersquos hypothesis the pre-diction would be inverted To evaluate these two hypotheseswe classified each endemic species into one of two evolutionaryage categories paleoendemics species that are either taxo-nomically distinct (ie well-differentiated monotypic gen-era) or considered to be the sister taxon of a large radiationin which one or more members also occur in the Cerrado re-gion and neoendemics species with sister taxa in adjacentSouth American regions Based on recent molecular work onSouth American birds (Bates et al 1999) we suggest thatneoendemics are younger than the PliocenendashPleistocene tran-sition (about 18 million years ago) whereas paleoendemicsoriginated before this period This is only a tentative classi-fication that must be evaluated in future molecular studiesIn addition we classified each avian endemic as either a for-est or nonforest species All 30 endemic species were analyzed(Table 1) We determined 12 species to be paleoendemicsand 18 to be neoendemics The majority (11 of 12) of the pa-leoendemic species are nonforest species whereas most of theneoendemics (12 of 18) are forest birds These results supportColersquos rather than Rizzinirsquos hypothesis

An additional aspect of these data concerns recent sug-gestions that ecotones or transition zones between two ormore distinctive types of vegetation provide important op-portunities for speciation (Smith et al 1997 Schlithuizen2000) For the Cerrado avifauna this does not appear to bewell supported with respect to the ecotone between savannaand forest Although 158 bird species inhabit both forest and

March 2002 Vol 52 No 3 BioScience 229

Articles

Table 1 Endemic bird species of the Cerrado regiona

Species Ageb Habitatc

Nothura minor P NTaoniscus nanusd P NPenelope ochrogaster N FColumbina cyanopis P NPyrrhura pfrimeri N FAmazona xanthops P NCaprimulgus candicans P NAugastes scutatus N NGeobates poecilopterusd P NSynallaxis simoni N FAsthenes luizae N NPhilydor dimidiatus N FAutomolus rectirostris N FHerpsilochmus longirostris N FCercomacra ferdinandi N FMelanopareia torquata N NScytalopus novacapitalis N FPhyllomyias reiseri N FSuiriri islerorum N FPolystictus superciliaris P NKnipolegus fraciscanus N FAntilophia galeata N FPoospiza cinerea N NEmbernagra longicauda N NCharitospiza eucosmad P NParoaria baeri N FSaltator atricollis P NPorphyrospiza caerulescensd P NBasileuterus leucophrys N FCyanocorax cristatellus P N

a This data set has been modified from Silva (1997) as Geobates

poecilopterus (a monotypic genus) has been moved from neoendem-

ic to paleoendemic and Antilophia galeata has been changed to

neoendemic because of the recent discovery of Antilophia bokerman-

ni (Coelho and Silva 1998) Suiriri islerorum is a newly described

endemic species (Zimmer et al 2001)

b P paleoendemic N neoendemic

c N nonforest F forest

d Monotypic genus

Figure 4 Three subareas of endemism are identified for birdswithin the Cerrado Espinhaccedilo Plateau (with four endemicspecies) Araguaia River Valley (with three endemic species)and Paranatilde River Valley (with two endemic species)

savanna no single case of intraspecific differentiation acrossthis transition has been reported so far In addition few gen-era have species in both forest and savanna In cases in whichavian genera do have species in both habitats those speciesare generally considered distantly related

Connections among Neotropical savannas SouthAmerican tropical savanna regions can be grouped into twomajor groups that are currently separated by the Amazon val-ley The northern block is formed by Llanos Roraima ParuMonte Alegre Amapaacute and Marajoacute whereas the second blockis formed by Cerrado Pantanal Llanos de Mojos and patchesof savannas located close to the transition between Amazo-nia and the Cerrado (Figure 5) Because South American sa-vannas exhibit biotic similarity to one another (Sarmiento1983) a consensus opinion is that these regions were con-nected in the recent past and the biotic disjunctions ob-served today are the result of vicariance rather than long-dis-tance dispersal across large expanses of Amazonian forest(Haffer 1967 Sarmiento 1983) Under this scenario SouthAmerican savannas are believed to have expanded and re-tracted their ranges during Quaternary climatic cycles Dur-ing the cold dry periods savannas expanded in Amazoniawhile humid forests retracted to peripheral ecological refugesDuring moist and warm periods humid forests spread againwhile savannas retracted to areas approximating their present-day ranges (Mayle et al 2000) These Quaternary cyclic cli-maticndashvegetational changes have been postulated by some bio-geographers to be the most important factor driving thespeciation process in both forest and savanna organisms intropical South America (Whitmore and Prance 1987 Haffer2001)

Three main corridors connecting northern and southernsavanna regions have been proposed (Haffer 1967 1974Webb 1991) (a) Andean connecting the southern block ofsavannas directly with Llanos and Roraima through the An-dean slopes (b) central Amazonian connecting the southernblock of savannas directly with some patches of savannaslocated north of the Amazon such as Monte Alegre andParu roughly following a belt of low precipitation acrosscentral Amazonia and (c) coastal connecting the southernand northern blocks through savanna patches located closeto the Atlantic coast such as Marajoacute and Amapaacute (Figure 5)Silva (1995a) examined the general distribution patterns ofsavanna-associated birds in the Cerrado region to determinehow well these patterns fit with these three proposed corri-dors It was assumed that if these corridors did exist currentpatterns of distribution of savanna-adapted birds might showevidence of their legacy Thus we predicted spatial congru-ence should exist between speciesrsquo ranges and the position ofthese past biotic corridors as mapped by their proponents (Fig-ure 5)

Present-day ranges of savanna birds support two of the threehypothesized biogeographic corridors (Figure 6) The Andeanconnection is supported by the distribution of five species andlinks the Cerrado region with the Llanos and Gratilde-Sabana The

Atlantic coast savanna corridor is consistent with the distri-butions of 33 species All these species occur in Amapaacute orMarajoacute with some of them extending their ranges in other sa-vannas of the northern block (Pinto Henriques and Oren1997 Silva et al 1997) No single species was recorded onlyin the savannas that follow the central Amazonian savanna cor-ridor (Silva 1995a) These two pieces of information suggestthat a savanna corridor following a belt of low precipitationacross the Amazon basin may not have existed and the mostrecent biotic connections between Amazonian savannas andthe Cerrado region were mostly through the savannas locatedalong the Atlantic coast (Silva 1995a Silva et al 1997) Aacutevila-Pires (1995) analyzed the distribution patterns of all lizardspecies within Amazonia and also found little support for acentral Amazonian savanna corridor Thus even though sa-vanna biotas have expanded and contracted their ranges dur-ing the Quaternary climatic cycles (Mayle et al 2000) thereis little evidence to suggest that these changes resulted in abroad savanna corridor across central Amazonia Penningtonand colleagues (2000) suggested that several lines of evidencepoint to tropical dry forests rather than savannas as replac-ing rain forests in Amazonia during the coolest periods of theQuaternary

Biogeography and conservationUntil very recently conservation of biological resources in theCerrado had received little international attention To date

230 BioScience March 2002 Vol 52 No 3

Articles

Figure 5 Major savanna corridors that have been hy-pothesized to historically connect northern and southernblocks of South American tropical savannas (a) Andeancorridor (b) central Amazonian corridor and (c) coastalcorridor

less than 2 of the total area of the Cerrado is protected inreserves (Mittermeier et al 2000) but there are important dif-ferences in how countries are approaching Cerrado conser-vation The Bolivian government created Noel Kempff Na-tional Park in 1988 thereby protecting the countryrsquos largestcerrado (Killeen and Schulenberg 1998) These roughly 42000ha of unaltered cerrado near the western edge of the biomeis one of the worldrsquos largest continuous protected parcels ofthis habitat (Figure 7) Because of the parkrsquos remoteness andBoliviarsquos comparatively low human population density (es-pecially in the countryrsquos eastern lowlands) the most sub-stantial human threats to this area come from burning asso-ciated with ranching operations in neighboring Brazil Boliviaalso possesses several smaller and more isolated plateaus withcerrado habitat these remain unprotected and largely unex-plored (Parker and Rocha 1991)

In Paraguay at the southern limit of the biome Cerradois protected by two national parks Serraniacuteas de San Luiacutes andCerro Coraacute (Clay et al 1998 Robbins et al 1999) Avian sur-vey work has been conducted at both sites however the re-gions are still not considered to be thoroughly surveyed andboth Clay (1998) and Robbins and their colleagues (1999) em-phasize that the rapid expansion of agricultural operationsthreaten all unprotected savanna regions in the country

In Brazil which possesses the vast majority of Cerrado thesituation is drastically worse There are few large reserves(more than 25000 hectares [ha]) and they are not distributedevenly across the biome (Figure 7) Consequently an im-portant part of the Cerradorsquos environmental diversity hasnot been incorporated in a network of protected areas Largeareas of Brazilian Cerrado and dry forests have been convertedto soybean and rice plantations Most of this large-scale habi-tat modification has not followed the most basic principlesof environmental conservation and problems of erosion anddeterioration of important streams and rivers are increasingInternational support for this alteration has come from ma-jor development agencies and some senior scientists whohave encouraged increased land use in the Cerrado to reducehuman pressure on Amazonian forests

The Brazilian governmentrsquos appreciation of the uniquenessof the Cerradorsquos biodiversity has developed slowly The firstmajor initiative occurred in 1997 when the Brazilian Ministryof Environment promoted a workshop which included over200 scientific experts on the region This group met to definepriority areas for conservation in the Brazilian Cerrado It fol-lowed the same methodology applied earlier for the AtlanticForest and Amazonia (Cavalcanti 1999) Several criteria forassessing biodiversity value were used such as number ofendemic species species richness presence of rare or en-dangered species or both and sites of unique communitiesor key areas for migratory species The group identified a to-tal of 87 priority areas The urgency for conservation actionin a priority area was determined by cross-referencing bio-diversity data with the human-pressure and land cover-change data (Cavalcanti 1999) As of August 1999 the Brazil-ian government has indicated that it will follow most of the

recommendations of this workshop and that new protectedareas and ecological corridors will be established In factsome important actions were taken during 2001 but twodeserve special comments The first was the expansion ofthe National Park of Chapada dos Veadeiros an important re-serve that includes pristine cerrados and patches of camposrupestres with several endemic plant species from 60000 to235000 ha The second was the creation of the Ecological Sta-tion of Serra Geral do Tocantins with 716306 ha close to theState Park of Jalapatildeo with 158885 ha Together these two ar-eas form the worldrsquos largest continuous area of protectedCerrado (Figure 7)

How can the biogeographic information discussed here beused to help protect and preserve the Cerradorsquos biodiversityFirst we know now that the Cerradorsquos biota is not homoge-neous so additional distribution data for different groups oforganisms must be collected and organized in a retrievable wayto help conservationists determine whether there are addi-tional unidentified subareas of endemism Second eventhough the three subareas of endemism identified here as har-boring restricted-range endemic species of birds are partiallycovered by one or more reserves (for comparison refer to Fig-ures 4 and 7 Silva 1997) most reserves need to be more for-mally implemented and connected through biodiversity cor-ridors to insure their preservation over the long term Analysesof other nonavian taxa are certain to uncover additional ar-eas that are not currently protected by reserves Third despiteoccupying a relatively small part of the total landscape

March 2002 Vol 52 No 3 BioScience 231

Articles

Figure 6 Examples of ranges of bird species supportingthe hypothesis (a b) of the Andean savanna corridor and(c d) of the coastal savanna corridor

ba

c d

galleryndashriverine forest cerradatildeo anddry forest have been shown to be keyhabitats for a substantial portion ofthe Cerradorsquos biodiversity Thus theirconversion into pastures and agricul-tural fields should be halted and theirconnectivity maintained Fourth newprojects in areas currently covered bycerrado or other types of open vegeta-tion should be temporarily banned un-til their impact on the flora and faunacan be rigorously assessed and conser-vation strategies for those areas de-signed Fifth it is critical that agricul-tural technology be developed andimplemented to help landowners in-crease productivity of lands alreadyunder cultivation (Macedo 1994 Silva1998) This would reduce pressure onlands still covered by natural vegetationFinally a special conservation actionplan must be developed to guaranteethe conservation of at least part of thelargest savannas in Amazonia and At-lantic Forest as they are important lab-oratories for ecological and evolution-ary studies in tropical South AmericaIt is urgent now because efforts are underway to establish soy-bean plantations in these enclaves

The Cerradorsquos conservation is a challenge from both sci-entific and political viewpoints Because important financialand political interests are involved in the destruction of theCerrado scientists and conservationists must be creative ingathering and synthesizing the best and most complete in-formation possible to generate a viable strategy to insureconservation of the worldrsquos richest tropical savanna

AcknowledgmentsWe thank Jon Fjeldsaring David C Oren Roberto B CavalcantiFernando C Novaes Bent O Poulsen Mary E PetersenTownsend Peterson Carsten Rahbek James A Ratter NeivaGuedes and Shannon J Hackett for discussions about Cer-rado biogeography and comments on the manuscript LuisBarbosa produced the figures Research support came fromConselho Nacional de Desenvolvimento Cientiacutefico e Tec-noloacutegico (grant no 30246488-3) National Geographic So-ciety (grant no 4964-93) Danish Natural Science ResearchCouncil (grant J no 11-0390) Museu Paraense Emiacutelio Goeldiand Universidade de Brasiacutelia

References citedAbrsquoSaber AN 1977 Os Domiacutenios morfoclimaacuteticos da Ameacuterica do Sul

Primeira Aproximaccedilatildeo Geomorfologia 52 1ndash21mdashmdashmdash 1983 O domiacutenio dos cerrados Introduccedilatildeo ao conhecimento Re-

vista do Servidor Puacuteblico 111 41ndash55Aacutevila-Pires TC 1995 Lizards of Brazilian Amazonia (Reptilia Squamata)

Zoologische Verhandelingen Leiden 299 1ndash706

Bates JM Hackett SJ Goerck J 1999 High levels of mitochondrial DNA dif-ferentiation in two lineages of antbirds (Drymophila and Hypocnemis)The Auk 1161093ndash1106

Brasil AE Alvarenga SM 1989 Relevo Pages 53ndash72 in Duarte AC ed Ge-ografia do Brasil Regiatildeo Centro-Oeste Rio de Janeiro IBGE

Cavalcanti RB ed 1999 Accedilotildees Prioritaacuterias para a conservaccedilatildeo da biodiver-sidade do Cerrado e Pantanal Belo Horizonte Conservation Internationaldo Brasil

Clay RP Cooper DR Mazar Barnett J Burfiled IZ Esquivel EZ Farintildea RKennedy CP Perrens M Pople RG 1998 White-winged NightjarsCaprimulgus candicans and cerrado conservation The key findings of pro-ject Aguaraacute Nu 1997 Cotinga 9 52ndash55

Coelho G Silva W 1998 A new species of Antilophia (Paseriformes Pipri-dae) from Chapada do Araripe Cearaacute Brazil Ararajuba 6 81ndash86

Cole MM 1986 The Savannas Biogeography and Geobotany LondonAcademic Press

Cracraft J 1985 Historical biogeography and patterns of differentiationwithin the South America avifauna Areas of endemism OrnithologicalMonographs 36 49ndash84

Eiten G 1972 The Cerrado vegetation of Brazil Botanical Review 38201ndash341

mdashmdashmdash 1990 Vegetaccedilatildeo Pages 9ndash65 in Pinto MN ed Cerrado Caracteri-zaccedilatildeo Ocupaccedilatildeo e Perspectiva Brasiacutelia Universidade de Brasiacutelia and SE-MATEC

Erwin TL Pogue MG 1988 Agra arboreal beetles of Neotropical forests Bio-geography and the forest refugium hypothesis (Carabidae) Pages 161ndash188in Vanzolini PE Heyer WR eds Proceedings of a Workshop on Neotrop-ical Distribution Patterns Rio de Janeiro Academia Brasileira de Ciecircn-cias

Furley PA Ratter JA 1988 Soil resources and plant communities of the cen-tral Brazilian cerrado and their development Journal of Biogeography15 97ndash108

Giullietti AM Pirani JR 1988 Patterns of geographic distribution of someplant species from the Espinhaccedilo Range Minas Gerais and Bahia Brazil

232 BioScience March 2002 Vol 52 No 3

Articles

Figure 7 Parks and ecological stations in the Cerrado region Only those reserveslarger than 25000 ha are shown in Brazil PE state park PN national park EEecological station

savanna no single case of intraspecific differentiation acrossthis transition has been reported so far In addition few gen-era have species in both forest and savanna In cases in whichavian genera do have species in both habitats those speciesare generally considered distantly related

Connections among Neotropical savannas SouthAmerican tropical savanna regions can be grouped into twomajor groups that are currently separated by the Amazon val-ley The northern block is formed by Llanos Roraima ParuMonte Alegre Amapaacute and Marajoacute whereas the second blockis formed by Cerrado Pantanal Llanos de Mojos and patchesof savannas located close to the transition between Amazo-nia and the Cerrado (Figure 5) Because South American sa-vannas exhibit biotic similarity to one another (Sarmiento1983) a consensus opinion is that these regions were con-nected in the recent past and the biotic disjunctions ob-served today are the result of vicariance rather than long-dis-tance dispersal across large expanses of Amazonian forest(Haffer 1967 Sarmiento 1983) Under this scenario SouthAmerican savannas are believed to have expanded and re-tracted their ranges during Quaternary climatic cycles Dur-ing the cold dry periods savannas expanded in Amazoniawhile humid forests retracted to peripheral ecological refugesDuring moist and warm periods humid forests spread againwhile savannas retracted to areas approximating their present-day ranges (Mayle et al 2000) These Quaternary cyclic cli-maticndashvegetational changes have been postulated by some bio-geographers to be the most important factor driving thespeciation process in both forest and savanna organisms intropical South America (Whitmore and Prance 1987 Haffer2001)

Three main corridors connecting northern and southernsavanna regions have been proposed (Haffer 1967 1974Webb 1991) (a) Andean connecting the southern block ofsavannas directly with Llanos and Roraima through the An-dean slopes (b) central Amazonian connecting the southernblock of savannas directly with some patches of savannaslocated north of the Amazon such as Monte Alegre andParu roughly following a belt of low precipitation acrosscentral Amazonia and (c) coastal connecting the southernand northern blocks through savanna patches located closeto the Atlantic coast such as Marajoacute and Amapaacute (Figure 5)Silva (1995a) examined the general distribution patterns ofsavanna-associated birds in the Cerrado region to determinehow well these patterns fit with these three proposed corri-dors It was assumed that if these corridors did exist currentpatterns of distribution of savanna-adapted birds might showevidence of their legacy Thus we predicted spatial congru-ence should exist between speciesrsquo ranges and the position ofthese past biotic corridors as mapped by their proponents (Fig-ure 5)

Present-day ranges of savanna birds support two of the threehypothesized biogeographic corridors (Figure 6) The Andeanconnection is supported by the distribution of five species andlinks the Cerrado region with the Llanos and Gratilde-Sabana The

Atlantic coast savanna corridor is consistent with the distri-butions of 33 species All these species occur in Amapaacute orMarajoacute with some of them extending their ranges in other sa-vannas of the northern block (Pinto Henriques and Oren1997 Silva et al 1997) No single species was recorded onlyin the savannas that follow the central Amazonian savanna cor-ridor (Silva 1995a) These two pieces of information suggestthat a savanna corridor following a belt of low precipitationacross the Amazon basin may not have existed and the mostrecent biotic connections between Amazonian savannas andthe Cerrado region were mostly through the savannas locatedalong the Atlantic coast (Silva 1995a Silva et al 1997) Aacutevila-Pires (1995) analyzed the distribution patterns of all lizardspecies within Amazonia and also found little support for acentral Amazonian savanna corridor Thus even though sa-vanna biotas have expanded and contracted their ranges dur-ing the Quaternary climatic cycles (Mayle et al 2000) thereis little evidence to suggest that these changes resulted in abroad savanna corridor across central Amazonia Penningtonand colleagues (2000) suggested that several lines of evidencepoint to tropical dry forests rather than savannas as replac-ing rain forests in Amazonia during the coolest periods of theQuaternary

Biogeography and conservationUntil very recently conservation of biological resources in theCerrado had received little international attention To date

230 BioScience March 2002 Vol 52 No 3

Articles

Figure 5 Major savanna corridors that have been hy-pothesized to historically connect northern and southernblocks of South American tropical savannas (a) Andeancorridor (b) central Amazonian corridor and (c) coastalcorridor

less than 2 of the total area of the Cerrado is protected inreserves (Mittermeier et al 2000) but there are important dif-ferences in how countries are approaching Cerrado conser-vation The Bolivian government created Noel Kempff Na-tional Park in 1988 thereby protecting the countryrsquos largestcerrado (Killeen and Schulenberg 1998) These roughly 42000ha of unaltered cerrado near the western edge of the biomeis one of the worldrsquos largest continuous protected parcels ofthis habitat (Figure 7) Because of the parkrsquos remoteness andBoliviarsquos comparatively low human population density (es-pecially in the countryrsquos eastern lowlands) the most sub-stantial human threats to this area come from burning asso-ciated with ranching operations in neighboring Brazil Boliviaalso possesses several smaller and more isolated plateaus withcerrado habitat these remain unprotected and largely unex-plored (Parker and Rocha 1991)

In Paraguay at the southern limit of the biome Cerradois protected by two national parks Serraniacuteas de San Luiacutes andCerro Coraacute (Clay et al 1998 Robbins et al 1999) Avian sur-vey work has been conducted at both sites however the re-gions are still not considered to be thoroughly surveyed andboth Clay (1998) and Robbins and their colleagues (1999) em-phasize that the rapid expansion of agricultural operationsthreaten all unprotected savanna regions in the country

In Brazil which possesses the vast majority of Cerrado thesituation is drastically worse There are few large reserves(more than 25000 hectares [ha]) and they are not distributedevenly across the biome (Figure 7) Consequently an im-portant part of the Cerradorsquos environmental diversity hasnot been incorporated in a network of protected areas Largeareas of Brazilian Cerrado and dry forests have been convertedto soybean and rice plantations Most of this large-scale habi-tat modification has not followed the most basic principlesof environmental conservation and problems of erosion anddeterioration of important streams and rivers are increasingInternational support for this alteration has come from ma-jor development agencies and some senior scientists whohave encouraged increased land use in the Cerrado to reducehuman pressure on Amazonian forests

The Brazilian governmentrsquos appreciation of the uniquenessof the Cerradorsquos biodiversity has developed slowly The firstmajor initiative occurred in 1997 when the Brazilian Ministryof Environment promoted a workshop which included over200 scientific experts on the region This group met to definepriority areas for conservation in the Brazilian Cerrado It fol-lowed the same methodology applied earlier for the AtlanticForest and Amazonia (Cavalcanti 1999) Several criteria forassessing biodiversity value were used such as number ofendemic species species richness presence of rare or en-dangered species or both and sites of unique communitiesor key areas for migratory species The group identified a to-tal of 87 priority areas The urgency for conservation actionin a priority area was determined by cross-referencing bio-diversity data with the human-pressure and land cover-change data (Cavalcanti 1999) As of August 1999 the Brazil-ian government has indicated that it will follow most of the

recommendations of this workshop and that new protectedareas and ecological corridors will be established In factsome important actions were taken during 2001 but twodeserve special comments The first was the expansion ofthe National Park of Chapada dos Veadeiros an important re-serve that includes pristine cerrados and patches of camposrupestres with several endemic plant species from 60000 to235000 ha The second was the creation of the Ecological Sta-tion of Serra Geral do Tocantins with 716306 ha close to theState Park of Jalapatildeo with 158885 ha Together these two ar-eas form the worldrsquos largest continuous area of protectedCerrado (Figure 7)

How can the biogeographic information discussed here beused to help protect and preserve the Cerradorsquos biodiversityFirst we know now that the Cerradorsquos biota is not homoge-neous so additional distribution data for different groups oforganisms must be collected and organized in a retrievable wayto help conservationists determine whether there are addi-tional unidentified subareas of endemism Second eventhough the three subareas of endemism identified here as har-boring restricted-range endemic species of birds are partiallycovered by one or more reserves (for comparison refer to Fig-ures 4 and 7 Silva 1997) most reserves need to be more for-mally implemented and connected through biodiversity cor-ridors to insure their preservation over the long term Analysesof other nonavian taxa are certain to uncover additional ar-eas that are not currently protected by reserves Third despiteoccupying a relatively small part of the total landscape

March 2002 Vol 52 No 3 BioScience 231

Articles

Figure 6 Examples of ranges of bird species supportingthe hypothesis (a b) of the Andean savanna corridor and(c d) of the coastal savanna corridor

ba

c d

galleryndashriverine forest cerradatildeo anddry forest have been shown to be keyhabitats for a substantial portion ofthe Cerradorsquos biodiversity Thus theirconversion into pastures and agricul-tural fields should be halted and theirconnectivity maintained Fourth newprojects in areas currently covered bycerrado or other types of open vegeta-tion should be temporarily banned un-til their impact on the flora and faunacan be rigorously assessed and conser-vation strategies for those areas de-signed Fifth it is critical that agricul-tural technology be developed andimplemented to help landowners in-crease productivity of lands alreadyunder cultivation (Macedo 1994 Silva1998) This would reduce pressure onlands still covered by natural vegetationFinally a special conservation actionplan must be developed to guaranteethe conservation of at least part of thelargest savannas in Amazonia and At-lantic Forest as they are important lab-oratories for ecological and evolution-ary studies in tropical South AmericaIt is urgent now because efforts are underway to establish soy-bean plantations in these enclaves

The Cerradorsquos conservation is a challenge from both sci-entific and political viewpoints Because important financialand political interests are involved in the destruction of theCerrado scientists and conservationists must be creative ingathering and synthesizing the best and most complete in-formation possible to generate a viable strategy to insureconservation of the worldrsquos richest tropical savanna

AcknowledgmentsWe thank Jon Fjeldsaring David C Oren Roberto B CavalcantiFernando C Novaes Bent O Poulsen Mary E PetersenTownsend Peterson Carsten Rahbek James A Ratter NeivaGuedes and Shannon J Hackett for discussions about Cer-rado biogeography and comments on the manuscript LuisBarbosa produced the figures Research support came fromConselho Nacional de Desenvolvimento Cientiacutefico e Tec-noloacutegico (grant no 30246488-3) National Geographic So-ciety (grant no 4964-93) Danish Natural Science ResearchCouncil (grant J no 11-0390) Museu Paraense Emiacutelio Goeldiand Universidade de Brasiacutelia

References citedAbrsquoSaber AN 1977 Os Domiacutenios morfoclimaacuteticos da Ameacuterica do Sul

Primeira Aproximaccedilatildeo Geomorfologia 52 1ndash21mdashmdashmdash 1983 O domiacutenio dos cerrados Introduccedilatildeo ao conhecimento Re-

vista do Servidor Puacuteblico 111 41ndash55Aacutevila-Pires TC 1995 Lizards of Brazilian Amazonia (Reptilia Squamata)

Zoologische Verhandelingen Leiden 299 1ndash706

Bates JM Hackett SJ Goerck J 1999 High levels of mitochondrial DNA dif-ferentiation in two lineages of antbirds (Drymophila and Hypocnemis)The Auk 1161093ndash1106

Brasil AE Alvarenga SM 1989 Relevo Pages 53ndash72 in Duarte AC ed Ge-ografia do Brasil Regiatildeo Centro-Oeste Rio de Janeiro IBGE

Cavalcanti RB ed 1999 Accedilotildees Prioritaacuterias para a conservaccedilatildeo da biodiver-sidade do Cerrado e Pantanal Belo Horizonte Conservation Internationaldo Brasil

Clay RP Cooper DR Mazar Barnett J Burfiled IZ Esquivel EZ Farintildea RKennedy CP Perrens M Pople RG 1998 White-winged NightjarsCaprimulgus candicans and cerrado conservation The key findings of pro-ject Aguaraacute Nu 1997 Cotinga 9 52ndash55

Coelho G Silva W 1998 A new species of Antilophia (Paseriformes Pipri-dae) from Chapada do Araripe Cearaacute Brazil Ararajuba 6 81ndash86

Cole MM 1986 The Savannas Biogeography and Geobotany LondonAcademic Press

Cracraft J 1985 Historical biogeography and patterns of differentiationwithin the South America avifauna Areas of endemism OrnithologicalMonographs 36 49ndash84

Eiten G 1972 The Cerrado vegetation of Brazil Botanical Review 38201ndash341

mdashmdashmdash 1990 Vegetaccedilatildeo Pages 9ndash65 in Pinto MN ed Cerrado Caracteri-zaccedilatildeo Ocupaccedilatildeo e Perspectiva Brasiacutelia Universidade de Brasiacutelia and SE-MATEC

Erwin TL Pogue MG 1988 Agra arboreal beetles of Neotropical forests Bio-geography and the forest refugium hypothesis (Carabidae) Pages 161ndash188in Vanzolini PE Heyer WR eds Proceedings of a Workshop on Neotrop-ical Distribution Patterns Rio de Janeiro Academia Brasileira de Ciecircn-cias

Furley PA Ratter JA 1988 Soil resources and plant communities of the cen-tral Brazilian cerrado and their development Journal of Biogeography15 97ndash108

Giullietti AM Pirani JR 1988 Patterns of geographic distribution of someplant species from the Espinhaccedilo Range Minas Gerais and Bahia Brazil

232 BioScience March 2002 Vol 52 No 3

Articles

Figure 7 Parks and ecological stations in the Cerrado region Only those reserveslarger than 25000 ha are shown in Brazil PE state park PN national park EEecological station

less than 2 of the total area of the Cerrado is protected inreserves (Mittermeier et al 2000) but there are important dif-ferences in how countries are approaching Cerrado conser-vation The Bolivian government created Noel Kempff Na-tional Park in 1988 thereby protecting the countryrsquos largestcerrado (Killeen and Schulenberg 1998) These roughly 42000ha of unaltered cerrado near the western edge of the biomeis one of the worldrsquos largest continuous protected parcels ofthis habitat (Figure 7) Because of the parkrsquos remoteness andBoliviarsquos comparatively low human population density (es-pecially in the countryrsquos eastern lowlands) the most sub-stantial human threats to this area come from burning asso-ciated with ranching operations in neighboring Brazil Boliviaalso possesses several smaller and more isolated plateaus withcerrado habitat these remain unprotected and largely unex-plored (Parker and Rocha 1991)

In Paraguay at the southern limit of the biome Cerradois protected by two national parks Serraniacuteas de San Luiacutes andCerro Coraacute (Clay et al 1998 Robbins et al 1999) Avian sur-vey work has been conducted at both sites however the re-gions are still not considered to be thoroughly surveyed andboth Clay (1998) and Robbins and their colleagues (1999) em-phasize that the rapid expansion of agricultural operationsthreaten all unprotected savanna regions in the country

In Brazil which possesses the vast majority of Cerrado thesituation is drastically worse There are few large reserves(more than 25000 hectares [ha]) and they are not distributedevenly across the biome (Figure 7) Consequently an im-portant part of the Cerradorsquos environmental diversity hasnot been incorporated in a network of protected areas Largeareas of Brazilian Cerrado and dry forests have been convertedto soybean and rice plantations Most of this large-scale habi-tat modification has not followed the most basic principlesof environmental conservation and problems of erosion anddeterioration of important streams and rivers are increasingInternational support for this alteration has come from ma-jor development agencies and some senior scientists whohave encouraged increased land use in the Cerrado to reducehuman pressure on Amazonian forests

The Brazilian governmentrsquos appreciation of the uniquenessof the Cerradorsquos biodiversity has developed slowly The firstmajor initiative occurred in 1997 when the Brazilian Ministryof Environment promoted a workshop which included over200 scientific experts on the region This group met to definepriority areas for conservation in the Brazilian Cerrado It fol-lowed the same methodology applied earlier for the AtlanticForest and Amazonia (Cavalcanti 1999) Several criteria forassessing biodiversity value were used such as number ofendemic species species richness presence of rare or en-dangered species or both and sites of unique communitiesor key areas for migratory species The group identified a to-tal of 87 priority areas The urgency for conservation actionin a priority area was determined by cross-referencing bio-diversity data with the human-pressure and land cover-change data (Cavalcanti 1999) As of August 1999 the Brazil-ian government has indicated that it will follow most of the

recommendations of this workshop and that new protectedareas and ecological corridors will be established In factsome important actions were taken during 2001 but twodeserve special comments The first was the expansion ofthe National Park of Chapada dos Veadeiros an important re-serve that includes pristine cerrados and patches of camposrupestres with several endemic plant species from 60000 to235000 ha The second was the creation of the Ecological Sta-tion of Serra Geral do Tocantins with 716306 ha close to theState Park of Jalapatildeo with 158885 ha Together these two ar-eas form the worldrsquos largest continuous area of protectedCerrado (Figure 7)

How can the biogeographic information discussed here beused to help protect and preserve the Cerradorsquos biodiversityFirst we know now that the Cerradorsquos biota is not homoge-neous so additional distribution data for different groups oforganisms must be collected and organized in a retrievable wayto help conservationists determine whether there are addi-tional unidentified subareas of endemism Second eventhough the three subareas of endemism identified here as har-boring restricted-range endemic species of birds are partiallycovered by one or more reserves (for comparison refer to Fig-ures 4 and 7 Silva 1997) most reserves need to be more for-mally implemented and connected through biodiversity cor-ridors to insure their preservation over the long term Analysesof other nonavian taxa are certain to uncover additional ar-eas that are not currently protected by reserves Third despiteoccupying a relatively small part of the total landscape

March 2002 Vol 52 No 3 BioScience 231

Articles

Figure 6 Examples of ranges of bird species supportingthe hypothesis (a b) of the Andean savanna corridor and(c d) of the coastal savanna corridor

ba

c d

galleryndashriverine forest cerradatildeo anddry forest have been shown to be keyhabitats for a substantial portion ofthe Cerradorsquos biodiversity Thus theirconversion into pastures and agricul-tural fields should be halted and theirconnectivity maintained Fourth newprojects in areas currently covered bycerrado or other types of open vegeta-tion should be temporarily banned un-til their impact on the flora and faunacan be rigorously assessed and conser-vation strategies for those areas de-signed Fifth it is critical that agricul-tural technology be developed andimplemented to help landowners in-crease productivity of lands alreadyunder cultivation (Macedo 1994 Silva1998) This would reduce pressure onlands still covered by natural vegetationFinally a special conservation actionplan must be developed to guaranteethe conservation of at least part of thelargest savannas in Amazonia and At-lantic Forest as they are important lab-oratories for ecological and evolution-ary studies in tropical South AmericaIt is urgent now because efforts are underway to establish soy-bean plantations in these enclaves

The Cerradorsquos conservation is a challenge from both sci-entific and political viewpoints Because important financialand political interests are involved in the destruction of theCerrado scientists and conservationists must be creative ingathering and synthesizing the best and most complete in-formation possible to generate a viable strategy to insureconservation of the worldrsquos richest tropical savanna

AcknowledgmentsWe thank Jon Fjeldsaring David C Oren Roberto B CavalcantiFernando C Novaes Bent O Poulsen Mary E PetersenTownsend Peterson Carsten Rahbek James A Ratter NeivaGuedes and Shannon J Hackett for discussions about Cer-rado biogeography and comments on the manuscript LuisBarbosa produced the figures Research support came fromConselho Nacional de Desenvolvimento Cientiacutefico e Tec-noloacutegico (grant no 30246488-3) National Geographic So-ciety (grant no 4964-93) Danish Natural Science ResearchCouncil (grant J no 11-0390) Museu Paraense Emiacutelio Goeldiand Universidade de Brasiacutelia

References citedAbrsquoSaber AN 1977 Os Domiacutenios morfoclimaacuteticos da Ameacuterica do Sul

Primeira Aproximaccedilatildeo Geomorfologia 52 1ndash21mdashmdashmdash 1983 O domiacutenio dos cerrados Introduccedilatildeo ao conhecimento Re-

vista do Servidor Puacuteblico 111 41ndash55Aacutevila-Pires TC 1995 Lizards of Brazilian Amazonia (Reptilia Squamata)

Zoologische Verhandelingen Leiden 299 1ndash706

Bates JM Hackett SJ Goerck J 1999 High levels of mitochondrial DNA dif-ferentiation in two lineages of antbirds (Drymophila and Hypocnemis)The Auk 1161093ndash1106

Brasil AE Alvarenga SM 1989 Relevo Pages 53ndash72 in Duarte AC ed Ge-ografia do Brasil Regiatildeo Centro-Oeste Rio de Janeiro IBGE

Cavalcanti RB ed 1999 Accedilotildees Prioritaacuterias para a conservaccedilatildeo da biodiver-sidade do Cerrado e Pantanal Belo Horizonte Conservation Internationaldo Brasil

Clay RP Cooper DR Mazar Barnett J Burfiled IZ Esquivel EZ Farintildea RKennedy CP Perrens M Pople RG 1998 White-winged NightjarsCaprimulgus candicans and cerrado conservation The key findings of pro-ject Aguaraacute Nu 1997 Cotinga 9 52ndash55

Coelho G Silva W 1998 A new species of Antilophia (Paseriformes Pipri-dae) from Chapada do Araripe Cearaacute Brazil Ararajuba 6 81ndash86

Cole MM 1986 The Savannas Biogeography and Geobotany LondonAcademic Press

Cracraft J 1985 Historical biogeography and patterns of differentiationwithin the South America avifauna Areas of endemism OrnithologicalMonographs 36 49ndash84

Eiten G 1972 The Cerrado vegetation of Brazil Botanical Review 38201ndash341

mdashmdashmdash 1990 Vegetaccedilatildeo Pages 9ndash65 in Pinto MN ed Cerrado Caracteri-zaccedilatildeo Ocupaccedilatildeo e Perspectiva Brasiacutelia Universidade de Brasiacutelia and SE-MATEC

Erwin TL Pogue MG 1988 Agra arboreal beetles of Neotropical forests Bio-geography and the forest refugium hypothesis (Carabidae) Pages 161ndash188in Vanzolini PE Heyer WR eds Proceedings of a Workshop on Neotrop-ical Distribution Patterns Rio de Janeiro Academia Brasileira de Ciecircn-cias

Furley PA Ratter JA 1988 Soil resources and plant communities of the cen-tral Brazilian cerrado and their development Journal of Biogeography15 97ndash108

Giullietti AM Pirani JR 1988 Patterns of geographic distribution of someplant species from the Espinhaccedilo Range Minas Gerais and Bahia Brazil

232 BioScience March 2002 Vol 52 No 3

Articles

Figure 7 Parks and ecological stations in the Cerrado region Only those reserveslarger than 25000 ha are shown in Brazil PE state park PN national park EEecological station

galleryndashriverine forest cerradatildeo anddry forest have been shown to be keyhabitats for a substantial portion ofthe Cerradorsquos biodiversity Thus theirconversion into pastures and agricul-tural fields should be halted and theirconnectivity maintained Fourth newprojects in areas currently covered bycerrado or other types of open vegeta-tion should be temporarily banned un-til their impact on the flora and faunacan be rigorously assessed and conser-vation strategies for those areas de-signed Fifth it is critical that agricul-tural technology be developed andimplemented to help landowners in-crease productivity of lands alreadyunder cultivation (Macedo 1994 Silva1998) This would reduce pressure onlands still covered by natural vegetationFinally a special conservation actionplan must be developed to guaranteethe conservation of at least part of thelargest savannas in Amazonia and At-lantic Forest as they are important lab-oratories for ecological and evolution-ary studies in tropical South AmericaIt is urgent now because efforts are underway to establish soy-bean plantations in these enclaves

The Cerradorsquos conservation is a challenge from both sci-entific and political viewpoints Because important financialand political interests are involved in the destruction of theCerrado scientists and conservationists must be creative ingathering and synthesizing the best and most complete in-formation possible to generate a viable strategy to insureconservation of the worldrsquos richest tropical savanna

AcknowledgmentsWe thank Jon Fjeldsaring David C Oren Roberto B CavalcantiFernando C Novaes Bent O Poulsen Mary E PetersenTownsend Peterson Carsten Rahbek James A Ratter NeivaGuedes and Shannon J Hackett for discussions about Cer-rado biogeography and comments on the manuscript LuisBarbosa produced the figures Research support came fromConselho Nacional de Desenvolvimento Cientiacutefico e Tec-noloacutegico (grant no 30246488-3) National Geographic So-ciety (grant no 4964-93) Danish Natural Science ResearchCouncil (grant J no 11-0390) Museu Paraense Emiacutelio Goeldiand Universidade de Brasiacutelia

References citedAbrsquoSaber AN 1977 Os Domiacutenios morfoclimaacuteticos da Ameacuterica do Sul

Primeira Aproximaccedilatildeo Geomorfologia 52 1ndash21mdashmdashmdash 1983 O domiacutenio dos cerrados Introduccedilatildeo ao conhecimento Re-

vista do Servidor Puacuteblico 111 41ndash55Aacutevila-Pires TC 1995 Lizards of Brazilian Amazonia (Reptilia Squamata)

Zoologische Verhandelingen Leiden 299 1ndash706

Bates JM Hackett SJ Goerck J 1999 High levels of mitochondrial DNA dif-ferentiation in two lineages of antbirds (Drymophila and Hypocnemis)The Auk 1161093ndash1106

Brasil AE Alvarenga SM 1989 Relevo Pages 53ndash72 in Duarte AC ed Ge-ografia do Brasil Regiatildeo Centro-Oeste Rio de Janeiro IBGE

Cavalcanti RB ed 1999 Accedilotildees Prioritaacuterias para a conservaccedilatildeo da biodiver-sidade do Cerrado e Pantanal Belo Horizonte Conservation Internationaldo Brasil

Clay RP Cooper DR Mazar Barnett J Burfiled IZ Esquivel EZ Farintildea RKennedy CP Perrens M Pople RG 1998 White-winged NightjarsCaprimulgus candicans and cerrado conservation The key findings of pro-ject Aguaraacute Nu 1997 Cotinga 9 52ndash55

Coelho G Silva W 1998 A new species of Antilophia (Paseriformes Pipri-dae) from Chapada do Araripe Cearaacute Brazil Ararajuba 6 81ndash86

Cole MM 1986 The Savannas Biogeography and Geobotany LondonAcademic Press

Cracraft J 1985 Historical biogeography and patterns of differentiationwithin the South America avifauna Areas of endemism OrnithologicalMonographs 36 49ndash84

Eiten G 1972 The Cerrado vegetation of Brazil Botanical Review 38201ndash341

mdashmdashmdash 1990 Vegetaccedilatildeo Pages 9ndash65 in Pinto MN ed Cerrado Caracteri-zaccedilatildeo Ocupaccedilatildeo e Perspectiva Brasiacutelia Universidade de Brasiacutelia and SE-MATEC

Erwin TL Pogue MG 1988 Agra arboreal beetles of Neotropical forests Bio-geography and the forest refugium hypothesis (Carabidae) Pages 161ndash188in Vanzolini PE Heyer WR eds Proceedings of a Workshop on Neotrop-ical Distribution Patterns Rio de Janeiro Academia Brasileira de Ciecircn-cias

Furley PA Ratter JA 1988 Soil resources and plant communities of the cen-tral Brazilian cerrado and their development Journal of Biogeography15 97ndash108

Giullietti AM Pirani JR 1988 Patterns of geographic distribution of someplant species from the Espinhaccedilo Range Minas Gerais and Bahia Brazil

232 BioScience March 2002 Vol 52 No 3

Articles

Figure 7 Parks and ecological stations in the Cerrado region Only those reserveslarger than 25000 ha are shown in Brazil PE state park PN national park EEecological station

Pages 39ndash69 in Vanzolini PE Heyer WR eds Proceedings of a Workshopon Neotropical Distribution Patterns Rio de Janeiro Academia Brasileirade Ciecircncias

Haddad CFB Andrade GV Cardoso AJ 1988Anfiacutebios anuros do Parque Na-cional da Serra da Canastra estado de Minas Gerais Brasil Florestal 649ndash20

Haffer J 1967 Zoogeographical notes on the ldquononforestrdquo lowland bird fau-nas of northwestern South America Hornero 10 315ndash333

mdashmdashmdash 1974 Avian Speciation in South America Publications of the Nuttall Ornithological Club 14 1ndash390

mdashmdashmdash 1985 Avian zoogeography of the Neotropical lowlands Ornitho-logical Monographs 36 113ndash146

mdashmdashmdash 2001 Hypotheses to explain the origin of species in Amazonia Pages45ndash118 in Vieira ICG Silva JMC Oren DC DrsquoIncao MA eds Biologi-cal and Cultural Diversity in Amazonia Beleacutem Museu Paraense EmiacutelioGoeldi

Harley RM 1988 Evolution and distribution of Eriope (Labiatae) and its rel-atives in Brazil Pages 71ndash120 in Vanzolini PE Heyer WR eds Pro-ceedings of a Workshop on Neotropical Distribution Patterns Rio deJaneiro Academia Brasileira de Ciecircncias

Killeen TJ Schulenberg TS eds 1998 A biological assessment of Parque Na-cional Noel Kempff Mercado Bolivia RAP working papers 10 Wash-ington (DC) Conservation International

Macedo J 1994 Prospectives for the rational use of the Brazilian cerradosfor food production Anais da Academia Brasileira de Ciecircncias 66 (supp1) 159ndash166

Mayle FE Burbridge BKilleen TJ 2000 Millennial-scale dynamics of south-ern Amazonian rain forests Science 290 2291ndash2294

Mendonccedila RC Felfili JM Walter BMT Juacutenior MCS Rezende AV FilgueirasTS Nogueira PE 1998 Flora vascular do Cerrado Pages 289ndash556 in SanoSM Almeida SP eds Cerrado Ambiente e Flora Brasiacutelia EMBRAPA

Mittermeier RA Myers N Mittermeier CG 2000 Hotspots Earthrsquos Biolog-ically Richest and Most Endangered Terrestrial Ecoregions Mexico CityCEMEX

Moojen J Locks MLangguth A 1997 A new species of Kerodon Cuvier 1825from the state of Goiaacutes Brazil (Mammalia Rodentia Caviidae) Boletimdo Museu Nacional Zoologia nova seacuterie 377 1ndash10

Muumlller P 1973 The Dispersal Centers of Terrestrial Vertebrates in theNeotropical Realm A Study in the Evolution of the Neotropical Biota andIts Native Landscapes The Hague Dr W Junk

Myers N Mittermeier RA Mittermeier CG da Fonseca GAB Kent J 2000Biodiversity hotspots for conservation priorities Nature 403 853ndash858

Nimer E 1979 Climatologia do Brasil Rio de Janeiro IBGEParker TA III Rocha O O 1991 The avifauna of Cerro San Simon an iso-

lated campo rupestre locality in the Department of Beni northern Bo-livia Ecologia en Bolivia 17 15ndash29

Pennington RT Prado DE Pendry CA 2000 Neotropical seasonally dry forestsand Quaternary vegetation changes Journal of Biogeography 27 261ndash273

Pinto Henriques M Oren DC 1997 The avifauna of Marajoacute Caviana andMexiana islands Amazon river estuary Brazil Revista Brasileira de Bi-ologia 57 357ndash382

Prado DE Gibbs PE 1993 Patterns of species distributions in the dry sea-sonal forests of South America Annals of the Missouri Botanical Gar-den 80 902ndash927

Ratter JA Dargie TCD 1992 An analysis of the floristic composition of 26cerrado areas in Brazil Edinburgh Journal of Botany 49 235ndash250

Ratter JA Askew GP Montgomery RF Gifford DR 1978 Observations onforests of some mesotrophic soils in central Brazil Revista Brasileira deBotacircnica 1 47ndash58

Redford KH Fonseca GAB 1986 The role of gallery forests in the zoo-geography of the cerradorsquos non-volant mammalian fauna Biotropica 18126ndash135

Ribeiro JF Walter BMT 1998 Fitofisionomias do bioma Cerrado Pages91ndash166 in Sano SM Almeida SP eds Cerrado Ambiente e Flora BrasiacuteliaEMBRAPA

Rizzini CT 1979 Tratado de Fitogeografia do Brasil Satildeo Paulo HUCITECand EDUSP

Robbins MB Faucett RC Rice NH 1999 Avifauna of a Paraguayan cerradolocality Parque Nacional Serrania San Luis depto Concepcion WilsonBulletin 111 157ndash302

Rodrigues MT 1988 Distribution of lizards of the genus Tropidurus inBrazil (Sauria Iguanidae) Pages 305ndash315 in Vanzolini PE Heyer WR edsProceedings of a Workshop on Neotropical Distribution Patterns Rio deJaneiro Academia Brasileira de Ciecircncias

Sarmiento G 1983 The savannas of tropical America Pages 245ndash288 inBourlieacutere F ed Ecosystems of the World Vol 13 Amsterdam Elsevier

Silva JMC 1995a Biogeographic analysis of the South American avifaunaSteenstrupia 21 49ndash67

mdashmdashmdash 1995b Birds of the Cerrado Region South America Steenstrupia21 69ndash92

mdashmdashmdash 1995c Avian inventory of the Cerrado Region South AmericaImplications for biological conservation Bird Conservation Interna-tional 5 291ndash304

mdashmdashmdash 1997 Endemic bird species and conservation in the Cerrado RegionSouth America Biodiversity and Conservation 6 435ndash450

mdashmdashmdash 1998 Integrating Biogeography and Conservation An examplewith birds and plants of the cerrado region Anais da Academia Brasileirade Ciecircncias 70 881ndash888

Silva JMC Oren DC Roma JC Henriques LMP 1997 Composition and dis-tribution patterns of the avifauna of an Amazonian upland savannaAmapaacute Brazil Ornithological Monographs 48 743ndash762

Schilthuize n M 2000 Ecotone Speciation-prone Trends in Ecology and Evo-lution 15 130ndash131

Smith TB Wayne RK Girman DJ Bruford MW 1997 A role for ecotonesin generating rainforest biodiversity Science 276 1855ndash1857

Webb SD 1991 Ecogeography and the Great American InterchangePaleobiology 17 266ndash280

Whitmore TC Prance GT 1987 Biogeography and Quaternary History inTropical America Oxford (UK) Clarendon Press

Zimmer KJ Whittaker A Oren DC 2001 A cryptic new species of fly-catcher (Tyrannidae Suiriri) from the Cerrado region of central SouthAmerica Auk 118 56ndash78

March 2002 Vol 52 No 3 BioScience 233

Articles