The most widely accepted classification system for Brazilian vegetation is that of Veloso et al. (1991), who separated the country into four biomes (see comment on biomes below): the Amazon Forest, the Atlantic Forest, the Savanna (= Cerrado), and the Steppe (= Caatinga + Campos sulinos). Joly et al. (1999) provide a good introduction to the phytogeographic divisions of Brazil, and Daly and Mitchell (2000) give a South American overview. We divided the country into five main phytogeographic domains: Amazon, Cerrado, Atlantic Forest (including the Araucaria forests), Caatinga, and Campos sulinos (Fig. 1). Here, instead of adopting a single Steppe biome for the northeastern Caatinga and the southern Campos sulinos (Veloso et al., 1991), we kept them separate due to their strong floristic and climatic differences. It is worth mentioning that the use of “biome” as a synonym of “phytogeographic domain” in Brazilian phytogeographic published works is incorrect (Coutinho, 2006). “Biome” refers to an area with physiognomic homogeneity regardless of the floristic composition, and the term “phytogeographic domain” implies physiognomic heterogeneity and takes the floristic composition as a very important component.
The Amazon forests extend over an area of approximately 6 million km2 in northern South America (Daly & Mitchell, 2000), and account for most of the world's remaining rainforests. The limits of these forests have been variously defined (Daly & Prance, 1989), but they are somewhat congruent with the Amazon basin. The basin, however, extends north and eastward beyond the forested area, and the forests become replaced by savannas in most of eastern Bolivia and central-northern Brazil (Goulding et al., 2003). Amazon rainforests are among the most diverse in the world (Gentry, 1988a; Valencia et al., 1994). Recent estimates of Amazonian plant diversity range from 25,000 angiosperm species in the entire basin (Goulding et al., 2003) to 30,000 in the Brazilian part alone (Gentry, 1982), but these may represent a crude underestimation of the total diversity (Hopkins, 2007). Based on data compiled from monographs, Gentry (1982) estimated 76% of the Amazonian flora as endemic at the species level.
The geological history of the Amazon basin is marked by a major shift in drainage orientation, caused by the uplift of the Andean mountain range (Hoorn, 1993). Before the Andean uplift, most of the Amazon basin drained westward to the Pacific and was covered by westward deposition of low-fertility sediments from the pre-Cambrian Brazilian and Guiana Shields (Hoorn, 1993). With the subsequent uplift of the northern Andes, starting approximately 15 mya, the direction of sediment deposition shifted eastward, and most of the ancient sands were covered by deep clay sediments coming from the Andes (Burnham & Graham, 1999). During the Late Miocene (8–10 mya), the Pebasian Sea transgression created a marine seaway uniting the Caribbean to the South Atlantic and resulted in a dramatic landscape change leading to the current configuration of the Amazon basin (Räsänen et al., 1995; Webb, 1995). The presence of rainforests in the basin largely colonizes relatively recent Andean sediments. Earlier sandy sediments from the Guiana Shield cover approximately 3% of the Amazon (ter Steege et al., 2000) and are found in patches mostly in the upper Rio Negro of northwestern Brazil (Anderson, 1981; Prance, 1996), neighboring areas in Venezuela and Colombia (Daly & Mitchell, 2000), and the Iquitos region of northeastern Peru (Alonso & Whitney, 2003). These areas of white-sand soils do not support tall rainforests but rather several physiognomies varying from open scrublands to low forests, depending on the proximity of the water table (Anderson, 1981; Huber, 1995; Prance, 1996; Daly & Mitchell, 2000).
The Amazon phytogeographic domain is broadly defined here to include the Guianan lowlands (Gentry, 1982; Granville, 1988) and the Guiana Shield (Huber, 1995) (i.e., the Amazonian subregion of Morrone, 2006), in spite of strong evidence to the contrary (Mori, 1991; Berry et al., 1995; Kelloff & Funk, 2004; Berry & Riina, 2005). Thus, in addition to the large extension of lowland rainforests and white-sand forests, several other physiognomies are found in scattered patches across this domain. The most prominent of these patches are the Amazonian savannas (Prance, 1996; Daly & Mitchell, 2000), which are floristically linked to the Cerrado (see section 2.2), and the tepuis, which are remnant sandstone plateaus of the Guiana Shield (Huber, 1995; Prance, 1996; Daly & Mitchell, 2000; Berry & Riina, 2005). Most tepuis are restricted to Venezuela, but some extend to northern Brazil, eastern Colombia and the Guianas (Daly & Mitchell, 2000; Berry & Riina, 2005). The tepui vegetation is very diverse physiognomically (Huber, 1995; Daly & Mitchell, 2000), and 42% of the flora from areas above 1500 m (the Pantepui region) has been estimated as endemic (Berry & Riina, 2005).
Of major importance to Brazilian biogeography are floristic connections between the Guiana Shield with white-sand Amazonian habitats and/or eastern Brazil, as exemplified by Bonnetia (Bonnetiaceae), Caraipa (Clusiaceae), Potalia (Gentianaceae), Humiria (Humiriaceae), Chamaecrista and Eperua (Leguminosae), Marcetia and Microlicia (Melastomataceae), Biophytum (Oxalidaceae), Rapateaceae, Pagamea and Sipanea (Rubiaceae), Barbacenia and Vellozia (Velloziaceae), and Xyridaceae. Among the few phylogenetic studies focused on white-sand plant lineages, Struwe et al. (1998) suggested a general pattern where endemic white-sand taxa represent older lineages from which more recent lineages have been derived. Although this hypothesis has not been sufficiently tested with phylogenetic data, Frasier et al. (2008) provided additional support for the older ancestry of endemic white-sand taxa. Other studies, however, point to more complex scenarios, involving several shifts to white-sand areas from ancestors growing in more fertile clay or terrace soils (Fine et al., 2005), back and forth movements between lowland forests and tepui summits (Steyermark, 1986; Givnish et al., 2000), and shifts to habitats with different light conditions, flooding regimes and altitude (Vicentini, 2007).
Based on the detection of areas of bird endemism in the Amazon, Haffer (1969) proposed a model to explain the diversification of the Amazonian biota in terms of allopatric speciation driven by forest fragmentation during glacial events in the Quaternary. This Pleistocene refuge model predicted that areas with a greater number of endemic species were likely to have acted as refuges for rainforest taxa during the expansion of savannas in glacial times. These isolated patches of rainforest in a landscape of savannas and dry forests would have provided the reproductive isolation required for allopatric speciation. The same (or slightly modified) areas of endemism reported for birds were believed to have also accounted for the diversification of frogs, lizards, butterflies and several groups of plants (Vanzolini, 1970; Brown, 1982; Prance, 1982). However, the initial excitement to detect such refuges for several groups of organisms was accompanied by many attempts to test the model, most of which have largely refuted the assumptions based on collection bias (Nelson et al., 1990), hypothesized paleoecological conditions (Colinvaux et al., 2000, 2001; Bush & Oliveira, 2006), and molecular systematic data (e.g., Moritz et al., 2000; Patton & Silva, 2005). Several additional models have been proposed to explain Amazonian biodiversity, but none have achieved the same credibility (for a review of these models see Haffer, 1997, 2008; Marroig & Cerqueira, 1997). Current views on Amazonian diversification suggest that it is unlikely for the pattern to be adequately explained by the model of vicariance alone (Bush, 1994), requiring a mixture of pre-Pleistocene speciation events (Patton & Silva, 2005) and recent radiations (Richardson et al., 2001; Erkens et al., 2007). Geological processes and marine transgressions throughout the Amazon basin have been invoked as potential causes for currently observed vicariance patterns (Räsänen et al., 1995; Patton & Silva, 2005; Rossetti et al., 2005), whereas recent diversification of some plant groups appear to have been promoted by ecological shifts related to habitat (Prance, 1982, 1994; Gentry 1988b, 1989; Tuomisto et al., 1995; Fine et al., 2005), pollinator specialization (Gentry, 1982; Kay et al., 2005), and fluctuating temperatures and precipitation (Graham, 1997).
In order to investigate diversification of the Amazonian biota, attempts to detect areas of endemism have been made for several groups of organisms (Haffer, 1969; Vanzolini, 1970; Prance, 1982; Brown, 1982; Cracraft, 1985; Cracraft & Prum, 1988), although most studies have relied solely on the detection of such areas and the possible causes for their existence. The following areas of endemism have been consistently identified in the Amazon: Guiana (northern Brazil, Guyana, Surinam and French Guiana); Imeri (southern Venezuela and neighboring areas in Brazil and Colombia); Napo (Upper Negro-Uaupés rivers in Brazil, and Colombian, Ecuadorian and northern Peruvian Amazon); Inambari (southwestern Brazilian Amazon, eastern Peru and northwestern Bolivia); Rondônia (Madeira-Tapajós interfluvial areas, into northern Bolivia); Pará (southern Pará and northern Mato Grosso states in Brazil); and Belém (northeastern Brazilian Amazon). Most information on the historical relationships among Amazonian areas of endemism based on phylogenetic data comes from studies of birds (Cracraft, 1985; Cracraft & Prum, 1988; Marks et al., 2002; Eberhard & Bermingham, 2005; Ribas et al., 2005), butterflies (Hall & Harvey, 2002) and dipterans (Nihei & Carvalho, 2007 and references therein) and, to our understanding, the issue of how such areas are related to each other based on plant phylogenies has never been addressed.
There is evidence disputing the recognition of the Amazon as a biogeographic unit (Garzón-Orduña & Miranda-Esquivel, 2007; Nihei & Carvalho, 2007). Some studies suggest a composite nature where southeastern Amazonian taxa are historically linked to eastern Brazilian taxa, and northwestern Amazonian taxa are more proximally related to Central American, Caribbean and Chocoan ones (Amorim & Pires, 1996; Amorim, 2001; Ribas & Miyaki, 2004; Nihei & Carvalho, 2007). Other studies are supportive of the Amazon as a single historical unit, that is, with southeastern and northwestern areas sister to each other (Silva & Oren, 1996; Bates et al., 1998; Eberhard & Bermingham, 2005). As far as we are concerned, the historical separation of the Amazon in these two blocks (southeastern and northwestern) has never been tested with plant phylogenetic studies, but seems to agree, at least spatially, with the extension of a middle Miocene marine transgression that separated the Guiana and Brazilian shields (Räsänen et al., 1995; Webb, 1995; Nores, 1999). The hypothesis that the biota of the southeastern portion of the Amazon region is more closely related to that of the central Brazilian forests than to the northwestern Amazonian ones has already been suggested by floristic analyses (e.g., Oliveira-Filho & Ratter, 1995; Ivanauskas et al., 2008), and remains to be tested with plant phylogenetic studies.
Regardless of the methodology used, raw distributions or taxon phylogenies, relationships among Amazonian areas of endemism seem to show some degree of congruence (Ron, 2000; Hall & Harvey, 2002). The summary area cladogram (Guiana + (Rondônia + (Pará+ Belém))) + (Imeri + (Napo + Inambari)) presented by Hall and Harvey (2002) seems to be a good working hypothesis, because it combines data from several unrelated groups of organisms. According to this hypothesis two Amazonian blocks (S/SE and W/NW) are sister groups and, as a whole, sister to Guiana. Alternative scenarios have been proposed, especially regarding the position of Rondônia as part of the W/NW block (Cracraft & Prum, 1988; Eberhard & Bermingham, 2005), and Guiana as part of the S/SE block (Racheli & Racheli, 2004; Eberhard & Bermingham, 2005; Garzón-Orduña & Miranda-Esquivel, 2007). Large-scale phylogenetic studies of plant taxa with widespread distribution in the Amazon basin, but with limited dispersal ability, could be designed to test if these area relationships are supported.
The Cerrado domain originally covered ca. 2 million km2 of the central Brazilian Plateau, extending west into Bolivia, south to Paraguay, and east to the Caatinga, and with some isolated patches found scattered across the Amazonian and Atlantic forests, and the Caatinga (Prance, 1996; Daly & Mitchell, 2000). Most of the cerrado vegetation is characterized by savanna physiognomies with a grass-rich ground layer growing on nutrient-poor soils with high aluminum content (Eiten, 1972; Ratter et al., 1997, 2006). The woody flora is mostly composed of sclerophyllous evergreen plants adapted to periodic fires, and with well-developed root systems reaching underground water tables. Depending on the density of the woody component, the structure of the vegetation can vary from open grassland to forest with a closed canopy (Silva & Bates, 2002; Ratter et al., 2003, 2006).
The Cerrado flora is very rich, with an estimated vascular plant diversity ranging from 6,429 to approximately 10,500 species (Mendonça et al., 1998; Ratter et al., 2006). Approximately 35% of the trees and 70% of the herbaceous and shrubby plants that grow in the Cerrado are endemic to this domain (Pennington et al., 2006a; Ratter et al., 2006), whereas most of the non-endemic species are associated with the Atlantic Forest domain (Méio et al., 2003; Ratter et al., 2006; but see Gonçalves, 2004 for an Amazonian connection). Besides the high levels of diversity and endemism, the Cerrado has sufferred a high degree of disturbance, especially due to agricultural expansion, cattle ranching, and charcoal production (Ratter et al., 1997; Silva & Bates, 2002). There are estimates that less than 20% of the cerrado vegetation remains undisturbed, which has resulted in its recommendation as a biodiversity hotspot (Myers et al., 2000; Mittermeier et al., 2005).
Floristic comparisons among several sites in the Cerrado domain have led to the recognition of seven floristic provinces based on the presence/absence of woody taxa (Ratter et al., 1996, 2003, 2006). These studies have pointed to a previously unrecognized floristic heterogeneity mostly associated with soil type and geographic location. A few taxonomic based biogeographic studies (e.g., Simon & Proença, 2000; Fiaschi & Pirani, 2008) seem to corroborate these patterns, and phylogeographic studies of endemic Cerrado plant species are suggestive of genetic structuring among these floristic provinces following Quaternary climatic changes (e.g., Ramos et al. 2007).
The highest levels of endemism in the Cerrado domain are found along the mountains of the Espinhaço Range (Minas Gerais and Bahia states), and the Chapada dos Veadeiros (Goiás state) (Prance, 1994; Simon & Proença, 2000; Silva & Bates, 2002; Fiaschi & Pirani, 2008). Most of the endemic species grow in areas above 900–1000 m along these mountains, which are covered by a low, mostly herbaceous or shrubby vegetation on sandy or stony soils called campos rupestres (Giulietti & Pirani, 1988; Harley, 1995; Alves et al., 2007). The high level of endemism in the campos rupestres flora has been recognized by several authors (Joly, 1970; Giulietti & Pirani, 1988; Harley, 1995; Rapini et al., 2002), but the explanations proposed for these diversity patterns remain very speculative because most of the relevant studies lack phylogenetic hypotheses for the endemic taxa.
Studies exploring diversification patterns among endemic Cerrado plants using dated molecular phylogenies are just beginning. A few studies suggest recent (3–4.7 mya) diversification events in high-altitude clades including Viguiera (Asteraceae), Microlicieae (Melastomataceae), and Minaria (Apocynaceae) (Schilling et al., 2000; Fritsch et al., 2004; and Rapini et al., 2007, respectively). Further phylogenetic studies of the endemic flora of the Cerrado could provide additional data to evaluate whether most of the endemic flora is the result of recent radiations, as suggested by Pennington et al. (2006b). Several angiosperm groups are good candidates for this goal, such as: Eremanthus, Lychnophora and Richterago (Asteraceae), Encholirium (Bromeliaceae), Kielmeyera (Clusiaceae), Eriocaulaceae, Pseudotrimezia (Iridaceae), Eriope (Lamiaceae), Chamaecrista and Mimosa (Leguminosae), Diplusodon (Lythraceae), Byrsonima (Malpighiaceae), Microlicia and Trembleya (Melastomataceae), Sauvagesia (Ochnaceae), Declieuxia (Rubiaceae), Barbacenia and Vellozia (Velloziaceae).
2.3 Atlantic forest
The Atlantic forests originally occupied approximately 1.5 million square kilometers, extending from Rio Grande do Norte to Rio Grande do Sul states along the Brazilian coast. The width of this forest strip is very variable, and it extends far inland in some areas of southeastern Brazil, eastern Paraguay, and Misiones Province of Argentina (Galindo-Leal & Câmara, 2003; Oliveira-Filho et al., 2006). The Atlantic coastal forests are separated from the Amazonian forests by a northeast–southwest diagonal swath of open or dry formations (Prado & Gibbs, 1993; Prado, 2000; Silva et al., 2004), which are believed to act as a current barrier to floristic exchange between these two forest blocks (Mori et al., 1981; but see Oliveira-Filho & Ratter, 1995; Costa, 2003). Because the Atlantic forests harbor a unique biota that is very rich in endemic species, and because of the high level of habitat destruction the region has been suffering – only approximately 7.5% of the original vegetation remains – it is considered one of the world's priorities for biodiversity conservation (Myers et al., 2000; Mittermeier et al., 2005).
The Atlantic Forest domain is characterized mostly by evergreen tropical forest but SDTFs (Oliveira-Filho et al., 2006) and subtropical forests (the Paranaense Province of Cabrera & Willink, 1973, including Araucaria forest) are also usually considered part of the domain (Oliveira-Filho & Fontes, 2000; the Parana Subregion of Morrone, 2006). In addition to forest physiognomies, mangroves and shrubby restinga vegetation are widespread in sea-level sandy areas (Scarano, 2002), and patches of high altitude grasslands and rocky outcrops are usually found above 2000 m along the Serra do Mar and Serra da Mantiqueira mountain ranges (Safford, 1999, 2007). The archipelago-like open formations found along the montane areas of the Atlantic forests harbor a highly endemic flora (over 20%) that has strong floristic connections with other South American montane areas such as the Andes (Safford, 2007), and the aforementioned campos rupestres of the Espinhaço Range (Giulietti & Pirani, 1988; Di Maio, 1996; Safford, 1999; Calió et al., 2008). The fragmented distribution of these open areas, coupled with episodic dispersal events, appear to have favored allopatric speciation in Sinningieae (Gesnericaeae) (Perret et al., 2007), and suggest that habitat heterogeneity may have played an important role in the diversification of the endemic Atlantic forest flora.
Vascular plant diversity and endemism in the Atlantic rainforests are among the highest in the world (Martini et al., 2007), but information on the geographic distribution of many taxa is lacking. There are approximately 20,000 species of vascular plants in the Atlantic forests (Myers et al., 2000), with estimated levels of endemism varying from 33% of the pteridophytes to more than 81% of the 803 species of bromeliads (Martinelli et al., 2008), and 41.6–44.1% of the total number of vascular plants from two reserves at southern Bahia (Thomas et al., 1998). Among the angiosperm genera, 159 are endemic to the Atlantic forests and approximately half of those are monotypic (Stehmann et al., in press). Many others groups are represented better there than anywhere else in the Neotropics, such as Hornschuchia (Annonaceae), many Bromeliaceae genera, Nematanthus and Sinningia (Gesneriaceae), Huberia and Pleiochiton (Melastomataceae), Dorstenia (Moraceae), Calyptrogenia and Myrceugenia (Myrtaceae), Oxalis subg. Thamnoxys (Oxalidaceae), Atractantha and Merostachys (Poaceae), Coccocypselum (Rubiaceae), Conchocarpus and Galipea (Rutaceae).
In addition to its highly endemic flora, the Atlantic forests harbor early diverging lineages of some angiosperm groups, such as the Poaceae subfamily Anomochlooideae (Judziewicz & Clark, 2007), Goniorrhachis and Barnebydendron, which correspond to the first diverging branches of the Detarieae clade of Leguminosae (Bruneau et al., 2008), and the Harleyi clade of Pagamea (Rubiaceae), which point to a possible arrival through dispersal from African ancestors (Vicentini, 2007). Based on the presence of presumably “primitive” species in the Atlantic forests, Gentry (1982) suggested that it could be a “source area” of Gondwanan taxa for other phytogeographic regions, such as the geologically recent Amazon lowlands and the Andes, although Prance (1982) discounted the floristic contribution of the Atlantic forests to the Amazon flora due to their early isolation. The basal placement of Atlantic forest taxa in several phylogenetic hypotheses of Neotropical organisms seems to support the contribution of early diverging lineages to the Atlantic Forest biota (Cracraft & Prum, 1988; Bates et al., 1998; Eberhard & Bermingham, 2005). In other cases, however, the presence of some lineages in these forests seems to result from more recent colonization from other South American source areas (Cracraft & Prum, 1988; Costa, 2003; Vicentini, 2007). Thus, as suggested by Silva and Castelletti (2003) and Pennington et al. (2006b), it seems more prudent to view the Atlantic Forest biota as having contributions from both old and recently diverged lineages.
Multiple centers of endemism based on several groups of organisms have been proposed for the Atlantic Forest domain (Prance, 1982; Cracraft, 1985; Soderstrom et al., 1988; Costa et al., 2000; Silva et al., 2004; Santos et al., 2007). The number of such centers varies depending on the organisms under consideration and the specific questions being addressed, both of which influence the selection of areas. Thus, although some studies point to a northern/southern separation in just two blocks (Cracraft, 1985; Soderstrom et al., 1988), finer-scale studies using low-vagility organisms suggest more numerous and smaller areas (e.g., Pinto-da-Rocha et al., 2005). Regardless of the study group and methodology used, most studies agree that there is an historical separation between the northern and southern parts of the domain, whose limits are more or less coincident with the Rio Doce valley (northern Espírito Santo state) (Cracraft & Prum, 1988; Amorim & Pires, 1996; Costa, 2003; Silva et al., 2004; Pinto-da-Rocha et al., 2005; Perret et al., 2006). Several examples of plant taxa restricted to either one of these areas are known, resulting in a strong floristic differentiation between the northern and southern Atlantic forests (Oliveira-Filho & Fontes, 2000; Oliveira-Filho et al., 2005).
The northern Atlantic forest (NAf) ranges from Rio Grande do Norte (ca. 5° S) to northern Espírito Santo (ca. 19° S) states, and comprises mostly a narrow strip of forest bounded to the west by the Caatinga domain (Thomas & Barbosa, 2008), as well as some inland areas, such as the “brejos nordestinos” (Rodal & Sales, 2008) and the Chapada Diamantina forests (Funch et al., 2008). Two centers of endemism are usually recognized in the NAf: Pernambuco (ca. 8° S), and Bahia (“central corridor”, from approximately 13° to 19° S) (Thomas et al., 1998). The NAf shows some floristic influence from the Amazonian forests, presumably due to historical connections through the Cenozoic (Rizzini, 1963; Andrade-Lima, 1966; Prance, 1979; Mori et al., 1981; Costa, 2003). Although the extension of these forest connections are unknown, three main routes for floristic exchange between the Amazonian and Atlantic rainforests have been proposed (Costa, 2003): a southern route through the Paraná River basin, a northeastern route through the Caatinga domain (Rizzini, 1963; Andrade-Lima, 1966), and a route by way of gallery forests across the central Brazilian cerrado (Oliveira-Filho & Ratter, 1995). The closer biogeographic relationship of the Pernambuco center plus the brejos nordestinos to the Amazon rather than to southern Atlantic forests is supportive of the northeastern route (Prance, 1979, 1982; Silva et al., 2004; Santos et al., 2007). In addition, evidence is provided by the disjunct occurrence of several genera in the NAf and the Amazonian forests that are lacking in the southern Atlantic forests such as Lacmellea and Macoubea (Apocynaceae), Anthodiscus (Caryocaraceae), Glycydendron (Euphorbiaceae), Gustavia and Lecythis (Lecythidaceae), Macrolobium and Parkia (Leguminosae), Roucheria (Linaceae), Adelobotrys and Graffenrieda (Melastomataceae), Anomospermum and Orthomene (Menispermaceae), Naucleopsis and Pseudolmedia (Moraceae), Aptandra (Olacaceae), Atractantha and Pariana (Poaceae), and Pagamea and Remijia (Rubiaceae).
The southern part of the domain [southern Atlantic forests (SAf)] ranges from Espírito Santo (ca. 19° S) to southern Santa Catarina (ca. 29° S), and it includes a large western extension of seasonally dry forests in southeastern Brazil, eastern Paraguay, and Misiones in Argentina (Oliveira-Filho & Fontes, 2000; Oliveira-Filho et al., 2006), and the Araucaria angustifolia Forest Province (Morrone, 2006). The seasonally dry forests of the Paranaense and Misiones nuclei are currently considered a distinct phytogeographic unit (Prado, 2000; Pennington et al., 2006a), whereas the subtropical Araucaria forests are sometimes considered on their own as the Paranaense Province (Cabrera & Willink, 1973) or the Araucaria angustifolia Forest Province (Morrone, 2006). The SAf block comprises an extensive center of endemism that seems to coincide well with the Serra do Mar and Serra da Mantiqueira mountain ranges (Prance, 1982; Silva et al., 2004; Pinto-da-Rocha et al., 2005). Here we adopt the delimitation of the Serra do Mar center of endemism as proposed by Silva et al. (2004), which is congruent with a group of historically related areas ranging from southern Espírito Santo to northern Santa Catarina (Pinto-da-Rocha et al., 2005). Instead of sharing a high number of taxa with Amazonia, the SAf seems to be influenced more strongly by elements of other regions. As an example, some Andean-centered taxa can be found in the SAf but are usually absent in the NAf (A. Amorim, pers. comm., 2009), such as Oreopanax (Araliaceae), Clethra (Clethraceae), Gaultheria (Ericaceae), Escallonia (Escalloniaceae), Gordonia (Theaceae), Macrocarpaea (Gentianaceae), Hypericum (Clusiaceae), Meriania (Melastomataceae), Calyptrogenia and Myrceugenia (Myrtaceae), Fuchsia (Onagraceae), Aulonemia, Chusquea and Colanthelia (Poaceae), Euplassa (Proteaceae), Meliosma (Sabiaceae), and Valeriana (Valerianaceae). As discussed in section 1, other floristic elements of the SAf are probably remnants of a southern Gondwanan land connection (Sanmartín & Ronquist 2004), such as A. angustifolia, Canellaceae, Weinmannia (Cunoniaceae), Crinodendron (Elaeocarpaceae, Crayn et al., 2006), Griselinia (Griseliniaceae), Podocarpus (Podocarpaceae), some Proteaceae (Barker et al., 2007), and Drimys (Winteraceae).
The floristic differences between the northern and southern blocks of the Atlantic forests are supported by the available phylogenetic data. Most biogeographic studies point to the Atlantic Forest domain as a composite biogeographic area where the southern and northern areas are not sister groups (e.g., Cracraft & Prum, 1988; Costa, 2003; Perret et al., 2006; Nihei & Carvalho, 2007; Santos et al., 2007), however, in others the same two blocks form a monophyletic Atlantic Forest (e.g., Amorim & Pires, 1996; Costa et al., 2000). Although rare, the exchange of floristic elements between the NAf and SAf has been proposed by Perret et al. (2006) and seems to account for the presence of typical elements of the SAf in montane areas of southern Bahia (A. Amorim, pers. comm., 2009) and the Chapada Diamantina (Funch et al., 2008).
The Caatinga domain of northeastern Brazil is the largest continuous area of SDTFs of South America, originally covering an area of approximately 850,000 km2 (Queiroz, 2006). The vegetation varies from an open thorny scrub to low dry forests; it is conditioned by a prevailing semiarid climate, with high evapotranspiration potential (1500–2000 mm/year) and low precipitation (300–1000 mm/year) concentrated during a short period of 3–5 months (Sampaio, 1995; Queiroz, 2006). Floristic diversity in the Caatinga is relatively low (Sampaio, 1995), especially when compared to that of the Atlantic rainforests and the Cerrado (e.g., Castro et al., 1999; Myers et al., 2000); however, 46% endemic species have been reported for the tree flora (Pennington et al., 2006a), and 52.5% for the family Leguminosae (144 of 274 species endemic to this domain, Queiroz, 2006). Likewise, Giulietti et al. (2002) have listed 18 angiosperm genera and 318 species as endemic to the Caatinga.
The floristic composition of the Caatinga shows strong links with other nuclei of South American SDTFs, such as Misiones, Piedmont, the Caribbean coast of Colombia and Venezuela, and the dry inter-Andean valleys, but not to the mostly subtropical South American dry forests of the Chaco domain (Prado & Gibbs, 1993; Prado, 2000; Pennington et al., 2000, 2006a). The recent recognition that SDTFs represent an archipelago-like biogeographic unit (Prado, 2000) has stimulated interest in the historical processes that may have shaped current distributions of SDTF-centered taxa (Pennington et al., 2000, 2004; Lavin, 2006; Ritz et al., 2007), not to mention their importance as repository areas for the conservation of the highly threatened Neotropical dry forest flora (Gentry, 1995).
Few studies have specifically focused on the Caatinga (Pennington et al., 2006a). In a first attempt to examine the historical biogeography of the Caatinga, Queiroz (2006) used distribution data of Leguminosae species, which account for approximately one-third of the total number of plant species found in the biome (Giulietti et al., 2002). By plotting the known geographic distributions of 274 species in the area occupied by the biome, he found that despite sharing an overall physiognomic similarity, the flora of the Caatinga could be divided into two distinct floristic blocks. One of these blocks is associated with soils derived from the crystalline basement and accounts for most of the floristic link of the Caatinga with the remaining areas of SDTFs (Prado, 2000). Genera characteristic of these areas include Amburana, Apuleia and Pterogyne (Leguminosae), as well as Balfourodendron (Rutaceae), Quiabentia (Cactaceae), Astronium (Anacardiaceae), and Patagonula (Boraginaceae) (Queiroz, 2006). The second group corresponds to plants growing on sandy sedimentary areas scattered across the domain, and these accounts for most of the endemic taxa of Leguminosae found in the Caatinga.
The historical development of this scenario was proposed as the result of a widespread process of pediplanation during the early Quaternary that uncovered the Precambrian crystalline bedrock in the region. According to this hypothesis, the earlier continuous sedimentary surfaces became dissected, leading to allopatric differentiation of their taxa. The exposed areas derived from the crystalline bedrock were occupied by elements typical of the SDTF flora (Queiroz, 2006), which were postulated as having evolved during the Miocene–Pliocene (Pennington et al., 2004; Saslis-Lagoudakis et al., 2008) and may have migrated later to the Caatinga. Phylogenetic analyses of taxa with multiple species centered in the Caatinga would be critical to test this scenario. Some good candidates for this purpose are groups within Croton and Jatropha (Euphorbiaceae), several Leguminosae genera such as Vachellia (Acacia s.l.), Aeschynomene, Bauhinia, Calliandra, Centrosema, Chamaecrista, Lonchocarpus, Macroptilium, Mimosa, Senna, Stylosanthes, and Zornia, as well as genera with several species endemic to the Caatinga, such as Melocactus and Pilosocereus (Cactaceae), Apodanthera (Cucurbitaceae), Hyptis (Lamiaceae) and Piriqueta Aubl. (Turneraceae) (Giulietti et al., 2002).
2.5 Campos sulinos (southern grasslands)
Extensive areas of southern Brazil are covered by open grassy formations generally called campos, which have been used as natural pastures (see Overbeck et al., 2007 for an overview on South Brazilian campos). In Brazil alone it is possible to distinguish between the campos do Planalto Meridional, which have a patchy occurrence within the Atlantic Forest domain and range from Paraná to northern Rio Grande do Sul states (Araucaria angustifolia Forest Province of Morrone, 2006), to the continuous pampas or campos da Campanha Gaúcha, which covers the largest part of Rio Grande do Sul and neighboring areas of Uruguay and Argentina (Pampa Province of Morrone, 2006). The patchy distribution of the campos do planalto is a consequence of its dynamics with the Araucaria Forest, which tends to advance over the campos– young Araucaria plants cannot grow in shade – with forest expansion being controlled by fire and other human activities (Klein, 1960; Behling et al., 2004; Overbeck et al., 2007).
Within the pampas, some authors further separate the southern Brazilian and Uruguayan grasslands (north of Rio da Prata) from the Argentinian pampas based on floristic differences (e.g., Soriano et al., 1992). The word “pampa” itself has a Quechua origin meaning “flat region.” In this sense, it includes all low, flat areas of the Rio da Prata basin northward to the Serra Geral. In this section we use the term pampas in this broad definition, which agrees with Morrone's Pampa Province (Morrone, 2006).
Floristic and physiognomic differences between the campos do planalto and the pampas are evident. Among the grasses, the relative importance of megathermic groups, such as Andropogoneae, Chlorideae, Eragrosteae and Paniceae in the campos do planalto is much higher than in the pampas, which in turn have a higher contribution of microthermic elements, including Agrostis, Aristida, Briza, Bromus, Calamagrostis, Danthonia, Piptochaetium, and Stipa (Burkart, 1975; Longhi-Wagner & Zanin, 1998; Boechat & Longhi-Wagner, 2000; Overbeck et al., 2007). In fact, the grass flora of the pampas consists of a mixture of both megathermic and microthermic grasses, which have distinct phenological phases (Burkart, 1975). Also noteworthy is the common presence of palms (mostly small species of Butia) in the campos do planalto, and its rarity or absence in the pampas.
Plant diversity in the Campos sulinos has been estimated as between 3,000 and 4,000 species (Overbeck et al., 2007), and some studies point to these grasslands as among the most species-rich in the world (Overbeck et al., 2006). The flora is dominated by species of sedges and grasses, but also includes shrubs and subshrubs of several families, such as Apiaceae (mostly Eryngium), Asteraceae (several species of Baccharis), Leguminosae, Myrtaceae, Malvaceae, Oxalidaceae and Rubiaceae (Joly, 1970; Soriano et al., 1992; Overbeck et al., 2007). The flora of the Campos sulinos in general has links to other open formations of South America. Several elements from the herbaceous flora of the Cerrado domain have their southern limits in the Campos sulinos (e.g., Boechat & Longhi-Wagner, 2000), where many elements of temperate/subtropical floras are fairly common. At the same time, some of these southern elements have their northernmost occurrence in this same region (e.g., Longhi-Wagner & Zanin, 1998). The apparent transitional nature of the Campos sulinos flora is even more evident in the state of Rio Grande do Sul, where the campos de planalto are replaced by the pampas at about 30° S (Smith, 1962; Burkart, 1975; Waechter, 2002; Ritter & Waechter, 2004; Overbeck et al., 2007).
Many examples of Andean-derived taxa are known from southern Brazil (Rambo, 1956; Smith 1962). A good example of connection with the Andean flora is provided by the presence of Gunnera manicata (Gunneraceae) in the wet hills and swamps of the high eastern planaltine areas of Santa Catarina and Rio Grande do Sul states. Dispersals of Andean genera both through Argentina to southern Brazil (Klein, 1960; Leite, 2002) and from the Paraná basin grasslands to the Andes (Katinas & Crisci, 2008) have been reported. The resulting disjunct pattern (see section 3.2.2.) is supported not only by similar mild climatic conditions, but also by former Quaternary connections that could have taken place during interglacial wetter periods (Ortiz-Jaureguizar & Cladera, 2006).