EDINB. J. BOT. 52(2): 141-194(1995) 141 A STUDY OF THE ORIGIN OF CENTRAL BRAZILIAN FORESTS BY THE ANALYSIS OF PLANT SPECIES DISTRIBUTION PATTERNS A. T. OLIVEIRA-FILHO* & J. A. RATTERf The floristic nature of central Brazilian forests, as well as their links to other main forest formations of eastern tropical South America, is assessed by means of multivariate analyses of 106 existing floristic checklists and by the analysis of a series of dot-maps showing the distribution of 55 woody species. Most species of central Brazilian forests seem to conform to two main distribution patterns: (1) species of deciduous and semideciduous forests are dependent essentially on the occurrence of patches of soils of intermediate to high fertility within the cerrado domain and tend to be distributed mostly along a northeast-southwest arch connecting the caatingas to the chaco boundaries; (2) considerable numbers of gallery forest species are dependent on high soil moisture and many appear to link the Amazonian and Atlantic rainforests by crossing the cerrado region in a northwest-southeast route via the dendritic net of gallery forests. Many gallery forest species are habitat generalists, some even occurring in the cerrado vegetation. Nevertheless, significant numbers of gallery endemics are also present. Galleries of the west and north of the Cerrado Province show stronger floristic links to the Amazonian rainforests, while those of the centre and south show stronger affinity with the montane semideciduous forests of southeastern Brazil. Present-day distribution patterns are discussed in the light of the current knowledge of palaeoenvironmental changes in the Neotropics. A natureza floristica das florestas do Brasil Central bem como seus lacos com outras grandes formacoes florestais da America do Sul tropico-oriental e avaliada por meio de analises multivariadas de 106 listagens floristicas existentes e pela analise de uma serie de mapas de distribuicao geografica de 55 especies lenhosas. A maioria das especies das florestas do Brasil Central parece ajustar-se a dois grandes padroes de distribuicao: (1) especies de florestas deciduas e semideciduas dependem essencialmente da ocorrencia de manchas de solos de media a alta fertilidade dentro do dominio dos ORIGIN OF CENTRAL BRAZILIAN FORESTS 143 Guayana Province Amazonian Province Atlantic Province Paranense Province Caatinga Province Chaco Province Brazilian savannas FIG. 1. Map showing the distribution of Brazilian savannas (cerrados), according to IBGE (1993), superimposed on the seven phytogeographic provinces of eastern tropical South America, according to Cabrera & Willink (1973). & Colinvaux, 1985; Colinvaux, 1989; Markgraf, 1989; Busch & Colinvaux, 1990; Busch et al., 1990; Absy et al., 1991). It was only recently that Ledru (1991, 1993) contributed the first palaeopalynological study for the Cerrado Province of central Brazil, carried out at Salitre in western Minas Gerais. She concluded from the analysis of pollen distribution and modern vegetation and climate that between 33,000 and 25,000 BP the region was moister than today and was covered by rainforests (Ledru, 1993). During the Pleistocene glacial maximum, from 17,000 to 13,000 BP, a drier climate predominated and the proportion of trees in the vegetation declined, although only in the final term did arboreal cover reduce to the extent of a forest retreat. In another recent study, based on the analysis of plant species distribution, Prado & Gibbs (1993) proposed that during the Pleistocene glacial maximum the present-day corridor of xeric vegetation was covered mostly by an extensive seasonal woodland formation. This agrees with the findings of Ledru (1993) at Salitre (right in the middle of the corridor) who suggested the predominance of a seasonal arboreal 144 A. T. OLIVEIRA-FILHO & J. A. RATTER vegetation during most of the Pleistocene, when the climate would be drier than now but not severely arid as previously thought (Brown & Ab'Saber, 1979; Bigarella & Andrade-Lima, 1982). According to Ledru (1993), the period following the last Pleistocene glacial maximum was characterized by the expansion of the Araucaria forest, indicating the predominance of a cooler and moister climate until 8500 BP. This period suffered a sudden and short-term interruption between 11,000 and 10,000 BP, corresponding to a dry interval accompanied by forest retreat. After 8500 BP the Araucaria forest was replaced by mesophytic forest, indicating an episode of moist climate with higher temperatures lasting until 5500 BP. The mid-Holocene dry interval which ensued at about 5000 BP caused a severe reduction in the forest area; since then forest expansion has once more occurred but not enough to recover the whole extent of the previous mesophytic forest period. Some authors believe that disturbances, both anthropogenic and natural (fire in particular), helped retard forest expansion within the Cerrado Province during the late Holocene, despite the fact that climate and soils certainly favour the establishment of forests in most of the region (Reis, 1971; Klein, 1975). The isolated areas of cerrado (savanna) that are found enclosed by the Amazonian forest (see Fig. 1) provide evidence both of the forest expansion through a previously cerrado-dominated landscape and of the maintenance of cerrado vegetation in areas where the climatic conditions could support forest vegetation (Bigarella et al., 1975). It is now clear that the effects of the late Quaternary climatic fluctuations on the open vegetation formations of central Brazil were not restricted to a simple expansion and contraction of cerrado vs. rainforest, but also included complex changes within the dry corridor region itself. Cerrados, seasonal forests, rainforests, and even Araucaria forests appear to have thrived in various epochs at least in certain areas of the region. In addition to climatic variation, other factors such as soils, biotic pressure and disturbance could also have had profound palaeoenvironmental effects, and further data are necessary to provide a reliable description of the past history of the vegetation cover of central Brazil. The dry corridor of open formations has been considered by many authors as an important impediment to species migration between the two main South American forest regions, thus explaining many of the differences in floristic composition between them. On the other hand, the disjunct distribution patterns of a considerable number of species in the coastal Atlantic and Amazonian rainforests suggest the possibility of past forest links (Rambo, 1961; Rizzini, 1963; Andrade-Lima, 1966). Andrade-Lima (1964) and Rizzini (1963) refer to a possible migration route through a postulated mesophytic forest corridor that would have traversed the present-day Caatinga Province of northeastern Brazil at certain periods since the late Tertiary. Andrade-Lima (1982) suggested that the montane forests ('brejos') that presently exist isolated within the semi-arid region are relics of an ancient and wider forest cover. Bigarella et al. (1975) argued that, in addition to this Northeast Bridge, the floristic similarity between the eastern Amazonian and southeastern Atlantic rainforests demonstrates that there must have been a second migration route, which they called the Southeast-Northwest Bridge. ORIGIN OF CENTRAL BRAZILIAN FORESTS 145 Although Bigarella et al. (1975) did not themselves suggest how the SoutheastNorthwest Bridge connected the two forest regions, one could postulate that such a connection could have run through central Brazil, either as a continuous forest corridor or as a series of forest patches between which 'island-hopping' occurred. The idea of the Cerrado Province acting as an effective obstacle to the migration of forest species seems rather doubtful when a detailed map of present-day vegetation is examined. The map of Brazilian savannas (cerrados) (Fig. 1), extracted from IBGE (1993), clearly shows that the distribution of cerrados in central Brazil is far from continuous. When this is superimposed on the Cerrado Province of Cabrera & Willink (1973) a great proportion of gaps is shown (Fig. 1). These correspond essentially to patches of semideciduous forests, deciduous forests and to the transitional mosaics between these forests and the cerrado, called 'areas of ecological tension' by IBGE (1993). In addition to this, the savanna (cerrado) vegetation formation in the IBGE's classification system includes gallery forests, which are ubiquitous throughout the cerrado region. Therefore, forests are a far from negligible component of central Brazilian vegetation and their role as migration routes for forest species cannot be neglected, especially when evidence exists of a wider forest cover during a number of episodes in the late Quaternary. In fact, even now, the occurrence of mesophytic forest (deciduous and semideciduous forests) within the cerrado biome is very extensive and generally underestimated; during a recent journey by road from the Distrito Federal to Estreito in Maranhao, crossing c. 1400km of the states of Goias and Tocantins, one of us (J.A.R.) estimated that probably nearly 50% of the vegetation traversed was degraded mesophytic forest or the closely related mesotrophic cerradao. The deciduous forests that occur within the Cerrado Province on patches of calcium-rich soils have already been regarded by Prado & Gibbs (1993) as an important 'island-hopping' migration route connecting the semi-arid caatingas to the semideciduous forests of the Parana basin and of the Andean flanks. If such a NE-SW route existed in central Brazil for deciduous calciphilous species, the existence of a NW-SE route for species associated with higher soil moisture is also conceivable through gallery forests. A number of authors have already suggested that the central Brazilian gallery forests represent floristic intrusions of the Amazonian and/or Atlantic forests into the cerrado domain. Cabrera & Willink (1973) mentioned that the galleries are actually a net connecting the forests of the Amazonian and Paranense Provinces. Based on floristic similarities, Rizzini (1979) stated that the galleries, together with other hinterland forests, are 'mediterranean extensions of the great Atlantic forest'. He also added that in Mato Grosso the continuous link to the Hylaea would give the galleries an Amazonian floristic nature. In fact, a considerable number of Amazonian species are known to penetrate the cerrados via gallery forests in that state, including species of rubber-tree (Hevea), a genus which is often reported to define the borders of the Amazonian forests (Veloso, 1966; Pires & Prance, 1977; Daly & Prance, 1989). Pires (1984) argued that if forest refuges did exist in the Amazon during the glacial ages they would not have had the postulated island 146 A. T. O L I V E I R A - F I L H O & J. A. RATTER configuration, but a dendritic pattern following river drainage. This author also suggested that present-day galleries within the cerrado region constitute hodiernal forest refuges. In the present contribution we investigate the floristic nature of central Brazilian forests by the analysis of present-day distribution patterns. Our work was based essentially on multivariate analyses of existing floristic checklists and on the analysis of a series of dot-maps showing the distribution of woody species. The links to other main forest formations were assessed, with special attention to the connections to the Amazonian and Atlantic forests. Although we are aware of the many risks of drawing conclusions from the still poorfloristicrecord, we believe that major patterns did emerge from our analyses and that these added an important contribution to the discussion of the past evolution of central Brazilian ecological spaces. In addition, we believe that the results contribute valuable information on the pattern of biodiversity in the region and that this type of information is essential in the planning of conservation areas. MATERIALS AND METHODS Selection offloristicchecklists One of the main purposes of this study was to assess the patterns offloristicdifferentiation of central Brazilian forests, both among the many forest types of the region and in relation to the forests of neighbouring vegetation provinces. To that end, a total of 106floristicchecklists were obtained from literature and other sources available in the Royal Botanic Gardens of Edinburgh and Kew in 1992-94. These were selected from floristic surveys of the woody flora of forests of eastern tropical South America, with special emphasis on central Brazilian forests. To ensure accuracy, only surveys with at least 85% of the taxa identified to species level were used. Unfortunately, standardization by area size proved impossible as this varied widely. However, 'regional lists' (those for areas of thousands of hectares) were included only for non-central Brazilian forests, which were incorporated in the dataset as paradigms. The 106floristicchecklists comprised only areas which could be classified physiognomically as forests or dense woodlands, i.e. predominantly arboreal, although sometimes with an incomplete canopy. Therefore, the array of vegetation types ranged from dense Amazonian rainforests to cerradoes (forested savannas). The 106 forest areas chosen extended from the state of Roraima, in northern Brazil, to Chaco and Santa Fe Provinces, in northeastern Argentina, and from the state of Pernambuco, in eastern Brazil, to Pando Province, in northern Bolivia. Their location is shown in Fig. 2 while Table 1 gives summarized information for each area. Classification of the vegetation types The environmental information available for the 106 forest areas was either very poor or hardly comparable due to the miscellaneous methods used (especially for ORIGIN OF CENTRAL BRAZILIAN FORESTS 147 FIG. 2. Map of Brazil and neighbouring countries showing the location of the 106 forest areas used in the analyses. The forest areas are identified by their number in Table 1, followed by the two-letter code corresponding to their classification according to the modified IBGE's classification system for Brazilian vegetation (see text, pp. 154-5). soil data). However, an environmental background was necessary to make sense of the results emerging from the multivariate analyses. So, we decided to apply to the forest areas a vegetation classification system highly dependent on both physiognomy and environmental conditions and to assess the classification categories in the multivariate analyses. The choice of a single well-defined classification system which suited our needs also allowed skirting the byzantine discussion about the best nomenclature for Brazilian vegetation. We chose the IBGE classification system for Brazilian vegetation (Veloso et al., 148 A. T. OLIVEIRA-FILHO & J. A. RATTER TABLE 1. Vegetation formation, locality, geographical coordinates, altitude, climatic type and source for each of the 106 forest areas used in the analyses. The two-letter codes after the identification numbers correspond to the classification of the forest areas according to the modified IBGE classification system for Brazilian vegetation (see text, pp. 154-5), which is also given in full in the following column. The double capitals after the localities are the official abbreviations of Brazilian states. Latitudes are South unless indicated. Altitudes are medians where a range was given in the source. An HS after the reference indicates that herbarium specimens were also used in the preparation of the checklist. Ident. no. Vegetation formation (modified IBGE system) Locality 9 Sm 10 Sd Gallery forest Submontane open rainforest Dystrophic cerradao Submontane deciduous forest Mesotrophic cerradao Gallery forest Submontane deciduous forest Submontane deciduous forest Mesotrophic cerradao Dystrophic cerradao Serra do Roncador (MT) Serra do Roncador (MT) Serra do Roncador (MT) Serra do Roncador (MT) Serra do Roncador (MT) Rio Suia-Missu (MT) Vale dos Sonhos (MT) Torixoreu (MT) Torixoreu (MT) Chapada dos Guimaraes (MT) 11 Gf Gallery forest Chapada dos Guimaraes (MT) 12 Sm 13 Sm 14 Sm 15 Cb 16 Fa 17 Cb 18 Fa 19 Sm 20 Cb 21 Cb Mesotrophic cerradao Mesotrophic cerradao Mesotrophic cerradao Lowland deciduous forest Alluvial semideciduous forest Lowland deciduous forest Alluvial semideciduous forest Mesotrophic cerradao Lowland deciduous forest Lowland deciduous forest Rio Claro, Cuiaba (MT) Pocone, Pantanal (MT) Faz. Acurizal, Pantanal (MT) Faz. Acurizal, Pantanal (MT) Faz. Acurizal, Pantanal (MT) Pocone, Pantanal (MT) Pocone, Pantanal (MT) Faz. Nhumirim, Pantanal (MS) Faz. Nhumirim, Pantanal (MS) Rio Abobral, Pantanal (MS) 22 Cb 23 Sm 24 Cs Lowland deciduous forest Mesotrophic cerradao Submontane deciduous forest Faz. Salina, Pantanal (MS) Faz. Salina, Pantanal (MS) Corumba (MS) 25 Cb 26 Fa 27 Sd 28 As 29 Da 30 Cs 31 Sm 32 Sm 33 Cm 34 Gf 35 Cm Lowland deciduous forest Alluvial semideciduous forest Dystrophic cerradao Submontane open rainforest Alluvial dense rainforest Submontane deciduous forest Mesotrophic cerradao Mesotrophic cerradao Montane deciduous forest Gallery forest Montane deciduous forest Porto Murtinho (MS) Rio Ivinheima (MS) Ilha do Bananal (TO) Ilha do Bananal (TO) Ilha do Bananal (TO) Ilha do Bananal (TO) Barrolandia (TO) Figueiropolis (TO) Padre Bernardo (GO) Silvania (GO) FERCAL(DF) 1 Gf 2 As 3Sd 4Cs 5 Sm 6Gf 7Cs 8Cs ORIGIN OF CENTRAL BRAZILIAN FORESTS 149 TABLE 1. (cont.) Lat. S* Long. W Alt. (m) Climate (Koppen) 12°49' 12°49' 12°49' 12°49' 12°49' 12°46' 15°00' 15°53' 15°53' 15°21' 51°46' 51°46' 51°46' 51°46' 51°46' 52°03' 54°30' 52°15' 52° 15' 55°49' 380 400 400 400 400 350 450 490 490 350 Aw/Am Aw/Am Aw/Am Aw/Am Aw/Am Aw/Am Aw Aw Aw Aw 15°21' 55°49' 350 Aw 15°20' 16°25' 17°45' 17°45' 17°45' 17°05' 17°10' 18°59' 18°59' 19°45' 55°51' 56°50' 57°37' 57°37' 57°37' 56°57' 56°57' 56°39' 56°39' 56°45' 270 95 85 85 85 85 85 89 89 95 Aw Aw Aw Aw Aw Aw Aw Aw Aw Aw 19°30' 19°30' 19°01' 56°10' 56°10' 57°39' 95 95 120 Aw Aw Aw/Bs 21°15' 22°47' 10°25' 10°25' 10°25' 10°25' 10°05' 12°04' 15°15' 16°40' 15°40' 57°40' 53°32' 50°20' 50°20' 50°20' 50°20' 48°55' 49°10' 48°30' 48°37' 47°56' 100 250 240 240 240 240 250 400 650 1050 1100 Aw/Bs Cw Aw/Am Aw/Am Aw/Am Aw/Am Aw Aw Aw Aw Aw Source Ratter et al. 1973, HS Ratter et al. 1973, HS Ratter et al. 1973 Ratter et al. 1973, 1977, 1978b Ratter et al. 1973, 1977 Ratter et al. 1978a, HS Ratter et al. 1977, 1978b Furley et al. 1988, Ratter et al. 1978b Furley et al. 1988 Oliveira-Filho & Martins 1986, Oliveira-Filho et al. 1990 Oliveira-Filho 1989, Oliveira-Filho et al. 1990 Oliveira-Filho & Martins 1986, 1991 Cunha 1990, Ratter et al. 1988 Prance & Schaller 1982 Prance & Schaller 1982 Prance & Schaller 1982 Cunha 1990 Cunha 1990 Ratter et al. 1988 Ratter et al. 1988 Cunha et al. 1986, Ratter et al. 1988, HS Dubs 1992 Dubs 1992 Ratter et al. 1988, Prado et al. 1992, HS Prado et al. 1992 Assis 1991 Ratter 1987 Ratter 1987 Ratter 1987 Ratter 1987 Ratter et al., unpubl. data Ratter et al., unpubl. data Ratter et al. 1977, 1978b Felfili 1990, 1993b, HS Ramos 1989 150 A. T. OLIVEIRA-FILHO & J. A. RATTER TABLE 1. (cont.) Went. no. Vegetation formation (modified IBGE system) 36 Gf 37 Gf Gallery forest Gallery forest 38 Sd Dystrophic cerradao 39 Fm 40 Gf 41 Cs 42 Cm 43 Sm 44 Gf 45 Gf 46 Gf 47 Gf 48 Fm 49 Fm Montane semideciduous forest Gallery forest Submontane deciduous forest Montane deciduous forest Mesotrophic cerradao Gallery forest Gallery forest Gallery forest Gallery forest Montane semideciduous forest Montane semideciduous forest Brasilia Botanic Garden (DF) APA Rio S. Bartolomeu (DF) Pandeiros, Januaria (MG) Sagarana (MG) Sagarana (MG) Paracatu (MG) Patrocinio (MG) EE Panga, Uberlandia (MG) PN Serra da Canastra (MG) ESAL, Lavras (MG) 50 Fm Montane semideciduous forest RB Poco Bonito, Lavras (MG) 51 Fm Montane semideciduous forest Bom Sucesso (MG) 52 Fm Montane semideciduous forest Camargos, Itutinga (MG) 53 Fm 54 Ds 55 Db 56 Db 57 Fs 58 Fm Montane semideciduous forest Submontane dense rainforest Lowland dense rainforest Lowland dense rainforest Submontane semideciduous forest Montane semideciduous forest Madre de Deus de Minas (MG) 59 Fs 60 Fs 61 Fs 62 Fm 63 Fs Submontane semideciduous forest Submontane semideciduous forest Submontane semideciduous forest Montane semideciduous forest Submontane semideciduous forest 64 Fs Submontane semideciduous forest 65 Fs Submontane semideciduous forest 66 Fs 67 Fs Submontane semideciduous forest Submontane semideciduous forest PE S. Rita Passa Quatro (SP) PE Teodoro Sampaio (SP) 68 Fm 69 Db 70 Db Montane semideciduous forest Lowland dense rainforest Lowland dense rainforest FR S. Jose dos Campos (SP) EE Ubatuba (SP) PE Ilha do Cardoso (SP) Locality PN Brasilia (DF) APA Gama-C. do Veado (DF) Faz. Agua Limpa (DF) Serra S. Jose, Tiradentes (MG) RF Rio Doce (MG) Southern Bahia (BA) RF CVRD, Linhares (ES) Anhembi (SP) PM Grota Funda, Atibaia (SP) RE Bauru (SP) APA Corumbatai, Ipeiina (SP) UNESP, Jaboticabal (SP) Serra do Japi, Jundiai (SP) RB Mogi Guacu (SP) RE Porto Ferreira (SP) Rio Claro (SP) ORIGIN OF CENTRAL BRAZILIAN FORESTS 151 T A B L E 1. (cont.) Lat. S* Long. W Alt. (m) Climate (Koppen) 15°40' 15°55' 47°56' 47°54' 1100 1100 Aw Aw 15°57' 47°56' 1100 Aw 15°53' 15°40' 15°28' 16°00' 16°00' 17° 10' 19°20' 19° 10' 20°30' 21°13' 21°20' 47°51' 47°56' 44°23' 47°00' 47°00' 46°57' 46°47' 48°24' 46°30' 45°00' 45°00' 1100 1100 350 550 550 900 950 800 950 925 1150 Aw Aw Aw/Bs Aw/Bs Aw/Bs Aw Aw Aw Aw/Cw Cw Cw 21°09' 44°54' 825 Cw 21°22' 44°37' 917 Cw 21°29' 44=22' 925 Cw 2i°or 19°40' 15°20' 19°18' 22°40' 23°10' 44°14' 42°35' 39°10' 40°04' 48°10' 46°25' 900 450 50 50 500 1200 Cw Aw/Cw Af Am Cw Cf 22° 19' 22°25' 21°15' 23° 11' 22° 16' 49°04' 48°45' 48°20' 46°52' 47°10' 570 600 560 1020 595 Cw Cw Cw Cf Cw 21°49' 47°25' 580 Cw 22°22' 47°28' 630 Cw 21°40' 22°30' 47°38' 52°20' 630 300 Cw Cw 23°O5' 23°27' 25°1O' 45°55' 45°04' 48°00' 840 105 100 Cf Af Af Source Felfili 1990, 1993b, Ramos 1994 Ratter 1980, 1991, Felfili & Silva Jr 1992, Felfili 1993a Ratter 1980, 1991, Felfili & Silva Jr 1992 Azevedo et al. 1990 Pereira et al. 1985 Ratter et al. 1977, 1978b Ratter et al., unpubl. data Ratter et al., unpubl. data Felfili 1990, 1993b, HS Felfili 1990, 1993b, HS Schiavini 1992 Mota 1984, HS Oliveira-Filho et al. 1994b Gavilanes et al. 1992a, Oliveira-Filho et al. 1994a CarvaJho et al. 1992, Oliveira-Filho et al. 1994c Vilela et al. 1995, Oliveira-Filho et al. 1994c Gavilanes et al. 1992b, Oliveira-Filho et al. 1994d Oliveira-Filho & Machado 1993 CETEC 1982, Silva Jr et al. 1994, HS Harley & Mayo 1980, Mori et al. 1983 Peixoto & Gentry 1990, HS Cesar & Leitao-Filho 1990a, 1990b Meira Neto et al. 1989, Grombone et al. 1990 Cavassan et al. 1984 Mantovani et al. 1986, Rodrigues 1991 Pinto 1989, Marchiori et al. 1992 Rodrigues 1986, Rodrigues et al. 1989 & Leitao Filho 1978, Gibbs Mantovani et al. 1989 Bertoni et al. 1982, Bertoni & Martins 1987 Pagano & Leitao Filho 1987, Pagano et al. 1987 Bertoni et al. 1988, Martins 1991 Campos & Heinsdijk 1970, Baitello et al. 1988 Silva 1989 Silva & Leitao Filho 1982 Barros et al. 1991 152 A. T. OLIVEIRA-FILHO & J. A. RATTER TABLE 1. (cont.) Ident. no. Vegetation formation (modified IBGE system) Locality 71 Fs 72 Mm 73 Cm 74 Cb Submontane semideciduous forest Montane mixed forest Montane deciduous forest Lowland deciduous forest PE Mata Godoy, Londrina (PR) FN Irati, Teixeira Soares (PR) Alto Rio Uruguai (SC/RS) Eastern Chaco (Argentina) 75 Cm Montane deciduous forest Western Chaco (Argentina) 76 Gf 77 Ds 78 Ds 79 As 80 Ds Gallery forest Submontane dense rainforest Submontane dense rainforest Submontane open rainforest Submontane dense rainforest Rio Beni savannas (Bolivia) Rio Ivon (Bolivia) Alvorada D'Oeste (RO) Presidente Medici (RO) Jiparana (RO) 81 Ds 82 Ds 83 Ds 84 Db 85 Db Submontane dense rainforest Submontane dense rainforest Submontane dense rainforest Lowland dense rainforest Lowland dense rainforest Presidente Medici (RO) Jam (RO) Ariquemes (RO) Manaus-Itacoatiara (AM) Reserva Ducke, Manaus (AM) 86 Db 87 Da 88 Ds 89 Da 90 Ds 91 Db 92 Db 93 Ds 94 Da 95 Db 96 Db 97 Db 98 Db 99 Ds 100 Dm Lowland dense rainforest Alluvial dense rainforest Submontane dense rainforest Alluvial dense rainforest Submontane dense rainforest Lowland dense rainforest Lowland dense rainforest Submontane dense rainforest Alluvial dense rainforest Lowland dense rainforest Lowland dense rainforest Lowland dense rainforest Lowland dense rainforest Submontane dense rainforest Montane dense rainforest EE INPA (AM) Rio Solimoes, Manaus (AM) EE Una de Maraca (RR) EE Ilha de Maraca (RR) Serra do Navio (AP) Belterra, Santarem (PA) O Deserto, Altamira (PA) Transamazonica, Altamira (PA) Rio Amazonas, Belem (PA) Mucambo, Belem (PA) Capitao Poco (PA) Rio Gurupi (PA) Maraba (PA) Rio Gelado, Carajas (PA) Iron ore, Carajas (PA) 101 Dm 102 Db 103 Gf 104 Sm 105 Db 106 Fm Montane dense rainforest Lowland dense rainforest Gallery forest Mesotrophic cerradao Lowland dense rainforest Montane semideciduous forest Copper ore, Carajas (PA) Aldeia Urutawy (MA) Carolina (MA) Rio Lajes (MA) Coastal Pernambuco (PE) Mata de Pau Ferro, Areias (PB) * Unless otherwise indicated ORIGIN OF CENTRAL BRAZILIAN FORESTS 153 TABLE 1. (cont.) Lat. S* Long. W Alt. (m) Climate (Koppen) 23°27' 25°26' 28°00' 28°30' 51°15' 50°24' 52°00' 59°30' 700 870 c.500 80 Cf Cf Cf Cw/Bs 24°45' 64°40' 1050 Cw/Bs 13°10' 11°45' 11°25' 1T10' 67°30' 66°02' 63°30' 62°20' 61°57' 180 200 180 200 230 Am/Cw Am Am Am/Aw Am/Aw 09°35' 09°20' 02°55' 02°55' 62° 15' 61°57' 62°00' 59°45' 59°35' 230 200 200 50 50 Am/Aw Am Am Am/Af Am/Af 02°30' 03°20' 03°22'N 03°22'N 00°55'N 02°40' 03°29' 03°35' 01°40' 01°35' 01°44' 02°20' 05°43' 06° 10' 06° 17' 60°05' 60°20' 61°26' 61°26' 52°01' 54°45' 51°40' 53°10' 48°30' 48°30' 47°09' 46°30' 49°00' 50°20' 50°23' 100 24 180 180 280 60 30 200 3 25 65 65 90 550 700 Am/Af Am/Af Am Am Am/Af Am Am/Aw Am Am Am Am Aw Aw Aw Aw 06°10' 03°10' 07°ll' 07°21' 08°30' 06°58' 50°50' 45°06' 47°25' 47°26' 35°00' 35°42' 650 75 150 170 30 550 Aw Aw Aw/Am Aw/Am As Bs/As iroo' lric Source Soares-Silva & Barroso 1992 Galvao et al. 1989 Klein 1972, Rambo 1980 Prado 1991, Prado et al. 1989, Franceschi & Lewis 1979 Brown et al. 1985, Prado 1991, Lewis & Pire 1981 Haase & Beck 1989 Boom 1986, 1987 Lisboa 1990, Lisboa & Lisboa 1990 Lisboa 1990 Salomao & Lisboa 1988, Lisboa 1989, 1990 Maciel & Lisboa 1989, Lisboa 1990 Absy et al. 1986/87 Absy et al. 1986/87 Prance et al. 1976 Higuchi et al. unpubl. data, Ribeiro & Hopkins (HS) Porto et al. 1976 Worbes et al. 1992 Milliken & Ratter 1989 Milliken & Ratter 1989 Rodrigues 1963 Carvalho et al. 1986a, 1986b Campbell et al. 1986 Dantas & Muller 1979 Cain et al. 1956, Pires & Koury 1958 Cain et al. 1956 Dantas et al. 1980 Balle 1987 Salomao 1991 Silva et al. 1987 Salomao et al. 1988, Salomao & Rosa 1989, Morellato & Rosa 1991 Silva & Rosa 1989 Balle 1986 Ratter et al., unpubl. data Ratter et al., unpubl. data Andrade-Lima 1960, HS Mayo & Fevereiro 1982, HS 154 A. T. O L I V E I R A - F I L H O & J. A. RATTER 1991), which met nearly all our requirements. However, a few modifications were necessary: (a) Gallery forests, which are part of the cerrado (savanna) formation in the IBGE system, were separated as a distinct category. However, this was applied only to the galleries occurring within the Cerrado Province, as defined by Cabrera & Willink (1973), and riverine forests of other provinces fell into other IBGE categories. (b) The altitudinal limit between the submontane and montane formations for latitudes between 16°S and 24°S was changed from 500 to 750m, as this latter figure was more effective in distinguishing the montane forests of the states of Sao Paulo (Torres et al., 1995) and Minas Gerais (Oliveira-Filho et al., 1994c). (c) The IBGE category 'forested savanna' (cerradao) was split into two, the 'mesotrophic cerradao' and the 'dystrophic cerradao', since an important ecological distinction has been found between the two formations (Ratter, 1971; Ratter et al., 1973, 1977; Ratter & Dargie, 1992). The map of Brazilian vegetation produced by the IBGE (1993) helped in classifying the forest areas. Information from the literature was also used, especially in dubious situations such as those of transitional areas, and when extending the classification system to the four areas in Argentina and Bolivia. The resulting classification of the 106 forest areas is given both in Table 1 and Fig. 2. A brief description of the 15 categories used follows: (a) Gallery forest (Gf): riverine forests occurring within the cerrado domain as more or less narrow strips along the valley bottoms and as rather larger areas around headwaters ('cabeceiras') often showing sharp boundaries with the neighbouring open formations. Number of areas: 13. (b) Dystrophic cerradao (Sd): dense savanna woodlands associated with basepoor, often sandy, soils; particularly common in the transition to the Hylaea and in association with galleries. Number of areas: 4. (c) Mesotrophic cerradao (Sm): dense savanna woodlands associated with baserich soils, often derived from calcareous material; frequently found in association with deciduous forests. Number of areas: 11. (d) Lowland deciduous forest (Cb): seasonal forests occurring at low altitudes (< 100m) on base-rich soils, especially in the Mato Grosso Pantanal and in the Chaco fringes. Number of areas: 7. (e) Submontane deciduous forest (Cs): seasonal calciphilous forests occurring on patches of base-rich soils at intermediate altitudes (100-7 50m) within the cerrado domain or in the transitional areas to the caatinga domain. Number of areas: 6. (f) Montane deciduous forest (Cm): seasonal calciphilous forests occurring on patches of base-rich soils at higher altitudes (>750m) within the cerrado domain or in the transitional area to the caatinga domain. This category also included the piedmont forests of central-western Argentina. Number of areas: 5. (g) Alluvial semideciduous forest (Fa): seasonal but moderately deciduous forests occurring on the floodplains of the major rivers of the Parana-Paraguay basin. These were the only riverine forests of central Brazil not classified as gallery forest since they are distinguished by their flood regime. Number of areas: 3. ORIGIN OF CENTRAL BRAZILIAN FORESTS 155 (h) Submontane semideciduous forest (Fs): seasonal subtropical forests occurring mainly in the mid-altitude hinterlands of southeastern and southern Brazil (100-750m), also appearing on soils of intermediate fertility in central Brazil. Number of areas: 10. (i) Montane semideciduous forest (Fm): seasonal but moderately deciduous forests, occurring in the high-altitude hinterlands of southeastern and central Brazil (>750m), often associated with soils of intermediate fertility. This category also includes the 'brejo' forests that cover the hinterland plateaux of northeastern Brazil. Number of areas: 11. (j) Montane mixed rainforest (Mm): subtropical Araucaria forests, occurring on the highlands of southern and southeastern Brazil. Number of areas: 1. (k) Submontane open rainforest (As): transitional forests found mainly in the southern fringes of the Hylaea, often on base-poor and sandy soils. A confused nomenclature has been applied to this physiognomy, e.g. evergreen seasonal forest, monsoon forest, and dry forest ('mata seca'). Number of areas: 3. (1) Alluvial dense rainforest (Da): evergreen rainforests occurring on the floodplains of the Amazonian rivers; also called 'varzea forests'. Number of areas: 4. (m) Lowland dense rainforest (Db): evergreen rainforests occurring at low altitudes both in the Amazonian ('terra firme' forests) and Atlantic coastal regions. Number of areas: 11 Amazonian and 4 Atlantic. (n) Submontane dense rainforest (Ds): evergreen rainforests occurring at midaltitudes both in the Amazonian ('terra firme' forests) and Atlantic coastal regions. Number of areas: 10 Amazonian and 1 Atlantic. (o) Montane dense rainforest (Dm): evergreen Amazonian rainforests ('terra firme' forests) occurring at higher altitudes in the Carajas mountains system. Number of areas: 2. Preparation and revision of the database The floristic information contained in the 106 checklists was organized in a database using Microsoft Excel 4. The first part of this database consisted of basic information about each area: geographical coordinates, altitude, climatic type, topographic position, soil type, vegetation formation and sources of information. Most of these are given in Table 1. The second part of the database was essentially a presence/ absence matrix for woody species registered in the 106 forest areas. This is not reproduced here because of space limitations. However, the printed version is available from the libraries of the Royal Botanic Garden Edinburgh and Universidade Federal de Lavras (Oliveira-Filho & Ratter, 1994). Before reaching its final form, the information contained in the database went through a detailed revision to check the 3528 species names cited in the checklists for synonymy and growth form. A minimum potential height of 3m was established as the criterion for inclusion and climbers were excluded. Checking synonymy was no trivial task: no less than 332 monographs and revisions were consulted, and we 156 A. T. OLIVEIRA-FILHO & J. A. RATTER were also helped by specialists from the Royal Botanic Gardens Kew, New York Botanical Garden, Missouri Botanical Garden and University of Sao Paulo. After eliminating 312 synonyms and 101 non-arboreal taxa, the number of species fell to 3114 (11.7% reduction). Another 118 synonyms were replaced by their presently accepted names without reducing the number of species. A total of 13.8% of synonyms was found in the original list of 3427 names for arboreal species. These figures demonstrate the level of caution necessary when dealing with vegetation survey data for this part of the world, and the risks of performing floristic analyses without a careful verification of all names. Details of literature used in the revision of the database as well as the list of synonyms eliminated are available from the first author. In addition to the records obtained from the literature, the database was amplified by examination and revision of the extensive collections from central Brazilian forests lodged in the herbarium of the Royal Botanic Garden Edinburgh (E). Multivariate analyses The floristic information (presence records) contained in the database was simplified by techniques of multivariate analyses in order to seek main patterns of data structure that could be associated with geographic/environmental variables. Special attention was given to the vegetation classification categories described above, which were treated here as synthetic geographic/environmental parameters. It is important to stress at this point that the whole procedure was an exploratory data analysis carried out with a hypothesis-generating purpose (Kent & Coker, 1992; Ratter & Dargie, 1992). Three techniques of multivariate analysis, corresponding to different methodological approaches, were used. The purpose was to seek patterns that could be accentuated in common by different analytical procedures. The techniques were: (1) an ordination by Detrended Correspondence Analysis (DCA) (Hill & Gauch, 1980), (2) a divisive hierarchical classification by Two-Way INdicator SPecies ANalysis (TWINSPAN) (Hill, 1979), and (3) an agglomerative hierarchical classification by UPGMA (Unweighted Pair-Groups Method using Arithmetic averages) using the Sorensen Coefficient of Community (CC) as a measure of similarity (Kent & Coker, 1992). For DCA and TWINSPAN, we used the versions contained in the package VESPAN II (Malloch, 1988). The floristic matrix analysed by both methods had 1423 species and 106 forest areas as a result of the elimination of 1691 species occurring in only one area, following the procedures described in Ratter & Dargie (1992). However, the full matrix with 3114 species and 106 areas was analysed by UPGMA. The CCs were calculated by database management in Microsoft Excel 4 and processed by the package NTSYS (Rohlf, 1992) in order to produce the clustering dendrogram. The floristic information contained in the database was eventually condensed by clustering the presence records within each of 12 groups of forest areas corresponding ORIGIN OF CENTRAL BRAZILIAN FORESTS 157 to the mainfloristicpatterns indicated by the multivariate analyses. The results were used to allow a direct quantitative assessment of the floristic links between the groups by both the number of species in common and the calculation of a similarity index: the Sorensen Coefficient of Community (CC). A consecutive clustering operation was carried out with the 12 groups and these were agglomerated into five wider groups: gallery forests, Amazonian and Atlantic forests, cerradSes and deciduous forests. These were compared in terms of number of species in common. Maps of geographical distribution A series of 52 maps of the geographical distribution of 55 tree species was produced to illustrate the distribution patterns of characteristic species of central Brazilian forest species. These concentrated on gallery forest species since the assessment of their floristic links to the Amazonian and Atlantic forests was one of the primary aims of this study. Besides, the distribution patterns of South American deciduous forest species, which represent another important component of central Brazilian forests, has already been investigated in a series of maps produced by Prado & Gibbs(1993). The selection of the 55 species for mapping followed two criteria: (1) the 13 most frequent species in the 13 gallery forests of the database were chosen regardless of their geographical or ecological distribution; (2) the remaining 42 species were deliberately chosen in order to illustrate patterns of geographical and ecological distribution of typical species of central Brazilian gallery forests. This choice was based both on the authors' field experience and on the 'indicator species' given by TWINSPAN. Unless we had valuable additional information, we also avoided producing maps of species already available in the recent literature (e.g. for Anadenathera colubrina, Sterculia striata and Astronium fraxinifolium given by Prado & Gibbs (1993)). The maps were produced using two sources of information: (1) herbarium specimens lodged at the herbaria of Kew (K) and Edinburgh (E), or cited in taxonomic literature, and (2) citations in reliable floristic checklists, always followed by verification in the literature for correspondence with the geographical distribution given by specialists. Following Prado & Gibbs (1993), different symbols are used in the maps to distinguish the two sources of information. MULTIVARIATE ANALYSES The multivariate analyses of the floristic data show a great deal of coincidence in the patterns arising from the three techniques used (Figs 3-5). Most of the vegetation classification categories we used appear as cohesive groups, demonstrating that the geographical and ecological variables underlying the classification system probably have a strong correlation with the mainfloristicpatterns detected. A detailed description and discussion of these results are given below. 158 A. T. OLIVEIRA-FILHO & J. A. RATTER Ordination (DCA) The ordination of the 106 areas in the first two DCA axes is given by Fig. 3. The distribution of the forest areas in the ordination space conforms basically to the shape of a triangle whose vertices are made up of: (a) Amazonian 'terra firme' rainforests (right vertex), the most distinct group in the whole array of forest areas, (b) semideciduous and mixed forests of southeastern Brazil (bottom left vertex), and (c) central Brazilian cerrad5es and deciduous forests (top left vertex). These three groups also correspond, respectively, to the main vegetation formations of the Amazonian, Paranense and Cerrado Provinces of Cabrera & Willink (1973). In the space defined by the three vertices lie other important vegetation formations. The Atlantic rainforests (**D) appear between the Amazonian rainforests and • Axis 2 Sm O Sd • Fm • Fs • Da A As X Cm V Fa o Db/s/m X Cs + Cb © Mm A Gf * Atlantic forest 400 350 - - 2 ° + 80,81,85,91,92,98,101 300 -- 82,83,84,86,90 250 - - 74 200 -- 150 100 ©• 50 -- 50 100 150 200 250 300 350 450 Axis 1 F I G . 3. Ordination of the 106 forest areas on the first two DCA axes. The forest areas are identified by their number in Table 1 while the symbols indicate their classification according to the modified IBGE classification system for Brazilian vegetation (see text, pp. 154-5). Lowland, submontane and montane dense rainforests are indicated by the same symbol for clarity. ORIGIN OF CENTRAL BRAZILIAN FORESTS 159 Paranense semideciduous forests (Fs, Fm), though closer to the latter. They also follow an interesting north-south sequence that suggests a decrease of Amazonian floristic influence with increasing latitude. The area containing all rainforests (D) in the ordination diagram can also be seen as that of tropical rain climates with lower seasonality (Koppen's Af and Am types), while marked seasonal climates, both tropical and subtropical (Koppen's Aw, Cw and Cf types), occupy the remaining space. Most lowland and submontane deciduous forests (Cb, Cs) appear in close association with mesotrophic cerradSes (Sm), while the dystrophic cerradoes (Sd) seem to link the former group to the transitional open rainforests (As). As mesotrophic cerradoes and deciduous forests are known to appear on soils of higher fertility, the top left vertex can be regarded as an area that corresponds to seasonal climate plus high soil fertility, while the space extending to the right would correspond to a transition to Amazonian forests on soils of lower fertility. Montane deciduous forests (Cm) appear in the space extending from the top vertex down to the Paranense semideciduous forests (Fs, Fm), so that this area of the diagram could be associated with cooler climates, either subtropical or montane (Koppen's Cw and Cf types). Nevertheless, DCA indicates little differentiation between submontane and montane semideciduous forests (Fm and Fs). In addition to this, two Fm areas, one in the Distrito Federal (39) and the other in Paraiba (106), are placed well apart from other semideciduous forests. However, this could be associated with geographical variables as they were rather isolated and distant from the others of their category. Area 68 is somewhat displaced towards the Atlantic rainforests, probably as a result of its geographical proximity to the coast. Gallery forests (Gf) are clearly the most widespread category in the ordination space. Nevertheless, two main groups can be detected. The first and more cohesive one (34, 36, 37, 40, and 44 to 47) is that of the centre and south of the Cerrado Province (Distrito Federal and states of Goias and Minas Gerais), which appears in a closer association with the Paranense semideciduous forests. The second and more widespread group (1, 6, 11, 76 and 103) is that of the north and west of the Cerrado Province (states of Mato Grosso and Tocantins), which appears amongst dystrophic cerradoes (Sm), transitional open rainforests (As) and alluvial forests (Da, Fa). The basic factor that appears to underline this differentiation is geographic, since the first group is basically on the Parana and Sao Francisco river basins, while the second lies on the Amazon and Paraguay river basins. The influence of rivers also seems to be important in the floristic links between the galleries and Amazonian forests, since most 'varzea' forests (Da) appear on the space between galleries and 'terra firme' forests (Db, Ds, Dm). On the other hand, the alluvial semideciduous forests (Fa) appear to connect the galleries to the Paranense forests. As an overall analysis of the patterns arising from the DCA ordination diagram we can say that the first ordination component has a strong latitudinal influence, especially discriminating the equatorial space, while the second is highly influenced by rainfall seasonality. Riverine forests (galleries and alluvial), together with hinter- 160 A. T. OLIVEIRA-FILHO & J. A. RATTER land montane forests and transitional open rainforests, appear to provide the floristic connections between rainforests, seasonal forests and cerradoes. Divisive hierarchical classification (TWINSPAN) The first level of division in the cladogram yielded by TWINSPAN (Fig. 4) separates the Amazonian forests (group 3) from the others (group 2), coinciding with the high floristic distinctness indicated for these forests by the first DCA axis. However, TWINSPAN indicates a high level of internal heterogeneity for the Amazonian forests which DCA fails to show in the first two ordination axes. This heterogeneity appears in the subsequent divisions of group 3, the first of which separates the two varzea (alluvial) forests from central and northern Amazonia (group 7) from the remainder (group 6). However, the other divisions make little sense in terms of topographic range (lowland, submontane and montane). In addition to this, none of the floristic divisions of the Amazonian forest into geographical sectors cited by Prance (1977) coincides clearly with the classification groups. This is probably a consequence of the very high level of floristic and environmental heterogeneity of the Amazonian forests, which cannot be analysed properly from the small number of existing surveys. Nevertheless, in spite of this heterogeneity, the Amazonian forests certainly form a well-defined floristic province and this is without doubt helped by a number of widespread and typically Amazonian species, such as Minquartia guianensis, Eschweilera coriacea, Brosimum guianense, Socratea exorrhiza and Jacaranda copaia (TWINSPAN's main indicator species for group 3). The division of group 2 indicates two important main groups. Group 4 includes essentially central Brazilian cerradSes (Sm, Sd), deciduous (Cs/Cb/Cm) and semideciduous forests (Fa/Fm) together with transitional open rainforests (As) and the western and northern gallery forests (Gf). Group 5 includes the Atlantic rainforests (Db/Ds), Paranense semideciduous forests (Fs/Fm), the mixed rainforest (Mm) and the central and southern gallery forests (Gf). This division has a clear influence of coastal-interior and vegetation seasonality factors, and a high correspondence to the main patterns indicated by the second DCA axis. It also places at a high level the same division between the two groups of gallery forests indicated by DCA. So, the same pattern arises for the galleries, indicating a mainly Atlantic/Paranense floristic link to the central-southern and a mainly Amazonian floristic link to the northern-western groups. The divisions within group 4 show some interesting patterns. Most mesotrophic cerradoes (Sm) and deciduous forests (Cs/Cb/Cm) fall into group 8, confirming the importance of the edaphic factor to those formations. The main indicator species given by TWINSPAN for this group were Sterculia striata, Anadenanthera colubrina, Tabebuia impetiginosa, Myracrodruon urundeuva, Dilodendron bipinnatum, Madura tinctoria, Cordia glabrata, Enterolobium contortisiliquum and Combretum leprosum, all of which have also been cited by Ratter (Ratter, 1971; Ratter et al., 1977, 1978a, b) as indicators of mesotrophic soils in central Brazil. Group 16 discriminates mainly FIG. 4. Cladogram yielded by TWINSPAN showing the hierarchical classification of the 106 forest areas after five levels of division. The forest areas are shown in the bottom boxes and identified by their number in Table 1, followed by the two-letter code corresponding to their classification according to the modified IBGE classification system for Brazilian vegetation (see text, pp. 154-5). The upper boxes contain the numbers that identify the groups and divisions following the standard TWINSPAN numbering. Numbers in parentheses indicate the number of areas in each group. 2 As 3Sd 5S 10 Sd 11 Gf 12Sm 27 Sd 31 Sm 32 S 38 Sd 76 Gf 79 As 103Gf 162 A. T. OLIVEIRA-FILHO & J. A. RATTER the forests of the Mato Grosso Pantanal (including the seasonally flooded ones (Fa) in subgroup 32) and of the fringing Chaco, while group 17 contains some central Brazilian deciduous forests (group 34) together with the Pantanal cerradoes (group 35). The subdivisions of group 9 are rather confusing, since northern-western galleries appear mixed with cerradoes (of both types), transitional open rainforests and some deciduous forests. Most of these areas appear in the same ordination space of the DCA dendrogram that links the central Brazilian seasonal formations to the Amazonian forest. A possible explanation is that the five western-northern gallery forests are in transitional regions where a considerable number of forest formations meet each other, as happens in the forests of the Ilha do Bananal (Ratter, 1987). This certainly increases the number of species shared by different forests surveyed in those regions and causes them to be clustered within the same groups. The divisions within group 5 are the most clearly organized of the whole cladogram. Most central and southern gallery forests are included in group 10 together with the only central Brazilian montane semideciduous forest (39), while group 11 contains essentially the Atlantic rainforests (Db/Ds) and Paranense semideciduous (Fs/Fm) and mixed forests (Mm). The only gallery to appear in group 11 is the southernmost (47), for which a strong floristic similarity with the Paranense forests could be expected. The main indicator species given by TWINSPAN for the centralsouthern galleries are Cheiloclinium cognatum, Virola sebifera, Sclerolobium paniculatum, Hirtella glandulosa, Ocotea spixiana, Xylopia emarginata and Mauritia flexuosa. The division of group 11 discriminates the northern Atlantic rainforests (group 23) from the remainder (group 22). The southern Atlantic rainforests are included with the montane semideciduous forests (Fm) and mixed rainforest (Mm) in group 44. Group 45 is composed mostly of submontane semideciduous forests (Fs). One important pattern arising from these is the notable floristic diversity of the Atlantic rainforests and their strong floristic links to the Paranense semideciduous forests, especially in the southern sector where the transition occurs via montane forests. Leitao-Filho (1987) has already pointed out that the Atlantic rainforests shelter a bewildering north-south floristic variation. In actual fact, there seem to be stronger floristic links between adjacent tracts of rainforests and semideciduous forests than between different sectors of the Atlantic rainforest along the Brazilian coast. Agglomerative hierarchical classification (UPGMA) Most patterns already indicated by both DCA and TWINSPAN also appear in the UPGMA dendrogram (Fig. 5). This is of particular methodological importance as UPGMA analysed the whole dataset, indicating that the elimination of the unicates had little or no effect on the main patterns arising via TWINSPAN or DCA. However, some important additional patterns arose. These are dealt with below and brief descriptions of the main groups are given: Group a: this includes all Amazonian 'terra firme' forests (Db/s/m) plus a tran- ORIGIN OF CENTRAL BRAZILIAN FORESTS 163 . 1 Gf . 6 Gf 2 As . 3 Sd . 10 Sd . 11 Gf 103 Gf . 76 Gf . 4 Cs ecs 42 Cm 7 Cs . 33 Cm . 41 Cs 35 Cm 28 As m 12 Sm 27 Sd 31 Sm , 32 Sm -CE 43 Sm . 13 Sm 19 Sm 23 Sm 14 Sm 104 Sm 1 k 87 Da 89 Da 77 Ds 81 Ds 78 Ds 80 Ds 98 Db 99 Ds 100 Dm 101 Dm 93 Ds 92 Db 62 Ds 83 Ds 91 Db 90 Ds 95 Db 96 Db 97 Db 102 Db 94 Da 85 Db . 86 Db , 79 As 1 i i 0.2 0.4 0.6 0.8 % similarity FIG. 5. Similarity dendrogram yielded by UPGMA, using S0rensen Coefficients of Community, showing the hierarchical classification of the 106 forest areas. The areas are identified by their number in Table 1, followed by the two-letter code corresponding to the modified IBGE classification system for Brazilian vegetation (see text, pp. 154-5). Other letters marking branches of the dendrogram identify groups discussed in the text. 164 A. T. OLIVEIRA-FILHO & J. A. RATTER sitional open rainforest (79As) and a 'varzea' forest (94Da). The heterogeneity of Amazonian forests is reflected by the high level of most clusterings. Group b: this is made up of highly heterogeneous riverine forests, all liable to seasonal floods, including 'varzea' forests of the Amazon basin (Da) and floodplain forests of the Mato Grosso Pantanal (Fa). This pattern only appeared with UPGMA. Group c: this contains the four northern Atlantic rainforests (Db/s) and the hinterland 'brejo' forest (54Ds). High floristic heterogeneity is also detected. Group d: this includes the two southern Atlantic rainforests (Db). Group e: this consists of the single area of montane mixed rainforest (72Mm), its distinctness being indicated only by UPGMA. Group f: this contains the montane semideciduous forests of the (Fm) Paranense Province plus the southernmost area of gallery forest (47Gf). Group g: this comprises the central and southern gallery forests (Gf) plus the montane semideciduous forest of the Distrito Federal (39Fm). The close floristic links between those galleries and the Paranense montane forests are shown by the low clustering level of groups f and g. Group h: this consists of 10 areas of submontane semideciduous forests of the Paranense Province (Fs) plus the southern Brazilian deciduous forest (Cm). Group i: this includes only a single area of alluvial semideciduous forest of the Parana river (26Fa). Groups e to i contain all deciduous and semideciduous forests, and the mixed rainforest that constitutes the Paranense Province, according to Cabrera & Willink (1973). These patterns indicate that the gallery forests of the central and southern part of the Cerrado Province are actually floristic expansions of the forests of the Paranense Province. Group j : this includes the deciduous forests occurring in the transition to the Chaco in the Mato Grosso Pantanal (24Cs, 25Cb), the Parana river valley (74Cb), and the piedmont forests (75Cm). A clearer discrimination of this group was made only by UPGMA. Group k: this consists of the five areas of lowland deciduous forests of the Mato Grosso Pantanal (Cb). The deciduous forests of the Pantanal and transitional Chaco (groups j and k) appear quite distinct from the other deciduous forests of central Brazil (group m), a pattern shown more clearly than by the other two analytical methods. Group 1: this contains areas of mesotrophic cerradao (Sm), although it also includes the dystrophic cerradoes of the Ilha do Bananal (27Sd) and the Distrito Federal (38Sd). Group m: this comprises the central Brazilian deciduous forests (Cs and Cm) and the transitional open rainforest of the Ilha do Bananal (28As). Group n: this consists of all five northern and western gallery forests (Gf) plus a transitional open rainforest (2As) and two dystrophic cerradoes (Sd). This group, which was entangled amidst other formations in the two previous methods, emerged quite clearly by UPGMA. The strongerfloristiclinks of the northern-western galleries to the dystrophic cerradoes and transitional open rainforests were once again demon- ORIGIN OF CENTRAL BRAZILIAN FORESTS 165 strated. The transition from Sd and As on dystrophic soils has already been described by Ratter et al. (1973) in the state of Mato Grosso, while the connections between Sd on the headwaters of gallery forests were referred to by Oliveira-Filho et al. (1990). Condensed floristic information As they were extracted from floristic checklists for particular forest areas, the condensed information given in Table 2 must be regarded as a means of quantitative assessment of thefloristiclinks between the main vegetation groups and not as their actual figures for number of species, either total or in common. The western-northern gallery forests showed highest floristic similarity, measured by the Sorensen Coefficient of Community (CC), in the comparisons with the dystrophic cerradoes, followed by the transitional open rainforests (dry forests) and the central-southern galleries. This confirms their strong links to the transitional formations that occur on base-poor soils in the southern fringe of the Amazonian Province. The central-southern gallery forests showed the highest CC values in the comparisons with the montane semideciduous forests, followed by those for the submontane semideciduous forests. Then follow those for the cerradoes (of both types), deciduous forests and the other galleries, all at a more or less similar level. This confirms that their stronger floristic links are with the Paranense semideciduous forests. Curiously enough, the comparisons with the southern Amazonian forests yielded similar figures for both groups of galleries. However, stronger floristic links were shown between the western-northern galleries, the central Amazonian rainforests, and the centralsouthern galleries and the Atlantic coastal rainforests. The transitional nature of the Amazonian open rainforests (dry forests) is clear from thefiguresobtained in the comparisons with central Brazilian forests (cerradSes, galleries and deciduous forests) which were significantly higher than those obtained in the comparisons with the other Amazonian rainforests. The riverine connection between Amazonian 'varzea' forests and the western-northern galleries is probably shown by the fact that although the CC value for this comparison is in itself not particularly high it is nevertheless higher than those with the Amazonian 'terra firme' forests. The floristic links between Amazonian and Atlantic rainforests appeared weaker than those between the latter and the semideciduous forests. The high affinity between dystrophic and mesotrophic cerradoes only confirms that they have much of the cerrado arboreal flora in common. However, the high CC value found for the comparison between deciduous forests and mesotrophic cerradoes indicates the wellknown affinity of these two communities. The Venn diagrams at the bottom of Table 2 give an assessment of the composition of the arboreal flora of gallery forests in relation to that of other main forest groups. The first Venn diagram makes clear that a high percentage (77%) of gallery species is made up of those shared with the main Amazonian rainforests and Atlantic forests sensu lato (including Paranense semideciduous forests). The proportion of non- Gallery forests n = 550 Gf/WN Gf/CS As Da D/Am/S D/Am/C D/Atl Sd Sm C/C Fs Fm N. of species Description: Groups: / - • ^ 393 0.242 - — 0.085 0.145 0.088 0.284 0.356 0.333 0.335 0.395 0.479 — Amazonian forests n « 1745 185 _ 0.124 0.149 0.116 0.107 0.444 0.312 0.239 0.129 0.117 83 70 Gallery forests. 113 As Transitional open forests. Gf/CS x—^ \ ^ 270 0.341 0.365 0.172 0.167 0.137 0.143 0.445 0.307 0.260 0.180 0.165 Gf/WN 50 30 31 • 865 0.487 0.200 0.088 0.092 0.116 0.117 0.101 95 91 78 145 Amaze nian D/Am/S Gallery forests n - 550 313 0.246 0.205 0.101 0.075 0.071 0.129 0.060 0.052 Da -XT 861 1173 109 109 90 20 48 48 55 220 0.557 0.274 0.176 0.186 Sd CerradOes n = 321 0.102 0.105 0.165 0.329 0.365 81 178 56 59 173 141 D/Atl 0.139 6.069 0.047 0.070 0.065 0.055 99 69 79 152 496 Amazonian terra firme D/Am/C 76 103 64 19 50 33 57 124 Gallery forests n - 550 226 0.419 0.204 0.195 Sm / ^ 452 ~xr—-• 276 0.163 65 0.286 0.257 forests 167 41 23 77 53 216 59 69 104 71 Fs 112 55 38 66 51 94 68 105 Brazil C/C Cerraddes n = 321 564 69 229 44 23 72 48 260 73 77 108 83 forest Fm TABLE 2. Syntheticfiguresfor the main forest types represented by the 106 checklists of the original database. Top: similarity matrix showing the Sorensen Coefficients of Community (CC) (lower left diagonal) and number of species in common (upper right diagonal). Bottom: Venn diagrams showing number of species shared by gallery forests and other main forest types. O R I G I N OF C E N T R A L B R A Z I L I A N FORESTS 167 shared species in galleries was only 22.9%, a figure that falls to 14.9% when the cerradao flora is included in the second diagram. The interface with seasonal formations is assessed in the third diagram, where a relatively weaker floristic link occurs. The stronger floristic links between galleries and moist forests is certainly explained by the similarity of their habitats depending on higher soil moisture in the valley bottoms within the Cerrado Province. This compensates to a large extent the seasonal droughts, allowing moist forest species to penetrate the cerrado region via gallery forests. However, soil moisture on the margins of galleries is not so high and this is where a dry transition to cerradoes occurs—thus explaining the floristic links between gallery forests and cerradoes, particularly of the dystrophic type. A link to deciduous forests is also present and this is probably related to transitions from gallery to upland forests on soils of higher fertility. The weaker overall floristic link between galleries and communities of mesotrophic as opposed to dystrophic soils undoubtedly derives from the simple fact that base-poor soils predominate in the Cerrado Province. DISTRIBUTION MAPS The species distribution maps given in Figs 6-18 were organized in a sequence starting with the most widespread distribution patterns and ending with the most restricted. Unless otherwise indicated, the ecological discussion of the species distribution is based mostly on the authors' field experience and previous publications. Important synonyms found in the checklists are given in the text, while footnotes explain some taxonomic complications. The 13 species with highest frequency (^9) in the 13 gallery forest areas were: Siparuna guianensis (13), Copaifera langsdorffii (12), Tapirira guianensis (11), Calophyllum brasiliense (11), Protium spruceanum (10), Endlicheria paniculata (10), Virola sebifera (10), Eugeniaflorida(10), Xylopia aromatica (9), Protium heptaphyllum (9), Maprounea guianensis (9), Cheiloclinium cognatum (9), and Emmotum nitens (9). Most of these are very widespread species and many are generalists in terms of habitat preference. Four examples of extremely widespread species are given in Fig. 6. Their geographical range extends virtually through the whole of the Neotropics, although Calophyllum brasiliense (but see footnote 2, p. 168) and Hieronyma alchorneoides are not found in the Caribbean islands. Madura tinctoria (L.) Don (= Chlorophora tinctoria (L.) Benth.) is one of only two mesotrophic soil indicator species represented in the maps.1 It occurs in a wide range of habitats from the arid semi-deserts of northern Mexico and northeastern Brazil to the humid Amazon valley and the seasonally flooded forests of the Mato Grosso Pantanal, only requiring soils of intermediate to high fertility. Its occurrence in gallery forests is thus probably dependent on higher soil fertility and not soil moisture. On the other hand, Margaritaria nobilis 1 However, the recent publication of Prado & Gibbs (1993) maps the distribution of a number of species of mesotrophic soils, such as Anadenanthera colubrina, Myracrodruon urundeuva, Combretum leprosum, Enterolobium contortisiliquum, Sterculia striata and Tabebuia impetiginosa. A. T. OLIVEIRA-FILHO & J. A. RATTER 168 Madura tinctoria (L.) Don ^v:,. Margaritaria nobilis L. f. -V 1000m) and their distribution follows the Atlantic coastal ridge and the hinterland mountains through the Espinhaco range and the Planalto Central. Both species show interesting disjunctions associated with high altitude. C. scabra also appears along the high Andean foothills while Drimys 180 A. T. OLIVEIRA-FILHO & J. A. RATTER Diospyros sericea A. DC. Hirtella hoehnei Pilger FIG. 18. Distribution of Diospyros sericea, Hirtella hoehnei, Vochysia pyramidalis and Ilex affinis. For explanation of symbols see Fig. 6. roraimensis (A. C. Smith) Ehrend. & Gottsb., which was formerly recognized as a subspecies of D. brasiliensis, appears in the Guyana highlands. In addition to these two, other gallery species known to follow an upper montane distribution pattern along hinterland mountain ranges are Weinmannia discolor Gardner and W. paullinifolia Pohl, Podocarpus lambertii Klotz., P. sellowii Klotz. and the already mentioned Hedyosmum brasiliense (Giulietti & Pirani, 1988). The remaining nine maps (Figs 16-18) represent gallery forest species with a distribution pattern concentrated in central Brazil. The first species, Diospyros hispida A. DC. (Fig. 16), is common in galleries with soils of intermediate fertility and is also found in cerradao and cerrado. Hence its distribution map conforms to that of the cerrado vegetation. Unonopsis lindmannii R. E. Fries and Callisthene major Mart. (Fig. 17) also appear occasionally in cerrado vegetation, but their typical habitat is ORIGIN OF CENTRAL BRAZILIAN FORESTS 181 the margin of galleries where they receive full sunlight. Pseudolmedia guaranitica Hassler,6 Virola malmei A. C. Smith, and Virola urbaniana Warburg are strict gallery species, with high preference for moist soils. Other typical riverside species found in central Brazilian galleries are given in Fig. 18. These are Diospyros sericea A. DC, Hirtella hoehnei Pilger, Vochysia pyramidalis Mart, and Ilex affinis Gardner. Of these D. sericea is also common in cerradao. CONCLUSIONS A long-lasting controversy has been associated with the factors determining the distribution of cerrado vegetation but in general the following are considered important: seasonal rainfall, soil fertility and drainage, fire (especially man-made), and the climaticfluctuationsof the Quaternary. These are, in fact, the same factors identified as important in maintaining the savanna biome worldwide, although in other areas, such as Africa, grazing also plays an important role (see papers in Werner (1991) and in Furley et al. (1992)). As pointed out by a number of authors, climatic seasonality by itself cannot explain the predominance of cerrado vegetation within its distribution range as the present climatic conditions would favour the establishment of forests in most of the Cerrado Province (e.g. Rizzini & Pinto, 1964; Reis, 1971; Klein, 1975; Van der Hammen, 1983). This is true, and actually forests are an extremely important constituent of the Cerrado Province in central Brazil and probably occur in all habitats where environmental factors allow their establishment and growth. Where more fertile mesotrophic soils occur in the Cerrado Province the climax vegetation is deciduous or semideciduous (mesophytic) forest. Such forests are seen in valleys where the topography has cut into more mineral-rich underlying rocks (e.g. silts and mudstones) or as great extensions in more fertile areas. One of the largest of the latter, the 'Mato Grosso de Goias', is estimated to have had an area of 40,000km2 (see in Fig. 1 the great gap in the distribution of cerrados in the state of Goias) while the anecdotal evidence of our road journey from the Distrito Federal to Maranhao (p. 145) demonstrates how much of the Cerrado Province was previously covered by this type of forest. The soils of these forests are particularly good for agriculture and because of this the vegetation has been devastated to such an extent that in many areas it is difficult to realize that forest was once an important, or even dominant, land cover. In our experience it is rare to encounter cerrado regions where at least some of this forest does not occur: even the most dystrophic chapadas have small fertile forested valleys! Such mesophytic forests are particularly common on the base-rich soils of the peripheral areas that connect the Cerrado to the Caatinga Province, in the East, and to the Chaco Province, in the western bound6 This species appears as Pseudolmedia laevigata Trecul in all checklists published for central Brazilian galleries and this is why both species were mapped. Berg (1972) in his revision of the genus regarded the two species as very similar. It would be no surprise if in the future they are united as a single species, thus completely changing the distribution pattern illustrated. 182 A. T. OLIVEIRA-FILHO & J. A. RATTER aries of the Mato Grosso Pantanal. They also appear on calcareous outcrops throughout the region, forming the so-called 'mata calcarea'. These have already been regarded by Prado & Gibbs (1993) as relics of a once even more extensive deciduous forest that during the glacial maxima would have connected the caatingas, in northeastern Brazil, to the semideciduous forests, in southeastern Brazil and southern Paraguay, and to the piedmont forests, in central-western Argentina. Palynological evidence is given by Ledru (1993) for the existence of such a forest during the Wiirm-Wisconsin glacial period in present-day cerrado areas. It is reasonable to think that an intense process of soil-leaching and acidification following the return of more humid climates to the cerrado region would have favoured the establishment of cerrado in most places and the isolation of deciduous forests on the present-day areas of mesotrophic and calcareous soils (Ratter et al., 1988). Although these forests have been almost entirely devastated by man, their original distribution given by the IBGE (1993) seems to indicate that they represent expansions of the Paranense semideciduous forests penetrating into the Cerrado Province via the basalt-derived fertile soils of western Minas Gerais and southern Goias. There seem to be no clear-cut ecological and floristic differences between deciduous and semideciduous forests. The level of deciduousness probably depends on the conjunction of soil moisture and chemical properties and the increasing role of a cool winter with increasing altitude and latitude. Often there are quite local differences in deciduousness: for instance, at Vale dos Sonhos, Mato Grosso, the same forest community is deciduous on the well-drained valley sides but semideciduous in the moister valley bottom (Ratter et al., 1978b). Transitional vegetation at the cerrado interface is interesting and widespread. The mesotrophic cerradao represents an intermediate community between mesophytic forest and cerrado associated with soils of intermediate fertility (Ratter 1971, 1992; Ratter et al., 1977; Furley et al., 1988). It usually occurs at the mesophytic forest/cerrado interface but also covers very large extensions of the Cerrado Province where such intermediate soils occur. Thus the mesophytic forests are an important element of central Brazilian vegetation associated with soil fertility and following a northeast-southwest geographical pattern (Prado & Gibbs, 1993). A second element is the dendritic net of gallery forests, highly dependent on soil moisture, extending in a northwest-southeast direction to connect the Amazonian and Atlantic forests. A considerable number of species shared by these two great forest provinces cross central Brazil via the gallery forests. Others extend their range into the Cerrado Province along the galleries but do not complete the crossing, while the majority of species remain strictly confined to their own rainforest region. Thus the movement of individual species along the Northwest-Southeast Bridge migration route seems to be very variable. Many of the species established throughout the galleries are habitat generalists (e.g. Schefflera morototoni, Casearia syhestris, Protium heptaphyllum, Tapirira guianensis, Virola sebifera, Copaifera langsdorffii and Hymenaea courbaril) and probably environmental conditions are not suitable for more specialized forest species. One could suggest a number of such limiting factors of which climate, with decrease of average temperature and increasing seasonality of rainfall, is probably the most important. ORIGIN OF CENTRAL BRAZILIAN FORESTS 183 Palynological evidence indicates that the levels of aridity in the central Brazilian region during most of the Wurm-Wisconsin glaciation were not as high as previously thought (Ledru, 1993). This could suggest that gallery forests remained in the region during this period, and other observations tend to support this. High plateaux predominate in central Brazil and these are known to concentrate rainfall and preserve humidity: for example, even within the present-day semi-arid Caatinga Province, the hinterland highlands have abundant water sources and both upland and gallery forests ('brejo' forests). In addition, the considerable number of endemic gallery species illustrated in the distribution maps also provide evidence of the survival of these forests during the glaciations. Nevertheless, there must have been a notable reduction in the extent of gallery forests during more arid periods, and where close observations are made it can be seen that they are at present in a phase of expansion (see Ratter, 1980, 1991). Probably this expansion has been accompanied by speciesenrichment with immigrants from the Amazon basin and southern mesophytic forests, which in turn has produced a strong floristic differentiation between the central-southern and western-northern galleries considered in the next paragraphs. The strongerfloristiclink between the central-southern galleries and the Paranense semideciduous forests, and especially its montane formations, gives evidence that lower temperatures were an important factor in the differentiation of those forests. Another source of evidence is given by the pollen record which indicates the establishment of the Araucaria forest in this region at the beginning of the last humid period (Ledru, 1993). The high altitude species found in the flora of galleries of the Espinhaco range and Planalto Central are probably the relics of the expansion of the Araucaria forest through central Brazil during the early Holocene. The pollen record also indicates the expansion of the mesophytic forests (deciduous and semideciduous) in the region following the retreat of the Araucaria forest. However, these forests were replaced by cerrado during the brief Holocenic dry episode and have not regained their area although rainfall apparently returned to previous levels. This failure of forest to re-establish may be man-induced, since certainly man and his fires were important elements in the region at least after the mid-Holocene. The occurrence of semideciduous forests on dry soils of very low fertility in the cerrado region of southern Minas Gerais (Oliveira-Filho et al., 1994b, c) demonstrates that the present-day environmental conditions favour the establishment of forests in much of the region and that factors unrelated to the physical environment are preventing the expansion of forest into cerrado. The western-northern galleries are set in a wide region where the complex transition between the central Brazilian cerrados and the Amazonian Hylaea takes place (Ratter et al., 1973, 1978a; Ackerly et al., 1989). All rivers of this part of the cerrado region flow northwards to the Amazon basin, so it is easy to understand the strong floristic links of the galleries to the Amazonian flora. An interesting floristic pattern that emerged from the analysis is that connecting galleries, dystrophic cerradoes and transitional open rainforests ('mata seca', or dry forest). Thisfloristicpattern apparently unites species related to very infertile, often sandy, soils of quite high moisture 184 A. T. OLIVEIRA-FILHO & J. A. RATTER content. Palynological records for Carajas, in eastern Amazonia (Absy et al., 1991) give evidence of forest retreat during the Wurm-Wisconsin glaciation period, while the work of Ratter et al. (1973, 1978a) demonstrates a surprisingly rapid presentday expansion of Amazonian forest in eastern Mato Grosso. As is known, the situation has been extremely dynamic and at times the galleries of Amazonia probably constituted important forest refuges (Pires, 1984). An important aspect of gallery forests is that they have interfaces with so many other types of vegetation, including rainforests, mesophytic forests and the cerrado itself. They are thus subjected to enormously different floristic influences resulting in great heterogeneity: the gallery forests undoubtedly harbour a remarkably diverse flora. In addition to this, they are an extremely important habitat for the cerrado fauna and act as migration corridors. Thus any policy aimed at conservation of biodiversity in Brazil must give high priority to protection of galleries and riparian forests. ACKNOWLEDGEMENTS The authors are very grateful to a number of people whose contribution to our work was of fundamental importance. Tania Sanaiotti (INPA, Manaus), Arnildo Pott (EMBRAPA/CPAP, Corumba), Manoela Silva and Ubirajara Maciel (both from Museu Emilio Goeldi, Belem) obtained checklists of difficult access. Jeanine Felfili and Manoel Claudio Silva Jr (both from the University of Brasilia), Paulo Ramos (IBAMA, Brasilia), Simon Mayo and Mike Hopkins (both from the Royal Botanic Gardens, Kew) kindly made available valuable unpublished checklists. We were helped during the taxonomic revision of the database by Washington MarcondesFerreira, from the University of Sao Paulo at Ribeirao Preto (Aspidosperma), Douglas Daly, from the New York Botanical Garden (Burseraceae), Carlos Reynel, from the Missouri Botanical Garden (Zanthoxylum), Carolyn Proenca, from the University of Brasilia (Myrtaceae), and by Gwil Lewis (Leguminosae), Terry Pennington (Inga and Sapotaceae) and Daniela Zappi (Cactaceae and Rubiaceae), all from the Royal Botanic Gardens, Kew. We received advice on multivariate analyses from Tom Dargie during an enjoyable stay with him and his family in Scotland's far-north. The facilities offered by the Royal Botanic Garden Edinburgh, and the kind help received from its staff during the period of this work, are gratefully acknowledged by the first author. The superb analysis of the dry seasonal forests of South America by Darien Prado and Peter Gibbs (University of St Andrews) was a permanent source of inspiration. REFERENCES AB'SABER, A. N. (1979). Os mecanismos da desintegra^ao das paisagens tropicais no Pleistoceno: Efeitos paleoclimaticos do periodo Wurm-Wisconsin no Brasil. Inter-Facies escitos e documentos 4: 1-11. ORIGIN OF CENTRAL BRAZILIAN FORESTS 185 A B ' S A B E R , A. N . (1980). Razoes da retomada parcial da semi-arides holocenica, por ocasi&o do otimum climaticum, primeiras ideias. Inter-Fades escritos e documentos 8: 1-7. A B ' S A B E R , A . N . (1982). The paleoclimate and paleoecology of Brazilian Amazonia. In: PRANCE, G. T. (ed.) Biological Diversification in the Tropics, pp. 41-59. New York: Plenum Press. A B S Y , M . L. (1982). Quaternary palynological studies in the Amazon Basin. In: PRANCE, G. T. (ed.) Biological Diversification in the Tropics, pp. 67-73. New York: Plenum Press. ABSY, M. L., CLEEF, A., FOURNIER, M., MARTIN, L., SERVANT, M., SIFFEDINE, A., SILVA, M. F. F., SOUBIES, F., SUGUIU, K., TURCQ, B. & VAN DER HAMMEN, T. (1991). Mise en evidence de quatre phases d'ouverture de la foret dense dans le sud-est de l'Amazonie au cours des 60000 derniere annees. Premiere comparison avec d'autres regions tropicales. Compt. Rend. Acad. Sci., Serie 2 t. 312: 673-678. ABSY, M. L., PRANCE, G. T. & BARBOSA, E. M. (1986/87). Inventario floristico de floresta natural na area da estrada Cuiaba-Porto Velho (BR 364). Acta Amazon. 16/17 (suppl.): 85-121. ABSY, M. L. & VAN DER HAMMEN, T. (1976). Some palaeoecological data from Rondonia, southern part of the Amazon Basin. Acta Amazon. 6: 293-299. ACKERLY, D. D., THOMAS, W. W., FERREIRA, C. A. C. & PIRANI, J. R. (1989). The forest-cerrado transition zone in southern Amazonia: results of the 1985 Projeto Flora Amazonica expedition to Mato Grosso. Brittonia 41: 113-128. ALLEM, A. C. (1976). Uma especie unica de Maprounea (Euphorbiaceae) na America do Sul. Acta Amazon. 6: 417-422. ANDRADE-LIMA, D. (1960). Estudosfitogeograficosde Pernambuco. Arq. Inst. Pesq. Agron. 5: 305-341. ANDRADE-LIMA, D. (1964). Contribuicao a dinamica da flora do Brasil. Arq. Inst. Ci. Terra 2: 15-20. ANDRADE-LIMA, D. (1966). Contribuicao ao estudo do paralelismo da flora Amazonico-Nordestina. Boletim Tecnico 19, Nova Serie do Instituto de Pesquisa Agronomica, Recife. ANDRADE-LIMA, D. (1982). Present-day forest refuges in Northeastern Brazil. In: PRANCE, G. T. (ed.) Biological Diversification in the Tropics, pp. 245-251. New York: Plenum Press. ASSIS, M. A. (1991). Fitossociologia de um remanescente de mata ciliar do Rio Ivinheima, MS. Unpublished Masters thesis, Universidade Estadual de Campinas. AZEVEDO, L. G., RIBEIRO, J. F., SCHIAVINI, I. & OLIVEIRA, P. E. A. M. (1990). Levantamento da vegetacSo do Jardim Botdnico de Brasilia - DF. Brasilia: Fundacao Zoobotanica. BAITELLO, J. B., PASTORE, J. A., AGUIAR. O. T., SERIO, F. C. & SILVA, C. E. F. (1988). A vegetacao arborea do Parque Estadual do Morro do Diabo, municipio de Teodoro Sampaio, Estado de Sao Paulo. Acta Bot. Brasil. 1: 221-230. BALLE, W. (1986). Analise preliminar de inventario florestal e a etnobotanica Ka'apor (Maranhao). Bol. Mus. Paraense Emilio Goeldi, serie Botdnica 2: 141-167. BALLE, W. (1987). A etnobotanica quantitativa dos indios tembe (Rio Gurupi, Para). Bol. Mus. Paraense Emilio Goeldi, serie Botdnica 3: 29-50. BARROS, F., MELO, M. M. R. F., CHIEA, S. A. C , KIRIZAWA, M., WANDERLEY, M. G. L. & JUNG-MENDACOLI, S. L. (eds) (1991). Flora fanerogdmica da Ilha do Cardoso, Vol. 1. Sao Paulo: Instituto de Botanica. 186 A. T. OLIVEIRA-FILHO & J. A. RATTER B E N S O N , W. W. (1982). Alternative models for infra-generic diversification in the humid tropics: Tests with passion vine butterflies. In: PRANCE, G. T. (ed.) Biological Diversification in the Tropics, pp. 608-640. New York: Plenum Press. B E R G , C. C. (1972). Olmedieae, Brosimeae (Moraceae). Flora Neotropica, Monograph No. 7, New York. B E R T O N I , J. E. A. & M A R T I N S , F . R . (1987). Composicao floristica de uma floresta riparia na Reserva Estadual de Porto Ferreira (SP). Ada Bot. Brasil. 1: 17-26. B E R T O N I , J. E. A . , M A R T I N S , F . R., M O R E S , J. L. & S H E P H E R D , G. J. (1988). Composicao floristica e estrutura fitossociologica do Parque Estadual de Vacununga, Santa Rita do Passa Quatro, SP - Gleba Praxedes. Bol. Teen. Inst. Florest. 42: 149-170. B E R T O N I , J. E. A . , S T U B B L E B I N E , W. H . , M A R T I N S , F . R. & L E I T A O F I L H O , H . F . (1982). Nota previa: comparacao fitossociologica das principais especies de florestas de terra firme e de varzea na Reserva Estadual de Porto Ferreira (SP). In: Anais do Congresso Nacional de Essencias Nativas, Campos do Jordao. Silvic. Sao Paulo 16A: 563-571. B E V E N , S., C O N N O R , E. F . & B E V E N , K. (1984). Avian biogeography in the Amazon basin and the biological model of diversification. J. Biogeogr. 11: 383-399. B I G A R E L L A , J. J. & A N D R A D E - L I M A , D . (1982). Paleoenvironmental changes in Brazil. In: PRANCE, G. T. (ed.) Biological Diversification in the Tropics, pp. 27-40. New York: Plenum Press. BIGARELLA, J. J., ANDRADE-LIMA, D. & RIEHS, P. J. (1975). Consideracoes a respeito das mudancas paleoambientais na distribuicao de algumas especies vegetais e animais no Brasil. In: Anais do Simposio Internacional sobre o Quaternario, Curitiba,1975. Anais Acad. Brasil. Ci. 47: 411-464. BOOM, B. M. (1986). A forest inventory in Amazonian Bolivia. Biotropica 18: 287-294. BOOM, B. M. (1987). Un inventario selvatico en la zona amazonica de Bolivia. Ecol. Bolivia 10: 1-14. BROWN, A. D., CHALUKIAN, S. C. & MALMIERCA, L. M. (1985). Estudo floristico-estrutural de un sector de selva semidecidua del noroeste argentine I. Composition floristica, densidad y diversidad. Darwiniana 26: 27-41. BROWN, K. S., JR & AB'SABER, A. N. (1979). Ice-age forest refuges and evolution in the Neotropics: correlation of paleoclimatological, geomorphological and pedological data with modern biological endemisms. Paleoclimas 5: 1-30. BUCHER, E. H. (1982). Chaco and Caatinga—South American arid savannas, woodlands and thickets. In: HUNTLEY, B. J. & WALKER, B. H. (eds) Ecology of Tropical Savannas, pp. 48-79. Berlin: Springer-Verlag. BUSCH, M. B. & COLINVAUX, P. A. (1990). A pollen record of a complete glacial cycle from lowland Panama. J. Veg. Sci. 1: 105-118. BUSCH, M. B., COLINVAUX, P. A., WIEMANN, M. C, PIPERNO, D. R. & LIU, K. B. (1990). Late Pleistocene temperature depression and vegetation change in Ecuadorian Amazonia. Quatern. Res. 34: 330-345. CABRERA, A. L. & WILLINK, A. (1973). Biogeografia de America Latina. Washington: Secretaria General de la Organization de los Estados Americanos. CAIN, S. A., CASTRO, G. M. O., PIRES, J. M. & SILVA, N. T. (1956). Application of some phytosociological techniques to Brazilian rain forest. Amer. J. Bot. 43: 911-941. CAMPBELL, D. G., DALY, D. C , PRANCE, G. T. & MACIEL, U. N. (1986). Quantitative ecological inventory of terra firme and varzea forest on the Rio Xingu, Brazilian Amazon. Brittonia 38: 369-393. O R I G I N OF C E N T R A L B R A Z I L I A N F O R E S T S 187 CAMPOS, J. C. C. & HEINDSDIJK, D. (1970). A floresta do Morro do Diabo. Silvic. Sao Paulo 7: 43-58. CARVALHO, D. A., OLIVEIRA-FILHO, A. T., VILELA, E. A. & G AVI LANES, M. L. (1992). Flora arbustivo-arborea das matas ciliares do Alto Rio Grande (MG). 1 - mata de Macaia (Bom Sucesso). In: Anais do 2° Congresso Nacional sobre Essencias Nativas, Sao Paulo. Revista Inst. Florest. 4: 274-282. CARVALHO, J. O. P., ARAUJO, S. M. & CARVALHO, M. S. P. (1986a). Estrutura horizontal de uma mata secundaria no planalto do Tapajos em Belterra, Para. In: Anais do Simposio do Tropico Umido, pp. 197-205. CARVALHO, J. O. P., SILVA, J. N. C , LOPES, J. C. A., MONTAGNER, L. H. & CARVALHO, M. S P. (1986b). Composicao floristica de uma mata secundaria no planalto de Belterra no Para. In: Anais do Simposio do Tropico Umido, pp. 207-215. CAVASSAN, O., CESAR. O. & MARTINS, F. R. (1984). Fitossociologia da vegetacao arborea da Reserva Estadual de Bauru, Estado de SSo Paulo. Revista Brasil. Bot. 7: 91-106. CESAR, O. & LEITAO F1LHO, H. F. (1990a). Estudo floristico quantitativo de mata mesofila semidecidua na Fazenda Barreiro Rico, municipio de Anhembi, SP. Revista Brasil. Biol. 50: 133-147. CESAR, O. & LEITAO FILHO, H. F. (1990b). Estudofitossociologicode mata mesofila semidecidua na Fazenda Barreiro Rico, municipio de Anhembi, SP. Revista Brasil. Biol. 50: 443-452. CETEC (1982). Levantamento da vegetacao do Parque Estadual do Rio Doce. Belo Horizonte: Fundacao Centro Tecnologico de Minas Gerais. COLINVAUX, P. (1989). The past and future Amazon. Sci. Amer. 260(6) (May): 102-108. CUNHA, C. N. (1990). Estudofloristicoefitofisionomicodas principals formacoes arboreas do Pantanal de Pocone - Mato Grosso. Unpublished Masters thesis, Universidade Estadual de Campinas. CUNHA, N. G., POTT, A. & GONCALVES, A. R. (1986). Solos calcimorficos da sub-regiao do Abobral, Pantanal Matogrossense. Circular Tecnica n° 19, EMBRAPACAP. DALY, D. C. & PRANCE, G. T. (1989). Brazilian Amazon. In: CAMPBELL, D. G. & HAMMOND, H. D. (eds) Floristic Inventory of Tropical Countries: The state of plant systematics, collections, and vegetation, plus recommendations for the future, pp. 402-426. New York: New York Botanical Garden. DANTAS, M. & MULLER, N.R.M. (1979). Estudosfito-ecologicosdo tropico umido brasileiro I - Aspectos fitossociologicos de mata sobre terra roxa na regiao de Altamira. In: Anais do 30° Congressso Nacional de Botdnica, pp. 205-208. Campo Grande. DANTAS, M., RODRIGUES, W. & MULLER, N. R. M. (1980). Estudos fitoecologicos do tropico umido brasileiro: aspectosfitossociologicosde mata sobre latossolo amarelo em Capitao Poco (PA). Boletim de Pesquisa da EMBRAPA/CPATU 9: 1-9. DICKINSON, R. E. & VIRJI, H. (1987). Climate change in the humid tropics, especially Amazonia, over the last twenty thousand years. In: DICKINSON, R. E. (ed.) The Geophysiology of Amazonia—Vegetation and climate interactions, pp. 91-101. New York: J. Wiley & Sons. DUBS, B. (1992). Observations on the differentiation of woodland and wet savanna habitats in the Pantanal of Mato Grosso, Brazil. In: FURLEY, P. A., PROCTOR, J. & RATTER, J. A. (eds) Nature and Dynamics of Forest-Savanna Boundaries, pp. 431-449. London: Chapman & Hall. 188 A. T. O L I V E I R A - F I L H O & J. A. RATTER E N D L E R , J. A . (1982). Pleistocene forest refuges: Fact or fancy? In: PRANCE, G. T. (ed.) Biological Diversification in the Tropics, pp. 641-657. New York: Plenum Press. FELFILI, J. M. (1990). (coord.) Biogeografia do bioma cerrado: Zoneamento dos recursos bioticos visando manejo e conservacao - Relatoriofinal:Chapada Pratinha. Scientific report. Brasilia: University of Brasilia. FELFILI, J. M. (1993a). Floristic composition and phytosociology of the gallery forest alongside the Gama stream in Brasilia, DF, Brazil. Revista Brasil. Bot. 17: 1-11. FELFILI, J. M. (1993b). Structure and dynamics of a gallery forest in Central Brazil. Unpublished doctorate thesis, University of Oxford. FELFILI, J. M. & SILVA, M. C , JR (1992). Floristic composition, phytosociology and comparison of cerrado and gallery forests at Fazenda Agua Limpa, Federal District, Brazil. In: FURLEY, P. A., PROCTOR, J. & RATTER, J. A. (eds) Nature and Dynamics of Forest-Savanna Boundaries, pp. 393-416. London: Chapman & Hall. FORERO, E. & GENTRY, A. H. (1987). Neotropical plant distribution patterns with emphasis on northwestern South America: a preliminary overview. In: VANZOLINI, P. E. & HEYER, W. R. (eds) Proceedings of a workshop on Neotropical distribution patterns, pp. 21-37. Rio de Janeiro: Academia Brasileira de Ciencias. FRANCESCHI, E. A. & LEWIS, J. P. (1979). Notas sobre la vegetation del valle santafesino del Rio Parana (Republica Argentina). Ecosur 6: 55-82. FURLEY, P. A., RATTER, J. A. & GIFFORD, D. R. (1988). Observations on the vegetation of eastern Mato Grosso, Brazil. III. The woody vegetation and soils of the Morro de Fumaca, Torixoreu. Proc. Roy. Soc. London B 235: 259-280. FURLEY, P. A., PROCTOR, J. & RATTER, J. A. (1992). (eds) Nature and Dynamics of Forest-Savanna Boundaries. London: Chapman & Hall. GALVAO, F., KUNIYOSHI, Y. S. & RODERJAN, C. V. (1989). Levantamento fitossociologico das principals associacoes arboreas da Floresta Nacional de Irati - PR. Revista Florest. 19: 30-49. GAVILANES, M. L., BRANDAO, M., OLIVEIRA-FILHO, A. T., ALMEIDA, R. J., MELLO, J. M. & AVEZUM, F. F. (1992a). Florula da Reserva Biologica Municipal do Poco Bonito, Lavras, MG. Ill - formacao florestal. Daphne 2: 14-26. GAVILANES, M. L., OLIVEIRA-FILHO, A. T., CARVALHO, D. A. & VILELA, E. A. (1992b). Flora arbustivo-arborea de uma mata ciliar do Alto Rio Grande, em Madre de Deus de Minas - MG. Daphne 2: 15-24. GIBBS, P. E. & LEITAO FILHO, H. F. (1978). Floristic composition of an area of gallery forest near Mogi Guacu, state of Sao Paulo, S.E. Brazil. Revista Brasil. Bot. 1: 151-156. GIULIETTI, A. M. & PIRANI, J. R. (1988). Patterns of geographic distribution of some species from the Espinhaco Range, Minas Gerais and Bahia, Brazil. In: VANZOLINI, P. E. & HEYER, W. R. (eds) Proceedings of a workshop on Neotropical distribution patterns, pp. 39-69. Rio de Janeiro: Academia Brasileira de Ciencias. GROMBONE, M. T., BERNACCI, L. C , MEIRA NETO, J. A., TAMASHIRO, J. Y. & LEITAO FILHO, H. F. (1990). Estrutura fitossociologica da floresta semidecidua de altitude do Parque Municipal da Grota Funda (Atibaia estado de Sao Paulo). Ada Bot. Brasil. 4: 47-64. HAASE, R. & BECK, S. G. (1989). Structure and composition of savanna vegetation in northern Bolivia: a preliminary report. Brittonia 41: 80-100. HAFFER, J. (1969). Speciation in Amazonian forest birds. Science 165: 131-137. HAFFER, J. (1982). General aspects of the refuge theory. In: PRANCE, G. T. (ed.) Biological Diversification in the Tropics, pp. 6-24. New York: Plenum Press. ORIGIN OF CENTRAL BRAZILIAN FORESTS 189 H A R L E Y , R. M . & M A Y O , S. J. (1980). Towards a Checklist of the Flora of Bahia. Royal Botanic Gardens, Kew. H I L L , M . O. (1979). TWINSPAN—a FORTRAN program for arranging multivariate data in an ordered two-way table by classification of individuals and attributes. Ithaca, New York: Cornell University. H I L L , M . O. & G A U C H , H . G . , J R (1980). Detrended correspondence analysis: an improved ordination technique. Vegetatio 42: 47-58. I B G E (1993). Mapa de vegetacao do Brasil. Rio de Janeiro: Fundacao Instituto Brasileiro de Geografia e Estatistica, Ministerio da Agricultura. K E N T , M . & C O K E R , P . (1992). Vegetation Description and Analysis, a Practical Approach. London: Belhaven Press. KLEIN, R. M. (1972). Arvores nativas da floresta subtropical do Alto Uruguai. Sellowia 24: 9-62. KLEIN, R. M. (1975). Southern Brazilian phytogeographic features and the probable influence of upper Quaternary climatic changes in the floristic distribution. Bol. Paran. Geoci. 33: 67-88. LEDRU, M. P. (1991). Etude de la pluiepollinique actuelle des forets du Bresil central: climat, vegetation, application a I'etude de V evolution paleoclimatique des 30.000 derniere annees. Unpublished doctorate thesis, Museum National d'Histoire Naturelle, Paris. LEDRU, M. P. (1993). Late Quaternary environmental and climatic changes in central Brazil. Quatern. Res. 39: 90-98. LEITAO-FILHO, H. F. (1987). Consideracoes sobre a floristica de florestas tropicais e subtropicais do Brasil. I.P.E.F. 35: 41-46. LEWIS, J. P. & PIRE, E. F. (1981). Resena sobre la vegetacion del Chaco santafesino. Serie Fitogeografica N° 18. Buenos Aires: Instituto Nacional de Tecnologia Agropecuaria. LISBOA, P. L. B. (1989). Estudo floristico da vegetacao arborea de uma floresta secundaria, em Rondonia. Bol. Mus. Paraense Emilio Goeldi, serie Botdnica 5: 145-162. LISBOA, P. L. B. (1990). Rondonia—colonizacao efloresta. Brasilia: CNPq, Programa Polonoroeste. LISBOA, P. L. B. & LISBOA, R. C. L. (1990). Inventarios florestais em Rondonia. I. Rodovia Presidente Medici - Costa Marques, Km 90. In: Anais do 35° Congressso Nacional de Botdnica, Manaus, 1984, pp. 205-229. Brasilia: IBAMA. LIU, K. B. & COLINVAUX, P. A. (1985). Forest changes in the Amazon Basin during the last glacial maximum. Nature 138: 556-557. MACIEL, U. N. & LISBOA, P. L. B. (1989). Estudo floristico de 1 ha de mata de terra firme no Km 15 da rodovia Presidente Medici - Costa Marques (RO-429), Rondonia. Bol. Mus. Paraense Emilio Goeldi, serie Botdnica 5: 25-37. MALLOCH, A. J. C. (1988). Vespan II, a computer package to handle and analyse multivariate species data and handle and display species distribution data. University of Lancaster. MANTOVANI, W., CATHARINO, E. L. M. & BORGES, H. B. N. (1986). Composicao floristica e fitossociologica, Sub-projeto 2. In: KAGEYMA, P. Y. (coord.). Estudo para implantacao de matas ciliares de protecao na bacia hidrogrdfica do Passa Cinco visando a utilizacao para abastecimento urbano, pp. 34-102. Scientific Report. Piracicaba: DAEE, USP, ESALQ. MANTOVANI, W., ROSSI, L., ROMANIUC NETO, S., ASSADLUDEWIGIS, I. W., WANDERLEY, M. G. L., MELO, M. M. R. & TOLEDO, C. B. (1989). Estudofitossociologicode areas de mata ciliar em MogiGuacu, SP, Brasil. In: BARBOSA, L. M. (coord.) Anais do Simposio sobre Mata Ciliar, pp. 235-267. Campinas: Fundacao Cargill. 190 A. T. O L I V E I R A - F I L H O & J. A. RATTER MARCHIORI, A. C. C , NASSAR, F.., JR & DEMATTE, M. E. S. P. (1992). Levantamento floristico efitossociologicodo estrato arboreo de mata residual do Viveiro Experimental da FCAV-UNESP, Jaboticabal, SP. In: Anais do VIII Congresso da Sociedade Botdnica de Sao Paulo, pp. 121-131. Campinas. MARKGRAF, V. (1989). Paleoclimates in Central and South America since 18000 BP based on pollen and lake-level records. Quatern. Sci. Rev. 8: 1-24. MARTINS, F. R. (1991). Estrutura de umafloresta mesofila. Campinas: Editora da UNICAMP—serie teses. MAYO, S. J. & FEVEREIRO, V. P. B. (1982). Mata de Pau-Ferro, A pilot study of the brejo forest of Paraiba, Brazil. Royal Botanic Gardens, Kew. MEIRA NETO, J. A. A., BERNACCI, L. C , GROMBONE, M. T., TAMASHIRO, J. Y. & LEITAO FILHO, H. F. (1989). Composicaofloristicada floresta semidecidua de altitude do Parque Municipal da Grota Funda (Atibaia, estado de Sao Paulo). Ada Bot. Brasil. 3: 51-74. MILLIKEN, W. & RATTER, J. A. (1989). The vegetation of the Ilha de Maraca. First report of the vegetation survey of the Maraca Rainforest Project. Royal Botanic Garden Edinburgh. MORELLATO, L. P. & ROSA, N. R. (1991). Caracterizacao de alguns tipos de vegetacao na regiao amazonica, Serra dos Carajas, Para, Brasil. Revista Brasil. Bot. 14: 1-14. MORI, S. A., BOOM, B. M., CARVALHO, A. M. & SANTOS, T. S. (1983). Southern Bahian moist forests. Bot. Rev. 49: 155-232. MOTA, A. L. P. (1984). Estudo dendrologico na mata da Casca d'Anta - Parque Nacional da Serra da Canastra, MG. Unpublished Masters thesis, Universidade Federal de Vicosa. OLIVEIRA-FILHO, A. T. (1989). Composicao floristica e estrutura comunitaria da floresta de galeria do corrego da Paciencia, Cuiaba (MT). Acta Bot. Brasil. 3: 91-112. OLIVEIRA-FILHO, A. T., ALMEIDA, R. J., MELLO, J. M. & GAVILANES, M. L. (1994a). Estruturafitossociologicae variaveis ambientais em um trecho da mata ciliar do corrego dos Vilas Boas, Reserva Biologica do Poco Bonito, Lavras (MG). Revista Brasil. Bot. 17: 67-85. OLIVEIRA-FILHO, A. T. & MACHADO, J. N. M. (1993). Composicao floristica de uma floresta semidecidua montana na Serra de Sao Jose, Tiradentes, Minas Gerais. Acta Bot. Brasil. 7: 71-88. OLIVEIRA-FILHO, A. T. & MARTINS, F. R. (1986). Distribuicao, caracterizacao e composicao floristica das formacoes vegetais da regiao da Salgadeira, na Chapada dos Guimaraes (MT). Revista Brasil. Bot. 9: 207-223. OLIVEIRA-FILHO, A. T. & MARTINS, F. R. (1991). A comparative study of five cerrado areas in southern Mato Grosso, Brazil. Edinb. J. Bot. 48: 307-332. OLIVEIRA-FILHO, A. T. & RATTER, J. A. (1994). Database: Woodyfloraof 106 forest areas of eastern tropical South America. Royal Botanic Garden Edinburgh. OLIVEIRA-FILHO, A. T., RATTER, J. A. & SHEPHERD, G. J. (1990). Floristic composition and community structure of a central Brazilian gallery forest. Flora 184: 103-117. OLIVEIRA-FILHO, A. T., SCOLFORO, J. R. & MELLO, J. M. (1994b). Composicao floristica e estrutura comunitaria de um remanescente de floresta semidecidua montana em Lavras (MG). Revista Brasil. Bot. 17: 169-184. OLIVEIRA-FILHO, A. T., VILELA, E. A., GAVILANES, M. L. & CARVALHO, D. A. (1994c). Comparison of the woody flora and soils of six areas of montane semideciduous forest in southern Minas Gerais, Brazil. Edinb. J. Bot. 51: 355-389. ORIGIN OF CENTRAL BRAZILIAN FORESTS 191 OLIVEIRA-FILHO, A. T., VILELA, E. A., GAVILANES, M. L. & CARVALHO, D. A. (1994d). Effect of flooding regime and understorey bamboos on the physiognomy and tree species composition of a tropical semideciduous forest in Southeastern Brazil. Vegetatio 113: 99-124. PAGANO, S. N. & LEITAO FILHO, H. F. (1987). Composicao floristica do estrato arboreo da mata mesofila semidecidua no municipio de Rio Claro (Estado de Sao Paulo). Revista Brasil. Bot. 10: 37-47. PAGANO, S. N., LEITAO FILHO, H. F. & SHEPHERD, G. J. (1987). Estudo fitossociologico em uma mata mesofila semidecidua no municipio de Rio Claro (Estado de Sao Paulo). Revista Brasil. Bot. 10: 49-61. PEIXOTO, A. L. & GENTRY, A. (1990). Diversidade e composicao floristica da mata de tabuleiro na Reserva Florestal de Linhares (Espirito Santo, Brasil). Revista Brasil. Bot. 13: 19-25. PEREIRA, B. A. S., MENDONCA, R. C , FILGUEIRAS, T. S., PAULA, J. E. & HERINGER, E. P. (1985). Levantamento floristico da area de Protecao Ambiental (APA) da Bacia do Rio Sao Bartolomeu, Distrito Federal. In: Anais do XXXVI Congresso Nacional de Botdnica, pp. 419-493. PINTO, M. M. (1989). Levantamentofitossociologicode uma mata residual: campus de Jaboticabal da UNESP. Unpublished Masters thesis, Universidade Estadual Paulista Julio de Mesquita Filho, Jaboticabal. PIRES, J. M. (1984). The Amazonian forest. In: SIOLI, H. (ed.) The AmazonLimnology and landscape ecology of a mighty tropical river and its basin, pp. 581-602. Dordrecht: Junk Publishers. PIRES, J. M. & KOURY, H. M. (1958). Estudo de urn trecho de mata de varzea proximo a Belem. Bol. Teen. Inst. Agron. N. 36: 3-44. PIRES, J. M. & PRANCE, G. T. (1977). The Amazon forest: a natural heritage to be preserved. In: PRANCE, G. T. (ed.) Extinction is forever—Threatened and endangered species of plants in the Americas and their significance in ecosystems today and in the future, pp. 158-213. New York: New York Botanical Garden. PORTO, M. L., LONGHI, H. M., CITADINI, V., RAMOS, R. F. & MARIATH, J. E. A. (1976). Levantamentofitossociologicoem area de mata de baixio na Estacao Experimental de Silvicultura Tropical - INPA - Manaus - Amazonas. Acta Amazon. 6: 301-318. PRADO, D. E. (1991). A critical evaluation of thefloristiclinks between Chaco and Caatingas vegetation in South America. Unpublished doctorate thesis, University of St Andrews. PRADO, D. E., FRANCESCHI, E. A. & BIANCHI, M. B. (1989). El bosque del escarpe occidental del Rio Parana (Argentina). Composition floristica y estrutura. (1). Anais do XL Congresso Nacional de Botanica. Acta Bot. Brasil. 3: 99-108. PRADO, D. E. & GIBBS, P. E. (1993). Patterns of species distribution in the dry seasonal forests of South America. Ann. Missouri Bot. Gard. 80: 902-927. PRADO, D. E., GIBBS, P. E., POTT, A. & POTT, V. J. (1992). The ChacoPantanal transition in southern Mato Grosso, Brazil. In: FURLEY, P. A., PROCTOR, J. & RATTER, J. A. (eds) Nature and Dynamics of Forest-Savanna Boundaries, pp. 431-470. London: Chapman & Hall. PRANCE, G. T. (1973). Phytogeographic support for the theory of Pleistocene forest refuges in the Amazon basin, based on evidence from distribution patterns in Caryocaraceae, Chrysobalanaceae, Dichapetalaceae and Lecythidaceae. Acta Amazon. 3: 5-28. PRANCE, G. T. (1977). The phytogeographic subdivisions of Amazonia and their 192 A. T. O L I V E I R A - F I L H O & J. A. RATTER influence on the selection of biological reserves. In: PRANCE, G. T. (ed.) Extinction is forever—Threatened and endangered species of plants in the Americas and their significance in ecosystems today and in the future, pp. 195-213. New York: New York Botanical Garden. PRANCE, G. T., RODRIGUES, W. A. & SILVA, M. F. (1976). Inventario florestal de urn hectare de mata de terra firme km 30 da Estrada Manaus - Itacoatara. Ada Amazon. 6: 9-35. PRANCE, G. T. & SCHALLER, G. B. (1982). Preliminary study of some vegetation types of the Pantanal, Mato Grosso, Brazil. Brittonia 34: 228-251. RAMBO, B. (1961). Migration routes of the south Brazilian rain forests. Pesquisas Botdnicas 12. Porto Alegre. RAMBO, B. (1980). A mata pluvial do Alto Uruguai. Roessleria 3: 101-139. RAMOS, P. C. M. (1989). Estudosfitossociologicosem umafloresta mesofitica semidecidua na FERCAL, Brasilia - DF. Unpublished Masters thesis, University of Brasilia. RAMOS, P. C. M. (1994). Vegetation communities and soils of Brasilia National Park. Unpublished doctorate thesis, University of Edinburgh. RATTER, J. A. (1971). Some notes on two types of cerradao occurring in northeastern Mato Grosso. In: FERRI, M. G. (ed.) Ill Simposio sobre o cerrado, pp. 100-102. Sao Paulo: Editora Universidade de Sao Paulo. RATTER, J. A. (1980). Notes on the vegetation of Fazenda Agua Limpa (Brasilia, DF, Brazil). Royal Botanic Garden Edinburgh. RATTER, J. A. (1987). Notes on the vegetation of the Parque Nacional do Araguaia (Brazil). Notes Roy. Bot. Gard. Edinburgh 44: 311-342. RATTER, J. A. (1991). Guia para a vegetacdo da Fazenda Agua Limpa (Brasilia, DF), com uma chave para os generos de dicotiledoneas do cerrado. Brasilia: Textos Universitarios, Editora Universidade de Brasilia. RATTER, J. A. (1992). Transitions between cerrado and forest vegetation in Brazil. In: FURLEY, P. A., PROCTOR, J. & RATTER, J. A. (eds) Nature and Dynamics of Forest-Savanna Boundaries, pp. 417-429 . London: Chapman & Hall. RATTER, J. A., ASKEW, G. P., MONTGOMERY, R. F. & GIFFORD, D. R. (1977). Observacoes adicionais sobre o cerradao de solo mesotrofico no Brasil Central. In: FERRI, M. G. (ed.) IV Simposio sobre o Cerrado, pp. 303-316. Sao Paulo: Ed. Univ. Sao Paulo. RATTER, J. A., ASKEW, G. P., MONTGOMERY, R. F. & GIFFORD, D. R. (1978a). Observations on the vegetation of northeastern Mato Grosso, II. Forests and soils of the Rio Suia-Missu area. Proc. Roy. Soc. London B 203: 191-208. RATTER, J. A., ASKEW, G. P., MONTGOMERY, R. F. & GIFFORD, D. R. (1978b). Observations on forests of some mesotrophic soils in central Brazil. Revista Brasil. Bot. 1: 47-58. RATTER, J. A. & DARGIE, T. C. D. (1992). An analysis of the floristic composition of 26 cerrado areas in Brazil. Edinb. J. Bot. 49: 235-250. RATTER, J. A., POTT, A., POTT, V. J., CUNHA, C. N. & HARIDASAN, M. (1988). Observations on woody vegetation types in the pantanal and at Corumba, Brazil. Notes Roy. Bot. Gard. Edinburgh 45: 503-525. RATTER, J. A., RICHARDS, P. W., ARGENT, G. & GIFFORD, D. R. (1973). Observations on the vegetation of northeastern Mato Grosso, I. The woody vegetation types of the Xavantina-Cachimbo Expedition Area. Philos. Trans. Roy. Soc. London B 226: 449-492. REIS, A. C. S. (1971). Climatologia dos cerrados. In: FERRI, M. G. (ed.) Ill Simposio sobre o Cerrado, pp. 15-26. Sao Paulo: Ed. Universidade de Sao Paulo. O R I G I N OF CENTRAL BRAZILIAN FORESTS 193 RIZZINI, C. T. (1963). Nota previa sobre a divisaofitogeograficado Brasil. Revista Brasil. Geogr. 25: 1-64. RIZZINI, C. T. (1979). Tratado defitogeografia do Brasil. 2° v. Aspectos sociologicos e floristicos. Sao Paulo: HUCITEC-EDUSP. RIZZINI, C. T. & PINTO, M. M. (1964). Areas climatico-vegetacionais do Brasil segundo os metodos de Thornthwaite e Mohr. Revista Brasil. Geogr. 26: 523-547. RODRIGUES, R. R. (1986). Levantamentofloristicoefitossociologicodas matas da Serra do Japi, Jundiai, SP. Unpublished Masters thesis, Universidade Estadual de Campinas. RODRIGUES, R. R. (1991). Andlise de um remanescente de vegetacao natural as margens do Rio Passa Cinco, Ipeuna, SP. Unpublished doctorate thesis, Universidade Estadual de Campinas. RODRIGUES, R. R., MORELATO, L. P. C , JOLY, C. A. & LEITAOFILHO, H. F. (1989). Estudo floristico efitossociologicoem um gradiente altitudinal de mata estacional mesofila semidecidua na serra do Japi, Jundiai. Revista Brasil. Bot. 12: 71-84. RODRIGUES, W. A. (1963). Estudo de 2,6 hectares de mata de terra firme da Serra do Navio, Territorio do Amapa. Bol. Mus. Paraense Emilio Goeldi 19: 1-22. ROHLF, F. J. (1992). NTSYS-pc, numerical taxonomy and multivariate analysis system, version 1.70. Setanket, New York: Exeter Software. SALOMAO, R. P. (1991). Uso de parcelas permanentes para estudos da vegetacao da floresta tropical umida. I. Municipio de Maraba, Para. Bol. Mus. Paraense Emilio Goeldi, serie Botdnica 7: 543-604. SALOMAO, R. P. & LISBOA, P. L. B. (1988). Analise ecologica da vegetacao de uma floresta pluvial tropical de terra firme, Rondonia. Bol. Mus. Paraense Emilio Goeldi, serie Botdnica 4: 195-233. SALOMAO, R. P. & ROSA, N. (1989). Analise da vegetafao de floresta pluvial tropical de terra firme pelo metodo dos quadrantes: Serra Norte, Carajas, PA. Ada Bot. Brasil. 2(1): 27-42 (suplemento). SALOMAO, R. P., SILVA, M. F. F. & ROSA, N. A. (1988). Inventario ecologico em floresta pluvial tropical de terra firme, Serra Norte, Carajas, Para. Bol. Mus. Paraense Emilio Goeldi, serie Botdnica 4: 1-46. SASTRE, C. (1975). Etude du genre Cespedesia Goudot (Ochnaceae). Boletin Cientifico n° 4 del Departamento del Valle del Cauca, Colombia. SCHIAVINI, I. (1992). Estrutura das comunidades arboreas de mata de galeria da Estacao Ecologica do Panga (Uberldndia, MG). Unpublished doctorate thesis, Universidade Estadual de Campinas. SILVA, A. F. (1989). Composicdofloristicae estruturafitossociologicado estrato arboreo da Reserva Florestal Professor Augusto Ruschi, Sao Jose dos Campos, SP. Unpublished doctorate thesis, Universidade Estadual de Campinas. SILVA, A. F. & LEITAO FILHO, H. F. (1982). Composicao floristica e estrutura de um trecho de mata atlantica de encosta no municipio de Ubatuba (Sao Paulo, Brasil). Revista Brasil. Bot. 5: 55-94. SILVA, M. C , JR, SCARANO, F. R. & CARDEL, F. S. (1994). Regeneration of an Atlantic forest formation in the understorey of a Eucalyptus grandis plantation in southeastern Brazil. J. Trop. Ecol. 11: 147-152. SILVA, M. F. F. & ROSA, N. A. (1989). Analise do estrato arboreo da vegetacao sobre jazidas de cobre na Serra dos Carajas - PA. Bol. Mus. Paraense Emilio Goeldi, serie Botdnica 5: 175-206. SILVA, M. F. F., ROSA, N. A. & OLIVEIRA, J. (1987). Estudos botanicos na 194 A. T. O L I V E I R A - F I L H O & J. A. RATTER area do Projeto Ferro Carajas. 5. Aspectos floristicos da mata do Rio Gelado, Para. Bol. Mus. Paraense Emilio Goeldi, serie Botdnica 3: 1-20. SOARES-SILVA, L. H. & BARROSO, G. M. (1992). Fitossociologia do estrato arboreo da floresta na porcao norte do Parque Estadual Mata dos Godoy, Londrina PR, Brasil. In: Anais do VIII Congresso da Sociedade Botdnica de Sao Paulo, pp. 101-112. Campinas. TORRES, R. B., MARTINS, F. R. & GOUVEA, L. S. K. (1995). Spatial partition of forest tree species in the state of Sao Paulo, Southeastern Brazil. Revista Brasil. Bot. (in press). VAN DER HAMMEN, T. (1974). The Pleistocene changes of the vegetation and climate in tropical South America. J. Biogeogr. 1: 3-26. VAN DER HAMMEN, T. (1982). Palaeoecology of tropical South America. In: PRANCE, G. T. (ed.) Biological Diversification in the Tropics, pp. 60-66. New York: Plenum Press. VAN DER HAMMEN, T. (1983). The palaeoecology and palaeogeography of savannas. In: BOURLIERE, F. (ed.) Tropical Savannas, pp. 19-35. Amsterdam: Elsevier. VANZOLINI, P. E. (1963). Problemas faunisticos do cerrado. In: FERRI, M. G. (ed.) Simposio sobre o Cerrado, pp. 307-320. Sao Paulo: Ed. Universidade de Sao Paulo. VELOSO, H. P. (1966). Atlasflorestaldo Brasil. Rio de Janeiro: Ministerio da Agricultura. VELOSO, H. P., RANGEL FILHO, A. L. R. & LIMA, J. C. A. (1991). Classificacdo da vegetacao brasileira adaptada a um sistema universal. Rio de Janeiro: IBGE. VILELA, E. A., OLIVEIRA-FILHO, A. T., CARVALHO, D. A. & GAVILANES, M. L. (1995). Flora arbustivo-arborea de uma mata ciliar do Alto Rio Grande em Itutinga - MG. Ada Bot. Brasil. (in press). WERNER, P. A. (ed.) (1991). Savanna Ecology and Management, Australian perspectives and intercontinental comparisons. London: Blackwell. WORBES, M., KLINGE, H., REVILLA, J. D. & MARTIUS, C. (1992). On the dynamics, floristic subdivision and geographical distribution of varzea forests in Central Amazonia. /. Veg. Sci. 3: 553-564.