Universidade Federal do Acre

Floristics and Economic Botany of Acre, Brazil
Florística e Botânica Econômica do Acre, Brasil

Botanists tend to use four types of floristic data to serve four distinct types of analyses related to the selection of priority areas for conservation:

• absolute floristic diversity
• diversity by spacial units (alpha-diversity)
• the occurrence and distribution of "habitat specialists", i.e., taxa restricted to certain vegetation types
• discrete and recurring patterns of geographic distribution, including endemism

Siting and policy for conservation depend particularly heavily on distribution patterns and endemism, but as Peres and Terborgh (1995) pointed out, the identification of priority areas for conservation in Amazonia has been based to a large extent on the biogeography of a limited number of taxa. Regions where the flora is inadequately sampled and documented, and where the taxonomy of some of the key groups of organisms is poorly resolved, present a special challenge for efforts to draw scientifically sound conclusions that will afftect regional conservation.

There is no question that some endemisms in Amazonia and elsewhere are actually artifacts of low collecting densities and intensities (e.g., Nelson et al. 1992), and the unfortunate reality is that this situation is not likely to change in the near future, as the level of botanical activity in Amazonia declined dramatically after Projeto Flora Amazônica effectively came to a close in 1987. The distributions cited herein will need to be re-examined on a yearly basis until the flora of southwestern Amazonia in particular is far better documented and represented in the world's herbaria. Meanwhile, regional efforts to guide conservation policy should combine all of the following three strategies, which we are in the process of implementing in Acre; the present analysis focuses on the first:

1. analyze available floristic data to ensure conservation of centers of endemism and centers of diversity, and maximum representation of components of the flora with different affinities;
2. use or develop vegetation maps to ensure maximum representation of the diversity of vegetation cover types (e.g., Oliveira and Nelson, in prep.);
3. calculate collecting densities in the region by political division (e.g., municipality), river basin, and vegetation cover type, and increase botanical survey (i.e., collecting) activity in under-represented areas in one or another category.

The systematic investigation of geographic distributions of the species that compose a flora requires a tremendous investment of time in the medium and long term. In advance of analyzing the entire flora as it is know to date, we employed a straightforward approach to reveal these patterns for an initial level of sampling intensity. First, we reviewed the distribution maps in the Flora Neotropica monograph series. Then, starting with those collections identified to species by taxonomic specialists, we used filters to select for taxa whose specific epithets might provide an indication of a non-widespread distribution:

• allusions to geographic locations unique to the region (e.g., juruana, purusana, acreana, acreensis, boliviana, peruana)
• locations that might indicate specific affinities outside of southwestern Amazonia (e.g., andina, cearensis, paraguariensis, tarapotana)
• references to naturalists and botanists who collected in SW Amazonia (e.g., ulei, tessmannii, poeppigii) or who collected only in regions that might indicate specific affinities outside of SW Amazonia (e.g., burchellii or sellowii for drier parts of Brazil, steinbachianus for northern Bolivia, klugii or schunkei for parts of Amazonian Peru)

We used this approach successfully in our 1997 report on floristic diversity in the Parque Nacional Serra do Divisor. This document was submitted to the SOS Amazônia/The Nature Conservancy/IBAMA consortium responsible for the national park's management plan.

Once the candidate species had been selected, their distributions were verified using recent taxonomic literature when available, the NYBG herbarium, and the Missouri Botanical Garden's on-line data-base TROPICOS, which provides lists of exsiccatae with localities for MO's holdings. NY houses the best most complete representation of the Amazonian flora, particularly for Brazil, and the MO herbarium is very strong in other parts of lowland northern South America. Information on vegetation types was gathered from herbarium label information, personal experience, vegetation maps, and correspondence with other botanists.

We selected for species with non-widespread distributions, i.e., we disregarded species widespread throughout Amazonia, the Neotropics, or beyond. Our conclusions are based on the analysis of the 171 species remaining from the more than 250 distributions examined.

Paradoxically, while the state as a whole stands firmly in the Southwestern Amazon Ecoregion, a number of factors divide Acre into two parts, and the distinctiveness of the two parts of the state adds to the region's importance as part of a biological corridor. The northwestern portion of Acre is drained by the Rio Juruá, while the southeastern portion is drained by the Rio Purus and by the Rio Acre sub-basin of the Purus. The two parts of the state are divided by the Fitzcarrald Paleoarch (Patton et al. 1997). The upper Juruá has essentially no dry season and relatively rich, young soils overall. The western frontier with Peru is marked by low mountains, the Serra do Moa and Serra do Divisor; they are home to some typically Andean species and genera occurring in specialized vegetation types that include cliffside communities and dwarf fire-climax ridge-top forests. Campina and Amazonian caatinga formations on white sand occur around at least six points on the upper Juruá in Acre; their flora is related to that of similar formations on the upper Rio Madeira, the Rio Negro, and elsewhere.

Acre is a center of diversity for palms. It is home to more species of palms than all of Bolivia, and at least 70% of the western Amazonian palm flora is found in the upper Juruá basin, including the Serra do Divisor and Serra do Moa (E. Ferreira, pers. comm.).

The upper Rio Purus/Rio Acre region has a pronounced dry season, poorer soils, and flat to undulating terrain where some upland forests are semi-deciduous and a number of species show affinities to dry forests and savannas, particularly south of Amazonia.

Many species occurring in Acre are found in the upper Purus or upper Juruá sectors of the state, but not both. An example of the former is the Brazil nut, Bertholletia excelsa H. B. K. (Lecythidaceae), while an example of the latter is the Western Amazonian palm Attalea tessmannii, found by botanists for the second time only in 1992.

Our analyses revealed nine rather discrete patterns:

• rare (5 collections or fewer) and relatively narrowly endemic to part of Acre plus contiguous portions of Bolivia (Pando, northern Beni) or Peru (Madre de Dios, Ucayali)
• rare and restricted to southwestern Amazonia - includes Amazonian parts of southeastern Peru (in Madre de Dios, Ucayali (formerly southern Loreto), San Martín, Huánuco, Junín, Cusco), Bolivia (Pando, northern Beni, and Amazonian parts of Santa Cruz and La Paz), and Brazil (contiguous Amazonas, parts of northern Mato Grosso and Rondônia)
• restricted to southwestern Amazonia, but not rare
• restricted to western Amazonia (western Amazonas, Brazil and including Amazonian Bolivia, Peru, Ecuador, and Colombia)
• rare species with apparently disjunct distributions
• affinities with (i.e., distributions in) the Andes and/or Andean piedmont
• affinities with formations on white sand: campinas, Amazonian caatingas (campinaranas), and hyperseasonal savannas
• affinities with drier formations (dry/semi-deciduous forests, cerrado vegetation, savannas, and caatinga)

In the following discussions, the authors of the species are provided only if the taxon does not appear in the tables.

Rare and relatively narrowly endemic (Table 1)

The most striking pattern in our analysis is formed by rare species apparently restricted to parts of Acre, sometimes immediately adjacent parts of Bolivia (Pando) or Peru (Madre de Dios, Ucayali [formerly southern Loreto]) or rarely Amazonas, Brazil. Each of the species listed in Table 1 is represented by five collections or fewer. These species correspond to 15.2% of the non-widespread distributions analyzed. This confirms the status of Acre as a floristic "hot spot" where an adequate sampling of the flora will reveal a large number of botanical novelties. Some are known only from the type collection, whether it was collected in 1901 (e.g., Eugenia pleurosiphonia) or 1996 (e.g., Piper sp. nov., Ehringhaus 290). The basins of the Rio Purus and Rio Juruá present comparable numbers of rare and endemic species. The information available to date suggests that the municípios most likely to yield new species in the near future are Marechal Taumaturgo, Tarauacá, and Porto Valter in the Juruá basin, and Assis Brasil, Brasiléia, and Sena Madureira in the Purus basin.

Among the rare, apparently narrow endemics is Psittacanthus amazonicus (Loranthaceae), first collected in the upper Rio Juruá by Ernst Ule in 1901, and for the second time in 1992, in the same locality. Aphelandra acrensis (Acanthaceae) is known only in Acre, near the upper Rio Acre (Purus basin) and the middle Tarauacá (Juruá basin).

The species that appear to extend only a short distance beyond Acre include two small trees found in the understory of terra firme forests. Guapira uleana (Nyctaginaceae) has been found outside Acre a short distance into the state of Amazonas, while Erythrochiton trichanthus (Rutaceae) is known outside of Acre only in bordering Madre de Dios, Peru,in the Parque Nacional de Manu. Similarly, the palm Astrocaryum faranae is both known only from the upper Rio Juruá in Acre and in contiguou Peru.

Restricted to southwestern Amazonia (Tables 1, 2, 3)

After the broadly defined Western Amazon pattern, the predominant non-widespread distribution pattern falls within the Southwestern Amazonia Eco-region, defined as comprising contiguous Brazil, Peru, and Bolivia (Dinerstein et al. 1995); this means that the Acre flora has stronger floristic affinities with southeastern Peru and northern Bolivia than with the remainder of the Brazilian Amazon. The percentages of species that are (1) rare and relatively narrowly endemic, (2) rare and endemic to southwestern Amazonia, and (3) restricted to southwestern Amazonia but not rare add up to (30.4%).

Species may have narrower or broader ranges within this Eco-Region. Disciphania cubijensis (Menispermaceae) is one of the latter, occurring in Acre in the Rio Tarauacá basin but also in Madre de Dios (Tambopata) in Peru and in Pando (Cobija) in Bolivia, extending into the Amazonian parts of Santa Cruz in Bolivia. Similarly, Byrsonima schunkei (Malpighiaceae) is found in Acre in the municípios of Porto Valter, Mâncio Lima, and Rio Branco, as well as in Peru in southern San Martín (Tocache Nuevo) and Huánuco (lowlands north of Tingo María), and in Boliva in Pando. Both Malmea dielsiana (Annonaceae) and Basistemon peruvianus (Scrophulariaceae) occur in Acre (Sena Madureira and Brasiléia) and in Peru (Madre de Dios, southern San Martín, and Ucayali). Anthurium croatii (Araceae) is a terrestrial herb rare in theforest understory in Peru (southern Loreto=Ucayali, Junín, and San Martin), Acre, and eastern La Paz in Bolívia.

Restricted to western Amazonia (Tables 1-4)

The number of endemic species represented in Acre, rare or not, greatly increases when one expands the focus to all of western Amazonia, including western Amazonas in Brazil, the rest of Amazonian Peru, northern Bolivia, and the Amazonian parts of Ecuador and Colombia. This pattern can also include Brazil's western Amazonas State, Rondônia, and (rarely) northern Mato Grosso. It is the predominant pattern encountered, accounting for 35.1% of the non-widespread distributions.

Fittonia albivenis (Acanthaceae), a small terrestrial herb usually in poorly drained soils, occurs in Ecuador (Morona-Santiago and Napo), Peru (Amazonas, Loreto, Madre de Dios, Ucayali) and Acre (Sena Madureira, Marechal Taumaturgo). Schoenobiblus peruvianus (Thymelaeaceae) is a small terra firme forest tree found in Acre (Marechal Taumaturgo and Brasiléia), Colombia (Amazonas), Ecuador (Napo), and Peru (Huánuco, Loreto, Madre de Dios, and Pasco). Eschweilera juruensis (Lecythidaceae) occurs in Acre (Sena Madureira), Amazonas State (Rio Envira, Rio Javari, and upper Rio Madeira), Colômbia (Amazonas), Ecuador (Napo), and Peru (Amazonas, Huánuco, Loreto, and Madre de Dios). Erythroxylum fimbriatum (Erythroxylaceae) is found in Acre (Brasiléia), Rondônia, Ecuador (Napo), and Peru (Amazonas, Loreto, Huánuco, and San Martín).

Rare species with apparently disjunct distributions (Table 5)

A non-trivial number of species in the Acre flora show surprisingly disjunct distributions. In the search for explanations, the first possibility that must be considered is that it may be an artefact of collecting; the floristic inventory of Amazonia is still woefully inadequate, and the distribution may turn out to be more continuous than appears now. Other than invoking long-distance dispersal, a given "real" disjunction may be explained in one of several ways, two of which are exemplified by these Acre disjuncts.

Guzmania vittata (Bromeliaceae) is disjunct between Acre and Caquetá in Colombia, and Pariana ulei (Chryosobalanaceae) between Acre, Pará, and the Rio Negro region of Amazônas. These two species may be white-sand habitat specialists, many of which are widespread but have interrupted distributions congruent with those of campinas and other formations on white sand.

Ficus gameleira (Moraceae) is disjunct between Acre and Maranhão and Taccarum ulei (Araceae) between Acre and Piauí. Maranhão has pre-Amazon forests and cerrado vegetation, while Piauí has cerrado and caatinga. These disjunctions may be relicts of southern Amazonian dry-formation distributions now interrupted during the current inter-glacial.

The remaining disjunctions are with other portions of Amazonia (although Pouteria polysepala (Sapotaceae) is also recorded from Esmeraldas in Ecuador), and a number of historic phenomena have been postulated that would have fragmented distributions of Amazonian biota in various patterns and via various mechanisms (see review in Oliveira & Daly 1999).

Affinities with the Andes and/or the Andean piedmont (Table 6)

Andean affinities are difficult to decipher, because a number of species in the Serra do Divisor and Serra do Moa occur not only above 900 m in the Andes but at elevations below 800 m in several portions of the Andean piedmont, a poorly defined region. Moreover, the labels of many collections indicate neither elevation nor habitat. We have been very conservative in what we have designated as "Andean" species; the distributions of species occurring in the Divisor and Moa ranges require more detailed study.

In the serras, rocky cliffs and other areas with shallow, unstable soils support a relatively high concentration of Andean elements. Dicranopygium cf. rheithrophilum (Harling) Harling (Cyclanthaceae) appears to be specialized in the humid, rocky margins of the Rio Moa. On cliffs above stream-beds one finds Cyathea bipinnatifida (Cyatheaceae), known elsewhere only from 850-2000 m in the Andes, from Venezuela to Bolivia.

Ladenbergia (Rubiaceae) is a predominantly Andean genus; it is represented in Acre by L. lambertiana (A. Braun ex Mart.) Klotzsch and L. oblongifolia. The latter, known in Acre from the Serra do Moa, is relatively widespread in the Andes of Ecuador, Peru, and especially Colombia up to 2100 m, but it also occurs at lower elevations at a few localities in western Amazonian Ecuador (Napo) and Peru (Loreto/Mishana). Cespedesia spathulata (Ruíz & Pavón) Planch. (Ochnaceae) is mostly Andean as well but it also occurs in sandy soils and rocky slopes in western Amazonia, around the Serra do Cachimbo in Pará, and in Mato Grosso, as well as near Manaus. Eschweilera andina (Rusby) J. F. Macbr. (Lecythidaceae) occurs in Acre (Rio Tarauacá) and typically in the piedmont of the Andes from Colombia to Bolivia. Aegiphila spicata (Verbenaceae) is a shrub known in Acre only Marechal Taumaturgo; elswehere, it occurs in Peru at middle elevations in Loreto, Ucayali, and Huánuco, but it climbs to 1500-2000 m in Pasco (Peru) and in La Paz (Bolivia). In the understory of the Acre serras on milder slopes, one finds Monolena primuliflora (Melastomataceae), a principally Andean species occurring also on the lower eastern piedmont in Ecuador and Peru.

Wettinia augusta Poepp. & Endl. (Arecaceae), found in the Serra do Moa in Acre, is an Andean species that occurs at lower elevations only near the Andean piedmont or near the upper reaches of rivers that drain the Andes. In addition to this gensus, Aiphanes, Aphandra, Chamaedorea, Chelyocarpus, Dictyocaryum, Iriartea e Phytelephas are also cited as typically sub-Andean palms. Another palm occurring in Brazil only in the vicinity of the Serra do Moa is Dictyocaryum ptarianum, whose principal distribution is at higher elevations, such as the tepuís of Venezuela but which also occurs in some lowlands. Similarly, Euterpe longevaginata is a palm that, although not considered strictly Andean, does occur in montane habitats in parts of the Andes and in Central America. (E. L. Ferreira, unpublished data).

Campinas, Amazonian caatingas (campinaranas), and hyperseasonal savannas (Table 7)

The most surprising affinity of the Acre flora is with formations on sandy soils, not only in the Rio Negro and upper Rio Madeira basins but also in other regions where these formations are less common and extensive.

The species associated with formations on white sand are often habitat specialists whose distributions adhere tightly but incompletely to the distribution of this complex of vegetation types. They are edaphic formations not separated climatically from the rest of Amazonia. These formations on extremely nutrient-poor podzols (white-sand soils) are still inadequately studied. The most developed of them, the "Amazon caatingas" or "campinaranas" ("varillal" in Peru), are forests consisting of closely packed slender trees with an often closed canopy to approximately 20 m high, with occasional large emergents. The understory has a thick layer of poorly decomposed litter. Epiphytes are common, but not lianas.

There is actually a continuum from campina forest to open, shrubby campina (called "bana" in Venezuela); this is an inverse function of the depth of the water table (Jordan 1985). The area covered by Amazon caatingas in Brazil has been estimated at 30,000 km² and a similar figure for "low campina" (Pires and Prance 1985), but these have yet to be confirmed by ground-truthing and interpretation of satellite images. White-sand formations occur sporadically in relatively wet portions of Amazonia and the Guayana region, except in the upper Rio Negro of Brazil, Colombia, and Venezuela, where they predominate.

These existence in Acre of these formations - and of the small black-water tributaries that signal their presence - was not documented in Acre until C. A. Cid Ferreira's Acre expedition in 1984 near Cruzeiro do Sul. Since then, these areas of white-sand formation have been found, all in the upper Rio Juruá basin:

• along the Igarapé Humaitá, which empties into the Juruá opposite Porto Valter
• along the Rio Bagé
• along the Rio Machadinho (tributary of the Rio Tejo)
• parts of the Serra do Moa
• near the old airport of Cruzeiro do Sul
• near Mâncio Lima

The latter is the southern end of a large white-sand area lying mostly in the state of Amazonas (B. W. Nelson, pers. comm.; maps in Projeto RADAMBRASIL 1975-1981). These formations in Acre are poorly collected, their extent not known, and they have not been mapped in any way. Moreover, it is highly likely that additional white-sand areas will be found as the vegetation of the state is more adequately mapped during the next several years. It should be pointed out that there is a shortage of active taxonomic specialists for a number of the herb groups that are so characteristic of white-sand formations, such as the Humiriaceae, some Bombacaceae, and some Euphorbiaceae.

The species strictly associated with white-sand formations in Acre correspond to 7.3% of the non-widespread distributions analyzed to date, but that percentage will rise as more of the collections from these areas are identified to species. We consider mapping and floristic surveys of these areas to be a high priority.

Formations on sand are physiognomically diverse as well as widely scattered, which complicates the detection and analysis of this type of affinity. The distribution of the terrestrial bromeliad Tillandsia paraensis is closely associated with the occurrence of formations on sand and exemplifies the spectrum of localities and vegetation cover associated with them:

• Acre: the campina of Humaitá on the upper Juruá
• Amazonas: principally in the upper Rio Negro and middle and upper Madeira, including the extensive "campo" near Humaitá and the campina of Aripuanã; also between Manaus and Caracaraí and the Rio Cuieiras (lower Rio Negro)
• Pará: Rio Tapajós (secondary forests on sand); Rio Curuá; Oriximiná; Serra dos Carajás (in canga vegetation)
• Mato Grosso: Sararé (cerrado vegetation)
• Rondônia: Pacaás-Novos (an area of sandy and rocky soils)
• Colômbia/Vaupés: upper Rio Negro
• Guiana: Atkinson (on white sand)
• Suriname: savannas
• Venezuela/Amazonas: Atabapo (Amazon caatinga); Yapacaná ("bosque enano")
• Venezuela/Bolívar: El Dorado (savanna)
• Venezuela/Sucre: (savanna)
• some localities with uncertain vegetation and substrate: Peru/San Martín (Tocache Nuevo); Bolívia/Santa Cruz (Ichilo); Bolívia/S. Yungas (Río Bopi); Bolívia/La Paz (Larecaja); Peru/Loreto (Mishuyacu)
• some localities where it is reported to occur in "várzea" (floodplain) but possibly on white sand: Brasil/Amazonas (near Lábrea); Peru/Loreto (Mariscal Castillo); Peru/Madre de Dios (Tambopata); Bolívia/Pando (Abunã).

The botanical data on white-sand formations in Acre are scarce but sufficient for a preliminary analysis of their affinities. Near Mâncio Lima, the few known areas of open campina vegetation are along side roads and tracks and therefore subject to disturbance, but they contain some of the characteristic taxa of campinas elsewhere in Amazonia, such as the foliose liquen Cladonia on exposed white sand, as well as numerous individuals of Pedicularis (Lentibulariaceae) and especially Xyris (Xyridaceae), plus some grasses. At the edge of the open areas, one finds small trees of Humiria (Humiriaceae), an unknown Bombacaceae, and Remijia ulei (Rubiaceae). The latter is common also known from the campinaranas of the upper Rio Negro and campinas of the upper Rio Madeira.

There are two other types of white-sand formations near Mâncio Lima. One has a high density of slender trees approx. 5 m tall, the most common species being a Dendropanax (Araliaceae). Other trees are Mollia (Tiliaceae), characterístic of forests inundated by black-water streams, and Ruizterania trichanthera (Vochysiaceae). The latter occurs in the upper Rio Negro, near São Paulo de Olivença on the upper Solimões, at Jenaro Herrera in the central Peruvian Amazon, and in Venezuelan Guayana, nordeste de Mâncio Lima, always in Amazon caatinga. The herbaceous stratum is dominated by the fern Danaea oblanceolata Stolze and by a terrestrial Anthurium (Araceae), both occurring on small hummocks formed by their own roots; this must maintain the plants above water during temporary inundations.

The third type of campina found near Mâncio Lima has a higher canopy (8-10 m) with emergents reaching 20 m. In addition to the same taxa present in the previous type, there are also trunked palms, Himatanthus sp. (Apocynaceae), Burseraceae, Trattinnickia burserifolia Mart. and Protium paniculatum Engl., and Marantaceae (Calathea and Ischnosiphon).

Sandy soils form the substrate for several vegetation types in the Serra do Moa, including the "matas de tabuleiro," home to species that have affinities with the upper Rio Negro or with the Andes. Rapatea spectabilis and/or R. muaju are found in low-lying parts of these terraces, as well as in the upper Rio Negro and in western Amazonia (principally Peru) where there are sandy soils overlying a perched water table. Rapateaceae are also characteristic of the Andes and especially of the Venezuelan Guayana, their center of diversity.

The Rubiaceae comprise a salient element of the Acre campinas. Platycarpum acreanum, already mentioned as a narrow endemic, is known only from the campinas of Mâncio Lima and Humaitá (upper Juruá). In Acre, Remijia firmula is found in the campinas near Cruzeiro do Sul, as well as in Venezuela (Amazonas) and Bolivia (Santa Cruz), while R. ulei occurs in the campinas of Mâncio Lima, as well as Colômbia (Guainía, Valle del Cauca), Ecuador (Morona-Santiago), Peru (Loreto), and Venezuela (Amazonas. Alibertia hispida has been collected in sandy soils in the Serra do Moa, and it is known from Amazonas (the campina of the Reserva Ducke near Manaus, also Livramento on the Rio Madeira) and from Peru (sandy soils near Iquitos).

Dry/semi-deciduous forests, cerrado vegetation, savannas, and caatinga (Table 8)

The drier southeastern portion of Acre, corresponding essentially to those parts of the state drained by the Rio Purus, is home to a significant number of species associated with drier formations peripheral to Amazonia or especially farther south. On Ernst Ule's second extended visit to Acre from late 1910 to early 1912, he traveled on the upper Rio Purus, mostly on the upper Rio Acre along Acre's borders with Bolivia and Peru, where he collected scores more of species new to science; he also worked on the Rio Xapuri. He noted that the upper Purus was rich in Cissus and Mikania, which saw as unusual for the Amazon. He thought this was reminiscent of the "dry outer region" of Martius and might indicate influence from south (Ule 1913).

In their seminal work on Amazonian phytogeography, first published in 1953, Adolpho Ducke and George Black also observed Acre's strong affinity with drier regions, as indicated by the occurrence of several species characteristic of peripheral or more southern distributions such as Phyllocarpus riedelii and Gallesia integrifolia.

Our analyses have confirmed these perceptive early observations. We have documented the occurrence of species with distributions in northeastern Brazil, the Cerrado region of central Brazil, and/or south into Paraguay and Argentina, as well as other regions with dry forests. These species account for 18.3% of those with non-widespread distributions.

These floristic as well as physiognomic affinities signal former physical connections and therefore genetic exchange among the drier formations that form at present a discontinuous arc around the southern periphery of Amazonia in what could be considered transition zones. Ratter (1987) pointed out the existence of areas of semi-deciduous forest among the humid Amazonian forests of northern Mato Grosso. We have observed that many trees lose their leaves during the dry season in the forests of southeastern Acre, but we have not yet quantified this to determine whether some of them are technically semi-deciduous.

Numerous species show distributions around part or all of the southern arc from Ceará in northeastern Brazil around and up to Acre in Brazil and even to Tarapoto in Peru (Prado and Gibbs 1993). Some also skirt the northern limits of Amazonia and may extend up to the dry forests of Central America and sometimes the Caribbean. The southern arc reaches its northwestern limit in the area defined by eastern Rondônia, Acre, Pando in Bolivia, and Madre de Dios and Ucayali in Peru, plus "islands" of dry forests farther west and north in Peruvian Amazonia.

A number of species in the Acre flora represent northern extensions of Cerrado or Paraguayan elements (e.g., Casearia gossypiosperma, Cariniana estrellensis, Couepia uiti, Samanea tubulosa, Aspidosperma ramiflorum), while others make up part of peri-Amazonian or southern peri-Amazonian distributions (e.g., Amburana cearensis, Poeppigia procera, Myroxylon balsamum, Rhamnidium elaeocarpum).

One of the most dramatic examples of dry-forest affinities is Amburana cearensis (Fabaceae). This species, the second most important timber tree in Acre after mahogany, occupies a large part of the arc of dry forest south of Amazonia that extends from Ceará through Central Brazil, extending south to Paraguay, and reaching northwet to Acre and Madre de Dios (Peru), plus a Peruvian disjunction in Pasco (see discussion below). Similarly, Passiflora tricuspis occurs in the Brazilian states of Piauí, Minas Gerais, Goiás, south to Paraná, and northwest to Rondônia, Acre, and into Peru in Madre de Dios (near Pucallpa); southwest to the Bolivian departments of Mapiri, Santa Cruz, La Paz, and Beni (Ivón); and south to Paraguay.

Celtis pubescens (Ulmaceae) extends even farther south, occurring in the Brazilian states of Bahia, Mato Grosso, the Distrito Federal, Minas Gerais, São Paulo and Acre; in Bolívia in Santa Cruz, La Paz, Pando, and Tarija; and south into both Paraguay and Argentina.

Platypodium elegans (Fabaceae) exemplifies a circum-Amazonian distribution: in addition to Acre, it is found in Colombia (Antioquia, Meta), Venezuela (Lara), northeastern Brazil, the Cerrado (Mato Grosso, Minas Gerais, Goiás, e São Paulo), Bolivia (Santa Cruz), and eastern Paraguay. Aspidosperma macrocarpon (Apocynaceae) is disjunct in Guyana and in Amapá but occurs principally in the southern arc in Brazil: Goiás, Mato Grosso, Rio de Janeiro, Distrito Federal, Acre, Minas Gerais, Rondônia, Bahia, Pará, Tocantins, Maranhão, and Amazonas.

A non-trivial number of species occupies only part of the more humid end of the southern arc and/or the "islands" of dry forest in Peru, which can camouflage their dry-forest affinities. For example, Gymnosporia ulei (syn.: G. magnifolia (Loes.) Lundell) (Celastraceae) occurs in Bolivia (Santa Cruz and Beni) and in such "islands" in Peru (Loreto/Pachitea, Pasco/Oxapampa, and San Martín/Tocache Nuevo). Dalbergia gracilis (Fabaceae) is found in Bolivia (Santa Cruz, Pando, Beni), Peru (Madre de Dios), and Brasil (Acre, Rondônia, Mato Grosso, e Amazonas). Samanea tubulosa (Benth.) Barneby & J. W. Grimes (Mimosaceae) is known from Paraguay, Argentina, drier regions of Ecuador (Guayas, Manabi), and dry-forest islands in Peru (San Martín/Tarapoto, Cusco/Tingo María), as well as in Bolivia (Beni, La Paz, Santa Cruz, Pando) and in the Brazilian states of Acre, Mato Grosso, Rondônia, and Pará, but this last part of its distribution falls in the so-called "transverse dry belt" (Pires-O'Brien 1997) , a swath cutting NW-SE across the Rio Trombetas and the lower Rio Xingu that receives only 1,750 mm of seasonally distributed rainfall and supports large extents of dry forest.

The relatively lower and more seasonal rainfall in southeastern Acre may provide only a partial explanation for its strong affinity with drier regions. Together with the high concentrations of fossils of ground sloths and other groups of animals characteristic of the Cerrado and other open formations (e.g., Oren 1993), this affinity may be seen as constituting indirect evidence of climatic changes in the past that affected possibly the vegetation cover but more certainly the floristic composition, as taxa from dry formations migrated into the region and persisted during more humid periods like the present. There may have been times in the past several thousand years when dry forests were the dominant type of vegetation in much of northern South America (Prado and Gibbs 1993; Oliveira Filho and Ratter 1995). We conclude that the affinities of southeastern Acre with drier regions are not anomalous, rather they represent relictual distributions in a zone of climatic transition. The conservation of the Acre must take into consideration areas of vegetation that represent this part of Acre's spectrum of affinities.

based on 191 non-widespread distributions