Seasonally Dry Tropical Forests
Ecology and Conservation
By Rodolfo Dirzo, Hillary S. Young, Harold A. Mooney, Gerardo Ceballos
ISLAND PRESS Copyright © 2011 Island Press
All rights reserved.
Neotropical Seasonally Dry Forests: Diversity, Endemism, and Biogeography of Woody Plants
REYNALDO LINARES-PALOMINO, ARY T. OLIVEIRA-FILHO, AND R. TOBY PENNINGTON
Neotropical seasonally dry forests are found from northwestern Mexico to northern Argentina and southwestern Brazil in separate areas of varying size (fig. 1-1). Their different variants have not always been considered the same vegetation type (e.g., Hueck 1978) or biogeographic unit (e.g., Cabrera and Willink 1980), but recent work has helped to define the extent, distribution, and phytogeography of seasonally dry tropical forest (SDTF) as a coherent biome with a wide Neotropical distribution (Prado and Gibbs 1993; Pennington et al. 2000; Pennington, Lewis et al. 2006). This unified interpretation is important both for biogeographic inference and for setting conservation priorities in Neotropical SDTF, which is the most threatened tropical forest type in the world (Miles et al. 2006).
Pleistocene climatic changes have been proposed as a possible force influencing the overall distribution of SDTF in the Neotropics (Prado and Gibbs 1993) and in driving evolution in SDTF plants (Pennington et al. 2000). Prado and Gibbs (1993) and Pennington et al. (2000) proposed a hypothesis in which during glacial times of cooler and drier climate, SDTFs were much more extensive than at present, perhaps forming contiguous forests across wide areas of the Neotropics. This view of current more-restricted areas of SDTF as "refugia" has been challenged by palynological studies that suggest the rain forests of Amazonia occupied hardly any less area than today and that the SDTFs of the Bolivian Chiquitano have been assembled recently (reviewed by Mayle 2004, 2006).
If there have been connections between some or all of the seasonal forests in the Neotropics during recent geological time, we would expect to find high floristic similarity among them. Prado and Gibbs (1993) and Pennington et al. (2000) highlighted a number of unrelated SDTF tree species that are widespread and found in several of the disjunct areas of Neotropical SDTF. They argued that these repeated distribution patterns were evidence of a once more widespread and perhaps continuous seasonal forest formation. These authors failed, however, to place these widespread species in the context of the entire woody flora of these areas, and no analyses of overall floristic similarity were presented.
In this chapter, we present a quantitative analysis of floristic similarity of the flora of the major areas of seasonal forests (SFs) in the Neotropics, including those of the floristically and ecologically unrelated but geographically adjacent vegetation of the Cerrados (savannas) and Chaco woodlands (fig. 1-1). This is the first quantitative analysis of the floras of these forests since Sarmiento (1975), who considered genera, and not species. Our species-level analysis provides a more fine-grained view of floristic variation. We use an ordination approach to analyze inventory data of woody plants from sites throughout Neotropical SDTF and examine the implications of the results for (1) patterns of diversity and endemism, (2) patterns of floristic relationships, (3) beta diversity, (4) biogeographic history, and (5) conservation prioritization.
Quantitative Analyses of Neotropical Seasonally Dry Tropical Floristic Nuclei
We define SDTFs following the broad concepts of Beard (1955) and Murphy and Lugo (1995), including tall evergreen SFs on moister sites, at one extreme of the series, to thorn woodland and cactus scrub at the other. We delimited 21 floristic nuclei of Neotropical SDTF, plus the Cerrado and Chaco areas (fig. 1-1). When nuclei are geographically isolated, this definition was straightforward, and in other cases we used previous phytogeographical studies that have revealed high affinities between some areas (e.g., Gentry 1995) for the equatorial Pacific SDTF in Peru and Ecuador. Published and unpublished but reliable tree inventory data from plots and sites for each of these regions were then aggregated to produce an initial species list for each of the floristic nuclei. Each nucleus' species list was enriched, whenever possible, with additional information of plants reported for the area (e.g., herbarium collections, checklists, and our own field experience). We considered plants that are woody and reach at least 3 meters during some stage of their life cycle, excluding woody lianas and climbers. Main sources of data were Ratter et al. (2003), Linares-Palomino et al. (2003), and Oliveira-Filho (unpublished data). The data were homogenized using relevant taxonomic literature and online databases (W3Tropicos, IPNI, IL-DIS) by checking for synonyms and misspellings. Doubtful identifications and records were excluded. The taxonomic treatment of families follows the Angiosperm Phylogeny Group II classification (APG 2003).
The final dataset included 3839 species from 806 floristic lists (table 1-1). Classification (UPGMA using group average and TWINSPAN) and ordination (nonmetric multidimensional scaling, MDS) analyses using standard settings in PC-ORD (McCune and Mefford 1999) were performed on a subset of 1901 species present in two or more floristic nuclei. The MDS ordination and the UPGMA cluster analysis were performed using the Sørensen distance. The same index was used to assess beta diversity among floristic nuclei, allowing comparison of our results with beta diversity studies of the Cerrado (Bridgewater et al. 2004).
Each species found in 10 or more floristic nuclei (i.e., widespread species) was then annotated as ecologically versatile if it occurred in several forest types, including SDTF (e.g., Maclura tinctoria Trema micrantha;table 1-2). We also annotated SDTF specialists (e.g., Anadenathera colubrina, Sid-eroxylon obtusifolium) and SDTF generalists—species that generally grow in SDTF but are occasionally found in other vegetation (e.g., Guazuma ulmifolia, Tabebuia impetiginosa). Annotation was based on bibliographic sources (e.g., Flora Neotropica Monographs) and our own field experience.
Patterns of Plant Species Diversity
Diversity and Endemism
The number of floristic lists per nucleus ranged from 2 to 376 (table 1-1). While this does not represent even geographic coverage of inventories, we do not believe our results are excessively biased, because nuclei covered by few studies often have many species and vice versa. For example, the Peruvian Eastern Andean SF nucleus has 101 species from just 2 lists, whereas 358 species are recorded from 376 lists in the Cerrado. This pattern of nuclei with more lists but lower overall species numbers (e.g., Coastal Caribbean SF, 19 lists, 135 species) and nuclei with a low number of lists but high numbers of species (e.g., Bolivian Chiquitanos SF) may reflect several historical and ecological factors, including the relative size of the nuclei and different rainfall regimes.
Species numbers ranged from 45 to 1602 per nucleus (table 1-1). The percentage of unique species present in each nucleus ranged from 1.9 percent in the Paraguay-Paraná SF to 77.5 percent in the Insular Caribbean SF (table 1-1). While these unique species are not strictly endemics (they may be present in other areas outside our nuclei), their numbers offer a reasonable proxy for levels of endemism.
Of the 3839 species, 457 were present in 5 or more nuclei, and only 55 (1.43 percent of the total; table 1-2, fig. 1-2A) have been recorded in 10 or more nuclei. Of the latter, 24 are ecologically versatile species, 22 are SDTF generalists, and only 9 are SDTF specialists (table 1-2).
The uneven geographic coverage and heterogeneous nature (from plots, transects, and floristic lists) of the basic data limit us from objectively comparing alpha diversity levels and total species numbers in the SDTF nuclei. Our data perhaps are more robust for analyzing patterns of endemism because some nuclei for which we have few inventories show high numbers of unicates (e.g., Caribbean, Mexico), and others for which we have sampled far more thoroughly show low numbers (e.g., Brejo and Peri-Caatinga). Though the percentage of unicate species varies widely from 1.9 to 77.5 percent, in general it is high, with 12 of 23 nuclei showing greater than 20 percent unicates, suggesting high endemism. While such high numbers of endemic species might be produced by recent, rapid evolution, it seems more likely that in many SDTF nuclei they represent the result of the considerable age of the biome, prolonged isolation, and limited arrival of immigrant lineages by dispersal (Pennington et al. 2009). This view is partly derived from the fossil record, which shows evidence for SDTF in the Ecuadorean Andes in the late Miocene, 10 to 12 million years ago (e.g., Burnham and Carranco 2004). Dated molecular phylogenies in general show patterns of speciation that predate the Pleistocene and high phylogenetic geographic structure where closely related species occupy the same geographic area (see Pennington, Lewis et al. 2006 and Pennington, Richardson et al. 2006 for reviews). This view of limited dispersal is corroborated by the contribution of Caetano and Naciri in this volume (chap. 2). Their population genetics approach to investigating the widespread distributions of two SDTF tree species, Astronium urundeuva and Geoffroea spinosa, shows high population genetic structure that is consistent with limited gene flow between major SDTF nuclei.
The three quantitative analyses we applied to the data consistently identified four major SDTF regions (fig. 1-2B–D): Caribbean/Mesoamerican, Andean (not including Bolivian Andes), Southern South American, and Brazilian. The position of the Bolivian Andes nucleus is intermediate, with affinities to both neighboring Andean and adjacent lowland sites.
The SDTFs in Mesoamerica have been considered a distinct phyto-geographic unit since the studies of Rzedowski (1978) and recent floristic data (e.g., Trejo and Dirzo 2002) have confirmed their remarkable plant diversity. Likewise, the Caribbean islands are also interesting because of their high endemism of vascular plant species (with more than 50 percent of approximately 12,000 species endemic) (Santiago-Valentin and Olmstead 2004). This fact is also reflected in the SDTF flora by high Sørensen distance values with the adjacent Mesoamerican SF (65 to 81 percent) and the highest unicates percentage (77.5 percent) in our analyses (table 1-1). There are, however, surprisingly few studies evaluating large-scale phytogeographic relationships in the entire Mesoamerican-Caribbean region (Santiago-Valentin and Olmstead 2004), apart from research on the floristic affinities between the vascular floras of the Yucatán peninsula and the greater Caribbean islands, particularly Cuba (Chiappy-Jhones et al. 2001). The only wide-ranging study evaluating the affinities of the SDTF floras of the region remains that of Gentry (1995), which was based on a rather small sample of transect inventories. Our analyses show strong evidence for a floristic connection between the Insular Caribbean SDTF (including the Greater and Lesser Antilles) and the Mexican and Central American SDTFs (highest Sørensen similarity was with the latter).
Gentry (1982c) noted a strong relationship between SDTF in northern Colombia-Venezuela and the Central American Pacific SDTF, suggesting that the wet Chocó forests in Colombia, which probably constituted a major rain forest refuge during glacial dry periods, had functioned as a barrier to the drier forests north and south of it (see also Simpson and Neff 1985). The Chocó has been suggested to have been a low and swampy area even before the Andean orogeny (Haffer 1970) and so a barrier to SDTF species. Our data support Gentry's view and also the high dissimilarity of Central American SDTF and the SDTF in coastal Ecuador that he discussed (Gentry 1982c).
Our analyses, placing all Brazilian sites (plus the Argentinean, Paraguayan, and Bolivian Chiquitano area) in the first major UPGMA and TWINSPAN divisions, support the floristic relationship of these areas within the "Pleistocenic Arc" of seasonal vegetation, as proposed by Prado and Gibbs (1993). They also suggested that the SDTF in the dry inter-Andean valleys of Peru might constitute remnants of a previously much wider expansion of this arc, but the Peruvian Andean areas are resolved separately in our analyses, just as Prado (2000) anticipated. Prado (2003) proposed several complex migration routes for the floristic elements that formed the caatinga, such as the Caribbean route (see also Sarmiento 1975) and the Andean route (see also Weberbauer 1914). Nevertheless, few species are disjunct between the caatingas and these areas, and our analyses show the caatinga to be firmly embedded in the Brazilian group.
Despite being situated adjacent to two major South American seasonal woodland ecosystems (the Chaco and Cerrado), the Chiquitano SF is unrelated floristically to either. Killeen et al. (1998) suggested that the Chiquitano SF had more in common with the semideciduous forests of the Andean piedmont of northwestern Argentina and the Misiones region of eastern Paraguay and northeastern Argentina, as well as with the Caatinga region of northeastern Brazil. More recently, Killeen et al. (2006) showed the transitional, albeit distinct nature of these forests. Our data, showing high Sørensen similarity values with the adjacent Pantanal, Argentinean Piedmont, and Paraguay-Paraná SF, provide evidence of strong floristic relationships. The low-level unicate species in these forests (table 1-1) provide some support for the view that the Chiquitano forests have been assembled recently (Mayle 2006). López (2003) argued that the Bolivian inter-Andean dry valley flora was more related to the Chaco and other Argentinean SFs. Of the 1156 species he reported for the Bolivian Andean dry valleys, more than half had their northernmost distribution in central Bolivia and parts of southern Brazil. More recently, Wood (2006) showed that the biogeographical relationships of the dry areas in the Bolivian Andes were variable and highly dependent on which family was studied: Labiatae showed an essentially Andean distribution, suggesting a pattern between SDTF areas similar to that shown by our UPGMA analysis. Asclepiadaceae and Acanthaceae instead showed stronger links with the lowland vegetation in southern South America (Argentinean SF, Chaco, and Cerrado), a pattern suggested by our TWINSPAN analysis. It seems that the Inter-Andean Bolivian SFs, due to their geologic and climatic history, as well as to their unique position close to major distinct biomes, are composed of plant species of variable biogeographic affinity, making generalizations difficult.
Weberbauer (1945) proposed that the xerophytic floras of Peru, Bolivia, and the Argentinean Chaco are remnants of a formerly homogeneous flora fragmented by Andean uplift. More recently, Sarmiento (1975) proposed the existence of a major disjunction, located somewhere in the Andes of southern Peru and northern Bolivia, separating the dry floras from northern Peru to Venezuela from those south and east of Bolivia to Argentina and Brazil. Evidence in support of this comes from Kessler and Helme (1999) and López (2003), who were unable to find strong connections between the Bolivian inter-Andean dry floras and southern Peru. Unfortunately, the unstable position of the Bolivian inter-Andean valleys in our analyses does not confirm or reject any of these theories.
Prado (2000) was able to find clear floristic differences between the Chaco (and closely related Chaquenian formations), the Cerrado, and South American SDTFs by quantitative floristic analysis. Our results agree that the Chaco vegetation has a peripheral position with respect to other seasonal forest formations in southern South America. The Cerrado woodlands, in contrast, seem to have closer floristic relationships with the adjacent Brazilian SF, particularly with those of the Peri-Caatingas, demonstrating the importance of ecological generalists shared between these forests. Several species characteristic of mesotrophic soils in the cerrados are also found in adjacent SDTF formations, such as Dilodendron bipinna-Tum (Sapindaceae) and Callistene fasciculata (Vochysiaceae).
Beta Diversity Levels in SDTFs and Savannas in The Neotropics
As expected, highest distance (or beta diversity) values were found between the Chaco and other seasonal forests (table 1-3). Highest similarity (or lowest beta diversity) was found between the Central-Atlantic and Austro-Atlantic SDTFs, the Central-Western and Brejo SFs, and the Central-Western and Austral-Atlantic SDTFs. There were 61 pairs of nuclei, out of 253 possible pairs, that had 90 percent or more dissimilarity, and 203 pairs had dissimilarity of more than 70 percent. In contrast, only 2 pairs of nuclei showed a similarity higher than 70 percent (table 1-3). Beta diversity estimates for vegetation units over large geographical areas are rare. One such study for the Brazilian Cerrado biome, an area covering some 2 million square kilometers (Bridgewater et al. 2004), compared floristic nuclei defined in a similar manner with those in this chapter. Sørensen distance values among 6 Cerrado nuclei were 0.38 and 0.74, indicating that the Cerrado flora is heterogeneous. Our data present higher distance values (table 1-3). Eighty percent of the pairwise comparisons had distances over 0.70. This high level of heterogeneity reflects both the continental scale of the study area and that SDTF exists as scattered areas in comparison to the continuous Brazilian Cerrado. However, in SDTF, it is clear that floristic similarity can be very low between geographically close areas of SDTF, even within some of the nuclei. For example, the similarity between the Mara-ñon and Mantaro inter-Andean dry valleys in Peru, separated by only about 400 kilometers, is only 14 percent, with only 16 species shared from a total of nearly 200 woody species, and Trejo and Dirzo (2002) showed the average Sørensen similarity between 20 Mexican SDTF sites (sampled using 0.1-hectare plots) to be only 0.09. The generally low floristic similarity argues for lack of recent floristic links and dispersal between isolated SDTF areas, as discussed under Biogeographic History below. (Continues...)
Excerpted from Seasonally Dry Tropical Forests by Rodolfo Dirzo, Hillary S. Young, Harold A. Mooney, Gerardo Ceballos. Copyright © 2011 Island Press. Excerpted by permission of ISLAND PRESS.
All rights reserved. No part of this excerpt may be reproduced or reprinted without permission in writing from the publisher.
Excerpts are provided by Dial-A-Book Inc. solely for the personal use of visitors to this web site.