Biogeography of the Antilles Based on a Parsimony Analysis of Orchid Distributions J
Total Page:16
File Type:pdf, Size:1020Kb
Journal of Biogeography, 28, 775±794 Biogeography of the Antilles based on a parsimony analysis of orchid distributions J. Carlos Trejo-Torres and James D. Ackerman Department of Biology, University of Puerto Rico-Rõo Piedras, PO BOX 23360, San Juan, Puerto Rico 00931-3360 Abstract Aim We obtain biogeographical patterns based on the distributions of shared orchid species of the Caribbean. These patterns are used to de®ne biogeographical zones. We then analyse the concordance between the distributional patterns with ecological and physical features of the islands. Location We use orchid species recorded on 49 islands of the Greater, Lesser, and southern Antilles, and the Bahamas. Three continental areas are included: Florida (North America), the Yucatan (Central America), and the Guianas (South America). Methods We use a parsimonious analysis of species distributions that produces the best arrangements of shared taxa among areas. The analysis uses 356 shared orchid species of the 863 species recorded for studied areas. The methodology has been used to infer historical relationships among areas but we interpret the results as static or ecological patterns of biogeographical af®nities. Results Two kinds of island groupings are revealed. (1) Groups with common geology and geomorphology: the Bahama Archipelago, the Virgin Islands, the Cayman Islands and the southern Dutch Antilles. (2) An aggregation of distant islands with a heterogeneous geology but a common physiography: the Greater Antilles/Trinidad/ Lesser Antilles/Margarita-Tobago. The Guianas are linked with the Greater Antilles, while the Yucatan and Florida are linked to the Bahamas. Main conclusions Groupings of islands are congruent with their gross ecological features either from similar geomorphology or common physiography. The strong af®nity among islands considerably distant among each other is explained by the high vagility of dust-seeded orchids. Then, ¯oristic af®nities seem determined by ecological characteristics of islands rather than by dispersal barriers. We predict that other plant groups with dust-like diaspores and animals with good vagility should show comparable biogeographic patterns. Parsimony analysis of distributions (PAD) is an alternative methodology to multivariate analysis to compare biotas, and a graphic complement to quantitative methods producing numerical values. Keywords Antilles, Caribbean, biogeography, Orchidaceae, parsimony analysis of endemicity, distributions, dispersal, islands. between Asia and Australia. The Antilles are also signi®cant INTRODUCTION because they represent one of the two connections, by way of The Antilles, or West Indies, have been frequently studied by an island chain, between two major biogeographical realms: biogeographers. These islands make up one of the largest the Neartic and the Neotropic. The majority of studies on tropical archipelagos in the world, second only to the islands Antillean biogeography concern the distributional patterns or the cladistic biogeography of animals, especially verte- brates and insects (e.g. Liebherr, 1988; Woods, 1989; Correspondence: J. Carlos Trejo-Torres, Department of Biology, University of Puerto Rico-RõÂo Piedras, PO BOX 23360, San Juan, Puerto Rico 00931-3360. Page & Lydeard, 1994; Hedges, 1996). Surprisingly, there 1 E-mail: [email protected] are only a few publications on the phytogeography of this Ó 2001 Blackwell Science Ltd 776 J. C. Trejo-Torres and J. D. Ackerman region (Howard, 1973; Samek, 1988; Adams, 1989; Lavin, more generic name: parsimony analysis of distributions 1993; Borhidi, 1996), although the ¯oristics of the area have (PAD). The method excludes single-site species (including been actively studied for decades (Zanoni, 1986; Liogier, single-site endemics), and takes into account shared species 1996). whether endemic or not to the Antilles or adjacent areas. The Caribbean region has been divided phytogeographi- We substitute the term distribution for endemicity to avoid cally in three subregions by Samek (1988). These main divi- confusion. sions are: (1) Mexico to Panama, (2) Colombia to Venezuela, Our goal here is to distinguish biogeographical patterns in and (3) an insular subregion including the Antilles proper. the Antilles based on one plant family, Orchidaceae. This The southern Antilles, from Aruba to Tobago are considered group of plants, of which there are approximately 700 part of the Colombian±Venezuelan subregion. Borhidi species in the Antilles, is relatively well known taxonomi- (1996) joins both continental subregions, from Mexico to cally and geographically because there are relatively recent Venezuela, into a single one. While these authors consider orchid treatments for the Cayman Islands (Proctor, 1984, south Florida as part of the Antillean subregion, Gentry 1996), Puerto Rico (Ackerman, 1995) and the Bahamas (1982) included only the Florida Keys in the Antillean region. (Sauleda & Adams, 1982; Sauleda, 1992). Furthermore, a Distinguishing biotic regions, or de®ning biological treatment for the Greater Antilles is underway (Ackerman, boundaries, has been among the major concerns of bioge- 1997; in press). The homogeneous wind-dispersal mechan- ographers. Traditionally, this has been performed with ism (except in bird-dispersed Vanilla) of the dust-seeds subjective methods (e.g. Gentry, 1982; Samek, 1988; among the species of this family, makes orchids a good focal Borhidi, 1996). However, more objective or analytical group for the study of distribution. Assuming that most approaches have been developed for the analysis of distri- orchids have the same dispersal capacity, their distributional butional data of organisms (e.g. Gauch, 1982; McCoy et al., patterns may be explained in terms of other ecological, 1986; Patterson & Atmar, 1986; Rosen & Smith, 1988; geographical and historical factors. Rosen, 1988; Vargas, 1991; Real et al., 1992; Scheiner, The questions we address are: (1) what are the phyto- 1992; Worthen, 1996; Puente et al., 1998). Here, we use an geographical relationships among the Antillean islands alternative method based on a parsimonious analysis of taxa based on shared orchid species? (2) are these relationships distributions (Rosen & Smith, 1988). This method repre- affected when neighbouring continental regions are consid- sents a direct way to search for the biogeographical af®nities ered? (3) what is the phytogeographical regionalization of among areas (Connor, 1988; Vargas, 1991), for the detec- the Antilles based on orchids, and (4) how do the distribu- tion of areas of endemism (Morrone, 1994a; Cardoso da tional patterns of orchids match with geography, geology, 2Silva & Oren, 1996; Bates et al., 1998), and for the physiography and ecology of the region? Apart from the delimitation of biological boundaries (Posadas, 1996; biogeographical analysis per se, we discuss the use of Posadas et al., 1997; Morrone, 1998). The parsimony parsimony analysis of naturally de®ned areas (i.e. islands) analysis presented here is a tool for searching the most with extremely dissimilar number of species, and the use of parsimonious arrangement of shared species among areas, as single-site species to look at ®ner degrees of biogeographical a means of revealing the biogeographical af®nities in a differentiation. hierarchical pattern (Rosen & Smith, 1988; Brady, 1994). The analysis presented here was originally called parsi- METHODS mony analysis of endemicity (PAE), and was suggested by 3Rosen (1985, cited in Rosen & Smith, 1988) and developed A total of ®fty-two areas (Table 1) were included in the by Rosen & Smith (1988). It was also independently study. Data on the distribution of species were taken from suggested by Legendre (1986) and Connor (1988). Since literature and from unpublished sources (Table 2). Synony- then it has been employed in the study of extant taxa of mies and valid species names were standardized mainly New Zealand (Craw, 1988), Australia (Cracraft, 1991), according to J.D. Ackerman (unpublished data). Southeast Africa (Morrone, 1994a), the Patagonia A presence/absence matrix of the 863 orchid species (Posadas, 1996), the Andes (Morrone, 1994b; Posadas reported for the studied areas was constructed in MacClade et al., 1997), the Amazonia (Cardoso da Silva & Oren, 3.01 (Maddison & Maddison, 1992). Using this program, 1996; Bates et al., 1998), the Austral region (Craw, 1989; areas were entered in the place of taxa, while taxa were Morrone, 1998; Glasby & Alvarez, 1999), Mexico (Luna entered in the place of characters. In the matrix, presence et al., 1999), and the entire world (Conran, 1995). The was indicated with a `1' and absence with a `0'. Once we units of comparison that have been used are sites, quad- constructed the matrix, we ran analyses of parsimony using rants or sections of regions, biogeographical areas, or PAUP 4 (Swofford, 2000). A hypothetical outgroup area natural geographical areas (e.g. islands, continents, ocean with all 0s (no species) were used in the analyses to root the basins). We use entire islands or groups of them, as the trees. General heuristic searches were carried out to look for units of study. While the method has been mainly used for the most parsimonious trees, which indicate the ¯oristic discovering the historical relationships among areas, we af®nities among studied areas. We obtained consensus trees give a static or non-historical interpretation to the patterns (Strict, Majority Rule and Adams) when more than one obtained (Rosen, 1992; Posadas, 1996; Posadas