Changes in Plant Community Diversity and Floristic Composition on Environmental and Geographical Gradients

Trends in community composition and diversity of neotropical forests as measured by a series of samples of (1) plants 2 2.5 cm dbh in 0.1 ha, (2) plants over 10 cm dbh in 1-ha plots, and (3) complete local florulas are analyzed as a function of various environmental parameters. These trends are also compared with those found in similar data sets from other continents. Altogether the basic 0.1-ha data sets are reported for 87 sites in 25 countries on six continents and several islands. New data from ten 1-ha tree plots in upper Amazonia are also compared with each other and with similar data from the literature. Some noteworthy trends include: (1) Lowland neotropical plant species richness is generally far more tightly correlated with precipitation than with edaphic factors. (2). The nearly linear increase of lowland neotropical plant species richness with precipitation reaches an asymptote (community saturation?) at about 4,000 mm of annual rainfall. (3) Although the species represented in adjacent forest types on different substrates may change dramatically, diversitytends to change relatively little in upper Amazonia. (4) The species present at different sites are very different but the families represented and their diversities are highly predictable from environmental parameters. (5) On an altitudinal gradient in the tropical Andes there is a sharp, essentially linear decrease in diversity from about 1,500 m to near the upper limit offorest above 3,000 m. (6) There is no indication of a "mid-elevation bulge" in diversity, at least not in the sampled habit groups. (7) Even near timberline, montane tropical forests are as diverse as the most diverse temperate forests. (8) Moist subtropical forests are markedly less diverse than their innertropical equivalents, but dry subtropical forests in Mexico are apparently richer in species than inner-tropical dry forests. (9) Central African forests are about as species rich as neotropical forests with similar amounts of precipitation, but forests in tropical West Africa are relatively depauperate. (10) Tropical Australasian forests are no more diverse than equivalent neotropical forests; the world's highest tree species diversityis in upper Amazonia, not Southeast Asia. (11) Contrary to accepted opinion, equivalent forests on the three continents are similar in plant species richness and (with a very few notable exceptions) fioristic composition but are markedly different in structure. The predictability of the fioristic compositions and diversities of tropical forest plant communities eems strong, albeit circumstantial, evidence that these communities are at ecological and perhaps evolutionary equilibrium, despite indications that certain aspects of their diversity are generated and maintained stochastically. Comparisons of the species richness (or other facets) of different forests or different vegetation types are often difficult because of the dissimilarity of the available data. In tropical Asia there is a wealth of data for trees in large sample areas (Ashton, 1964, in press; Whitmore, 1984; Proctor et al., 1983; Kartawinata et al., 1981) but few published data on nontrees. In the Neotropics there are several local florulas (Croat, 1978; Dodson & Gentry, 1978; Janzen & Liesner, 1980; Dodson et al., 1985), but until recently there have been no tree-plot data from high diversity regions based on reliable identifications. Africa has far more extensive coverage by regional and country-wide floras but no local florulas nor large-plot data from high-diversity regions. Recently, a series of 0.1-ha samples of many of the world's most diverse extra-tropical plant communities has been accumulating (e.g., Naveh & Whittaker, 1979; Cowling, 1983; Peet & Christensen, 1980; Rice & Westoby, 1983; Eiten, 1978). Elsewhere we have reported the first comparable data set for tropical forests (Gentry & Dodson, 1987a, b). A standardized sampling technique that includes only plants 2.5 cm in diameter in 0.1 ha has also been developed and applied to a series of tropical forests (Gentry, 1982b, 1986b; Lott et al., 1987; Stallings et al., in press; Lorence & Sussman, 1988); the methodology for obtaining these O. -ha samples, each the sum of ten 2 x 50 m belt transects, is discussed in detail elsewhere (Gentry, 1982b, in prep.). The primary data set on which this paper is based are these 0.1-ha samples, which are now available for 38 lowland neotropical sites, 1 1 montane neotropical sites, and 13 subtropical and 9 temperate-zone sites in the Americas. Similar data sets are available from 6 sites in tropical Africa, 3 sites in tropical Australasia, 2 sites in Europe, and from several tropical islands: New Caledonia, Madagascar, Mauritius, Jamaica (Tables 1, 2; Fig. 1). Supplementary data are taken from local florulas in the Neotropics (Dodson & Gentry, 1978,1988; Croat, 1978; Volume 75,-Number 1 Gentry 3 1988 Plant Community Diversity FIGURE 1. Locations of study sites. Dots = 0.1-ha samples (see Tables 1, 2). Arrows = local fiorulas. For location of 1-ha tree plots see Gentry, 1988. Janzen & Liesner, 1980; Dodson et al., 1985; Hammel, pers. comm. (La Selva, Costa Rica)) and from the Makokou region of northwestern Gabon (Halle, 1964, 1965; Halle & Le Thomas, 1967, 1970; Florence & Hiadik, 1980; iladik & Halle, 1973; Hladik & Gentry, in prep.). A supplemental data set is provided by a series of 1-ha tree plots in various parts of the Neotropics (Gentry, 1988; Prance et al., 1976; Campbell et al., 1986; see also Gentry, 1982b) and Paleotropics (e.g., Ashton, 1964, 1977, in press; Gartlan et al., 1986). Here I first review how the species richness of plant communities changes on five different environmental gradients: latitudinal, precipitational, edaphic, altitudinal, and intercontinental. Observations on a few noteworthy trends in forest structure are also included. Second, I analyze some patterns of floristic change along the same environmental gradients. Finally, I use these analyses to examine briefly the question of why some plant communities have so many more species than others. In all of these analyses I will use number of species as the simplest and most appropriate measure of diversity, as suggested by Whittaker (1977). Shannon-Wiener H' values are reported in Tables 1 and 2, but are so tightly correlated (R2 = 0.93) with the absolute number of species that their use would add little to the analysis. Moreover, the wetforest H' values of 7 to 8 are far above the levels at which H' has been statistically analyzed (cf. May, 1975). TEMPERATE-TROPICAL PATTERNS Figure 2 summarizes the latitudinal trends in species richness, based on the 74 lowland (= < 1,000 m) 0.1-ha sites for which comparable samples are available. It is well known that tropical forests are generally far richer than temperate forests in species (e.g., see Richards, 1952; MacArthur, 1972). Figure 2 indicates that for vascular plants speciesrich tropical forests are typically an order of magnitude more diverse. Also apparent in Figure 2 are several much less well-known corollaries to the general latitudinal diversity gradient. 1) The difference in species richness between different tropical forests is far greater than the difference between temperate zone and species-poor tropical forests. Whereas the temperate forest samples have 15-25 species and tropical dry forest ones mostly 50-60 species, the samples of moist and wet tropical forests average about 150 species and pluvial forests over 250 species (Gentry, 4 Annals of the Missouri Botanical Garden TABLE 1. Site characteristics for 0.1-ha samples. NumAltiPrecipiber of Grid tude tation FamiNumber of Site Coordinates (i) (mm) lies Species H' Reference Temperate North America Burling Tract, Virgin38055'N 30 1,053 12 21 3.54 Givnish et al., ia 77010'W unpubl. Northwest Branch, 39002'N 20 1,060 14 20 3.22 Maryland 77002'W Tyson Reserve, Mis38030'N 150 932 12 23 3.26 Zimmerman & souri (oak woods) 9003 lW Wagner, 1979 Tyson Reserve, Mis38030'N 150 932 11 25 3.68 Zimmerman & souri (chert glade) 90031'W Wagner, 1979 Babler State Park, 38032'N 150 930 13 21 3.61 Missouri 90040'W Cuivre River State 39?01'N 140 930 15 26 3.46 Park, Missouri 91000'W Valley View Glades, 38015'N 225 930 14 22 3.68 Missouri 90037'W Indian Cave State 40030'N 320 900 12 23 3.74 Tate, 1969 Park, Nebraska 95043'W Great Smoky Moun21-30 White, 1983 tains National (upper 5%) Park, Tennessee/ N.C.