Does neutral theory explain community composition in the Guiana shield forests?

Abstract

Recent years have seen a fierce debate about the causes of beta-diversity, the change in species composition over landscapes, in tropical rainforests. The introduction of neutral models has spurred this debate, by assuming that species composition changes randomly over landscapes with dispersal limitation as the driving ecological process. In this study the interaction between local and regional diversity of trees is investigated. We inventoried 61 1-ha plots in forests on brown and white sands, and on bauxite mountains in Guyana and Suriname. These data combined with earlier data led to the first biogeographical study on a species level involving 156 forest plots ranging from Venezuela to French Guiana. Forest composition changed dramatically between forest types, and at geographical distances. In general, forest composition followed geological patterns in the Guianas. Differences in plant attributes between the forest groups revealed a spectrum in covariation. On the one end of this spectrum a high dominance of trees with large seeds in dry fruits was combined with high wood densities in several forest types. On the other end a high dominance of trees and species with smaller seeds in fleshy fruits was combined with relatively low community averaged wood density. This spectrum could reflect different colonizing strategies of trees in the forest types, and could have an effect on potential dispersers. With neutral-like dynamical community models, different hypotheses on the low diversity of the white sand forest were evaluated. Forests on white and brown sands draw their species from different regional species pools. We provide strong evidence that asymmetric environmental filtering and a lower regional diversity (possibly due to the smaller size of the white sand area) could influence the lower diversity found in white sand forests compared with neighboring brown sand forests. White sand forests were not necessarily more dispersal limited than brown sand forests. Still within the white sand forests apparent changes in species abundance patterns occurred along geographical distances. Thus, we showed that a gradient in beta-diversity could occur in forests under similar climatic and soil conditions. The conservation value in terms of tree diversity of three bauxite mountains in Northeastern Suriname, was evaluated by placing 23 1-ha plots on the plateaus, slopes, and in the lowlands. These mountains are all destined for open-pit mining for aluminum ore, and may constitute a rare and endangered landscape type. Habitat type (e.g. the plateaus) explained more floristic similarity between plots than geographical distance. The plateaus also had the highest tree alpha-diversity currently found in Suriname. However, forest composition on the plateaus was also partly a random draw of species from the lowlands, and each of the mountains had its own specific species composition. Forests in the Guianas show evident differences in tree responses and composition, yielding little support for a neutral view. Most patterns are caused by small sets of common species that may dominate the forests, especially where the alpha-diversity of the forests is low. Striking biogeograpical patterns in the rainforests of the Guianas however remain unexplained by (current) environmental conditions.