Understanding the geographic distribution of species: An evaluation of different methods for modeling species distributions and a test of the niche characteristics hypotheses.

摘要

Achieving a better understanding of factors that limit the geographic distribution of species is a key objective in different disciplines. Environmental variables (EV; e.g., temperature) are frequently used to help describe distributions of species because survival and reproduction can occur only within a certain range of EV. Thus, EV are used to predict the potential distribution of species through the use of species distribution modeling (SDM) methods and to test mechanisms that underlay variation in range sizes of different species. This study evaluated the accuracy of seven different SDM methods (Bioclim, CART, DOMAIN, GLM, GARP, MARS and MaxEnt) in predicting the geographic distribution of birds and Bursera trees in Mexico. In addition, this study tested the hypothesis that niche characteristics (niche breadth and niche position) explain inter-specific variation in species' range sizes for Bursera trees. SDM performance was evaluated as a function of the (1) type of EV included in the model; (2) the species taxonomic group and their range size; and, (3) local abundance and environmental variation. Significant differences were found among models with MaxEnt, DOMAIN, MARS performing significantly better than Bioclim, GARP, GLM, and CART, but performance was influenced by range size, taxonomic group, abundance, environmental variation, and environmental variable type. All models performed less well for widely distributed species, but some, like DOMAIN, were particularly challenged relative by species that occur over broad geographic areas. Birds were harder to model accurately when compared to trees, and direct variables were significantly more predictive of species' ranges than indirect variables, especially for plants. Using a combination of direct and indirect variables did not increase model performance, and in fact, lowered average performance for trees, although not significantly so. Models based on narrowly restricted and common species had lower performance than models based on narrowly restricted and uncommon species. Finally, the niche breadth hypothesis, which suggests a positive relationship between niche breadth and range size, was strongly supported. Thus, Bursera tree species that have broad environmental tolerances and that are able to utilize a wide range of resources are also widely distributed.