Probability distributions of extinction times, species richness, and immigration and extinction rates in neutral ecological models

Abstract

In community ecology, neutral models make the assumption that species are equivalent, such that species abundances differ only because of demographic stochasticity. Despite their ecological simplicity, neutral models have been found to give reasonable descriptions of expected patterns of biodiversity in communities with many species. Such patterns include the expected total number of species and species-abundance distributions describing the expected number of species in different abundance classes. However, the expected patterns represent only the central tendencies of the full distributions of possible outcomes. Thus, ecological inferences and conclusions based only on expected patterns are incomplete, and may be misleading. Here, we address this issue for the spatially implicit neutral model, by using classic results from birth–death processes to derive (1) the probability distribution of extinction time of a species with given abundance for the metacommunity; (2) the probability distributions of total species richness and number of species with given abundance for both the metacommunity and local community; and (3) the probability distributions of the average immigration and extinction rates in the local community, across different values of total species richness. We illustrate the utility of these probability distributions in providing greater ecological insight via statistical inference. Firstly, we show that under the neutral metacommunity model, there is only 2.65×10−9 probability that the age of a common tree species in the Amazon is ≤ 3 × 108 yr, which is approximately the oldest estimated age of the first angiosperm. Thus, species ages from the model are unrealistically high. Secondly, for a tree community in a 50 ha plot at Barro Colorado Island in Panama, we show that the spatially implicit model can be fitted to observed species richness and an independent estimate of the immigration parameter, with the fitted model predicting a species-abundance distribution close to the observed distribution. Our results complement those using sampling formulae that specify the multivariate probability distribution of species abundances from neutral models.