COS 86-5 - Predicting biodiversity in a variable world: pairwise analyses

Background/Question/Methods

Studies on spatiotemporal patterns of biodiversity have extensively employed presence-absence data. Over the years, ecologists have gathered evidence that environmental heterogeneity is an important source of spatial variations of biodiversity. Beta diversity, that is, the changes in community composition across space is commonly evaluated through different measures that rely on null models that, in general, do not consider heterogeneity or variability in: (i) the size of species pool or landscape, (ii) species incidence probabilities, and (iii) changes in environmental conditions and how species respond to them. In addition, beta diversity measures are often static and neglect the temporal dynamics of biodiversity. Alleviating these constraints is central for advancing beta diversity research. Here, we develop a framework that incorporates heterogeneities of different kinds to analyze and make predictions about species presence-absence data across space and time.

Results/Conclusions

Our framework is based on the species independence assumption, i.e., species incidence in a given location is not influenced by the presence of other species. Although our approach is neutral in this respect, it can deal with spatial heterogeneity and various species responses to variable environments across a landscape. Our approach represents an extension of MacArthur and Wilson's (1967) island biogeography theory that incorporates colonization-extinction processes with both temporal dynamics and the spatial dimension. In particular, we provide an exact calculation of the expected Jaccard dissimilarity index between two sites that may differ in their environmental conditions under the hypothesis of species-independent colonization-extinction dynamics. Although our examples use simple colonization-extinction models, our formula for the expected value applies for any species distribution model predicting species incidence probabilities across a landscape as long as species independence is assumed. In addition, we also calculate segregation and co-occurrence indices for two local sites inhabited by a variable number of species. Our framework allows us to make quantitative predictions about compositional dissimilarity. We also found that the steeper the environmental gradient is, the steeper the expected compositional change will be. We conclude with some applications and further generalizations of the framework beyond pairwise analyses.