Methodological advances in the construction and application of environmental niche models

Environmental niche models are a class of methods that use georeferenced locality data in conjunction with environmental data to build mathematical models of species environmental tolerances. These models are frequently used in evolutionary and ecological studies as estimates of some aspect of the species ecological niche. They are appealing as niche estimates due to their ease of construction and the ready availability of appropriate data, but there remain significant methodological and conceptual questions about their use and development. My research objectives were to (1) develop methods for making interspecific comparisons of environmental niche models; (2) demonstrate the application of those methods in a study of niche conservatism, an issue of fundamental importance in evolutionary ecology; (3) develop and compare methods for measuring uncertainty in niche models, and (4) develop and distribute a software package that makes all of these analyses freely available to members of the scientific community.;The first chapter introduces the use of two similarity metrics for environmental niche models and uses resampling procedures to test hypotheses about patterns of niche evolution. These metrics are also mapped onto a phylogeny for Cuban Anolis lizards to examine patterns of niche divergence within the clade. The results of this study and the accompanying literature review show that closely related species are frequently more ecologically similar than expected under some null hypotheses, but are rarely identical. The second chapter demonstrates the application of the nonparametric bootstrap and delete- d jackknife to niche models built using maximum entropy and shows that significant uncertainty can exist in projections of habitat suitability even on fairly large data sets, and also that models can predict habitat suitability with high confidence while still containing considerable uncertainty in the underlying mechanistic model of environmental tolerances. It is also shown that the delete-d jackknife can overestimate uncertainty when sample sizes are small and d is large. The third chapter serves as a brief introduction to the ENMTools software written to perform these analyses.