Doctor of Philosophy (PhD)
Most of the earth’s biodiversity is unknown to science. With global climate change set to drastically alter its distribution, it is imperative to catalogue it and understand its function in order to preserve it and better understand how this change will impact humanity. Recent technological and statistical advances have in theory made possible increasingly rapid discovery and description of diversity. The statistical properties and performance of these new approaches are still poorly known, however, their integration with complementary methods from disparate disciplines has not been achieved. In this dissertation we present three chapters of original research that advance these areas of biodiversity science. The first introduces a new implementation of the GMYC, a statistical model used for species delimitation. This implementation fully accounts for uncertainty in model parameters, and we test its performance under various historical scenarios. We find that the model generally performs well, but that failing to account for uncertainty in nuisance parameters inflates confidence in species limits. The second introduces a method to examine the fit of empirical data to a multispecies coalescent model commonly used in phylogenetic inference. Systematic and phylogeographic studies are generating ever-larger datasets which often range up to the genome scale and often wish to use coalescent models to infer parameters such as phylogenies, divergence times and effective population sizes. Though the multispecies coalescent can infer these parameters, it is unclear the extent to which it is a good fit for these new empirical datasets. We employ our new approach to 25 published datasets and find that a majority of them show poor fit to the data and that for some of them, that poor fit affects inference. In the last chapter we integrate statistical approaches from both ecology and systematics to infer species limits, phylogeny, population genetic structure and ecological community structure in a study of a poorly known tropical alpine fly fauna. We find that we can effectively describe patterns of diversity in the absence of a low-level taxonomic framework, but that inference of the processes structuring that diversity remains difficult. We also find that some of our inferences of community structure are sensitive to uncertainty in species limits and phylogeny.
Document Availability at the Time of Submission
Release the entire work immediately for access worldwide.
Reid, Noah M., "Model-based approaches to discovering diversity : new implementations, tests of adequacy and an empirical application to central American Diptera" (2013). LSU Doctoral Dissertations. 18.