Using Freshwater Fish to Explore Some Major Assumptions of Species Distribution Models

In a wide range of conservation applications such as predicting invasion or restoration success and species responses to land use and climate change, or in GAP analysis or delineating protected areas for species of concern, species distribution models (SDMs) are inevitable.  These models however rely on a wide range of statistical and ecological assumptions that are often ignored or not rigorously tested. I use a dataset of the distribution of 20 diffusively rare freshwater fish species to explore how assumptions and modeling methods affect inferences in predictive SDMs. Biological data for this effort was obtained from the USGS NAWQA program, whereas the environmental data were obtained from the USGS Water Resources Division and the NHDplus products. Some specific questions I investigate include: 1) Are environmental determinants of fish occurrences coincident with determinants of diffusive rarity? 2) How much do pure spatial versus environmental factors determine predictions of occurrences and abundances? 3) How do factors such as sample size, prevalence, range extent, and generalist versus specialist traits affect predictive accuracy and precision of fish distribution models? Statistical model choice, validation, and comparison of null hypothesis testing and information-theoretic inferential approaches are also explored.