Development of a Transferable Species Distribution Modeling Approach

Understanding the distributions of species is fundamental to ecology, especially within the marine environment where biotic and abiotic variables are often dynamic over space and time. Ecological models have been used to make inferences on species-environment relationships in addition to predicting past and future patterns of species distributions. However, these models are typically developed for a single location, time period, spatial scale, or species life stage, resulting in poor predictive accuracy beyond the study domain. This project will provide a new transferable modeling approach that is applicable to a wide range of species, which will be used to test predictions of current and future distributions of green turtles (Chelonia mydas) in the Gulf of Mexico, Brazil, and Qatar. Since there is a lack of information on how marine turtles in general will respond to climate change impacts, these results will provide a crucial evaluation of projected range shifts that could be used by both conservationists and decision makers. This project will foster the integration of research and education through mentoring undergraduate students that identify as underrepresented minorities, teaching marine ecology and conservation at Florida State University, hosting workshops on the analysis of animal movement data, as well as hosting public outreach events. These efforts to disseminate knowledge and results beyond traditional means within the scientific community will be used to engage future scientists and the general public on pressing ecological topics.

The overall aim of this project is to develop a transferable modeling approach that can provide accurate and precise estimates of species distributions under changing conditions. This study addresses some of the challenges of model transferability through three hypotheses: 1) a model informed by biological processes and physiological constraints will exhibit greater transferability than those that do not; 2) the analysis of environmental variables at the scale of species perception will result in greater model transferability; 3) a model that accounts for differences among life stages will result in greater model transferability. These hypotheses will be tested using a large dataset of green turtle occurrences at three distant locations (Gulf of Mexico, Brazil, Qatar) with a flexible Bayesian species distribution model. The use of independent datasets (from Brazil and Qatar) to validate the species distribution model (developed on data from the Gulf of Mexico) provides a unique opportunity to critically evaluate the spatial transferability of the model. The findings from this proposed work will provide a roadmap for the further development of transferable ecological models to predict biodiversity, disease outbreaks, and risks of biological invasions across global regions and possibly under future climate change impacts. Elucidating the importance of biological mechanisms, scale, and life stage differences to ecological modeling is transformative because these factors have been largely untested with regard to model transferability. Since predictions remain a major frontier in ecology, products of this work will be of broad interest to a diverse group of scientists, such as marine ecologists, biological oceanographers, fisheries scientists, and conservationists.

Josh Cullen, PhD
Josh Cullen, PhD
NSF Postdoctoral Research Fellow

My research interests include animal movement ecology, Bayesian modeling, and R stats.