Species Distribution Models

Developing methods to model where species are, where they aren't, and why

Ecologists and managers use species distribution models to make predictions about where species are likely to be. My work in this area focuses on the development, implementation, and evaluation of new methods for species distribution modeling. In particular, I've focused on the comparative performance of different methods of density-ratio estimation for estimating the relative suitability of a habitat for a given species. I am finishing up development of an R package, in collaboration with my co-PI John Drake and Cory Merow at the University of Connecticut, which provides easy access to these methods as well as a number of others developed in the Drake Lab over the past few years.

We want to know where parasites are too

In addition to the development of new methods I strive to apply existing methods to new contexts, in particular to the spatial prediction of infectious disease risk. So far this has spanned from the application of tried and true machine learning methods to a complicated and sensitive data set on Guinea worm infection in the dogs of Chad, to the identification of transmission dominant host species in an ungulate community on the National Bison Range. I also lead two separate groups in this area at the Center for the Ecology of Infectious Diseases at UGA. In our Disease Mapping Working Group we aim to provide a setting in which center members can develop their spatial modelling and mapping skills. Currently this involves developping a species distribution model for the parasite Echinoccoccus multilocularis. In our CDC Aedes Prediction Challenge team, we use stacked species distribution models to submit monthly predictions of the probability of detection of Aedes aegypti and Aedes albopictus mosquitoes in over 150 counties across the USA.