EEB Student Thesis Defense - A Composite Likelihood Approach for Inferring Isolation by Environment on Spatially Autocorrelated Landscapes

Summary: Understanding the factors that govern the distribution of genetic variation across a species is a fundamental goal of population genetics, and has important implications for conservation. For example, identifying environmental variables that influence gene flow can help us predict how populations may respond to rapidly shifting ecologies in the face of climate change. However, when nearby environments tend to be similar, it is hard to tell whether observed patterns of gene flow are actually controlled by environmental variables, or whether they are just controlled by distance, which is the default expectation. In this talk, I will introduce an update that I developed to the R package BEDASSLE, which is designed to distinguish between these scenarios, called isolation by environment (IBE) and isolation by distance (IBD). My update modifies the underlying model, making the package faster and compatible with larger datasets, and adds a model selection component that directly tests whether IBE, IBD, or both combined, best explains the data. The updated method performed well on my simulated data, suggesting that it could be a useful tool for understanding what shapes genetic structure in populations.

Advisor: Gideon Bradburd