Population genetics methods are commonly used to evaluate migration of individuals among subpopulations. However, inferring migration rates can be difficult when subpopulations are closely related (low genetic differentiation) or sample sizes are small (< 30 individuals). We developed a population genomic method suitable for these challenges. From observed allele frequencies among sparsely sampled and genetically similar subpopulations, we simulated parental populations, F1 hybrids, and backcrosses (F1BCs). We then assigned each individual to a parental population using a leave-one-individual-out (LOO) approach with ADMIXTURE in a supervised framework in which we iteratively queried the genetic origin of a single individual while all other individuals served as reference samples for the population in which they were collected. We applied the LOO approach to the observed data and classified individuals as residents, migrants, F1 hybrids, or F1BCs by comparing observed assignments among populations to assignment distributions characterized with the simulated dataset using permutation tests.
We applied this method to evaluate migration rates of invasive wild pigs (Sus scrofa) among counties in Missouri, USA. We used our methods and Bayesian models to evaluate the environmental and anthropogenic covariates associated with the movement of invasive wild pigs.
Results/Conclusions
We found that our simulation and assignment methods create an effective strategy for evaluating migration when populations are genetically similar and sample sizes are small. For example, we were able to reliably detect backcrossed individuals’ ancestry up to two generations after migration. We also found our LOO assignment method to be more reliable than traditional assignment tests.
Bayesian models revealed that migration of wild pigs among counties in Missouri was positively associated with the recreational hunting industry and wildlife violations, suggesting that humans illegally move wild pigs to support populations for hunting. Covariates in the models allowed us to predict the probability of wild pig movement into each county in Missouri. We are in the process of extending these methods to the national level. The ability to predict movement of invasive species into new locations can provide managers and policymakers with the information necessary to prevent such movement in the future.