To ensure that individuals clustered according to their described species, we assessed nuclear genetic structure through Bayesian model-based clustering in structure v.2.3 . structure determines the most likely number of differentiated clusters (K) represented by the sample and assigns the sampled genotypes to the inferred clusters. Using a random subset of 1,000 markers, we estimated the log likelihood of the data, given different numbers of genetic clusters K, using an admixture model with correlated allele frequencies, without sampling locations as priors and with all other parameters as defaults. For each k value of 1 through 6, we ran ten replicates (50,000 burn-in cycles, 100,000 MCMC iterations), from which we calculated ΔK following . We used the program structure harvester to identify the number of populations K with the best support.