2.5 Population genomic summary statistics
A series of diversity and demographic statistics were estimated from the folded site frequency spectrum (SFS) with ANGSD . Genome-wide heterozygosity was estimated per sample by first computing the folded site allele frequency likelihood using the reference genome as ancestral state and then calculating the folded SFS. The folded SFS was calculated independently for each sampling site after removing admixed (less than 70% ancestry to any cluster under K4) and migrant individuals (individuals whose ancestry was different from the prevalent cluster of their sampling location). Then we estimated both Watterson’s theta and Tajima’s D, using a sliding-window approach with window size of 50 kb and a step size of 10 kb. Individual inbreeding coefficients (F ) were estimated with the software ngsF (v1.2.0) (Vieira et al., 2013). First, approximate F were obtained in an initial run using the -aprox_EM method, with a maximum root mean squared difference between iterations of 1x10-5(-min_epsilon ) and random initial values. From the output of this first run, the initial parameters for the final run were derived, where the -min_epsilon value was decreased to 1x10-7 to assume convergence. To avoid convergence to local maxima, this two-step analysis was repeated ten times, as suggested by the authors (Vieira et al., 2013).