Patterns of genetic diversity.
We estimated changes in patterns of genetic diversity between the three datasets using a variety of metrics using Excel add-on GenAlex V6.5 (Peakall & Smouse 2006; Peakall & Smouse 2012). These include the percentage of monomorphic loci, expected heterozygosity (He), observed heterozygosity (ho) and the diversity q-profile of “effective-number” diversity metrics qD as Box1 in Sherwin et al. (2017). These are the effective number of alleles on three scales: allelic richness ((\({}^{0}D\)) or q = 0), Shannon’s information ((\({}^{1}D\)) or q = 1), and heterozygosity ((\({}^{2}D\)) or q = 2). \({}^{2}D\) from Shannon’s information index is mathematically intermediate between the number of different alleles, and the effective number of alleles derived from heterozygosity (Sherwinet al. 2017, 2021); it therefore avoids either the former’s heavy emphasis on rare alleles (and resulting problems of estimation due to the likelihood of undetected alleles), or the latter’s very heavy dependency on common alleles (thus largely ignoring rare alleles which can be very important for adaptation). Differences ofqD metrics between datasets were assessed using a Wilcoxon test, with each locus being used as an independent assessment of genetic diversity. While it is unlikely that all loci were truly independent, the effect of this non-independence should not be large enough to impact the findings (Waples et al. 2021).