Detection of Outlier Loci
Prior to assessing population structure and adaptation, we identified outlier loci (i.e., loci putatively under the influence of selection) and divided our datasets into (1) putatively adaptive and (2) putatively neutral markers for each species. To detect outlier loci, we used four population differentiation (PD) methods and two environmental association (EA) methods that implement different methodologies (Table 2, Appendix D).
The use of both PD and EA methods allows for the conservative detection of outlier loci by comparing complementary methods with different assumptions and biases. PD methods use measures of differentiation (FST or in the case of ordination approaches, multivariate distances) to identify outlier loci, where extreme levels of differentiation (i.e., those outside of neutral expectations) are likely candidates for selection (Liggins et al. 2019). EA methods use correlations between environmental variables and allele frequencies to identify loci putatively under selection (Rellstab et al. 2015). While both methods are prone to false positives, combining approaches is an effective strategy for identifying true outlier loci (de Villemereuil et al. 2014; François et al. 2016; Liggins et al. 2019). Here, we retained only loci identified as outliers by at least three methods and divided each of our datasets into neutral and outlier loci.