2.4.2 Identifying potential outliers
A modified version of POWSIM (Ryman & Palm, 2006) was used to investigate whether the distribution ofF ST values in our Pool-seq data was consistent with the expectation for selectively neutral loci, or if there was deviation from such a distribution which could reflect directional selection (cf. Lamichhaney et al., 2012). We compared the observed genome-wide distribution ofF ST with a simulated “expected” distribution under drift only. The observed distribution was based on per SNP estimates (1 bp; n =12,177,462; meanF ST=0.083). We used a conservative approach defining outliers as SNPs with an observed F STabove the largest value of the expected distribution (F ST>0.864; details in Appendix S3). We also defined outliers as those SNPs with anF ST above a thresholds corresponding to the 99.5th percentile (e.g. Fabian et al., 2012; Keehnen, Hill, Nylin, & Wheat, 2018) of the genome-wide F ST distribution. A threshold of the 97.5th percentile was used forF ST based on 5 kb windows (e.g. Gagnaire et al., 2018; Keehnen et al., 2019).