Candidate loci associated with climate
The univariate approaches identified a total of 975 loci associated with
climate variables. Bayenv2 identified a total of 631 outlier loci; of
these, 283 loci were associated with climate-based PC1 and 354 loci were
associated with climate-based PC2. To run LFMM, we first determined the
appropriate number of latent factors using snmf. The snmf analysis
returned the lowest CE value for K =10 (0.519), followed byK =11 (0.520) and K =12 (0.521) (Figure S3). As higher
values of K resulted in a higher number of outlier loci, we only
report outlier loci detected using K =10 as a conservative
approach. LFMM identified a total of 497 outlier loci, among which 134
loci were associated with climate-based PC1 and 377 with climate-based
PC2. Of the outlier loci identified by LFMM, 152 were also identified by
Bayenv2, corresponding to 15.6% overlap between the two methods. The
multivariate pRDA identified 485 outlier loci associated with the first
2 RDA axes. Among these, 69 loci were also identified by univariate
approaches. Thus, 5.0% of loci were identified by both approaches. The
RDA without correcting for population structure identified 108 loci as
outliers, with only 8 loci identified by univariate approaches as well
(0.8% overlap with univariate methods) and an overlap of only four loci
with the RDA with correction for population structure. Overall, a total
of 1,392 outlier loci corresponding to 1,003 RAD-tags were associated
with climate using all methods (Venn Diagrams Figure S1). We considered
213 loci detected by at least two methods as strong candidates.