“…More specifically, a standard Bayesian population genomic outlier screen (BAYESCAN, Foll & Gaggiotti, 2008) was complemented with individual-based latent factor mixed modeling (LFMM, Frichot et al, 2013), and multivariate redundancy analyses (RDA, Borcard et al, 2011), to identify (i) SNPs putatively under selection and (ii) the main environmental drivers of population differentiation while accounting for spatial autocorrelation in allele frequencies. Although the use of candidate SNPs has been the standard method to find loci putatively under selection (for example, Namroud et al, 2008;Alberto et al, 2013;Guichoux et al, 2013;Olson et al, 2013), they may compromise the relevance of classical outlier detection methods due to overestimation of neutral distributions. This stresses the importance of integrating landscape genomic methods to reliably reveal loci putatively under selection and the selective agents underlying the observed patterns (De Villemereuil et al, 2014;Weinig et al, 2014).…”