Data S1: Sample list, collection site details, and summary of individual assignments to genetic clusters in ADMIXTURE and DAPC analyses
Data S2: ENM modeling results summary, including model selection results and diagnostic statistics from ENMeval (Muscarella et al.2014) and MaxEnt model metrics from ENMwizard (Heming et al.2018)
Table S1: Per-locus genetic summary statistics for P. tremuloides and P. trichocarpa individuals in this study
Table S2: Pairwise FST estimates between P. tremuloides genetic clusters, and their 95% confidence intervals (CIs)
Table S3: Hierarchical analysis of genetic variance among P. tremuloides SNPs
Results are given for variance partitioning within genetic clusters (Fclust/total), within subpopulations by cluster (Fpop/clust), and within individual trees relative to populations. Ranges given in parentheses are 95% confidence intervals.
Figure S1: Plot of ADMIXTURE cross-validation error versus K, showing that K= 3 is the best fit for the full P. tremuloidesand P. trichocarpa dataset of 34,796 SNPs (sensuAlexander & Lange 2011)
Figure S2: Plot of Bayesian information criterion (BIC) scores for k-means clustering solutions over a range of K, from the first step of DAPC, with K = 3 being the best solution
Figure S3: Results of DAPC cross-validation in Restablishing that the appropriate number of principal components to retain ranges from 20–100 with similarly high (>90%) prediction success
Figure S4: DAPC loading values plotted for all 34,796 SNPs, with SNP name labels beside the SNPs with the highest loadings
Figure S5: Genetic patterns of heterozygote and singleton alleles within and among P. tremuloides genetic clusters, calculated while excluding putatively admixed edge populations (Qmax < 0.75)
Results are analogous to corresponding panels of Fig. 3 (see Fig. 3 caption for details), but based on different sampling.
Figure S6: Heatmap of interindividual Nei’s D estimates
Results were reordered by row and column means, and flanked by dendrograms of the values. Color key and histogram at top left show the distribution of mean D-values.
Figure S7: Heatmap of unordered interpopulation FST estimates flanked by clustering dendrograms from the distances
Color key and histogram at top left show the distribution of mean FSTvalues.
Figure S8: Results of isolation by distance tests
Results are based on generalized linear modeling analyses of linearized FST versus log[geographic distance (km)] of P. tremuloides (top left) and its genetic clusters 1 (top right), 2 (bottom left), and 3 (bottom right).
Figure S9: Unrooted maximum-likelihood tree topology from the ingroup-only TreeMix analysis
Results are show the tree graph estimated when a single migration event was allowed (A), and residual plot of the graph (B). Scale bars and legends same as in Fig. 2.
Results of TreeMix analyses on the full dataset conducted over varying levels of k block sizes (10 to 5000 bp) accounting for linkage disequilibrium. Scale bars and legends same as in Figs. 2 and S9.
Figure S11: Projection of the final present-day ecological niche model of P. tremuloides onto three late Pleistocene climatic scenarios
The ENM was built with MaxEnt and bioclimatic variables obtained from WorldClim v1, and the three Pleistocene time-slices (B–D) correspond to scenarios described in the Fig. 4 caption and Table 1. Model projections show continuous suitability values obtained using the cloglog format of MaxEnt after the application of a 10th-percentile threshold. Extent of LGM ice sheets is indicated in white.