Population structure: individual-based clustering and pairwise differentiation
The DAPC of posterior genotype probabilities assigned to their respective population provenance revealed high overlap between populations from different forest patch sizes in the BDFFP landscape in both species (Figure S5). When the DAPCs was performed assigning populations to patch size groups (1-, 10, 100-ha, and continuous forests), we observed that populations of R. flaccida in 1-ha fragments tend to diverge from the main cluster along Axis 1, which explained >60% of the genetic variation (Figure 3A). With the reduced dataset, the population in the 1-ha fragment showed little overlap with the population clusters of 10-, 100-ha, and continuous forests. The population of C. surinamensis in the 100-ha fragment using the complete dataset, genotypes diverged along the Axis 1 from the clusters with the populations in 1-, 10-ha, and continuous forests. Between the 1-, 10-ha, and continuous forests clusters, genotypes tend to diverge along Axis 2 which explained 25.9% of the genetic variation. When we analyzed the reduced dataset the four clusters that correspond to different forest patch sizes, diverged along both Axis, with little overlap (Figure 3B).
Pairwise FST comparison of population andFST comparison of the population over loci are presented in Tables S7 & S8. Global FST measures across populations in small and large patches were 0.064 for R. flaccida and 0.191 for C. surinamensis with the complete dataset. With the reduced dataset FST measures were lower: 0.014 for R. flaccida and 0.128 for C. surinamensis . For populations of R. flaccida , mean pairwise comparison FST values tend to be higher among small patches, than between small-large patches and within large patches (Figure 4). Higher values of FST were also observed for pairwise comparison among small patches for C. surinamensis , while between small-large patches and within large patches tended to be lower (Figure 4).
Using the statistical test of AMOVA we group populations of the two species by patch size (1-, 10-, 100-ha, and continuous) and by size class (small and large patches). In both species, the major proportion of the genetic variation observed was explained by variation within populations. The geographic location of the populations across the landscape explained to a minor extent the genetic variation observed (Table 2). The species R. flaccida FST values tend to be slightly higher among small patches comparison and between small-large patches comparison, indicating some degree of differentiation. However, this genetic difference was not significant (Table 2). On the contrary, genetic differences between small and large patches were significant for the species C. surinamensis (Table 2).