Spatial autocorrelation and migration patterns
Results for the spatial autocorrelation of all population pairs in small patches (1- and 10-ha), and large patches (100-ha, and continuous forests), for the complete dataset, are summarized for the two species (Figure 5; Table S9). When we examine all population pairs, irrespective of the patch size, pairwise genetic and geographic distances show no correlation based on the Mantel test for both species (R. flaccida : R2 = 0.074, p-value = 0.269;C. surinamensis R2 = -0.014,p-value = 0.514). Using the dataset with 20% of missing data, the same pattern was observed (R. flaccida :R2 = 0.006, p-value = 0.5; C. surinamensis R2 = 0.175, p-value = 0.079). Correlation within pairs of populations of R. flaccida in small patches showed a positive slope indicating that populations in small patches as they are farther distant geographically, are more genetically dissimilar (Figure S8A). This was more pronounced when we use the reduced dataset (Figure S8D) with a significant correlation based on the computed Mantel test (R2 = 0.46,p-value = 0.02). For C. surinamensis , a non-significant positive slope was observed (R2 = 0.389,p-value = 0.068) within populations in small patches. When pairwise comparisons were computed for populations within large patches, the genetic distances were not correlated to geographic distances in both species (Figure S8B, E, H, K). R. flaccida presents a slightly positive slope (R. flaccida : R2= 0.186, p-value = 0.341), while populations of C. surinamensis showed a non-significant negative slope (C. surinamensis: R2 = -0.11, p-value = 0.65). The reduced dataset with >75% of missing data supports the same pattern (R. flaccida : R2 = 0.056,p-value = 0.44; C. surinamensis:R2 = -0.055, p-value = 0.56).
The relative migration levels (Nm ) estimated between populations based on neutral SNPs using the GST and Nmparameters showed similar patterns (Table S10). The network graph using the Nm parameters suggests a high migration between R. flaccida populations irrespective of patch size and isolation (Figure 5C, Table S10). Populations in small patches act as source populations with high relative migration to other forest fragments and continuous forests across the landscape. While populations in large patches showed signs of asymmetrical migration among them, significant symmetrical migration was observed between the populations in the Florestal continuous forests and the Dimona 100-ha fragment. The migration network of the species C. surinamensis was less complex with fewer patches connected by network vertices, suggesting low or no migration between populations across the landscape (Figure 5D, Table S10). Population in small patches showed significant relative migration among them, at a higher rate than with large forest patches. Populations in large patches were completely isolated (Dimona 100-ha and Dimona continuous), where no migration was observed. Significant migration was observed but at a relatively low rate from the Florestal continuous forest to the closest Km 41 and Cabo Frio continuous forests.