Population structure within Sporophila corvina
We used three complementary methods to estimate patterns of genetic differentiation and population structure. The Estimated Effective Migration Surfaces (EEMS) approach found that the continental divide between Caribbean and Pacific slopes forms the most important barrier where gene flow is 100 times lower than the average despite relatively short geographic distance (Fig. 1B). This continental divide extends from the Central Valley of Costa Rica along the Talamanca mountain range throughout western Panama, reaching the Canal area in central Panama (map areas colored dark orange). Conversely, individuals on either side of both slopes showed high genetic similarity, likely due to strong connectivity. Our results were consistent among deme densities (200, 300, and 400 demes), and between runs (Figure S2).
Further, we found that S. corvina individuals from Costa Rica and Panama were best grouped into three genetic clusters (Figure 1C), as this clustering presented the lowest cross validation error (population assignments for K = 2–5 are available in Figure S3). The three genetic groups that we identified were largely consistent with the known distributions of the three subspecies present in the region (Figures 1A,C and S4). However, we also identified a large proportion of individuals with mixed ancestries, suggesting pervasive gene flow between subspecies. From the 255 individuals included, 107 were identified as pure parental subspecies (53 corvina , 37hoffmanni , and 17 hicksii ) and 148 showed some evidence of admixture (79 corvina-hicksii , 16 corvina-hoffmanni , 46hoffmanni-hicksii, and 7 three-way hybrids; Figure 1C). We found relatively low genetic differentiation between subspecies pairs; the highest differentiation occurred between black and pied subspecies (corvina vs hicksii : FST = 0.074,corvina vs hoffmanni : FST = 0.066), while the two pied subspecies showed the lowest genetic differentiation (hoffmanni vs hicksii : FST = 0.039). Finally, the PCA analysis identified roughly three genetic clusters; the first axis of variation explained 4.44% of the genetic variation and primarily separated corvina individuals from the two pied subspecies and the second PC axis explained 2.10% of the genetic variation, and divided hicksii from the other two subspecies (Figures 3 and S5).