Figure 1. Sampled localities (red dots, grey circles give population ID and clustering as estimated from admixture analyses (Fig. 3)) depicted on maps of habitat suitability(IBRHabitat, 0 – unsuitable, 1 – highly suitable). Red lines represent least-cost distances between study localities, white areas on maps represent elevations below/above possibly suitable distribution ranges (for habitat suitability across full distribution, see Fig. S3). Per-species min/max FST is given, as well as the pairwise FST for the two adjacent populations. I-spline plots show the variable most strongly explaining population genetic differentiation as estimated through GDM (Table 2). Black bars represent 100 km, black outlines represent coast lines of Costa Rica and Ecuador. A: Bee-pollinated Ad. adscendens , lowland rainforests, Costa Rica; B: Bee-pollinated M. maxima, montane cloud forests, Ecuador; C: Hummingbird-bat-pollinated M. phlomoides , montane cloud forests, Costa Rica; D: M. sanguinea with hummingbird-rodent-pollinated southern populations and hummingbird-bat-pollinated northern population, cloud forest-Páramo ecotone, Ecuador; note that while rodents (ground-dwelling mammals) are considered as relativly immobile pollinators, we found high connectedness among localities, possibly attributable to traplining hummingbird pollinators; E: hummingbird-bat-pollinated M. tomentosa , cloud forests, Ecuador; F: passerine-pollinated A. costaricensis, cloud forests, Costa Rica.