Ecological niche divergence
ENMs sufficiently predicted habitat suitability within each lineage’s buffered minimum convex polygon, as indicated by mean AUC scores:S. atlantis (mean = 0.809, s.e. = 0.084), northern S. hesperis (mean = 0.810, s.e. = 0.063), and southern S. hesperis(mean = 0.803, s.e. = 0.085) (Fig. 4). Visual inspection of predicted habitat suitability surfaces across the entire study landscape suggests that the three genomic lineages are divergent in their respective habitat associations and ecological niches, and that the highest density of suitable habitat for each lineage is generally found within and adjacent to their buffered minimum convex polygon. Relative contributions of geographic and environmental predictor variables to ENMs, measured as the drop in AUC scores after each variable was randomly permuted, are reported in Table 2. For each of the three lineages, land cover and growing degree days had the greatest contribution to ENMs. Contributions of other variables varied considerably among lineages.
Linear mixed effects models indicated there was a significant reduction in the power of each lineage’s ENMs when predicting the localities of each of the other two lineages. The coefficient of the “within-vs. between-lineage” binary predictor variable was significantly negative for each pairwise comparison: S. atlantis and northernS. hesperis (β = -0.292; p = 0.028), S. atlantisand southern S. hesperis (β = -0.6811; p < 0.00), and northern and southern S. hesperis (β = -0.333;p = 0.023). Together, these results indicate that all three lineages are significantly divergent in their habitat associations and ecological niches.
These analyses were repeated using null ENMs built with randomly generated localities confined to each lineage’s buffered minimum convex polygon. Linear mixed effects models addressing resulting AUC scores indicated no significant reduction in the power of each lineage’s null ENMs when predicting the random localities generated for the other lineages. Specifically, the coefficient of the “within- vs.between-lineage” binary predictor variable was non-significant forS. atlantis and northern S. hesperis (β = -0.012; p= 0.822), S. atlantis and southern S. hesperis (β = 0.020;p = 0.751), and northern and southern S. hesperis (β = 0.015; p = 0.810). These results indicate that observed differences in habitat associations and ecological niches cannot be attributed to biases arising from differences in available habitat among the three lineages.