Heritable genetic variation of dispersal
In our house sparrow metapopulation data set there were 2118 resident adults and 484 adults that had dispersed between islands prior to recruitment. Among the dispersers, 399 individuals dispersed to an island of the same habitat type as their natal island, whereas 85 individuals either dispersed from a farm habitat island to a non-farm habitat island (N = 42) or in the opposite direction (N = 43). Proportions of dispersing recruits produced by adult sparrows was considerably higher on non-farm habitat islands than farm habitat islands (Figure 2; Supplementary Table S1), and within each habitat type there was a tendency for disperser parents of both sexes to produce a somewhat higher proportion of dispersing recruits than parents that were residents (Figure 2). The interchange of individuals between islands of different habitat types enabled us to use genetic groups animal models to separate heritable genetic causes from environmental causes of spatial variation in individual dispersal propensity, because house sparrows with genomes partially originating from the farm genetic group were present in the non-farm habitat and vice versa (Supplementary Figure S1).
The basic GGAM analysis showed that the dispersal probability in our house sparrow metapopulation had a heritable genetic basis, with additive genetic variance explaining approximately 10% of the observed variation in dispersal probability (Table 1). Furthermore, the basic GGAM indicated that a considerable portion of the observed variation in dispersal probability among individuals was explained by environmental differences between natal islands (ca. 25%) and hatch years (ca. 1 %), and provided strong evidence that females had a higher probability to disperse than males (Table 1). Finally, there was strong evidence from our basic GGAM that the estimated mean genetic value (i.e. mean breeding value) for dispersal was lower for the non-farm habitat than the farm genetic group (Table 1).
Our extended GGAM analysis showed that the farm genetic group had a higher additive genetic variance for dispersal probability than the non-farm genetic group (Figure 3, Table 1). The posterior difference in additive genetic variance between the farm and non-farm genetic groups had a mode of 0.556, with a 95% credible interval (CI) ranging from 0.045 to 1.134 (Supplementary Materials, Figure S2), providing clear evidence that the additive genetic variances were different. Correspondingly, the heritability of dispersal was higher in the farm habitat than the non-farm habitat (h2 = 0.124 and 0.017, respectively; Table 1). In contrast, the proportions of variation in dispersal probability explained by differences between hatch years and natal islands were similar in the two habitat types and similar to estimates from the basic GGAM (Table 1). In agreement with the basic GGAM, there was also strong evidence from the extended GGAM for a sex difference in dispersal probability, and that the estimated mean genetic value (i.e. mean breeding value) for dispersal was lower for the non-farm than the farm genetic group (Figure 3, Table 1).