Differences among latitudinal regions (Q2)
Adding the factor latitudinal region to models with the four life history traits notably increased fit to the data (models 5–7, AIC=–488.6 to –503.9, Table 1). This is particularly evident when comparing the best-fit models for each instance (models 4 vs. 6, ΔAIC=21.6). Model performance was indistinguishable for models 6 vs. 7 (ΔAIC=1), which only differed in the addition of seed dispersal mode. Finally, in models 5 and 7 the factor seed dispersal mode was no longer a significant predictor of FST (Table 1 and 2). Below we focus on results from model 7, as it is the most inclusive model of the factors we tested with the best fit to the data.
Figure 1 shows how the levels of each factor affect population differentiation as measured by FST values (after transformation). The effect of each factor is depicted after accounting for the effect of the other independent variables in model 7. For mating system, outcrossers tend to have lower population differentiation than mixed-mating plants (Fig. 1a). Trees tend to have significantly lower population differentiation relative to non-woody plants and shrubs, while the latter two growth forms did not differ between each other (Fig. 1b). Pollination by small insects leads to significantly greater differentiation compared to large insect, vertebrate and wind pollination, while the latter three pollination modes did not differ between each other (Fig. 1c). Temperate zones have significantly lower FST values than tropics and subtropics, and the latter two regions did not differ from each other (Fig. 1e). Finally, seed dispersal mode was not a significant predictor of population genetic differentiation. FST values associated with gravity dispersal were highly variable, and although gravity dispersal results in higher FST values compared to wind dispersal, this difference was not significant. Animal dispersal also resulted in highly variable FST values that did not differ from other dispersal modes (Fig. 1d).