Moderators of the dispersal-range size relationship
The final model, including both plants and animals, explained 41% of the variation (50% considering random effects) and identified four moderators as main factors explaining the differences in the dispersal-range size relationship between studies (Fig. 3). Specifically, studies that included multiple dispersal-related traits, that were carried out in temperate or subtropical regions or that included a larger number of species, showed significantly more positive dispersal-range size relationships than studies including fewer dispersal-related traits, that were carried out in tropical regions (or multiple regions), or that included fewer species in the analysis (Fig. 3, Table S2, S3 Suppl. Mat.). Furthermore, studies carried out at higher taxonomic breadth, such as ‘phylum’ or ‘division’ showed fewer positive relationships between dispersal and range size than studies carried out for a particular ‘class’, ‘order’ or ‘family’.
When examining the relationship between dispersal and range size for animals and plants separately, we found that the factors responsible for a positive dispersal-range size relationship differed between those broad clades (Fig. 3, Table S1, S2 Suppl. Mat.). In animals, similar results as for the global dataset were found, with the number of dispersal-related traits and number of species positively contributing to the dispersal-range size relationship. In addition, we found that studies on endotherms showed fewer significant, positive associations between dispersal and range size than on ectotherms (Fig. 3, Table S1, S2 Suppl. Mat.). In plants, taxonomic breadth remained important as a variable to explain the positive dispersal-range size relationship, with studies carried out at ‘division’ level or higher showing fewer positive dispersal-range size relationships than studies carried out at the ‘family’ level (Fig. 3, Table S1, S2 Suppl. Mat.).
It should be noted that a large sample size may increase the probability of detecting false positives, which could explain our finding that number of species and number of dispersal-related traits affects the dispersal-range size relationship. Although, theoretically, this could go in both a positive or negative direction in terms of dispersal-range size relationships. The distribution of species numbers used in the studies is strongly right-skewed: more than 90% of the studied relationships included less than 500 species (380 relationships, Fig S2 Suppl. Mat.). We performed a sensitivity analysis to explore the effect of outliers for both the number of species and number of dispersal-related traits on our findings. Nevertheless, the positive effect of species number on the dispersal-range size relationship remains when including only studies with less than 500 species, but not when only including studies with less than 100 species (Table S3 Suppl. Mat.). Importantly, the positive effect of including multiple dispersal-related traits remained after excluding possible outliers (studies with 10 dispersal-related traits; Table S3 Suppl. Mat.). Alternatively, our result may reflect serendipity, because the more traits a study includes, the higher the chance becomes that one of those traits relates to range size. We tested whether this is the case by comparing the relationship between the overall dispersal-range size relationship and the number of dispersal-related traits vs. the relationship between the overall dispersal-range size relationship and the number of dispersal ‘factors’ (one factor can be composed of several dispersal-related traits) included in the models (Table S4 Suppl. Mat.). Interestingly, although the effect of the number of dispersal-related traits on the dispersal-range size relationship is significantly positive, the number of dispersal ‘factors’ included in each relationship does not have a significant effect (Table S4 Suppl. Mat.). This means that using multiple dispersal-related traits likely results in a better approximation of the dispersal process and/or a higher probability to capture traits that are relevant for dispersal, increasing the chance to find a positive dispersal-range size relationship.