Biogeographers strive to disentangle the relative contribution of the different processes behind species distribution. Although methods to predict species distribution have been developed for over two decades, there is still an ongoing debate concerning their real efficiency. Many assumptions upon which these approaches are built remain poorly tested. A common assumption is that biotic interactions are unimportant beyond the scale of a local community. On the other hand, Diamond hypothesized more than 40 years ago that species interacting by competition should avoid each other in space, leading to a 'checkerboard' distribution. This hypothesis was contested at several times and it led to the developmment of null model analysis. Despite the huge amount of literature studies investigating co-distribution at the community level, the hypothesis that ecological interactions such as competition, predation and mutulism, impact co-distribution has never been tested explicitly. Here we directly address the question: do ecological interaction impact the geographic distribution of species?
Results/Conclusions
We have developed a general theory leading us to few simple and intuitive predictions regarding the relationship between species co-distribution and ecological interactions. We have shown that: i) interactions impact pairwise co-occurrence; ii) the strength of an association decreases with the length of the shortest path between two species; iii) the strength of an association decreases with the number of interactions a species is experiencing. We have perform analysis of co-distribution of four datasets for which we had a priori knowledge of ecological interactions: leaf gallers and their parasitoids inhabiting the leaves of European salix, trees of Eastern North America, the microbes inhabiting the leaves of the purple pitcher plant and hummingbirds pollinating flowers of Caribbean islands. While we find significant spatial associations between pairs of interacting species, in most cases it tends to disappear once we remove the effect of shared environmental requirements. We discuss our results in the light of recent methodological improvements of species distribution models.