Plant community structure has been studied using a diversity of species and habitats for many decades. Major questions about plant communities have been addressed in these individual experiments, including understanding the prevalence of competition and facilitation, the characteristic patterns of interactions (e.g. asymmetric vs. symmetric interactions, hierarchies vs. deviations from hierarchies), and the distribution and density of interaction intensities. I have systematically reviewed the literature to identify plant interaction experiments that measure performance in pairwise species combinations. I have represented these plant interaction experiments as networks, in which species pairs are joined by competitive and facilitative interactions. While interactions in plant communities have not yet been described as networks, network theory has been applied to many other ecological communities. Networks allow for the calculation of metrics that make it possible to compare community structure quantitatively. I have performed a Bayesian random effects meta-analysis of network metrics relevant to plant community structure, including asymmetry, transitivity, strength distribution, and connectance, calculated from 31 plant interaction experiments from the literature and from my own large-scale field study. A meta-analysis of network metrics is a novel approach, and is a useful method to incorporate the underlying variation from multiple studies.
Results/Conclusions
Competition is the prevalent force in plant communities (mean interaction strength: -0.11; 95% credible interval: -0.17, -0.05), although some networks are facilitative on average. Species interactions are less asymmetric than expected (mean asymmetry: -0.20; 95% CI: -0.35, -0.03), given the predicted positive feedback from initial differences in resource capture ability between plant species. Intransitivity, or deviation from a strict hierarchy, is a feature of most plant interaction networks. This intransitivity could promote stability in plant communities, as a strictly hierarchical community would eventually be overtaken by the competitive dominant. The distribution of interaction strengths in plant communities is exponential, similar to what has been found in other ecological networks and indicative of an unequal distribution of centrality among species in the community, in which only a few species are highly important. Plant interaction networks are more densely connected than other ecological networks (mean connectance: 0.47; 95% CI: 0.40, 0.53). Representing plant communities as networks of interactions has provided quantitative evidence in support of an alternative view of plant communities that include facilitative as well as competitive interactions, that are not strictly hierarchical, and that have a higher density of interactions relative to other ecological systems.