Some ecological interactions are easy to observe and identify: a seal predating on a fish is antagonistic and a clownfish protecting its anemone home is mutualistic. Yet, many interactions are more obscure. This can be because they are hard to observe (e.g. the multitude of symbiotic bacteria surrounding the roots of plants) or because the nature of the relationship is complicated (e.g. an insect predator which consumes both the herbivores of a plant and the plant itself).
As it is in general much simpler to detect the presence of an interaction than to identify the type of interaction observed, it would be a useful addition to ecological methodology to be able to use the former to infer the latter. We utilize the group model (Allesina & Pascual 2009) to divide an ecological network into similarly interacting groups of species. These groups can be modules -- species which coexist in space or time leading to a high density of interactions between group members when compared to the density of their interactions with species outside of the group -- or trophically defined groups for which there is a high density of interactions between two distinct sets of species, e.g. herbivores and plants.
Results/Conclusions
We find that the groups identified by the model resemble the unique roles species fill in the ecosystem. That is, groups based solely upon the structure of the binary or weighted network of interactions have statistically non-random clustering of species based on the types of interactions each species participates in. For instance, in food-webs, the group model is able to distinguish parasites from other species, tending to assign parasites and non-parasites to different groups. Similarly, in a highly-structured, tripartite networks, such as a system of plants, herbivores, and parasitoids of the herbivores, the model is able to a priori classify species roles.
Simplifying ecological networks to the composite groups identified by the model (and analyzing the average interaction strengths between groups), we are further able to categorize which aspects of network structure are important for defining a species role. This has important implications for identifying the role of unclassified species and their interactions, which can lead to more informed management policies and conservation efforts.