Ecological interaction networks: prospects and pitfalls

Abstract

Interaction networks are a tool to visualize and to study the relationships between interacting species across and within trophic levels. Recent research uncovered many properties of such networks that remained undetected in previous food web studies. These patterns could be related to evolutionary and ecological processes. The study of interaction networks promises therefore progress in the study of constraints that act on the coevolution of interacting species and on food webs. However, there are still many pitfalls associated with the statistical analysis, the properties of the metrics involved and the appropriate null model choice. Here I review the mechanisms that shape interaction matrices, the possible internal structures and their ecological interpretation, and the analytical tools to identify matrix structure. Progress in the field needs critical meta-analytical and comparative studies that indentify the best suited null models (low type I and II error probabilities and high power to disentangle statistical from ecological processes) and clarify the interdependence of different concepts and metrics associated with network approaches. It is not improbable that many patterns recently associated with ecological and evolutionary processes might turn out to be simple side effects of the sampling from the underlying metacommunity distributions.