Historically, food webs have been the primary tool for understanding and comparing structure, energy transfer and species interactions in ecosystems. However, food webs depict only predator-prey interactions, whereas species can interact in many different ways. Here, we use an empirical dynamic approach to construct mutual influence webs for five ecosystems and evaluate the importance of predator-prey interactions versus indirect interactions. Unlike other frameworks exploring species interactions, this approach accounts for the dynamic nature of interactions. Here we use Convergent Cross Mapping (CCM), a test for causation between two dynamic variables acting within the same dynamic system, to explore interactions between abundance time series of taxa in five distinct ecosystems: two marine and one freshwater planktonic system, a kelp forest system and a desert system. We use this approach to identify the strength of all causal predator-prey and indirect interactions and to construct mutual influence webs for each system. By analyzing and comparing these interaction webs, we quantitatively test if predator-prey interactions are indeed the most important interactions for shaping the population dynamics of taxa.
Results/Conclusions
We find that predator-prey interactions account for 43% of the strongest interactions in our systems and their relative importance can change depending on both the type and variability of environmental forces present. To demonstrate these secondary hypotheses, we used a 5-species coupled food chain model with varying degrees of stochasticity (ep), observational error (eo), and a combination of both (ep and eo). When observational noise was added to the model, interaction strengths decreased regardless of interaction type. Simply put, as sampling becomes more inaccurate, the ability to make predictions between time series decreases and the apparent strength of interactions also decreases. When stochasticity was added to the model, the strength of predator-prey and trophic cascade interactions remained relatively unchanged while competition and higher order indirect interaction weakened. We also found that our mutual influence webs were highly connected with an average connectance across systems of C = 0.54. This suggest that high connectivity is not a distinct trait of a select few systems, but rather a common trait, shared across diverse ecosystems.