ABSTRACT Dynamic heterogeneous metacommunities can be analyzed as complex networks. However, the interplay of species interactions, local environmental conditions, and spatiotemporal dispersal remains poorly understood. We assess the relative importance of these drivers in structuring river‐floodplain metacommunities based on a spatiotemporal eDNA dataset. We applied Bayesian network learning to infer species interactions, spatiotemporal dynamics, and the importance of environmental factors, and used graph theory to summarize and analyze the patterns in the derived networks. Our analysis revealed distinct sub‐communities linked by interacting species, spanning a gradient from environmentally filtered to dispersal‐driven. Top predators and invasive species are identified as key connectors, being most important in regulating network dynamics and cohesion. Our findings highlight that combining Bayesian networks with graph theory has high potential to uncover the causal structure of metacommunities and provide a mechanistic understanding of community assembly in dynamic ecosystems, informing ecosystem management in dynamic landscapes.
Andrea et al. (Thu,) studied this question.