From Metawebs to Realised Food Webs: A Process-Based Framework for Network Representation

Tanya Strydom, Alexander M Dunhill, Jennifer A Dunne, Timothée Poisot , Andrew P Beckerman 

Ecological networks are widely used to understand biodiversity and predict ecosystem responses to environmental change, yet different food web representations encode different assumptions about how feeding interactions arise. We develop a process-based framework that links global metawebs of evolutionarily feasible interactions to regional metawebs constrained by spatiotemporal overlap, and ultimately to realised food webs shaped by local ecological conditions. Five processes govern these transitions: evolutionary compatibility and spatiotemporal co-occurrence constrain which interactions are possible and available, while species abundance, consumer diet choice, and non-trophic interactions influence which interactions are realised. We connect these processes to four approaches to network construction (empirical, feasibility-based, process-based, and structure-based) and derive testable hypotheses for evaluating when incorporating ecological processes improves predictions beyond those based on co-occurrence or aggregate network structure. By distinguishing the assumptions underlying each representation, the framework clarifies what different food webs can reveal, where their inferences are limited, and how these limitations can be tested using independent interaction records, trait and distribution data, and repeated observations across sites or time. Ultimately, it provides a basis for matching food web representations to ecological questions and for testing whether explicitly accounting for the processes that generate interactions improves our ability to anticipate ecological change.