Food webs without the food: Interaction-neutral community models reveal biotic interactions

Background/Question/Methods . Community models fall along a spectrum of assumptions about species and their interactions. At one end, neutral biodiversity models do not require information about individual species’ biology. At the other end, food web models define species largely in terms of biotic interactions that may or may not be understood. Interaction-neutral models present an alternative community modelling approach that falls in between. Species are highly-specified by their unique vital rates (fecundity, mortality, growth rate, etc.) under a range of environmental conditions, and these vital rates are incorporated into individual matrix population models or other suitable modelling structures. A dependency assumption (finite space for recruitment, upper limits on aggregate biomass, etc.) links individual matrices together, and cross-species sensitivity analysis can then be used to discover pairwise biotic interactions. Importantly, these interactions arise only as a consequence of individual species’ vital rates expressed in the context of the environment, and do not depend on prior knowledge of biotic interactions.

Results/Conclusions . To explore the utility of interaction-neutral models, we developed a time-varying logistic r-K model for aquatic invertebrate populations and a stochastic stage-structured model for riparian vegetation. Both models incorporate disturbance dynamics in the form of flood and drought events that have strong effects on vital rates. Model analysis identified distinct threshold boundaries, where small changes in disturbance regime resulted in major changes to population dynamics. These thresholds were associated with traits conferring resistance and/or resilience to extreme flooding or drought events. Models recovered field-observed community patterns, suggesting that abiotic forcing by disturbance explains a significant amount of population dynamics. Cross-species sensitivity analysis identified species that had strong biotic effects on other community members (keystone species), as well as species with high bidirectional effects (mutual) or little influence on other species (passive). In many cases, the biotic interactions were highly dependent on the environmental state, suggesting that at least some biotic interactions might highly context-dependent. These results show promise for forecasting climate-induced population shifts in river ecosystems. In general, interaction-neutral models could be useful for modeling communities where among-species biotic interactions are poorly understood or highly dependent on environmental context.