Populations and communities are defined by the potential for their constituents to interact. Canonical models of consumer-resource dynamics assume that these interactions are independent and identically distributed across individuals—an assumption that has led to highly successful models, that also significantly misrepresent the way in which individual behavior affects population trajectories. How can this be? More specifically—How and when do behaviorally generated correlations in interaction rates among individuals scale up to significantly affect population dynamics?
We investigated this question by testing the influence of collective mobility patterns on the dynamics of simple consumer-resource cycles. We adapted an individual-based spatially explicit implementation of the Lotka-Volterra consumer-resource model to include simple flocking behavior in the predators, wherein nearby individuals tended to align their velocities. Absent collective behavior, the dynamics of this model are exactly described by the standard pair of differential equations. The effects of “switching on” collective behavior are then measurable in terms of the departure of the system from the predictions of the mean field model.
Collective behavior significantly altered the equilibrium densities and stability of the consumer-resource system. Roving flocks of predators created local, rather than global, depressions in prey density, lead to spatiotemporal refuges for the prey population. Dynamic fluctuations in the size and location of these refuges led to temporal fluctuations, and long-range autocorrelations, in prey recruitment, and in the rate at which predators consumed and converted prey biomass. Collective behavior thus altered the “observed” parameters of the consumer-resource system, and generated stochastic fluctuations in their values. The central prediction of this in silico experiment is that collective behavior, which occurs at the level of individuals, causes significant biases and fluctuations in apparent vital rates of interaction, when these are fitted to population-level abundance data. This is testable in field systems that yield concurrent time series on individual behavior and population abundance.