Using network models to reveal underlying processes leading to the organization of individual variation in resource use
Intrapopulational variation is a common pattern found in nature. It can be caused by many factors such as differences related to sex, morphology, environment and genetics. As a consequence, individuals within a population can exploit resources differently. Individual-resource networks represent the feeding interaction between individuals and the resources they consume and are used to explore distinct ways in which individuals vary in resource use. Our goal is to generate quantitative predictions about the structure of individual-resource networks, connecting the observed structural patterns with possible mechanisms leading to variation in resource use within populations. We investigated three models describing individuals exploiting resources: individuals sharing the same rank preference for prey, but differing in their willingness to accept new resources; individuals sharing the same top-ranked resource and relying upon different alternative resources; or individuals having distinct preferences on their rank sequence. We studied the network structure that would arise under each model using individual-resource networks from two populations of the snail Nucella sp., two populations of the snail Thais sp., and one population of the Cocos finch Pinaroloxias inornata. We compare each models’ performance in reproducing features of empirical networks, such as nestedness and modularity, and evaluated the models’ fit.
The models investigated generate different network topologies. For instance, a nested pattern of interaction, in which the preys consumed by specialist individuals are a subset of the preys consumed by generalists, is found when individuals share the same rank preference. On the other hand, a modular pattern, in which few individuals form groups with similar diets, is found when individuals differ in their alternative prey choices and when they differ in their rank sequence. The model that presented the worst performance was the one that predicts shared preferences among individuals. For P. inornata and the two populations of Thais sp., the model with best performance was the one that predicts the same top-ranked prey and different alternative prey choices. For the two populations of Nucella sp. the model with best performance was the one that predicts different rank preference for preys. Our approach highlights the importance of generating quantitative predictions in order to accurately define and differentiate possible mechanisms leading to variation in resource use.