Herbivore Optimal Foraging: A Comparative Test of Three Models

Abstract

Three models of herbivore foraging are tested in this paper: contingency models (Owen-Smith and Novellie 1982), nutrient content weighted by food abundance (Stenseth and Hansson 1979), and linear programming (Belovsky 1978). The models are tested using data from two microtine populations, kudu, and moose. In all four cases, the linear programming model solved for nutrient maximization best predicted the diet composition by food classes (grass, forb, etc.) and energy intake. Foraging time and digestive capacity are the two most important constraints in the linear programming model. Contingency models do not appear to predict herbivore diets for several reasons. First, either they often do not contain constraints or the constraints are incorrectly applied. Second, simultaneous search is assumed which may not hold for plant foods. The results presented here contradict Owen-Smith and Novellie's (1982) claim that linear programming is a special case and the contingency model is a representation of herbivore diet choice. Finally, the model of food energy content weighted by abundance might not be expected to work a priori, because it is not an optimal foraging model. In addition, this model's diet predictions are dominated by food abundance. Furthermore, Stenseth's (1981) claim that linear programming is similar to this model is shown to be incorrect.