Ecological consequences of body size: a model for patch choice in desert rodents

Summary Recent models exploring the ecological consequences of body size have assumed that its primary effect is to determine how easily individual prey of different sizes can be pursued or handled. However, for predators that eat small, particulate food, size-related costs associated with finding and harvesting prey should be at least as important as those associated with consuming individual prey once thay have been harvested. Such predators should have generalized diets, and body size differences would not be expected to influence substantially the sizes of prey eaten. The effect of body size on spatial patterns of foraging could, however, be substantial for these predators if prey have a patchy distribution.I develop a simple model for a particle feeder foraging in patchy environments and use it to examine the special case of patch choice by seed-eating desert rodents. The model implies that for most parameter values large and small animals should specialize to different extents on the most profitable patches. Size differences among coexisting desert rodents therefore can be expected to promote partitioning of food by differential patch choice. Preliminary observations of desert rodent seed dispersion and microhabitat preferences indicate that interspecific differences in patch choice do exist.The model predicts that the nature of the relationship between size and patch choice depends on the values taken by certain model parameters. Thus, although the model predicts that patch choice generally should vary with body size, the spatial scale of patchiness and the way in which within-patch harvest rates and between-patch travel velocities scale with size determine whether, and in what way, body size should affect patch choice. As yet estimates of these parameters for heteromyid rodents are not precise enough for us to have much confidence in specific model predictions about this system. However, it will only be a matter of time before we can derive better estimates; in principle the model is testable, and when suitably modified should be applicable to many systems.