Mary V. Price and Rachel A. Correll. 2000. Depletion of seed patches by Merriam's kangaroo rats: Are GUD assumptions met? Ecological Society of America Annual Meeting, Snowbird, UT.

Abstract. Because the foraging behavior of granivores is often difficult to observe directly, it has become popular to study their foraging ecology indirectly by observing the extent to which experimental seed patches are depleted. An extension of the Marginal Value Theorem suggests that the density of seeds an optimal forager leaves behind (the Giving Up Density, or GUD) can provide a quantitative assay of its perception of fitness costs and benefits associated with the patch, among other things. Interpreting GUDs as more than qualitative indications of overall foraging activity or patch preference rests, however, on a set of assumptions that have rarely been tested. We used direct observation of foraging kangaroo rats to test two assumptions: That the curve relating cumulative harvest to time spent within a patch (Gain Curve) is smoothly decelerating, and that animals leave seed patches when harvest rates have fallen to a threshold level. Gain curves were characterized in the laboratory for 6 experienced individual Dipodomys merriami, using seed trays similar in size and design to ones used in many previous GUD studies. Polynomial regression indicated that all gain curves were linear until 60-80% of seeds had been harvested (ca. 200 s of foraging time), and declereated only thereafter. Animals searched more systematically than random expectation. In the field, single individuals depleted seed trays in multiple visits and harvested less during each successive visit. Amounts removed during the first visit (66%; 1.98g )were smaller than maximum cheek pouch capacit yand corresponded to depletion levels achieved in the laboratory during the linear portion of the gain curve. These results indicate that kangaroo rats deplete patches and use patch-leaving rules than are not as simple as those assumed by the GUD model. This suggests caution in using the GU approach before its assumptions are tested.