1.Herbivores frequently have to make tradeoffs between two basic needs: the need to acquire forage and the need to avoid predation. One manifestation of this tradeoff is the “landscape of fear” phenomenon – wherein herbivores avoid areas of high perceived predation risk even if forage is abundant or of high quality in those areas. Although this phenomenon is well-established among invertebrates, its applicability to terrestrial large herbivores remains debated, in part because experimental evidence is scarce. 2.This study was designed to (a) experimentally test the effects of tree density – a key landscape feature associated with predation risk for African ungulates – on herbivore habitat use, and (b) establish whether habitat use patterns could be explained by tradeoffs between foraging opportunities and predation risk-avoidance. 3.In a Kenyan savanna system, replicate plots dominated by the tree Acacia drepanolobium were cleared, thinned, or left intact. Ungulate responses were measured over four years, which included years of moderate rainfall as well as a severe drought. 4.Under average rainfall conditions, most herbivores (primarily plains zebra, Grant's gazelle, and hartebeest) favored sites with fewer trees and higher visibility – regardless of grass production – while elephants (too large to be vulnerable to predation) favored sites with many trees. During the drought, however, herbivores favored sites that had high grass biomass, but not high visibility. Thus, during the drought, herbivores sought areas where food was more abundant, despite probable higher risk of predation. 5.These results illustrate that the “landscape of fear”, and the associated interactions between top-down and bottom-up effects, is not static, but rather shifts markedly under different conditions. Climate thus has the potential to alter the strength and spatial dynamics of behaviorally-mediated cascades in large herbivore systems.

All_Data_Density_ExptExcel file containing data on herbivore use (dung counts, camera trap data), grass production, grass composition and nutrients, and visibility in a replicated manipulation of tree density.