Is there a 'browse trap'? Dynamics of herbivore impacts on trees and grasses in an African savanna

Despite widespread acknowledgement that large mammal herbivory can strongly affect vegetation structure in savanna, we still lack a theoretical and practical understanding of savanna dynamics in response to herbivory. Like fire, browsing may impose height-structured recruitment limitations on trees (i.e. a 'browse trap'), but the demographics of herbivore effects have rarely been considered explicitly. Evidence that cohorts of trees in savannas may establish during herbivore population crashes and persist long term in savanna landscapes is anecdotal. Here, we use an experimental approach in Hluhluwe iMfolozi Park in South Africa, examining the response of grass biomass and tree populations to 10 years of graduated herbivore exclusion, and their subsequent response when exclosures were removed. We found that grazer exclusion increased grass biomass and that, despite presumable increases in fire intensity and grass competition, herbivore - especially mesoherbivore, including impala and nyala - exclusion resulted in increases in tree size. After herbivore reintroduction, grazers reduced grass biomass over short time-scales, but tree release from browsing persisted, regardless of tree size. Synthesis. This work provides the first experimental evidence that release from browsing trumps grazer-grass-fire interactions to result in increases in tree size that persist even after browser reintroduction. Escape from the 'browse trap' may be incremental and not strictly episodic, but, over longer time-scales, reductions in browsing pressure may lead to tree establishment events in savanna that persist even during periods of intense browsing. Explicitly considering the temporal demographic effects of browsing will be the key for a much-needed evaluation of the potential global extent of herbivore impacts in savanna. This work provides the first experimental evidence that release from browsing trumps grazer-grass-fire interactions to result in increases in tree size that persist even after browser reintroduction. Escape from the 'browse trap' may be incremental and not strictly episodic, but, over longer time-scales, reductions in browsing pressure may lead to tree establishment events in savanna that persist even during periods of intense browsing. Explicitly considering the temporal demographic effects of browsing will be the key for a much-needed evaluation of the potential global extent of herbivore impacts in savanna.