Severe simulated herbivory constrains Sclerocarya birrea saplings regardless of resource availability.

Allocation of carbon to growth and defences in plants is affected by resource availability and herbivory, two determinants of woody plant abundance in savannas. Both factors are critical at the sapling stage of woody plants, yet their combined effects on sapling growth and defence have been poorly studied in savannas. Sclerocarya birrea is an iconic keystone species in southern African savannas, widely used by herbivores and humans. We hypothesised that low resource availability or high herbivory intensity would reduce growth and nutritional value for herbivores, but high resource availability or low herbivory intensity would increase growth and nutritional value. Support for our hypothesis would imply that any level of browsing intensity (relative to unbrowsed plants), regardless of resource availability, would potentially hold saplings in a browser trap because of reduced growth or improved nutritional value. We explored responses of S. birrea saplings to five levels of nutrient fertiliser and six levels of simulated browsing intensity. Low fertiliser application rates did not affect growth, but the highest fertiliser rate increased canopy growth. The most severe intensities of simulated browsing negatively affected growth and nutritional value. Our results suggested that either resource limitation throughout the growing season, or extremely intense browsing early in the growing season, is potentially able to retard long‐term growth of saplings and delay recruitment of adults. Although saplings that are severely browsed early in the growing season might not be re‐browsed later in the same season because of their reduced nutritional value, their weakened nitrogen status would be an additional limitation to regrowth. No significant interactions between resource availability and herbivory intensity were detected, which we suggest resulted from responses to resource availability and herbivory operating at different temporal scales. We recommend that future experiments of this type include carefully timed measurements to detect transient responses. [ABSTRACT FROM AUTHOR]