Increasing drought frequency is consistently predicted to be a major feature of climate change in the tropics and subtropics. In savannas, rainfall variability most obviously translates into variation in grass productivity, with implications for grazers populations, including possible grazer population crashes. However, interactions between drought, grass productivity, and grazing are poorly studied, especially in the tropics. To compound this, most work has focused at experimental scales (from meters to even hectares), but has ignored how these drought effects play out in the landscape context in interaction with grazing and especially grazer movement. Here, we ask how rainfall variability, including drought, drives variability in the grass layer in a large conservation area in South Africa savanna (Kruger National Park), in interaction with population dynamics and movement of large grazer populations. We use grass biomass data from 533 Veld Condition Assessment (VCA) sites established in 1989 along a rainfall gradient (~ 300-750 mm mean annual rainfall) on contrasting soils, combined with herbivore census data from fixed-wing aerial population counts.
Results/Conclusions
We find that grass biomass was highly responsive to climate variability, especially rainfall, and that recovery from extreme events occurred on the time-scale of years, not decades. Notably, grass declines during drought were especially severe – even more severe than previously documented elsewhere, perhaps because grazing dramatically reduced in grass biomass during drought, even in parts of the landscape that experienced less severe drought. Large-scale animal movement to these drought refugia may have spatially extended the effects of drought and mitigated grazer mortality, preventing drought-driven population crashes. Landscape processes, including animal movement, appear to be critical in determining savanna drought responses, suggesting that evaluations of drought impact must expand to include this larger scale of analysis. That experimentation may not be possible at this larger scale should not prevent more comprehensive observational analysis.