Herbivore data
The presence of animal species was estimated by camera traps located in
the same plots where the vegetation sampling was performed (one in each
plot, n = 60). Bushnell Essential E3 camera traps with low glow IR flash
were used to collect data on herbivores. The camera traps were set to
take three images in the interval
of three seconds once the camera was triggered, from which only one was
used for the analysis. The next
trigger was delayed by one minute after the last snapshot. When more
individuals of the same species occurred in one image, we considered
them a single record. In this paper, we use data from a total of 140
days from both dry (June – August 2018) and rainy (December – February
2019) seasons.
From this camera-based dataset, we extracted records of large herbivores
(elephants, equids, rhinoceroses, and bovids) that are reported to
influence grass species composition (Cumming 1982, du Toit 2003). We
considered both grazers and mixed feeders, i.e. species that feed on
grass (Table 1); for simplicity, we use the term ‘herbivore’ for both
groups. For each plot, we calculated (i) herbivore species
richness , expressed as the total number of herbivore species recorded
over the monitoring period, and herbivore species abundance,
corresponding to the total number of records of a given species. The sum
of records of all grazing species (Table 1) served as a proxy for (ii)herbivore abundance. Further, we also estimated the total
herbivore biomass per plot, calculated as species abundance × mean
species biomass summed up across all species. The species biomasses were
taken from Kingdon et al. (2013) and Kingdon and Hoffmann (2013a, b) and
calculated as a mean of the ranges given for male and female biomass. We
are aware that the term ‘abundance’ as we use it does not correspond to
the exact numerical abundance of species populations per plot, but we
suggest that it is an informative measure of the total herbivore feeding
pressure in a plot.