Study Area and Dataset
The study was carried out in nine protected areas in the grassland biome in the summer rainfall region of central South Africa (supplementary Fig. S1). This is an open, mostly treeless landscape, spread over a generally flat topography (masl ~ 1 300m), although three of the protected areas sampled are located in the eastern escarpment (masl ~ 2 000m). Mean annual rainfall varies between 300 and 700 mm per year, increasing from west to east. The protected areas studied are fairly small (between 2 200 and 22 000 ha), exclude large carnivores, and dispersal is limited to infrequent translocation events, although animals are harvested in some years. Resident herbivore populations live close to equilibrium, with average population growth rates approaching 0 even after accounting for harvesting events (Table 1), as is typical for many wildlife populations (Fryxell et al. 2014).
Stable isotope data are from herbivore faeces collected during excursions made over two years (2014 and 2015), during the dry season months (May-October) when resources were more limiting and we expected more pronounced patterns of niche partitioning. We collected faeces by locating animals from within a vehicle and then following on foot, which ensured that specimens were fresh and hence uncontaminated by soil and insects, and had correct species designations. We aimed to collect faeces from a minimum of 20 individuals per population (i.e. per species per protected area), but ultimately included all populations for whichn ≥ 10. For simplicity, we treated both species of wildebeest (Connochaetes taurinus and C. gnou ), and both species of zebra (Equus quagga and E. zebra zebra ), as conspecifics, because only one of each species was present in any one protected area. The final dataset comprised 12 taxa, ranging in body size from ~40 to 600 kg, and included eight grazers, two browsers, and two intermediate-feeders.
Faecal specimens were stored in brown paper bags, and air-dried indoors to constant weight. Faeces were then ground in a coffee grinder, and weighed separately in tin capsules. Specimens were combusted individually in a Flash2000 elemental analyser and the resultant CO2and N2gases introduced to a Delta V Plus isotope ratio mass spectrometer via a Conflo IV gas control unit (all instruments by Thermo Scientific, Bremen, Germany).13C/12C and15N/14N ratios are presented in standard delta (δ) notation relative to Vienna PeeDee Belemnite and atmospheric N2 standards, respectively. Standard deviations of repeated measurements of laboratory standards were less than 0.2 ‰. Faecal δ13C data for most the populations included here have previously been reported elsewhere (Malindie et al. 2020). All data are included in supplementary Tables S1 and S2. The analyses that follow were done in R v 3.5.2 (R_Core_Team 2015).