Isotopic Niche Breadths and Overlap
Isotopic niche breadths, measured as standard ellipse areas (SEAc ) in isotopic bi-space, varied widely between species, and between populations of the same species (Fig. 1; Table 1). While SEAc’s were generally smaller amongst species identified as grazers or browsers than intermediate-feeders, there were several grazer populations that had surprisingly broad niches, especially zebra (Equus spp.), warthog (Phacochoerus africanus ), buffalo (Syncerus caffer ), and one red hartebeest (Alcelaphus buselaphus ) population. Niche widths of some intermediate-feeders were relatively small, and, in at least one protected area, bimodally distributed (Fig. 1c).
Grazer and browser species were distinct along the δ13C axis, and most intermediate-feeders overlapped more with browsers than with grazers (Fig. 1). Nevertheless, some grazer populations overlapped strongly with intermediate-feeders, and visually there were also cases where grazer species differed from each other. The most notable example of this separation is seen in Fig. 1f, representing the one protected area where intermediate-feeders are absent, and where browsers occur in abundances too small for us to have obtained a sufficient sample. Displacement of grazer niches in this instance hints at a release effect in the absence of the competing intermediate-feeder guild. Overall, most of the variation in δ13C data was explained by species-level differences (Table 2). In the best-fit model (lowest AICc ), Locality did not have a significant effect, and inter-specific differences accounted for >75% of the total variance in the data. Accordingly, intra-population variation (residual error) was much smaller than inter-species variation. Even when analyzing grazers only, inter-species differences accounted for >50% of the total variance. Interspecific differences were less pronounced in δ15N data, with ‘Locality’ accounting for >40% of total variance, and the difference in magnitude of the intra- and inter-specific effects was not as obvious as in the case of δ13C. Patterns in δ15N data probably reflect local differences in climate, geology, or other environmental conditions (Schulze et al. 1998; Aranibar et al. 2008; Codron et al. 2013), but within any one locality there is simply less opportunity for herbivores to partition this niche.
Given the prevalence of interspecific differences, pairwise isotope niche overlaps were unsurprisingly low. The medianOij was 0.19, with an interquartile range from 0.03 to 0.41, and reached 0 in 20 of 166 pairwise comparisons. At the other end of the scale, Oij seldom approached 1 (> 0.95 in four instances), and values larger than 0.8 were observed in only 13 cases. Pairwise overlaps were similarly low when treating either δ13C or δ15N as individual niche axes (supplementary Table S5). The average overlap across populations was not systematically related to factors like protected area size, or number of species in the system, but was consistently lower than predicted by null models, even when only a single guild (grazers) was considered in the analysis (Fig. 2a). By contrast, standard deviations of Oij within any one protected area were larger than expected by random chance (Fig. 2b), with the exception of one assemblage (two when only grazers were considered), implying clustering of species in niche space that is consistent with heuristic concepts separating browsers, intermediate feeders and grazers into discrete guilds (supplementary Figs S3 and S4).