Environmental context shapes the relationship between grass consumption and body size in African herbivore communities

Though herbivore grass dependence has been shown to increase with body size across herbivore species, it is unclear whether this relationship holds at the community level. Here we evaluate whether grass consumption scales positively with body size within African large mammalian herbivore communities and how this relationship varies with environmental context. We used stable carbon isotope and community occurrence data to investigate how grass dependence scales with body size within 23 savanna herbivore communities throughout eastern and central Africa. We found that dietary grass fraction increased with body size for the majority of herbivore communities considered, especially when complete community data were available. However, the slope of this relationship varied, and rainfall seasonality and elephant presence were key drivers of the variation—grass dependence increased less strongly with body size where rainfall was more seasonal and where elephants were present. We found also that the dependence of the herbivore community as a whole on grass peaked at intermediate woody cover. Intraspecific diet variation contributed to these community-level patterns: common hippopotamus (Hippopotamus amphibius) and giraffe (Giraffa camelopardalis) ate less grass where rainfall was more seasonal, whereas Cape buffalo (Syncerus caffer) and savanna elephant (Loxodonta africana) grass consumption were parabolically related to woody cover. Our results indicate that general rules appear to govern herbivore community assembly, though some aspects of herbivore foraging behavior depend upon local environmental context.

Methods
We quantified local grass dependence from a dataset of stable isotope data collected from across eastern and central Africa, Cerling et al. (2015). We identified which species were missing local data using a dataset of herbivore community composition, Rowan et al. (2020). We then computed regional species-level averages for those community members missing local diet data, also from the aforementioned stable isotope dataset. 

We extracted species-level body mass estimates from EltonTraits (Wilman et al. 2014).

We downloaded shapefiles for each community from the World Database on Protected Areas (WDPA). Using these shapefiles, we extracted environmental covariates for each herbivore community: we calculated mean annual rainfall, mean rainfall seasonality, mean annual temperature, and mean temperature seasonality for each community from climatic data layers downloaded from WorldClim 2.0 (Fick and Hijmans 2017); we extracted estimates of woody cover for each community from a data layer of fractional woody cover for sub-Saharan Africa (Venter et al. 2018).