Sympatric large mammalian herbivore species differ in diet composition, both by eating different parts of the same plant and by eating different plant species. Various theories proposed to explain these differences are not mutually exclusive, but are difficult to reconcile and confront with data. Moreover, whereas several of these ideas were originally developed with reference to within-plant partitioning (i.e., consumption of different tissues), they may analogously apply to partitioning of plant species; this possibility has received little attention.

Plant functional traits provide a novel window into herbivore diets and a means of testing multiple hypotheses in a unified framework. We used DNA metabarcoding to characterize the diets of 14 sympatric large-herbivore species in an African savanna and analyzed diet composition in light of 27 functional traits that we measured locally for 204 plant species.

Plant traits associated with the deep phylogenetic split between grasses and eudicots formed the primary axis of resource partitioning, affirming the generality and importance of the grazer-browser spectrum. A secondary axis comprised plant traits relevant to herbivore body size. Plant taxa in the diets of large-bodied species were lower on average in digestible energy and protein, taller on average (especially among grazers), and tended to be higher in tensile strength, zinc, stem-specific density, and potassium (and lower in sodium, stem dry matter content, and copper). These results are consistent with longstanding hypotheses linking body size with forage quality and plant height, yet they also suggest the existence of undiscovered links between herbivore body size and a set of rarely considered food-plant traits. We also tested the novel hypothesis that the leaf economic spectrum (LES), a major focus in plant ecology, is an axis of resource partitioning in large-herbivore assemblages; we found that the LES was a minor axis of individual variation within a few species, but had little effect on interspecific dietary differentiation.

Synthesis. These results identify key plant traits that underpin the partitioning of food-plant species in large-herbivore communities and suggest that accounting for multiple plant traits (and tradeoffs among them) will enable a deeper understanding of herbivore-plant interaction networks.

See methods in Potter et al. Journal of Ecology.

This dataset contains plant trait data (two versions) and herbivore diet data (two versions), as well as keys to interpreting plant and herbivore taxa (three files). A README file is included.

To generate the diet-weighted trait averages, matrix multiply the diet data by the plant trait data. For this to work, the columns of the left matrix (diet) must exactly match the rows of the right matrix (traits). Using the 'TOTU' versions of the diet and trait data accomplishes this. In R, the following code works: as.matrix(diet) %*% as.matrix(traits) . 

Please consult the README file for important details.