2.4. Multidimensional nutritional niche

We evaluated changes in the seasonal diet composition of bison using nutritional geometry, a multidimensional method of assessing an animal's dietary nutrition in the context of ecological niche theory (Machovsky‐Capuska et al., 2016). We assessed the bison's dietary, macronutrient, and fiber niches. These niche estimates quantified the macronutrient compositions of plants seasonally foraged by bison thus accounting for the limitations of availability (Coogan et al., 2018). We plotted diet, macronutrient, and fiber content on right‐angled mixture triangles (RMTs); a three‐dimensional simplex, that uses the implicit z‐axis to geometrically display the space (i.e., niche) of three components of an animal's diet (Raubenheimer, 2011). For diet composition RMTs, we used percent content of browse (x‐axis), forbs (y‐axis), and graminoids (z‐axis) in the diet (Spitzer et al., 2020). In these RMTs, niches closer to the origin represent grazing behavior and niches at the z‐axis represent browsing. We used percent metabolizable energy for carbohydrates (x‐axis), lipids (y‐axis), and protein (z‐axis) to create macronutrient RMTs (Machovsky‐Capuska et al., 2016). For fiber RMTs, we used percent content of lignin (x‐axis), hemicellulose (y‐axis), and cellulose (z‐axis) (Aryal et al., 2015). To determine if changes in macronutrient composition were significant between seasons, we calculated the mean percent metabolizable energy of all plants consumed within each season and generated a 95% confidence ellipse around the mean (Monnette, 1990). If the 95% confidence ellipse from one season encapsulated the mean of another season, then those two seasons did not significantly differ (Fox, 2016; Monnette, 1990). We calculated confidence intervals around the means for macronutrient, fiber, and diet components to represent the nutritional components within each plant each season. We then calculated weighted means (using the RRA as the weighting factor) to represent how the components were consumed (i.e., realized niches). Finally, we calculated niche breadth and overlap for seasonal realized diet (at the taxonomic unit and forage group levels), macronutrient, and fiber niches. Niche breadths were calculated as the diversity of taxonomic units, macronutrient concentrations, and fiber concentrations while accounting for the relatedness of taxonomic units using R package indicspecies (De Cáceres et al., 2011; R Core Team, 2017) and compared them using Mann–Whitney–Wilcoxon tests (Ramsey & Schafer, 2002). We assessed differences in individual macronutrients and fiber components in bison diets using one‐way ANOVAs. Then, we used post hoc Tukey's HSD tests for the three seasons and four functional forage groups (Ramsey & Schafer, 2002).