In the new study, the team again collaborated with scientists at Yellowstone, who tracked animals from five herbivore species and collected fecal and plant samples. The species included pronghorn, bighorn sheep, mule deer, elk and bison. At Brown, researchers analyzed the samples using DNA metabarcoding, which helped to identify what plants the animals had consumed, and simple AI techniques to figure out how many different diet types exist within Yellowstone’s vast herds of wildlife and whether each species has its own unique diet type.

Study co-author Hannah Hoff, a Ph.D. candidate studying ecology, evolution and organismal biology at Brown, brought expertise in botany and data science to the project. Inspired by a seminar led by Brown University Professor of Biology and Data Science Sohini Ramachandran, Hoff incorporated machine learning in combination with genetic techniques to better understand herbivore behavior.

Dietary differences between species turned out to be smaller than previously assumed. Instead, researchers found that members of different species could have a lot of overlap in their diets and the amount of overlap depended on where and when they were feeding.

One of the team’s insights was that opportunity may drive foraging behavior more than animal species. In the summer, many animals from all species converged on a diet of nutritious wildflowers in the meadows of the summer range, while in the winter, a diet concentrated on coniferous trees and shrubs became more common. 

During the winter, bison in particular, but not exclusively, tended to keep looking for grasses and similar types of food even when they were frozen under snow, while some of the smaller herbivores, like mule deer and pronghorn antelope, tended to switch more dramatically toward a diet of evergreen trees. 

“It turns out the appropriate question is not, ‘Does that species eat grass?’” Kartzinel said, “but rather, ‘Is it eating grass right now?’” 

As a plant ecologist, Hoff took a plant-forward approach to understanding this ecological community.

“There’s sometimes a tendency to treat vegetation as a static ‘habitat type,’ instead of as a dynamic set of interacting species with their own individual ecologies,” Hoff said. “Centering our analysis of diet groupings on the plant species that distinguished them allowed us to examine how seasonality, nutrition and spatial distribution influence herbivore foraging — insights that may be obscured by broad diet classifications.” 

Kartzinel said the findings offer a lesson for scientists as well as iconoclastic animals.

“Imagine a herd of bison who are all supposed to be grazers, with one or two who want to eat like browsers,” Kartzinel said. “The traditional way that scientists would tell this story could be to dismiss the difference as aberrant or unimportant. But findings like this show us that dietary diversity actually is normal, and we should tell the story of the browsing bison, as well.”

This research was funded by the National Science Foundation (DEB-2046797, OIA-2033823) and the National Parks Service Cooperative Research and Trainings Program (P22AC00332-00, P23AC00378).