Introduced large herbivores have partly filled ecological gaps formed in the late Pleistocene, when many of the Earth’s megafauna were driven extinct. However, extant predators are generally considered incapable of exerting top-down influences on introduced megafauna, leading to unusually strong disturbance and herbivory relative to native herbivores.

We report on the first documented predation of juvenile feral donkeys (Equus africanus asinus) by cougars (Puma concolor) in the Mojave and Sonoran Deserts of North America. We then investigated how cougar predation corresponds with differences in feral donkey behavior and associated effects on desert wetlands.

Focusing on a feral donkey population in Death Valley National Park, we compared donkey activity patterns and impacts between wetlands with and without cougar predation.

Donkeys were primarily diurnal at wetlands with cougar predation, thereby avoiding cougars. However, donkeys were active throughout the day and night at sites without predation. Donkeys were ~87% less active (measured as hours of activity a day) at wetlands with predation (p<0.0001).  Sites with predation had reduced donkey disturbance and herbivory, including ~46% fewer access trails, 43% less trampled bare ground, and 192% more canopy cover (PERMANOVA, R² = 0.22, p=0.0003).

Our study is the first to reveal a trophic cascade involving cougars, feral equids, and vegetation. Cougar predation appears to rewire an ancient food web, with diverse implications for modern ecosystems. Our results suggest that protecting apex predators could have important implications for the ecological effects of introduced megafauna.

These data include camera trap detection and effort files, wetland vegetation data, and R scripts to analyze them. Camera trap data consists of raw detections, identified to species, with an ‘eventID’ indicating whether detections of a species at a site were within 30 minutes of each other. The effort file indicates whether a camera trap was operating on a given day and includes daily covariates (maximum daily temperature). 

All scripts required to run analyses are included in repository. All scripts were written in R.