Accounting for imperfect detection in observational studies: modeling wolf sightability in Yellowstone National Park

Abstract

Abstract Imperfect detection is ubiquitous among wildlife research and is therefore commonly included in abundance estimation. Yet, the factors that affect observation success are largely unknown for rare and elusive species, such as large carnivores. Here, we took advantage of intensive ground‐based monitoring effort and an extensive GPS data set (2000–2018) and developed a winter sightability model for gray wolves (Canis lupus) in northern Yellowstone National Park, Wyoming, USA. Our resulting sightability model indicated that observation success was positively affected by the topographic nature of where wolves were in relation to observer locations (viewshed), areas being less forested (openness), and wolf group size, and negatively affected by distance from observer locations. Of these, viewshed had the strongest effect on the probability of observing a wolf, with the odds of observing a wolf being four times more likely when wolves were in the predicted viewshed. Openness was the next most influential covariate, and group size was the least influential. We also tested whether a wolf being harvested from a pack when they were outside of Yellowstone National Park had an effect on wolf sightability. We did not, however, find support for human‐induced mortality affecting wolf sightability inside of Yellowstone National Park. Our results indicate that the ability to observe wolves was greatly affected by ecological and landscape‐level factors, a finding that is likely to generally extend to other large carnivores. As such, our sightability model highlights the importance of considering landscape structure and variation in large carnivore use of the landscape when conducting observational‐based studies.

Keywords

• Canis lupus
• harvest
• resource selection probability function
• wildlife observation
• wolves