Graduation Year


Graduation Month


Document Type


Degree Name

Bachelor of Science

School or Department

Wildlife Biology


Wildland Restoration – Terrestrial

Faculty Mentor

Mark Hebblewhite

Faculty Mentor Department

Wildlife Biology


wolf, wolves, yellowstone, sightability, model, Canis lupus

Subject Categories

Animal Sciences | Ecology and Evolutionary Biology


Imperfect detection is ubiquitous among wildlife research and can affect research conclusions and management. Detection probability is often included in observation-based models. We leveraged research of gray wolves (Canis lupus) in northern Yellowstone National Park (YNP) to evaluate how the probability of sighting radio collared wolf packs from ground-based observation locations was affected by the characteristics of each spatial location (i.e., distance from the road, visibility (from a viewshed analysis), habitat openness, carcass presence, and wolf group size). We used two complementary approaches focusing on sightings during early (mid-November to mid-December) and late (March) winter periods between 1995 and 2017. First, we used 2,681 unique, daily observations of 17 wolf packs collected during 44 unique 30-day winter monitoring periods. We then compared these ground observations to the same number of random locations, each sampled from within wolf pack home ranges. Using this dataset, we used conditional logistic regression to estimate the probability of observing a group of wolves. Second, we used information on continuous observations of wolves collared with Global Positioning System (GPS) radio collars. We developed a similar probability of observing a group of wolves using logistic regression, but we compared GPS locations where wolves were observed from the ground crews to location where wolves were known to not be observed. We termed the first analysis a used-available model, and the second a used-unused model in accordance with the field of resource selection functions. Using the used-available model, we found that the probability of wolf sightings declined as wolves were farther from the road and increased when wolves were in open, visible areas and when wolves were in larger groups. These results were very similar to the used-unused model developed with only GPS-collared wolf locations. The top model included the same covariates, which each had the same directional effect on the probability of seeing wolves. We used our results to build spatial predictions for seeing wolves in YNP. These predictions are useful to managers for identifying “hot-spots” of wolf observations and can be incorporated into research related to wolf ecology and predator-prey dynamics that relies on observations of wolves.

Honors College Research Project




© Copyright 2018 Jeremy SunderRaj