Identifying individual cougars (Puma concolor) in remote camera images – implications for population estimates

Authors

Peter D. Alexander, Utah State UniversityFollow
Eric M. Gese, USDA National Wildlife Research CenterFollow

Document Type

Article

Date of this Version

2018

Citation

Alexander, P.D., and E.M. Gese. 2018. Identifying individual cougars (Puma concolor) in remote camera images — implications for population estimates. Wildlife Research 45(3):274-281. doi: 10.1071/WR17044

Comments

U.S. government work.

Abstract

Context. Several studies have estimated cougar (Puma concolor) abundance using remote camera trapping in conjunction with capture–mark–recapture (CMR) type analyses. However, this methodology (photo-CMR) requires that photo-captured individuals are individually recognisable (photo identification). Photo identification is generally achieved using naturally occurring marks (e.g. stripes or spots) that are unique to each individual. Cougars, however, are uniformly pelaged, and photo identification must be based on subtler attributes such as scars, ear nicks or body morphology. There is some debate as to whether these types of features are sufficient for photo-CMR, but there is little research directly evaluating its feasibility with cougars.

Aim. We aimed to examine researchers’ ability to reliably identify individual cougars in photographs taken from a camera-trapping survey, in order to evaluate the appropriateness of photo-CMR for estimating cougar abundance or CMR-derived parameters.

Methods. We collected cougar photo detections using a grid of 55 remote camera traps in north-west Wyoming, USA. The photo detections were distributed to professional biologists working in cougar research, who independently attempted to identify individuals in a pairwise matching process. We assessed the level to which their results agreed, using simple percentage agreement and Fleiss’s kappa. We also generated and compared spatially explicit capture–recapture (SECR) density estimates using their resultant detection histories.

Key results. There were no cases where participants were in full agreement on a cougar’s ID. Agreement in photo identification among participants was low (n = 7; simple agreement = 46.7%; Fleiss’s kappa = 0.183). The resultant SECR density estimates ranged from 0.7 to 13.5 cougars per 100 km² (n = 4; s.d. = 6.11).

Conclusion. We were unable to produce reliable estimates of cougar density using photo-CMR, due to our inability to accurately photo-tag detected individuals. Abundance estimators that do not require complete photo-tagging (i.e. mark–resight) were also infeasible, given the lack of agreement on any single cougar’s ID.

Implications. This research suggested that there are substantial problems with the application of photo-CMR to estimate the size of cougar populations. Although improvements in camera technology or field methods may resolve these issues, researchers attempting to use this method on cougars should be cautious.