EVALUATION OF MARK-RESIGHT MODEL ASSUMPTIONS FOR ESTIMATING MOUNTAIN SHEEP NUMBERS

Mark-resight methods have been used to estimate numbers of mountain sh eep (Ovis canadensis canadensis), but no evaluation has been made of p otential bias or precision of estimates. We simulated the mark-resight procedure, including simulations with violations of several assumptio ns, using the joint maximum likelihood estimator from a hypergeometric distribution (JHE). We also conducted a field study with a mountain s heep herd near Trickle Mountain, Colorado. These field data, particula rly estimates of heterogeneity of individual sighting probabilities an d group sizes, were used to provide realistic simulation distributions . Aggregation of animals did not bias the estimate but did decrease pr ecision and confidence interval (CI) coverage. The potential for overe stimation increased with aggregation. From the Trickle Mountain field study, mean group size was 6.3 (SD = 6.0), and animals that were seen together spent about 19.2% of the time together. Violating the assumpt ion of equal sighting probabilities among individuals may have biased the estimates, especially with smaller populations (e.g., 50), but the bias was not large (approx 8% max.). Heterogeneity of individual sigh ting probabilities lowered coverage and decreased precision. Individua l sighting probabilities of the Trickle Mountain herd ranged from 0.33 to 0.86, with a mean of 0.58 (SD = 0.2). A modified equation was used to account for immigration/emigration in the population. The Minta-Ma ngel estimator, which could be used when heterogeneity of individual s ighting probabilities exists, did not perform as well as JHE. We recom mend mark-resight and the JHE as a population estimation method for mo untain sheep where demographic closure exists within a well-defined st udy area.