Closure violation in DNA-based mark-recapture estimation of grizzly bear populations

We use methods in the program MARK to explore the effects of closure violation when DNA-based mark-recapture methods are used to estimate grizzly bear (Ursus arctos) populations. Our approach involves the use of Pradel models in MARK to explore the relationship between recruitment, apparent survival rates, recapture rates, and distance between mean bear-capture locations and the edge of the sampling grid. If the population is demographically closed, it can be assumed that apparent survival estimates the fidelity of bears to the grid area and recruitment estimates rates of addition of bears to the grid area. A core bear population is defined from the Pradel analysis and is used to approximate the grid-based population size. The Huggins closed-population model in MARK is used to provide robust superpopulation estimates by explicitly modeling the relationship between capture probability and distance of bear-capture location from the grid edge. Data from a grizzly bear DNA-based mark-recapture inventory conducted in British Columbia is used to illustrate this method. The results of the Pradel analysis suggest that bears with mean capture locations within 10 km of the grid edge exhibit reduced fidelity rates and higher addition rates. Using the population of bears captured more than 10 km from the grid edge, a core-extrapolated estimate is derived, which is substantially lower than naive CAPTURE superpopulation estimates. The Huggins model superpopulation estimate displays superior precision compared with CAPTURE model estimates. Our results illustrate the danger of naive interpretation of closed-model estimates. This method allows further inferences to be made concerning the spatial causes of closure violation, and the degree of bias caused by closure violation to be explored.