Mark-resight population estimation with imperfect observations

Minta and Mangel (1989) developed a Monte Carlo method for estimating population size from mark-resight data in which there is considerable variability in the resighting frequencies among individuals and irregular census surveys. Their method assumed that sightings were perfect, so that all animals present (marked or unmarked) were resighted. We describe an updated version that allows a proportion 1 - p of the individuals present to be overlooked during censusing. We compare confidence interval coverage for the original method and the updated method using simulated data sets, describe a method to estimate p, and evaluate how error in estimating p affects confidence interval coverage. The updated method for estimating a confidence interval performs considerably better than the original estimate. We find that p can be overestimated by 30% or underestimated by 10% and the confidence interval generated still includes the true population size 90% of the time. This technique may improve confidence interval estimates for small and threatened populations.