Research overview: N-mixture models for unmarked populations

N-mixture models estimate abundance from repeated counts of unmarked animals at multiple sites by separating true abundance from imperfect detection, assuming counts arise from a binomial sampling of a latent site abundance drawn from a specified distribution. This overview describes how the approach applies to camera-trap count data, its appeal for species that cannot be individually identified, and the well-documented sensitivities that demand caution: estimates can be biased or unstable when assumptions about the abundance distribution, closure, or independence are violated, and detection and abundance can be difficult to separate. It surveys diagnostics, model checking, and debates over the method's reliability. For practitioners, the overview clarifies when N-mixture models are a reasonable choice and when alternative recognition-free estimators may be safer for inferring felid and prey abundance.