Do high recapture rates indicate representative sampling? The relationship between recapture probability, risk-taking, and personality

ABSTRACT Context. Monitoring programs provide valuable information on wildlife populations, thereby underpinning strategies for conservation and control. For threatened species, where every animal represents a substantial portion of the population, representative sampling is vital. One fundamental challenge during sampling is understanding drivers of survey bias; for instance, behavioural heterogeneity in trap response. Methods such as capture–mark–recapture have long been used to estimate capture and recapture heterogeneity; yet, this method, like many others, is able to gather data only from the trappable and re-trappable portion of the population; a problem that presents a particular challenge for small or vulnerable populations. A greater understanding of why biases arise can result in improved survey methods, more reliable survey data and increased modelling accuracy. Aims. We focus on an endangered species with unusually high recapture probabilities (0.78–0.92), namely, the mountain pygmy-possum (Burramys parvus). Specifically, we examine whether, within a single trapping session, a recapture bias exists either as a function of past trapping experience or personality. Methods. We tested whether recapture probability differs among cohorts with different capture histories (‘known’ animals captured during trapping sessions in previous years vs ‘new’ animals trapped for the first time in this study). We also tested for individual personality, general risk-taking behaviour during foraging, and subsequent links to recapture probability. Key results. Recapture probability was significantly affected by cohort. New animals had lower probabilities of recapture and took fewer risks during foraging than did known animals. Although personality did not significantly influence recapture probability, it did influence risk-taking during foraging. Conclusions. Despite high recapture probability within the populations, captures were significantly skewed towards a subset of the population, likely being due to different perceptions of risk among individuals. Implications. Understanding potential sources of bias during live-capture surveys is the initial step towards modifying and improving surveys to reduce sampling biases and to ensure representative population sampling.