Estimation of Demographic Parameters from Live-Encounter Data: a Summary Review

Abstract

Estimation of demographic parameters is central to research questions in wildlife management, conservation, and evolutionary ecology. I review the 7 major classes of mark–recapture models that investigators can use to estimate apparent survival and other parameters from live-encounter data. Return rates are the product of 4 probabilities: true survival (S), site fidelity (F), site propensity (δ), and true detection (p*). Cormack-Jolly-Seber (CJS) models improve upon return rates by separating apparent survival (φ = S × F) from the probability of encounter (p = δ × p*). The main drawback to mark–recapture models based on live-encounter data is that the complement of apparent survival (1 − φ) includes losses to mortality and to permanent emigration, and these 2 ecological processes are difficult to disentangle. Advanced mark–recapture models require additional sampling effort but estimate apparent survival with greater precision and less bias, and they also offer estimates of other useful demographic parameters. Time-since-marking or transient models control for individuals not encountered after the occasion they are first marked, a common feature of wildlife populations. Temporal symmetry models combine forward- and reverse-time modeling to estimate recruitment (f) and the finite rate of population change (λ). Multi-strata models include dynamic categorical information and offer state-specific estimates of apparent survival and encounter rates, as well as probabilities of changing states (Ψ). Robust design models subdivide sampling occasions into shorter periods, and they partition encounter rates (p) into estimates of temporary emigration (γ = 1 − δ) and true detection (p*). Joint models combine live encounters with other sources of information, including dead-recovery data, and decompose apparent survival into estimates of true survival (S) and site fidelity (F). Cormack-Jolly-Seber and multi-strata models have a large literature, but many of the advanced models have not yet received widespread use. In the future, wildlife ecologists should design field studies that take advantage of the best possible statistical procedures now that a range of alternative models and software tools are available.