Recent Models for Mark-Recapture and Mark-Resighting Data

Abstract

A topic of current interest in the analysis of capture-recapture data is the development of models that allow biologically meaningful constraints on survival and capture rates. Of particular interest are models in which survival is a function of environmental variables. Thus, in an important recent paper, Clobert and Lebreton (1985) described analysis of data from resighting of marked birds, where survival was modelled as a function of winter temperature. Clobert and Lebreton (1985) have, however, developed their methodology (including necessary computer routines) in the context of data from resighting samples only (thus assuming no losses on capture), whereas with appropriate interpretation of formulae their methods can be applied to recapture data that frequently include losses on capture. The purpose of this note is to indicate the modifications needed to make the Clobert-Lebreton methods applicable to recapture data. Also, to prevent future duplication of effort, relationships between models that have been developed separately for resighting data and recapture data are noted. Procedures in Clobert and Lebreton (1985) and Clobert et al. (1985) are based on the model for estimating time-specific survival and sighting rates from resighting data, proposed by Cormack (1964). The year after Cormack's model appeared, Jolly (1965) and Seber (1965) described what is now called the Jolly-Seber model for estimating survival rates (and population size) from recapture data, allowing for the possibility that physical recapture may result in loss of some animals. With a view to achieving parsimony and improving estimator precision, parallel efforts (including development of necessary computer routines) have since specialized the original Cormack and Jolly-Seber models to cases where survival and/or recapture (resighting) rates are assumed constant. Thus, Sandland and Kirkwood (1981) and Clobert et al. (1985) describe reduced-parameter versions of Cormack's (1964) model, and Jolly (1982) and Brownie, Hines, and Nichols (1986) describe analogous reduced-parameter versions of the Jolly-Seber model. Of the corresponding computer programs, the Clobert-Lebreton program (Clobert et al., 1985; Clobert and Lebreton, 1985) permits the greatest flexibility for modelling survival and capture rates. [For other somewhat parallel efforts, not immediately relevant to the discussion here, see e.g., Cormack (1981) and Crosbie and Manly (1985).] The Cormack and Jolly-Seber models differ in that new releases are viewed as uninformative constants in the former, and as random variables (providing information about population size) in the latter. The models yield equivalent formulae for obtaining estimates of survival and recapture rates, but lead to different interpretations of expectations in variance formulae (Seber, 1982; Brownie et al., 1986). When there are no losses on capture, the Cormack and Jolly-Seber models in fact produce identical point and interval estimates for survival and capture rates (Seber, 1982, pp. 214-215; Brownie and Robson, 1983). Reduced-parameter versions of the Cormack and Jolly-Seber models, which make the same assumptions about survival and capture rates, should yield numerically identical