Statistical inference for capture-recapture experiments

Abstract

This monograph presents a detailed, practical exposition on the design, analysis, and interpretation of capture-recapture studies. The Lincoln-Petersen model (Chapter 2) and the closed population models (Chapter 3) are presented only briefly because these models have been covered in detail elsewhere. The Jolly- Seber open population model, which is central to the monograph, is covered in detail in Chapter 4. In Chapter 5 we consider the "enumeration" or "calendar of captures" approach, which is widely used by mammalogists and other vertebrate ecologists. We strongly recommend that it be abandoned in favor of analyses based on the Jolly-Seber model. We consider 2 restricted versions of the Jolly-Seber model. We believe the first of these, which allows losses (mortality or emigration) but not additions (births or immigration), is likely to be useful in practice. Another series of restrictive models requires the assumptions of a constant survival rate or a constant survival rate and a constant capture rate for the duration of the study. Detailed examples are given that illustrate the usefulness of these restrictions. There often can be a substantial gain in precision over Jolly-Seber estimates. In Chapter 5 we also consider 2 generalizations of the Jolly-Seber model. The temporary trap response model allows newly marked animals to have different survival and capture rates for 1 period. The other generalization is the cohort Jolly-Seber model. Ideally all animals would be marked as young, and age effects considered by using the Jolly-Seber model on each cohort separately. In Chapter 6 we present a detailed description of an age-dependent Jolly-Seber model, which can be used when 2 or more identifiable age classes are marked. In Chapter 7 we present a detailed description of the "robust" design. Under this design each primary period contains several secondary sampling periods. We propose an estimation procedure based on closed and open population models that allows for heterogeneity and trap response of capture rates (hence the name robust design). We begin by considering just 1 age class and then extend to 2 age classes. When there are 2 age classes it is possible to distinguish immigrants and births. In Chapter 8 we give a detailed discussion of the design of capture-recapture studies. First, capture-recapture is compared to other possible sampling procedures. Next, the design of capture-recapture studies to minimize assumption violations is considered. Finally, we consider the precision of parameter estimates and present figures on proportional standard errors for a variety of initial parameter values to aid the biologist about to plan a study. A new program, JOLLY, has been written to accompany the material on the Jolly-Seber model (Chapter 4) and its extensions (Chapter 5). Another new program, JOLLYAGE, has been written for a special case of the age-dependent model (Chapter 6) where there are only 2 age classes. In Chapter 9 a brief description of the different versions of the 2 programs is given. Chapter 10 gives a brief description of some alternative approaches that were not considered in this monograph. We believe that an excellent overall view of capture- recapture models may be obtained by reading the monograph by White et al. (1982) emphasizing closed models and then reading this monograph where we concentrate on open models. The important recent monograph by Burnham et al. (1987) could then be read if there were interest in the comparison of different populations.