Investigations of potential bias in the estimation of λ using Pradel's (1996) model for capture-recapture data

Pradel's (1996) temporal symmetry model permitting direct estimation and modelling of population growth rate, lambda(i), provides a potentially useful tool for the study of population dynamics using marked animals. Because of its recent publication date, the approach has not seen much use, and there have been virtually no investigations directed at robustness of the resulting estimators. Here we consider several potential sources of bias, all motivated by specific uses of this estimation approach. We consider sampling situations in which the study area expands with time and present an analytic expression for the bias in (lambda) over cap (i) We next consider trap response in capture probabilities and heterogeneous capture probabilities and compute large-sample and simulation-based approximations of resulting bias in (lambda) over cap (i). These approximations indicate that trap response is an especially important assumption violation that can produce substantial bias. Finally, we consider losses on capture and emphasize the importance of selecting the estimator for lambda(i) that is appropriate to the question being addressed. For studies based on only sighting and resighting data, Pradel's (1996) (lambda) over cap (i) is the appropriate estimator.