Movement patterns and study area boundaries: influences on survival estimation in capture  mark  recapture studies Gregg E. Horton and Benjamin H. Letcher G. E. Horton (gregg.horton@sbcglobal.net) and B. H. Letcher, S.O. Conte Anadromous Fish Research Center, USGS/Leetown Science Center, One Migratory Way, Turners Falls, MA 01376, USA. Present address for GEH: 6489 Dry Creek Road, Healdsburg, CA 95448, USA. The inability to account for the availability of individuals in the study area during capturemarkrecapture (CMR) studies and the resultant confounding of parameter estimates can make correct interpretation of CMR model parameter estimates difficult. Although important advances based on the CormackJollySeber (CJS) model have resulted in estimators of true survival that work by unconfounding either death or recapture probability from availability for capture in the study area, these methods rely on the researcher’s ability to select a method that is correctly matched to emigration patterns in the population. If incorrect assumptions regarding site fidelity (non-movement) are made, it may be difficult or impossible as well as costly to change the study design once the incorrect assumption is discovered. Subtleties in characteristics of movement (e.g. life history-dependent emigration, nomads vs territory holders) can lead to mixtures in the probability of being available for capture among members of the same population. The result of these mixtures may be only a partial unconfounding of emigration from other CMR model parameters. Biologically-based differences in individual movement can combine with constraints on study design to further complicate the problem. Because of the intricacies of movement and its interaction with other parameters in CMR models, quantification of and solutions to these problems are needed. Based on our work with stream-dwelling populations of Atlantic salmon Salmo salar, we used a simulation approach to evaluate existing CMR models under various mixtures of movement probabilities. The Barker joint data model provided unbiased estimates of true survival under all conditions tested. The CJS and robust design models provided similarly unbiased estimates of true survival but only when emigration information could be incorporated directly into individual encounter histories. For the robust design model, Markovian emigration (future availability for capture depends on an individual’s current location) was a difficult emigration pattern to detect unless survival and especially recapture probability were high. Additionally, when local movement was high relative to study area boundaries and movement became more diffuse (e.g. a random walk), local movement and permanent emigration were difficult to distinguish and had consequences for correctly interpreting the survival parameter being estimated (apparent survival vs true survival). In capturemarkrecapture (CMR) studies, animal move- ment can affect estimates and interpretations of survival. An important reason is the difficulty in distinguishing unavail- ability for capture from death in studies of population ecology. Evidence for this is the numerous studies that report estimates of apparent survival (f) (the probability of survival given availability in the study area) rather than true survival (S) (Dinsmore and Collazo 2003, Miller et al. 2003, Hagen et al. 2005, Petty et al. 2005, Sandercock et al. 2005). Unavailability in the study area can result from physical absence (emigration from the study area) or presence in a state that renders the individual impossible to capture (e.g. an underground burrow) and either can affect the interpretation of parameter estimates. When a tagged individual is temporarily unavailable for capture for at least one occasion but has a non-zero probability of again being available for capture in the study area sometime in the future, this temporary unavailability is confounded with recapture probability (p). Alternatively, unavailability for capture is confounded with mortality when a tagged individual is unavailable for capture and permanently remains unavailable even though the animal may be alive (Burnham 1993, Kendall et al. 1995). In CMR studies, both temporary and permanent unavailability is often due to emigration from the study site. Emigration in CMR models Differences in which pairs of parameters are confounded highlights the importance of correctly identifying or assuming which emigration pattern is present in a sample from a study population so that the parameter estimate of interest (often, the probability of survival) can be correctly interpreted. There are three reasons why individuals tagged on previous occasion i may not be captured on some future occasion: 1) death on the study area; 2) both alive and Oikos 117: 11311142, 2008 doi: 10.1111/j.2008.0030-1299.16686.x, # 2008 The Authors. Journal compilation # 2008 Oikos Subject Editor: James Roth. Accepted 10 March 2008 1131