Survival estimates strongly depend on capture-recapture designs in a disturbed environment inducing dispersal

Capture-recapture (CR) approaches are extensively used to estimate demographic parameters. Their robustness relies on the selection of suitable statistical models, but also on the sampling design and effort deployed in the field. In colonial or territorial species showing breeding site fidelity, recurrent local perturbations, such as predation-induced breeding failure, may lead individuals to disperse locally or regionally. This might induce heterogeneity in individual CR histories and biases in demographic parameter estimates. Here, we assessed the effect of buffer areas and multi-site designs on annual survival estimate accuracy and precision. First, using simulated data with and without a multi-site design, we compared survival estimates for different scenarios involving contrasted local and regional dispersal. Then, using data from a local multi-site long-term monitoring survey carried out in a colony of black-legged kittiwakes, we tested the homogeneity of individual CR histories and compared survival estimates from three datasets including one or several breeding cliffs with or without buffer areas. Results from simulations highlighted that a single-site design consistently led to underestimated or less precise survival values compared to a multi-site design. Similarly, a multi-site design allowed detecting significant temporal trends while a single-site one did not. In the field-based study, adding buffer areas reduced individual heterogeneity, but all three datasets suffered from strong trap-dependence possibly due to individual breeding success affecting nest-site fidelity. Implementing a multi-site design significantly improved the accuracy and precision of annual survival estimates, regardless of CR models applied. Adapting the spatial scale of sampling design to the local environment and species behaviour is essential to provide robust parameters of key relevance for population monitoring and management. We show that implementing buffer areas and/or multi-site designs may be especially beneficial for long-lived species facing regular local perturbation events leading to local dispersal.