Using radio-tracking data to predict post-release establishment in reintroductions to habitat fragments

Dispersal from release areas is a critical problem for reintroductions. Reliable methods are therefore needed for analysing post-release monitoring data to guide further releases. Radio tracking can greatly improve data quality by distinguishing dispersal from mortality. However, fates of animals continue to be uncertain if transmitters have short battery life and detection range, as is typically the case with small animals. We present an approach for simultaneously modelling probabilities of fidelity (remaining in release area), survival, detection and transmitter failure from post-release monitoring data, and illustrate how it was applied to translocations of North Island robins (Petroica longipes) to 17 forest fragments (5-56 ha) over 5 years. The modelling showed that fidelity probability depended on the sex (higher in females) and translocation date (higher in winter than autumn), and that variation among fragments was well explained by the “cost distance to nearest neighbour” (an index reflecting the amount of pasture and shrubland needing to be crossed to reach another forest area) and the area of the release fragment (higher in larger fragments). Combined with survival, the estimated probability of a bird remaining in its release fragment the next breeding season ranged from 0.02 to 0.39. As these estimates were refined, they could be used to assess suitability of fragments for further releases and numbers of each sex needing to be released to compensate for dispersal. The Bayesian framework underlying the approach potentially allows application to any amount of data by using informative priors derived from previous translocations or expert opinion.