Using prior data to improve models for reintroduced populations: A case study with North Island Saddlebacks



This article is corrected by:

  1. Errata: Gedir, J. V., J. M. Thorne, K. Brider, and D. P. Armstrong. 2013. Using prior data to improve models for reintroduced populations: a case study with North Island saddlebacks. Journal of Wildlife Management 77:1114–1123 Volume 77, Issue 8, 1678, Article first published online: 24 October 2013

  • Associate Editor: Sarah Converse


Monitoring of reintroductions is essential, but often requires significant resources. If data are available from previous reintroductions, they can be used as priors in a Bayesian framework, potentially reducing the amount of data needed from a new reintroduced population before reasonable projections can be made. We applied this approach to the reintroduction of North Island saddlebacks (Philesturnus rufusater) to Bushy Park, New Zealand, an 87-ha forest reserve within a fenced exclosure where exotic mammalian predators had been eradicated. We released 34 birds in June 2006, and conducted intensive monitoring of survival and fecundity until January 2010. We modeled the data in an integrated framework whereby we generated population projections concurrent with estimations of survival and fecundity parameters, meaning estimation of uncertainty was propagated into the projections. We compared projections from 1, 2, and 3 years of monitoring data, using either non-informative priors (i.e., inferences from Bushy Park only) or informative priors generated from 2 previous saddleback reintroductions where 5–8 years of data had been collected. The Bushy Park saddleback population grew rapidly over the monitoring period (approx. 80% per annum), and projections from 3 years of data with non-informative priors gave a median equilibrium breeding population of 255 with a 95% prediction interval of 122–1,238. Projections from 2 years of data gave a lesser mean equilibrium population of 93 (50–360). Projections from 1 year of data revealed uncertainty about the success of the reintroduction at this stage (8.4% of simulated populations declined), and equilibrium population size could not be projected without the density effect. Using informative priors eliminated the uncertainty about whether the population would grow, allowed equilibrium population size to be projected from 1 year of Bushy Park data, and made the projections from 2 years of data more consistent with those from the full 3-year dataset. We discuss the implications of monitoring requirements on decision making. © 2013 The Wildlife Society.