An Integrated Approach for Predicting Fates of Reintroductions with Demographic Data from Multiple Populations

Abstract

We devised a novel approach to model reintroduced populations whereby demographic data collected from multiple sites are integrated into a Bayesian hierarchical model. Integrating data from multiple reintroductions allows more precise population-growth projections to be made, especially for populations for which data are sparse, and allows projections that account for random site-to-site variation to be made before new reintroductions are attempted. We used data from reintroductions of the North Island Robin (Petroica longipes), an endemic New Zealand passerine, to 10 sites where non-native mammalian predators are controlled. A comparison of candidate models that we based on deviance information criterion showed that rat-tracking rate (an index of rat density) was a useful predictor of robin fecundity and adult female survival, that landscape connectivity and a binary measure of whether sites were on a peninsula were useful predictors of apparent juvenile survival (probably due to differential dispersal away from reintroduction sites), and that there was unexplained random variation among sites in all demographic rates. We used the two best supported models to estimate the finite rate of increase (λ) for populations at each of the 10 sites, and for a proposed reintroduction site, under different levels of rat control. Only three of the reintroduction sites had λ distributions completely >1 for either model. At two sites, λ was expected to be >1 if rat-tracking rates were <5%. At the other five reintroduction sites, λ was predicted to be close to 1, and it was unclear whether growth was expected. Predictions of λ for the proposed reintroduction site were less precise than for other sites because distributions incorporated the full range of site-to-site random variation in vital rates. Our methods can be applied to any species for which postrelease data on demographic rates are available and potentially can be extended to model multiple species simultaneously.

Combining data‐derived priors with postrelease monitoring data to predict persistence of reintroduced populations

“…The final point depends on the priors being valid, as they may otherwise bias parameter estimates, especially for small data sets. This potential problem is avoided if fully data‐derived priors are used, meaning not only that the prior data are as equally objective as the new data (Morris et al., ), but also that the uncertainty involved in extrapolating prior data to a new site is properly accounted for (Parlato & Armstrong, ). Although we were able to adopt this approach, this will be difficult with most reintroductions due to lack of high‐quality data for multiple sites, and hence, some degree of expert judgment will be needed to derive priors (Canessa et al., ).…”

“…NI robins were widespread throughout the North Island and nearby offshore islands at the time of European settlement, but disappeared from most of their original range following forest clearance and introduction of mammalian predators. They have now been reintroduced to many sites where mammalian predators have been controlled or eradicated, with these reintroductions having mixed success (Miskelly & Powlesland, ; Parlato & Armstrong, ). This scenario is ideal for assessing the usefulness of data‐derived priors because multiple reintroductions have taken place, postrelease data have been collected using fairly consistent methodology, the biology and threats are well understood, but there remains considerable uncertainty about whether many reintroductions will be successful or not.…”

“…We report the first case study illustrating how fully data‐derived priors can be combined with postrelease monitoring data when making predictions for reintroduced populations. Parlato and Armstrong () earlier showed how data on vital rates for multiple reintroduced populations of North Island (NI) robin ( Petroica longipes ) populations could be integrated using Bayesian hierarchical modeling, allowing random site‐to‐site variation to be accounted for. This meant that prior distributions could be derived for the finite rate of increase (λ) expected at proposed reintroductions, meaning that site selection could be improved over time through passive adaptive management (McCarthy et al., ).…”

Strategic Rat Control for Restoring Populations of Native Species in Forest Fragments

“…The species originally occurred throughout forested parts of the North Island, but was extirpated from >90% of its original range due to forest clearance and predation by non‐native mammals. The key limiting factor appears to be predation by ship rats ( Rattus rattus ), particularly on nests and nesting females (Brown ; Armstrong et al ; Parlato & Armstrong ). Although the species is not classified as threatened, many local restoration programs are attempting to recover robin populations through translocation and predator management (Parlato & Armstrong ).…”

“…The key limiting factor appears to be predation by ship rats ( Rattus rattus ), particularly on nests and nesting females (Brown ; Armstrong et al ; Parlato & Armstrong ). Although the species is not classified as threatened, many local restoration programs are attempting to recover robin populations through translocation and predator management (Parlato & Armstrong ). The small territory sizes (approximately 0.5 ha) and limited movements of robins and their key predator means that populations may be managed on small fragments on private land.…”