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

Monitoring is an essential part of reintroduction programs, but many years of data may be needed to obtain reliable population projections. This duration can potentially be reduced by incorporating prior information on expected vital rates (survival and fecundity) when making inferences from monitoring data. The prior distributions for these parameters can be derived from data for previous reintroductions, but it is important to account for site‐to‐site variation. We evaluated whether such informative priors improved our ability to estimate the finite rate of increase (λ) of the North Island robin (Petroica longipes) population reintroduced to Tawharanui Regional Park, New Zealand. We assessed how precision improved with each year of postrelease data added, comparing models that used informative or uninformative priors. The population grew from about 22 to 80 individuals from 2007 to 2016, with λ estimated to be 1.23 if density dependence was included in the model and 1.13 otherwise. Under either model, 7 years of data were required before the lower 95% credible limit for λ was > 1, giving confidence that the population would persist. The informative priors did not reduce this requirement. Data‐derived priors are useful before reintroduction because they allow λ to be estimated in advance. However, in the case examined here, the value of the priors was overwhelmed once site‐specific monitoring data became available. The Bayesian method presented is logical for reintroduced populations. It allows prior information (used to inform prerelease decisions) to be integrated with postrelease monitoring. This makes full use of the data for ongoing management decisions. However, if the priors properly account for site‐to‐site variation, they may have little predictive value compared with the site‐specific data. This value will depend on the degree of site‐to‐site variation as well as the quality of the data.监测是动植物再引入项目的核心环节之一，但要获取可靠的种群预测结果，往往需要积累多年的监测数据。若在基于监测数据开展统计推断时，纳入预期生命率（存活率与繁殖力）的先验分布（prior distributions），则有望缩短数据采集周期。此类参数的先验分布可源自过往再引入项目的相关数据，但需重点考量不同放归地点间的差异。我们评估了此类信息性先验（informative priors）是否能够提升对新西兰塔瓦拉努伊区域公园再引入的北岛知更鸟（*Petroica longipes*）种群有限增长率（finite rate of increase，λ）的估算精度。我们通过逐年添加放归后的监测数据，对比了使用信息性先验与无信息先验（uninformative priors）的模型的精度提升情况。该种群在2007至2016年间从约22只增长至80只；当模型纳入密度制约（density dependence）项时，λ的估算值为1.23，未纳入时则为1.13。无论采用哪种模型，均需积累7年的监测数据，方可使λ的95%可信下限（credible limit）大于1，从而确认种群能够持续存续。本次研究中，信息性先验并未缩短这一数据需求周期。基于实测数据推导的先验分布在再引入项目开展前具有较高应用价值，可用于提前估算λ；但在本次研究的案例中，一旦获取针对该放归地点的专属监测数据，先验信息的作用便会被后续实测数据所覆盖。本文提出的贝叶斯方法（Bayesian method）对于再引入种群的研究具有逻辑合理性：其可将用于放归前决策的先验信息与放归后的监测数据相结合，充分利用全部数据支撑持续的种群管理决策。但若先验信息已充分考量了不同地点间的异质性，则相较于针对目标放归地点的专属监测数据，其预测价值将较为有限，这一价值的高低取决于不同地点间的异质性程度以及监测数据的质量。