Incorporating data‐based estimates of temporal variation into projections for newly monitored populations. (22nd May 2021)

• Record Type: Journal Article
• Authors: Parlato, E. H.; Ewen, J. G.; McCready, M.; Gordon, F.; Parker, K. A.; Armstrong, D. P.
• Is Part Of: Animal conservation. Volume 24:Number 6(2021)
• Page Start: 1001
• Page End: 1012
• Publication Date: 2021-05-22
• Subjects: population modelling -- population viability -- annual variation -- environmental stochasticity -- prior information -- Bayesian inference -- hihi -- reintroductions; Conservation biology -- Periodicals; Wildlife conservation -- Periodicals; Conservation de la biodiversité; Conservation de la faune; Périodique électronique (Descripteur de forme); Ressource Internet (Descripteur de forme); 333.95416
• DOI: 10.1111/acv.12702
• Languages: English
• ISSNs: 1367-9430

Abstract

Abstract: The importance of accounting for temporal variation in vital rates when modelling population dynamics is well recognized. However, long‐term (usually >5 years) datasets are needed to estimate this variation. Consequently, models for newly monitored populations typically assume no temporal variation or use default values provided in software programmes, both of which can give misleading inferences about population dynamics. The goal of this study is to improve estimation of dynamics in the initial years of conservation programmes by incorporating data‐based estimates of temporal variation from other populations with longer‐term data available. We show how data‐derived priors can be generated using estimates of temporal variation in vital rates from other populations, providing information about expected variation until sufficient population‐specific data are available. We specifically evaluated whether data‐derived priors improved our ability to estimate temporal variation for a reintroduced population monitored for 3 years postrelease. We first predicted population growth and probability of extinction assuming no temporal variation in vital rates, then compared projections to those obtained when temporal variation was estimated with uniform priors or with the data‐derived priors. Both types of priors were constrained to plausible ranges, and we also assessed sensitivity of model outputs to widening those ranges. Median projected population size was similar under all three models, but extinction probability was higher with inclusion of temporal variability, reiterating the importance of incorporating this source of uncertainty. Projections with temporal variation were similar irrespective of whether data‐derived priors or uniform priors were used. However, the data‐derived priors generated more precise estimates of annual variation that were less sensitive to relaxation of prior constraints. The approach we present can facilitate management decisions at the outset of conservation programmes when risk assessments that account for all relevant uncertainties can be crucial for determining optimal management strategies. Abstract : Conservation biologists are frequently required to make predictions about population persistence from short‐term data. However, analyses of data from short‐term studies typically do not account for temporal variation because longer‐term data are needed to estimate that variation. This paper presents an approach for improving inferences in the initial years of monitoring programmes by incorporating data‐based estimates of temporal variation from other populations with longer‐term data available. We show how data‐derived priors can be generated using estimates of temporal variation in vital rates from other populations, providing information about expected variation until sufficient population‐specific data are available. The approach we present can facilitate management decisions that account for all relevant uncertainties. Photo credit: Ian Douglas