LAUSR

Abstract Many species of conservation interest exist solely or largely in isolated populations. Ideally, prioritization of management actions among such populations would be guided by quantitative estimates of extinction risk, but conventional methods of demographic population viability analysis (PVA) model each population separately and require temporally extensive datasets that are rarely available in practice. We introduce a general class of statistical PVA that can be applied to many populations at once, which we term multiple population viability analysis or MPVA. The approach combines models of abundance at multiple spatial locations with temporal models of population dynamics, effectively borrowing information from more data-rich populations to inform inferences for data-poor populations. Covariates are used to explain population variability in space and time. Using Bayesian analysis, we illustrate the method with a dataset of Lahontan cutthroat trout ( Oncorhynchus clarkii henshawi ) observations that previously had been analyzed with conventional PVA. We find that MPVA predictions are similar in bias and higher in precision than predictions from simple PVA models that treat each population individually; moreover, the use of covariates in MPVA allows for predictions in minimally-sampled and unsampled populations. The basic MPVA model can be extended in multiple ways, such as by linking to a sampling and observation model to provide a full accounting of uncertainty. We conclude that the approach has great potential to expand the use of PVA for species that exist in multiple, isolated populations.