Cross Validation of Quasi-Extinction Risks from Real Time Series: An Examination of Diffusion Approximation Methods

摘要

Population viability analysis (PVA) has quickly become a widespread component of conservation planning for at-risk species. However, the quantitative utility of the PVA toolkit remains a strongly contested issue, with recent efforts turning from pure simulation studies to include direct assessments of the predictability of the dynamics of real populations. To explore further the efficacy of such extinction risk modeling, we employed a cross validation approach with long-term population censuses from 271 time series representing 46 taxa, most of which were of past or present conservation concern (including 7 populations that actually went extinct). We used diffusion-approximation methods to estimate quasi-extinction risks for each population via two parameter-estimation techniques: the classical Dennis approach and the recently developed Dennis-Holmes approach. The latter technique can partition process error (environmental stochasticity) from non-process error (such as observation error).