Incorporating landscape stochasticity into population viability analysis

• Published in: Ecological applications Vol. 17; no. 2; pp. 317 - 322
• Main Authors: Chisholm, R.A, Wintle, B.A
• Format: Journal Article
• Language: English
• Published: United States 01.03.2007
• Subjects: animal ecology; Animals; Certhia americana; computer software; DLMP model; dynamic landscape metapopulation model; Ecosystem; Environmental Monitoring - methods; landscapes; Passeriformes - physiology; Population Dynamics; population ecology; population viability analysis; Software; Stochastic Processes; wild birds; wildlife management
• ISSN: 1051-0761; 1939-5582
• DOI: 10.1890/05-1580

The importance of incorporating landscape dynamics into population viability analysis (PVA) has previously been acknowledged, but the need to repeat the landscape generation process to encapsulate landscape stochasticity in model outputs has largely been overlooked. Reasons for this are that (1) there is presently no means for quantifying the relative effects of landscape stochasticity and population stochasticity on model outputs, and therefore no means for determining how to allocate simulation time optimally between the two; and (2) the process of generating multiple landscapes to incorporate landscape stochasticity is tedious and user-intensive with current PVA software. Here we demonstrate that landscape stochasticity can be an important source of variance in model outputs. We solve the technical problems with incorporating landscape stochasticity by deriving a formula that gives the optimal ratio of population simulations to landscape simulations for a given model, and by providing a computer program that incorporates the formula and automates multiple landscape generation in a dynamic landscape metapopulation (DLMP) model. Using a case study of a bird population, we produce estimates of DLMP model output parameters that are up to four times more precise than those estimated from a single landscape in the same amount of total simulation time. We use the DLMP modeling software RAMAS Landscape to run the landscape and metapopulation models, though our method is general and could be applied to any PVA platform. The results of this study should motivate DLMP modelers to consider landscape stochasticity in their analyses.