PS 33-154 - Modeling extinction in experimental populations of Daphnia

Tuesday, August 7, 2012
Exhibit Hall, Oregon Convention Center
Jake M. Ferguson, Center for Modeling Complex Interactions, University of Idaho, Moscow, ID and Jose M. Ponciano, Department of Biology, University of Florida, Gainesville, FL
Background/Question/Methods

Models of animal population variability are of fundamental importance for determining species persistence. Typical models of stochastic population dynamics include a deterministic density-dependent component and stochasticity due to demographic and environmental factors. Despite the recognized significance of stochastic components to population dynamics there is still confusion on how these quantities should scale with population abundance. We focused this study on how the extinction process in experimental Daphnia pulex populations depends on demographic and environmental stochasticity. We examined whether stochastic population time series models accurately assess population risk, and to what degree model assumptions affect these predictions.We used first passage times (defined as the time it takes for an initial population to reach some threshold population) to characterize model predictions. A number of assumptions associated with population dynamics models were tested by fitting models that varied assumptions about the density dependence, variance structure, and probability distribution used to describe abundance transitions.

Results/Conclusions

We found that models incorporating both demographic and environmental stochasticity predict extinction performed better than models with only one term and that demographic stochasticity was found to scale as a function of both the population size and the form of density dependence. It is shown how sensitivies about the probability distributiion used to model abundance transitions make distributional assumptions an important component of model building.  Additionally, we found that model selection procedures such as AIC do not necessarily select models that predict time to extinction best and discuss some alternatives for performing model selection.