PARENT SESSION
Poster Session # 2: Statistical Ecology and Ecological Modeling.

Monday, August 4 Presentation from 5:00 PM to 6:30 PM. SITCC Exhibit Hall B.

Using Bayesian and frequentist statistics to develop individual trait-based models of population dynamics.

Ellison, Aaron^*,1, Gotelli, Nicholas², ¹ Harvard Forest, Petersham, MA² University of Vermont, Burlington, VT

ABSTRACT- We used frequentist and Bayesian methods to estimate parameters for a population growth model of Sarracenia purpurea, a perennial carnivorous plant. Individual, trait-based models typically use fixed parameters estimated from data with frequentist statistics. Variability (stochasticity) in these models is applied uniformly across individuals by using a common stochatic error term. In a Bayesian analysis, the parameter estimates are not assumed to be fixed, but instead vary among individuals and across time. We compared two classical models (without and with the stochastic error term) with two models whose parameters were estimated with Bayesian analysis. These models are hierarchical: the classical models are nested within the Bayesian models. Bayesian models outperformed frequentist models in recovering observed variability in plant traits and population size structure. Although individual plants in Bayesian models had lower forecasted survivorship than those in frequentist models, population growth rates were higher and more variable in the Bayesian models; this led to a lower probability of extinction predicted by the Bayesian models relative to the frequentist models. Our results illustrate that population projections are sensitive to the underlying assumptions regarding parameter estimation and the sources of measurement and process error. Bayesian modeling makes these assumptions explicit and could be especially useful in guiding data collection and model development for the management of threatened and endangered species.

Key words: Sarracenia, trait-based models, Bayesian statistics, demography