Friday, August 7, 2009

PS 89-151: The influence of informed versus uninformed priors on forecasts of growth rates and extinction risks of a New England population of northern pitcher plants (Sarracenia purpurea L.)

Sydne Record, University of Massachusetts Amherst, Aaron M. Ellison, Harvard Forest (Harvard University), and Nicholas J. Gotelli, University of Vermont.


Ecologists have a long-standing interest in modeling population dynamics.  When applying the results of population growth models to the management of endangered species, it is critical that the uncertainty in the model predictions is low.  One limitation of demographic matrix models is that they require difficult to collect, long-term data sets to help reduce uncertainty in their forecasts.  One possible solution to reducing uncertainty in these models is to include information from similar studies as prior data in a Bayesian demographic model.  There are few examples of studies using Bayesian demographic models, and all of them specify uninformed priors.  In this study, we compare the results of stochastic Bayesian demographic matrix models for northern pitcher plants (Sarracenia purpurea), using uninformed and informed priors.  The data we use in the models come from demographic studies on S. purpurea at two locations: Molly Bog in Vermont and Hawley Bog in Massachusetts.  We run the models on the Molly Bog data and use the Hawley Bog data to construct the informed priors.  To incorporate environmental stochasticity, we forecast nitrogen deposition, an important driver of pitcher plant extinction risk, and use experimental data on the effects of variable amounts of nitrogen on vital rates to determine our sampling of transition matrices for “good” versus “bad” years.   


Sensitivity and elasticity analyses showed that the influence of particular matrix elements on population growth for transition matrices derived from uninformed and informed priors was similar.  Population growth (λ) for transition matrices was most sensitive to the survival of juvenile and non-flowering adults.  Instantaneous growth rates and extinction probabilities estimated from the analysis using informed priors were more optimistic than those estimated from the analysis using uninformed priors.  Instantaneous growth rates estimated at Molly Bog were r = log(λ) = -0.067 individuals·individual-1·year-1 with uninformed priors and r = -0.051 individuals·individual-1·year-1 with informed priors.  The mean probability of extinction, where a population falls below one individual, after 100 years estimated at Molly Bog with uninformed priors was 0.6466 and 0.331 with informed priors.  The higher growth rate and lower extinction risk from the analysis with informed priors may be because on average Hawley Bog has lower annual nitrogen deposition than Molly Bog.  These results suggest that ecologists should take care when using informed priors because even if the prior data comes from a nearby population of the same species, there may be site-specific demographic differences introduced.