|HOME SCHEDULE AUTHOR INDEX SUBJECT INDEX|
Modeling population dynamics with a mechanistic description of stochasticity.
Gross, Kevin*,1, Ives, Anthony1, 1 University of Wisconsin-Madison, Madison, WI
ABSTRACT- Analyzing population dynamics data in an ecologically meaningful way presents challenges that may not be met by conventional time series methods. Our study system, pea aphids and parasitoid wasps that attack them, poses two of these challenges. First, we have multiple (and sometimes discrepant) sources of data about population densities that must be reconciled. Second, information about the vital rates governing population dynamics is difficult to extract from the observed data. To accommodate these complexities, we have developed a hierarchical, hidden state-space model that allows us to assess the nature and volume of information about the aphid vital rates contained in the data. Our model uses a three-level hierarchy (the data, the true population densities, and the underlying vital rates) that separates sampling variability (measurement error) from fluctuations in population densities. We approach the model from a Bayesian perspective. The main product of our analysis, a posterior distribution of the time series of aphid vital rates, allows us to ask if the data do or do not contain evidence of aphid population regulation by parasitism. Additionally, we discuss general issues in Markov chain Monte Carlo (MCMC) and the relative merits of different sampling protocols.