Background/Question/Methods Understanding and predicting fluctuations in the abundance of natural populations is a central activity of ecology.  Natural populations are inevitably under the influence of stochastic processes (weather, floods, fire); this environmental variation causes variation in vital rates (survival, fecundity) which may be positively or negatively correlated across the life cycle.  We used twenty years of data and a hierarchical Bayesian model to estimate vital rates and their covariation for an endangered plant, Dicerandra frutescens ssp. frutescens (Lamiaceae), as a function of time-since-fire and year effects.  Samples from the joint posterior distribution of model parameters were then used to simulate population dynamics as a function of fire regime and year variation.  The design of the simulations allowed us to partition variation in the stochastic population growth rate due to process variability (fire, year effects, and demographic stochasticity) vs. parameter uncertainty (finite amounts of data). 

Results/Conclusions Time-since-fire had negative effects on recruitment and fecundity: germination and the number of flowering branches declined with time-since-fire, and reproductive plants were increasingly likely to become non-reproductive.  Time-since-fire had little effect, or possibly a positive effect, on survival.  Model comparison strongly supported inclusion of time-since-fire effects and weakly supported inclusion of year effects.  Simulations of population growth suggested populations of Dicerandra frutescens ssp. frutescens are least likely to go extinct if the average time between fires is 24 to 28 years.  About half of the variation in the stochastic population growth rate was due to parameter uncertainty.  This worked example illustrates a general analytical framework for 1) estimating vital rates as a function of an environmental factor, 2) accounting for covariation among vital rates due to fixed and random effects, and 3) simulating population dynamics as a function of stochastic environmental processes while taking into account uncertainty about their effects.  We discuss future areas of development for hierarchical Bayesian population modeling.