A hierarchical Bayesian model of time-since-fire effects on the demography of the Florida scrub endemic, Dicerandra frutescens ssp frutescens (Lamiaceae)

Managing populations, either for conservation, harvesting, or control, requires a mechanistic or semi-mechanistic understanding of population dynamics.  Here we report on how time-since-fire affects demographic transitions in an endangered plant, Dicerandra frutescens ssp. frutescens (Lamiaceae), which is specialized to gaps created by fire.  We used a hierarchical Bayesian model to estimate transition probabilities (i.e. the elements of population projection matrices) as a function of time-since-fire and random effects, while accounting for process variability and sampling uncertainty.  The data come from a long-term study (13 years) of marked individuals in five populations.  We used a Bayesian model comparison criterion to evaluate different models of time-since-fire effects.  The best models indicate that death becomes increasingly probable and progression increasingly improbable with time-since-fire.  The magnitude of some of the time-since-fire effects is substantial: death is 3 to 5 times more likely for flowering plants >6 years vs. 3-6 years post-fire, 3-step progression is almost 7 times less likely, and large flowering plants are more than six times more likely to stop flowering.  Our approach is quite general.  It can be easily adapted to examine how any environmental covariate affects demographic transitions.  It also forms the basis for a Bayesian population viability analysis.