Return on investment: A hierarchical model of the pathways from bud production to seed dispersal in an understory herb

The fate of female reproductive efforts after flower production in herbaceous plants can have a profound impact on spatial population structure, gene flow, and population viability. For an ovary to become a viable, dispersed seed, it must be pollinated and the developing seed must not be aborted by the plant or predated by seed-eating insects. Challenges to the accurate assessment of seed fate include the hierarchical scale of many plant systems, uncertainty in estimates of important parameters that describe these processes, and non-ignorable missing data, which results from the difficulties in estimating processes that render potential seeds unobservable (such as seed abortion and predation). Using hierarchical Bayesian methods in a multilevel context, I modeled seed fate in Eurybia chorolepis (Asteraceae), an understory perennial herb found in the southern Appalachians. Missing data were modeled from within the MCMC loops, and thus the data as well as the parameters were jointly estimated. The hierarchical design of the experiment enabled a careful inventory of the variation in floral production, pollination, seed abortion and seed predation at the scale of the stem (ramet) and that of the patch. Most of the variation in this study occurred at the stem-level, indicating fine scale relationships between the insect community and seed fate. Environmental variables, such as soil moisture, light, and soil chemicals showed no association with key processes determining seed fate. This talk presents the model and results, as well as a discussion on the importance of scale to the study of plant-insect interactions.