COS 83-8
Predicting invasive plant performance with a single abiotic factor: a hierarchical Bayesian approach
Invasive plant fecundity is a key demographic parameter underlying propagule pressure. Thus, fecundity is critical for range expansion and establishment success across habitats. However, as a function both of plant traits and environmental conditions, it can vary significantly across ecological habitats and gradients. Moreover, measuring invasive plant fecundity in the field is often logistically difficult and ethically questionable, thereby inhibiting the development of accurate predictive models of invasive plant spread. Here, we demonstrate how to probabilistically model the seed production of Microstegium vimineum, a widespread and problematic invasive grass, as a function of an abiotic variable, by combining experimental datasets with a Bayesian Hierarchical Model (BHM). First, we analyzed seed production as a function of plant biomass in a mesocosm experiment that manipulated light levels. We then modeled biomass data from a field introduction experiment across 21 common garden sites that differed widely in available light. Finally, we combined these models and data with a BHM to predict seed production as a function of available light in the field.
Results/Conclusions
Our BHM demonstrates a saturating relationship between seed production and light availability. However, seed production is surprisingly high across the gradient of light availability. At full sun sites, seed production is virtually certain to maintain population growth (p = 99%). Moreover, at realistic forest shading levels (e.g. 10-20% available light), there is a substantial probability that the seed production of a single plant is adequate to overwhelm forest litter establishment resistance and maintain a stand (p = 74%). Even at very low light levels (e.g. 5% available light), the probability is still significant (p = 54%). Based on our model results, we suggest that M. vimineum is not inhibited by low light conditions, even at densely shaded interior forest sites, and is predominantly dispersal limited throughout much of its invasive range. In conclusion, we show that it is logistically and ethically tractable to model a key plant demographic process as a function of a single abiotic variable by combining manipulative experiments with field data. Moreover, we show that these models are useful for testing ecological hypotheses, including predictions of invasive plant spread, and can inform land management decisions.