MODELING INVASIVE PLANT SPREAD - THE ROLE OF PLANT-ENVIRONMENT INTERACTIONS AND MODEL STRUCTURE

Alien plants invade many ecosystems worldwide and often have substanti al negative effects on ecosystem structure and functioning. Our abilit y to quantitatively predict these impacts is, in part, limited by the absence of suitable plant-spread models and by inadequate parameter es timates for such models. This paper explores the effects of model, pla nt, and environmental attributes on predicted rates and patterns of sp read of alien pine trees (Pinus spp.) in South African fynbos (a medit erranean-type shrubland). A factorial experimental design was used to: (1) compare the predictions of a simple reaction-diffusion model and a spatially explicit, individual-based simulation model; (2) investiga te the sensitivity of predicted rates and patterns of spread to parame ter values; and (3) quantify the effects of the simulation model's spa tial grain on its predictions. The results show that the spatial simul ation model places greater emphasis on interactions among ecological p rocesses than does the reaction-diffusion model. This ensures that the predictions of the two models differ substantially for some factor co mbinations. The most important factor in the model is dispersal abilit y. Fire frequency, fecundity, and age of reproductive maturity are les s important, while adult mortality has little effect on the model's pr edictions. The simulation model's predictions are sensitive to the mod el's spatial grain. This suggests that simulation models that use matr ices as a spatial framework should ensure that the spatial grain of th e model is compatible with the spatial processes being modeled. We con clude that parameter estimation and model development must be integrat ed procedures. This will ensure that the model's structure is compatib le with the biological processes being modeled. Failure to do so may r esult in spurious predictions.