Abstract
Lespedeza cuneata is one of the worst invasive plants across much of the North American Great Plains. In this dissertation, I describe the combination of field experimentation, statistical modeling and spatially explicit computer simulations that I used to investigate the persistence and spread of an L. cuneata invasion in a native prairie site. Management treatments reduced local infestations, but failing to treat led to large increases of L. cuneata. Density-based models with even simple spatial information were more predictive than presence/absence models of untreated conditions. The effect of herbicide was so strong that it negated the predictive value of model variables. The simulations indicated that imperfect detectability can reduce the overall treatment effectiveness when treatment intensity was low. The effect of imperfect detectability on spread could be minimized with intensive treatment. Taken together, these results indicate that continual management efforts over multiple years are necessary to control established infestations.