Submitted to: Weed Science
Publication Type: Peer reviewed journal
Publication Acceptance Date: 2/15/2005
Publication Date: 9/1/2005
Citation: Rinella, M.J., Sheley, R.L. 2005. Models that predict invasive weed and grass dynamics: 1. Model development. Weed Science 53:586-593. Interpretive Summary: In order to select effective methods for controlling invasive weeds, managers must be able to predict how weeds, and the plants that they grow with, will respond to herbicides, reseeding, grazing, biological control, and other management strategies. We used field experiments to develop mathematical models that predict how various plants respond to weed control strategies.
Technical Abstract: Invasive weed managers are presented with a complicated and ever-enlarging set of management alternatives. Sifting through these alternatives and identifying the optimal weed management strategy for a given set of conditions requires predicting how the alternatives will affect plant community composition. Field experiments have advanced our predictive capability in this regard, but several extrapolation problems impose limits on the prediction accuracy that can be achieved by simply interpreting treatment means and variances as predictions. These and other extrapolation problems also cause experiments to mischaracterize prediction confidence. Examples of the extrapolation problems include nonlinear relationships between competing plants, site-to-site variation in equilibrium biomass production of desired species and weeds and varying random error variances. Our objective was to develop models that alleviate a subset of these problems. We developed the models using data from two experiments in which four Kentucky bluegrass (Poa pratensis) densities and six western wheatgrass (Pascopyrum smithii) and leafy spurge (Euphorbia esula) densities were combined in field plots. Predicted versus observed graphs indicate that the models predict our data accurately, which suggests models might predict plant community response to weed management actions more accurately than do means and variances from experiments that are not adjusted to reflect local conditions. However, truly rigorous tests of prediction accuracy will require testing the models on data not used in model development. Such independent tests are the focus of the companion to thi