Published online by Cambridge University Press: 20 January 2017
The impact of invasive weed management on plant community composition is highly dependent on location-specific factors. Therefore, treatment means from experiments conducted at a given set of locations will not reliably predict community response to weed management elsewhere. We developed a model that rescales treatment means to better match local conditions. The goal of this paper was to determine if this rescaling improves predictions. We used our model to predict leafy spurge stem length density and grass biomass data from field experiments. The experiments consisted of herbicide-treated plots, untreated controls, and, in some cases, grass seeding treatments. When herbicides suppressed leafy spurge, the model explained 21 to 48% more variation in grass response than did mean grass response to the same or similar herbicide treatments applied at other sites. When herbicides killed grass, the model explained 53% more variation in leafy spurge response than did mean leafy spurge response to the same herbicide treatment applied at other sites. We regressed model predictions against observed data and tested the null hypothesis that resulting slope terms were equal to 1.0. Because the null hypothesis was rejected in two of four tests, the model may systematically over- or underpredict in some situations. However, measurement error in the observed data, unintended herbicide injury, or an inaccurate allometric relationship may account for a major proportion of the systematic deviations, and these factors would not cause prediction error in some management applications. Because the model tends to be better than the means from experiments at predicting plant community composition, we conclude that the model could advance managers' ability to predict plant community responses to invasive weed management.