Weed Research, 199S, Volume 35, 42t~+28 Evaluation of three empirical models depicting Ambrosia artemisiifolia competition in white bean D. CHIKOYE AND C. J. SWANTON Department of Crop Science, University of Guelph, Guelph, Ontario, Canada NIG2W1 Received 29 June 1994 Revised version accepted 5 March 1995 Summary The performance of three empirical models des- cribing white bean yield loss (YL) from common ragweed competition was compared using field experiments from Staffa and Woodstock, both in Ontario, Canada, in 1991 and 1992. One model was based upon both weed density and relative time of emergence. The other two models described yield loss as a function of weed leaf area relative to the crop. The model based on both weed density and relative time of emergence best described the data sets. The pre- dicted maximum yield loss (A) and the para- meter for relative time of weed emergence (C) varied across locations and years whereas the yield loss at low weed density (/) was relatively more consistent across locations and years. Use of thermal time (base temperature=10°C) rather than calendar days did not change the overall fit of the model, but reduced the value of the parameter for the relative time of weed emergence (C). The two parameter leaf area model accounting for maximum yield loss (m) gave a better fit to the data compared with the one parameter model. The relative damage coefficient (q) varied with time of leaf area as- sessment, location and year. Values of q calcu- lated from relative leaf area growth rates of the crop and weed were similar to observed values. © 1995 European Weed Research Society The relationship between q and accumulated thermal time was linear but varied with location and year. As management tools, models based upon relative leaf area have advantages over models based on dereity and relative time of emergence since the level of weed infestation needs only to be assessed once, whereas density and emergence time require frequent observa- tions. The ability to assess accurately and quickly both the crop and weed leaf area, how- ever, may limit the practical application of models based on leaf area. The inability of em- pirical models to account for year-to-year vari- ation in environmental conditions was observed. Introduction Models that quantify crop yield losses based upon eariy observations of weed infestation are essential for assessing the profitability of weed control options. Many empirical or descriptive models have been developed to describe the res- ponse of crop yield to one or more parameters that characterize weed infestation. Most models describe crop yield loss from a given weed den- sity (Coble etal., 1981: Cousens, 1985; Weaver et al, 1987; Chikoye et al., 1995). Weed density alone may not be a good predictor of crop yield loss. It does not account for weed size and con- sequently does not account for relative time of weed emergence, a factor known to be critical in determining the competitive effects of weeds (O'Donovan et al., 1985; Kropff et at., 1992; Knezevic^ra/., 1994). Cousens et d. (1987) incoiporated the influence of time of weed emergence into the hyperbolic damage function. Regression models which account for both weed density and time of