Competition between maize and Echinochloa crus-galli analysed by a hyperbolic regression model

Yield reduction of maize in relation to naturally established populations of Echinochloa crus-galli and Chenopodium album was studied in field experiments over 2 years in which the maize was grown at a wide range of weed densities. Both the crop and the weeds were harvested at intervals during the season. The competitive relations were described accurately by a model based on a hyperbolic relation between yield and plant density. The model can be linearised by considering the reciprocals of the average weight per plant. However, estimating the regression coefficients by linear regression introduced a severe bias due to heterogeneity of variances. Estimation was improved by applying non-linear regression, using a logarithmic transform of the yield equation. Fitted regressions were used to interpolate the yield data to standard weed densities. At a density of 100 Echinochloa plants m-², maize yields were reduced by 8% and 82% in 1982 and 1983, respectively, illustrating the problems in generalising the results of one competition experiment to the other. Three possible fields of application of the competition model are discussed, i.e. adjustment of experimental plot yield for variation in weed population, prediction of expected crop yield losses, and prediction of long-term changes in weed seed populations.