LIMITS OF A SIMPLE-MODEL TO PREDICT YIELD LOSSES IN MAIZE

In order to estimate yearly variations and accuracy of prediction of m aize yield losses due to weeds, we used supplementary maize individual s that mimic weeds and reduce the experimental variability. Biomass de crease at various growth stages and grain yield losses of maize were i nterpreted through regression analyses in terms of density and biomass of supplementary individuals. Comparison of losses due to weeds and s upplementary maize does not allow the calculation of equivalent weed-m aize density or biomass at the same competitive effect. In most cases regression curves obtained in maize-maize experiments were different f rom one year to another. The use of results from several trials, previ ously published, on the effect of plant spacing on maize yield over a number of years also showed a large variation in predicted values, whi ch confirmed the inaccuracy of an empirical predictive model using wee d density and biomass.