Comparison of empirical models for predicting yield loss of irrigated rice (Oryza sativa) mixed with Echinochloa spp.

Crop yield losses due to weeds can be described by empirical models. The objective of this study was to compare empirical models to predict interference by a mixed population of Echinochloa colona and E. crusgalli. in irrigated rice. Three experiments, one under field conditions and two under greenhouse conditions, were set up during the growing season of 2005/2006. The treatments tested in the field were 6 rice cultivars: BRS Atalanta, IRGA 421, IRGA 416, IRGA 417, Avaxi; and BRS Fronteira and 10 Echinochloa spp. densities, naturally present in the soil seedbank under field conditions. The variables soil cover, shoot dry weight of the Echinochloa spp. and their relative mass were evaluated in relation to the rice cultivars. The greenhouse experiments were carried out both in monoculture and replacement series to evaluate shoot dry mass and leaf area production, both for the crop and the weed species. The data obtained for the variables were analysed using linear and non-linear regression models. The fitting of the data to the empirical models varied as a function of the rice cultivars and variables tested. Among the models describing crop yield loss as a function of the evaluated variables, the single-parameter empirical model presented a better prediction than the two-parameter model. For the two-parameter models, yield loss estimation was obtained by the evaluated variable. In general, shoot dry mass of the weed was a better predictor of irrigated rice yield loss than soil cover.