Accurate weed population estimation and yield loss prediction are important components of integrated weed management. Field experiments using Italian ryegrass as a weed in broccoli were conducted from 1994 to 1997 to compare weed density to other methods of weed population estimation, to evaluate the performance of weed population estimates in yield description models, and to study the affect of environmental variability on the predictive ability of models. A strong linear relationship was obtained between Italian ryegrass density and direct leaf area (r
2 = 0.60 to 0.99). For Italian ryegrass, density and estimates of canopy from the optical device (crosswire device) had a hyperbolic relationship with high coefficients of determination (r
2 > 0.72). Both direct leaf area and canopy estimates described broccoli yield as well as or better than Italian ryegrass density. The Li-Cor LAI-2000 Plant Canopy Analyzer (PCA) provided poor estimates of Italian ryegrass population (r
2 from 0.00 to 0.63) that failed to describe broccoli yield. No relationship was observed between estimates of light interception through the plant canopy obtained with the Li-Cor LI-191-S Line Quantum Sensor (LQS) and either Italian ryegrass density or broccoli yield. The low performance of the PCA and lack of performance of the LQS were likely due to the smaller size of the plants and larger gaps in the plant canopy caused by wide bed spacing. At similar densities, Italian ryegrass competition with broccoli was stable from year to year. Under high Italian ryegrass density, water supply affected competition. This may limit construction of robust yield prediction models, especially in areas where water is mainly from rainfall.