It is widely agreed that competition is an important factor that regulates plant populations and shapes communities in agricultural landscapes. Many studies have suggested that crop and grassland competition can be used for cost-effective sustainable weed control. However, effective weed management requires a precise knowledge of the effects of agronomic practices and there is a lack of quantitative indicators to compare and predict the success of weed control by competition. Here, we studied weed abundance dynamics over a 12-year period in crop-grassland rotations (rotation treatments consisted of maize, wheat and barley crops, alternating with temporary grassland maintained for three or six years in the rotation and fertilised with similar to 30 or similar to 230 kg ha(-1) year(-1) of nitrogen). In addition to classical statistical analysis of the different rotation treatments, we modelled weed abundance as a function of the crop and grassland competition, expressed here by biomasses harvested in the preceding years. We show that weed abundance decreases over the years in grassland and subsequent crops only if the grassland receives sufficient nitrogen fertiliser. Our model had a greater explanatory power than the rotation treatments. This model estimates a critical biomass level above which weeds are suppressed in subsequent years, and below which they tend to thrive. This critical biomass level was 24.3 and 4.7 tonnes ha(-1) of dry matter for crops and grassland, respectively, highlighting the greater competitiveness of grasslands than of crops. Several clear differences between weed functional groups emerged. This new modelling approach directly links the interannual dynamics of weed populations to current and previous biomass production levels. This approach facilitates the development of environment-friendly weed management strategies and paves the way for comparisons of the competitiveness against weeds of crops and grassland under various pedoclimatic conditions and agronomic practices.