A descriptive model of the quantitative evolution of the annual weed flora, the principles of which have been exposed in earlier work (DEBAEKE & SEBILLOTTE, 1988), has been applied to 3 broad-leaved weeds, common but with contrasting dynamics, coming from natural populations of the experimental area in Grignon (Yvelines, France). The parameter values used were derived from a survey of 64 plots, spread over 2 successive years and covering a large diversity of cropping systems, and were also taken from published results. The comparisons between predicted and observed seed banks, after 17 years of cropping on 30 ploughed plots, led to a satisfactory classification of the plots tested ; differences between observed and expected values were due to some seed imports, to bad estimation of recent seed production and to the hctcrogeneity of the initial seed banks. There was fairly good agreement in seed-bank annual growth rates simulated in Grignon and determined by sampling on an experimental field at Dijon. This confirms the value of the results on the long-term kinetics of seed banks and suggests that the empirical parameters used may be more generally applicable.