Abstract Data from five 3‐year experiments comparing different strategies of non‐chemical weed control were analysed in order to set up a model that could be used for simulating further strategies. The treatments included different frequencies and effects of weed control using flaming, steaming and weed brushing. Each plot consisted of about 100 m of sidewalk. The amount of weed on the sidewalks was recorded several times during the 3‐year period using a system of cameras mounted on a Terrain Vehicle moving on the sidewalk and post processing using image analyses of extracted images. The percentage of sidewalk covered by weed was estimated on the basis of the image analyses. The model used for describing the weed coverage comprises three components. The first component describes the growth of weeds; the second describes reduction caused by wearing and decay of weeds in the autumn and the third describes the reduction caused by weed control using weed brushing, steaming and flaming. The parameters of the model were estimated from the data for later use for simulating additional strategies. Copyright © 2004 Society of Chemical Industry