Numerous studies have quantified weed-related yield and economic losses in wheat cropping. This study evaluates results of the first four years on-farm experimentation in 94 winter wheat fields (47 farms) across three regions of Switzerland. A diversified six-year crop rotation (2020 to 2025) was implemented with reduced pesticide use (-75%), aiming at a maximum 10% yield loss compared with the national average. Each farm has one field conventionally cultivated, and another strictly restricting pesticides and implementing agroecological crop protection strategies. The objectives were to determine (i) how management influenced weed biomass and grain yield, and (ii) how weed biomass and grain yield interacted with two climatic variables, four soil-linked variables, and eight variables describing the used management. The analysis approach consisted of four steps. (1) The nonlinear Cousens model was fitted to yield against increasing weed biomass to cluster farms fitting the expected yield-reduction response (in), and farms above (upper) or below (lower) its 98% confidence limits. (2) Principal component analysis (PCA) grouped variables associated with weed biomass and yield, considering separately outcomes forin, upper andlower clusters. (3) Random Forest (RF) classification highlighted the variable importance at the three clusters, using a weed dry biomass < 50 g m-2 as indicator of poor weed pressure. (4) Kruskal-Wallis and Chi-square tests described the extent to which the most important variables influenced yield and weed biomass. Results showed a substantial number of farms not fitting the Cousens-in cluster, suggesting weeds are not the driving yield loss agent. Climatic data, nutrient input, and mechanical soil cultivation were the key variables explaining higher yields at poor weed pressure (i.e., upper cluster) and at a lesser degree, herbicide use. Growing degree days (GDD) and rainfall were the two decisive variables explaining low yields in the events of poor or no weed pressure (i.e., lower cluster). Reducing mechanical weeding activities was somewhat related to lower yield effects under poor weed pressure. Nevertheless, it is too early to draw definitive conclusions and recommendations, because the six-year rotation is undergoing and requires a similar analysis approach.
