Data was used from 1691 field trials testing herbicides to explore large-scale patterns in yield loss in cereals due to weeds in Sweden. More specifically, we evaluated the relative importance of differences between regions, crops, soils, and years. In a negative hyperbolic function, weed biomass explained 31% of the variation in yield loss due to weeds (calculated from yield in herbicide-treated control plots and in weedy control plots). Variation in the residuals was then partitioned between groups of categorical, environmental variables. Geographic region (8) and crops (5) accounted for most of this variation. Crops which ranked from the most to the least affected by weeds, were spring-barley, spring-wheat, oats, winter-wheat and rye. When adjusted for differing weed abundance, clay soil suffered the smallest and organogenic soil suffered the largest yield losses due to weeds. Differences between years were non-significant. The large unexplained variation in yield loss was the likely result of spatial heterogeneity of weeds, other pests and soil within trials. Considering this large variation, it might be difficult to combine in the same field trial, the two aims of most weed trials: estimating the beneficial/detrimental effect of a treatment and the species-wise responses towards it. We described the association between weed communities and situations with high or low losses with ordination methods. In this way, the weeds could be ranked from the most benign to the most detrimental for cereal yields in Sweden. The worst weed species in autumn-sown crops were Capsella bursa-pastoris and Matricaria perforata, and in spring-sown Polygonum spp. and Galeopsis spp. © 2003 Elsevier B.V. All rights reserved.
2004. Vol. 86, no 2-3, 199-209 p.