Soil compaction is a serious problem in modern agriculture. State-of-the-art soil compaction models derive from foundation engineering and consider soils as continuous and homogeneous media. However, agricultural soils are structured at different scales. Modelling mechanical strength as a probability-distribution function will account for anisotropy and scale-dependency of soil physical properties. Discrete elements models can integrate this information. This project evaluated the new approach, and hence contributed to more precise predictions, i.e. prevention of soil compaction.