A simple model to estimate change in precompression stress as a function of water content on the basis of precompression stress at field capacity

Precompression stress is an important criterion in soil mechanics and is often determined at a water content equivalent to a matric potential of − 6 kPa. In German-speaking countries, this matric potential corresponds to field capacity. Yet in order to assess the risk of compaction in arable soils, it needs to be known for a wide range of soil water content levels. The site-specific determination of relationships between precompression stress and matric potential or water content is, however, highly labour intensive. Furthermore, previous regression models can only deduce changes in precompression stress depending on water content to a limited extent and not for all values. Alternatively, these models do not directly include precompression stress at a matric potential of − 6 kPa as the basis of calculation. Thus the derivation and validation of a simple model are to be presented, which can be used to predict any precompression stress for decreasing soil water content levels. This requires only an initial precompression stress for a matric potential of − 6 kPa and the respective soil water content as a percentage of field capacity. The model is based primarily on an analysis of numerous studies in which precompression stress was determined for various matric potentials. Relationships between precompression stress at a matric potential of − 6 kPa and the relative water content as a percentage of field capacity at a matric potential of − 30 kPa were also derived in the laboratory. These data were used to develop a mathematical model for four soil texture classes, as well as “All texture classes” collectively. This model was tested by way of soil compression tests and the determining of precompression stress at 25 sites. All soil compression tests were initially carried out with a matric potential of − 6 kPa. Tests were carried out in parallel to this with greater matric potentials (− 10 to − 1500 kPa). The accuracy of the modelling approach presented here is good, both in terms of the use of systems of equations for “All texture classes” and for differentiated soil texture classes. In comparison to the regression model for all texture classes, calculation according to soil texture class causes a reduction of the mean absolute errors from 0.15 to 0.11 and of the RMSE from 0.19 to 0.14. Simultaneously, the coefficient of determination and the index of agreement (IA) increase, from 0.54 to 0.67 and 0.92 to 0.95 respectively. Calculation according to different soil texture classes is therefore particularly recommended in the case of applications with high accuracy requirements.