Estimating precompression stress of structured soils on the basis of aggregate density and dry bulk density

In soil mechanics, precompression stress is an essential parameter for estimations of the compaction risk of cultivated land. In order to determine this factor, regression equations were developed. They require various input variables of water and air regime, dry bulk density as well as the shear strength parameters c and p. In this paper, we propose a regression model, which estimates the precompression stress from the two parameters dry bulk density (BD) and aggregate density (AD). The experiments were conducted on various structured arable soils in Germany. Altogether 25 natural soils and seven disturbed substrates were examined with three to seven replications. On all sites, precompression stress (log sigma(p)) was determined by means of stress-strain measurements under drained conditions and a matric potential of -6 kPa. The same samples were used for estimating the dry bulk density. Parallel to this, density measurements of aggregates with a diameter of 8-10 mm were made at a matric potential of -6 kPa. Aggregate density and dry bulk density were put into a relation (AD/BD ratio). This quotient shows the state of the inter-aggregate pore system and thus the load-support strength between the aggregates. A multiple linear regression equation of simple design allows to determine the level of precompression stress using the input variables AD/BD ratio and dry bulk density. Precompression stress rises with increasing dry bulk density. An increasing AD/BD ratio leads to a decline of precompression supposing the density values remain constant. The model produced good agreement with the measured values. The determination coefficient of the regression function was 0.84, the mean absolute error (MAE) 0.12 and the root mean square error (RMSE) 0.14. The index of agreement according to Willmot [Willmot, C.J., 1982. Some comments on the evaluation of model performance. Bull. Am. Meteorol. Soc. 63 (11), 1309-1313] was 0.95. (c) 2006 Elsevier B.V. All rights reserved.