Prediction of soil solid thermal conductivity from soil separates and organic matter content: computational micromechanics approach

Evaluation of soil thermal conductivity at a given saturation usually requires the value of conductivity in the saturated state, which is commonly estimated by the geometric mean from the conductivities of water and soil solids (lambda(s)). Therefore, an accurate estimate of lambda(s) has a marked effect on soil conductivity within the full range of saturation. In this research, we modelled the soil solid as a multi-phase medium comprising individual isotropic phases with conductivities drawn randomly from an assumed probability density function. The variation in the thermal conductivity of minerals is associated with the corresponding soil separates: clay, silt and sand. A computational micromechanics approach was used to evaluate the homogenized thermal conductivity of the soil solids with different fractions of soil separates and organic matter contents. A new formula to predict lambda(s) is proposed by fitting a four parameter, non-linear function to the numerical results. Thirty experimental results of soil conductivities at saturation (our investigations and literature data) were used to validate the proposed model. We show that the new approach improves analytical predictions in terms of the root mean square error and bias. The proposed model to estimate lambda(s) is suitable for practical purposes because it requires information on the fraction of sand and organic matter contents only.