A hyperbolic model for the thermal conductivity of freezing soils

Thermal conductivity is a key parameter characterizing heat and water transfer in soils. It is difficult to measure thermal conductivity under freezing conditions. For this reason, some models have been proposed to estimate the thermal conductivity of freezing soils, but most of them failed to mimic the variation of thermal conductivity with temperature. In this study, a hyperbolic model with three parameters (λ(0°C), α and β) was proposed to model the relationship between thermal conductivity and subzero temperature. The parameters λ(0°C), α and β of the hyperbolic model were estimated via artificial neural networks. The performance of the new model and 5 widely used models was evaluated with measured thermal conductivities. The results show that the new model accurately estimates the thermal conductivity of soils during a freezing process. In addition, the proposed model performs the best among the 6 models, as indicated by RMSE (0.105 Wm−1K−1). The new model accurately mimics the variation of thermal conductivity with subzero temperature and can be incorporated into numerical algorithms for coupled heat and mass transfer in cold regions.