Canadian Field Soils IV: Modeling Thermal Conductivity at Dryness and Saturation

The thermal conductivity data of 40 Canadian soils at dryness \((\lambda _{\mathrm{dry}})\) and at full saturation \((\lambda _{\mathrm{sat}})\) were used to verify 13 predictive models, i.e., four mechanistic, four semi-empirical and five empirical equations. The performance of each model, for \(\lambda _{\mathrm{dry}}\) and \(\lambda _{\mathrm{sat}}\), was evaluated using a standard deviation (SD) formula. Among the mechanistic models applied to dry soils, the closest \(\lambda _{\mathrm{dry}}\) estimates were obtained by MaxRTCM \((\textit{SD} = \pm ~0.018\,\hbox { Wm}^{-1}\cdot \hbox {K}^{-1})\), followed by de Vries and a series-parallel model (\(\hbox {S-}{\vert }{\vert }\)). Among the semi-empirical equations (deVries-ave, Advanced Geometric Mean Model (A-GMM), Chaudhary and Bhandari (C–B) and Chen’s equation), the closest \(\lambda _{\mathrm{dry}}\) estimates were obtained by the C–B model \((\pm ~0.022\,\hbox { Wm}^{-1}\cdot \hbox {K}^{-1})\). Among the empirical equations, the top \(\lambda _{\mathrm{dry}}\) estimates were given by CDry-40 \((\pm ~0.021\,\hbox { Wm}^{-1}\cdot \hbox {K}^{-1}\) and \(\pm ~0.018\,\hbox { Wm}^{-1}\cdot \hbox {K}^{-1}\) for18-coarse and 22-fine soils, respectively). In addition, \(\lambda _{\mathrm{dry}}\) and \(\lambda _{\mathrm{sat}}\) models were applied to the \(\lambda _{\mathrm{sat}}\) database of 21 other soils. From all the models tested, only the maxRTCM and the CDry-40 models provided the closest \(\lambda _{\mathrm{dry}}\) estimates for the 40 Canadian soils as well as the 21 soils. The best \(\lambda _{\mathrm{sat}}\) estimates for the 40-Canadian soils and the 21 soils were given by the A-GMM and the \(\hbox {S-}{\vert }{\vert }\) model.