Response of soil thermal conductivity to various soil properties

Soil thermal conductivity is an important property of soil that has various applications in simulation of land-surface water and energy processes, and geothermal energy calculations. In this study we conduct laboratory experiments to evaluate the impacts of various soil properties on soil thermal conductivity. The experiments included three soil types (i.e., sand, sandy loam, silt loam), four soil salinity levels, and a wide range of soil water content. To find the most effective factors we utilized different statistical methods (i.e., linear regression, information theoretic sensitivity index, Pearson's correlation coefficient, and random forest). Moreover, we developed and tested three soil thermal conductivity models (i.e., linear regression, random forest and quadratic equations). The results show that soil water content has the highest level of influence on thermal conductivity. Also, the quadratic relationship between thermal conductivity and independent variables provide the most accurate predictions among the models evaluated in this study.