Research progress of soil thermal conductivity and its predictive models

Abstract

Thermal conductivity is one of the significant thermal parameters for estimating heat flux and thermal regime of soil in cold regions. This paper summarizes the test methods of thermal conductivity for a variety of soils, analyzes the influence of different factors on thermal conductivity and its mechanism, and evaluates the applicability of different predictive models of thermal conductivity. The transient and steady-state methods are two main test methods of thermal conductivity, and the transient method is mostly applied to measure soil thermal conductivity due to the short test time. Moreover, the test of the thermal conductivity for the remolded soil indoor is more common. Various factors affect thermal conductivity, mainly including the inherent and environmental factors. And compared with other factors, quartz content and moisture content have more significant effect on thermal conductivity. There are many theoretical, empirical and other predictive models of thermal conductivity, but few unified predictive models of thermal conductivity for various types and states of soils. However, machine learning (ML) algorithms are applicable to different types and states of soils, and which have relatively higher predictive accuracy than other common models. Future research should focus on the thermal conductivity for the coarse-grained and undisturbed soils, especially for the soils with the effect of freeze-thaw, and the predictive models considering the influence of phase change, unfrozen water content and frost heave on the thermal conductivity for the coarse-grained and undisturbed soils. This research can provide significant information for scholars to further study the variation and predictive models of thermal conductivity.