Prediction of soil thermal conductivity based on multivariate probability distribution models

Abstract

Soil thermal conductivity (λ) is an important parameter for determining the thermal properties of rock and soil materials. In this study, multivariate probability distribution (MPD) models were established based on the factors influencing λ. The performance of the MPD models was evaluated by testing parameters and comparing them with the traditional empirical relationship model of λ to verify the effectiveness of the MPD models. According to the research results, MPD models can accurately predict λ. With the increase in influencing factors considered by the MPD models, the prediction accuracy significantly improved, the correlation coefficient (R2) increased from 0.7125 to 0.9248, the E(ε) value was reduced to 1.0208, and the COV(ε) value was reduced to 0.2336. Among the established MPD models, the performance of the λ-{w, ρd, n, Sr, c, sa, qc} model was the best, and the prediction accuracy of the MPD models was better than that of the traditional empirical relationship model. The results of this study suggest that different types of MPD model should be chosen to estimate the thermal conductivities of different types of soil with significant differences in engineering properties and complex sedimentary environments.