Abstract:
Objective To comprehensively consider various factors through evaluating prediction models, make full use of each model’s advantages and disadvantages within the scope of applicable conditions, play its advantages, acquire concise, fast and accurate prediction of soil thermal conductivity and realize quantitative research on its complexity degree.
Method The advantages, disadvantages, application conditions and influencing factors of the previous 16 soil thermal conductivity models are analyzed and summarized. The predicted data of 14 models are compared with their measured data collected from the literature. The model evaluation is realized through linear regression analysis and root mean square error analysis.
Result Soil thermal conductivity is greatly affected by moisture content and quartz content. The thermal conductivity of quartz is about 7.9 W·m-1·K-1, which is the highest in all soil minerals. The thermal conductivity of soil in humid state is much higher than that in dry state.Under normal temperature condition, the regression coefficients of Wiener model are 0.133 and 2.208, and the decision coefficients are 0.393 and 0.820, which deviates significantly from other models; Geo-Mean model shows the lowest regression coefficient of 0.668 and the highest root mean square error of 0.598, the prediction values deviated significantly from the measured values; The regression coefficients of the models of Zhang et al, Chen and Haigh are 0.994, 0.919, 0.891 respectively, and the root mean square errors are 0.280, 0.315, 0.394 respectively, showing relatively high prediction accuracy.The regression coefficient of the model of Lu et al is 0.850, the determination coefficient is 0.976, the prediction accuracy of soil thermal conductivity is general, while the improved model of Su et al based on model of Lu et al shows the highest regression coefficient of 0.997, the highest determination coefficient of 0.980, showing the best performance.
Conclusion In the case of soil texture, improved model of Lu et al is recommended. This model can describe the effects of basic parameters of soil physics on soil thermal conductivity in more detail.