A Normalized Empirical Prediction Model of Soil Thermal Conductivity with Three Parameters

Soil thermal conductivity is a fundamental thermophysical property that characterizes the soil’s ability to conduct heat. It plays a critical role in applications such as geothermal energy development and thermal energy storage. However, existing prediction models for soil thermal conductivity often suffer from complex functional forms and difficulties in obtaining the required input parameters. To address these challenges, this investigation developed an empirical prediction model based on the relationship between soil saturation and thermal conductivity. The model’s performance was evaluated using the coefficient of determination (R²) and root mean square error (RMSE) as statistical metrics. The proposed model was compared with three theoretical models and two existing empirical models using both published datasets and laboratory measurements. Results showed that predicting the thermal conductivity of sandy soils is more challenging for classical model. Among the three empirical models evaluated, the new model consistently achieved R² values greater than 0.85 and RMSE values below 0.20 W·m⁻¹·k⁻¹ across all three datasets. This suggests that the new model offers lower predictive uncertainty and better adaptability to different soil types, providing a new approach for estimating soil thermal conductivity. It should be noted, however, that the new model was developed based on data from unfrozen mineral soils under room temperature conditions. In practical applications involving other soil types such as organic-rich, frozen, or contaminated soils, alternative predictive models may be more appropriate.