Accurate data of thermal conductivity are required in many agricultural, meteorological and engineering applications. New regression equations for predicting thermal conductivity based on easily measured quantities such as penetration resistance and water content or air-filled porosity are presented. The thermal conductivities from the equations are compared with those from a statistical-physical model of a good estimation capability. The measurements of the quantities were done on silt loam in a sloping vineyard (Italy) at various times and locations to get a wide range of measured values. It is shown that the performance of the equations relating the thermal conductivity with penetration resistance and air-filled porosity is greater (R-2=0.94) than with penetration resistance and volumetric water content (R-2=0.77). Therefore, the equations based on measured penetration resistance and air-filled porosity are recommended for predicting the thermal conductivity of the soil. Adding sand content and transformation of strength values to root squares somewhat improved the predictions. To minimize the effects of spatial variability of the measured quantities on the thermal conductivity and to reduce measurement time and soil disturbance, systems for combined measurements of penetration resistance and water content at the same place need to be used in further studies. (c) 2006 Elsevier Ltd. All rights reserved.