To investigate the heat transfer characteristics of different soils, a non steady thermal probe technology was used to measure thermal resistivity of typical soils in Nanjing area with different moisture content and dry density. The effect of moisture content, dry density, saturation, mineral composition and other factors on thermal resistivity were studied and the generalized thermal conductivity prediction models for soils at different states were established. The correction factors for prediction model to estimate thermal resistivity of both Nanjing and India soils were proposed. The test results indicate that thermal resistivity of soil decreases with the increase of moisture content and dry density and trends to the constant value when moisture content exceeds a certain value. The relationship between thermal resistivity and saturation is similar to that of moisture content. Thermal conductivity of soil particle is determined by itsmineral composition and the fractions of quartz have a significant effect on the heat transfer characteristics of particles. Geometric mean method can be used to estimate the thermal conductivity of saturated soils. A modified normalized model for predicting thermal conductivity of unsaturated soils at unfrozen state is proposed, and the prediction results for sandy soil are of high accuracy, but for fine particle soils are needed to be adjusted for considering the difference in soil deposited areas.