Comparative analysis of nine categories of soil freezing characteristic curve models for simulating frost heave

The relationship between unfrozen water content and subzero temperature is defined as the soil freezing characteristic curve (SFCC), which is a fundamental relation for frozen soil and is an input parameter of great importance in numerically modeling frost heave. This study makes a systematic comparison of nine common SFCC models to evaluate their performances in simulating frost heave. The nine SFCC models are divided into empirical models and SWCC-based models, and two approaches are used for frost heave computation: the void ratio method, which quantifies soil volume changes, and the ice lens method, which accumulates ice lens growth over time. The results show that XU2001 (power function-based) ranks highest for its simplicity, deviation control within 2 mm, and a correlation coefficient above 80 %, making it ideal for empirical simulations. AN1973 (a hybrid approach integrating multiple function types) follows, with MK2007 (exponential function-based) and JA1977 (piecewise linear function-based) showing moderate performance. TI1976 (power function with ice content) ranks lowest due to parameter sensitivity. Compared to empirical SFCC models, SWCC-based models determine parameters solely through test data rather than soil properties, resulting in greater stability and higher accuracy. The average evaluation metrics show that the frost heave computed by the void ratio approach can better reflect the changing trend of frost heave, but the frost heave computed by the integration of the ice lens thickness approach is more stable in the calculation. The results also indicate that the frost heave evaluated by the approach of integrating ice lens thickness is more accurate than that by analyzing void ratio for the soil with a lower permeability coefficient like clay, while they have little difference for the silty and sandy soil that has a higher permeability coefficient.