Frost damage on infrastructure in seasonally frozen regions is mainly caused by the coupled water-heat transfer during freeze-thaw processes. Because of complex geological deposition and weathering, the properties of seasonally frozen soil are spatially variable. In this study, based on random field theory, heat transfer process, and frozen soil physics, a water-heat coupling model is developed to explore the impact of non-uniform thermal parameters on soil water-heat behavior. The statistical characteristics of the water-heat behavior and frozen depth of a slope are analyzed. The simulation results show that the water-heat coupling process of the soil exhibits obvious seasonal differences. The uncertainty in thermal conductivity has a greater effect on soil waterheat state than the uncertainty in volumetric heat capacity. The maximum frozen depth (MFD) from the traditional deterministic analysis is slightly smaller than the mean value of analysis result considering the nonuniformity of thermal parameters; as such, the deterministic analysis is likely to underestimate the MFD, which may result in local frost damage to infrastructure in cold regions. To ensure the safety of infrastructure in cold regions, the most unfavorable conditions need to be considered, and the upper bound of the MFD based on the random analysis can serve as the guideline for frost protection design.

来源平台：INTERNATIONAL JOURNAL OF HEAT AND MASS TRANSFER