We evaluated hydraulic properties of a frozen soil based on the observed temperature, water content, and pressure head near the freezing front in a one-dimensional freezing column experiment for an unsaturated silty loam soil. Amount of unfrozen water near the freezing front was more than equilibrium amount of water according to the retention curve at the equivalent pressure head because freezing was a slow process. We proposed a modified-&theta model to describe nonequilibrium amount of unfrozen water in the retention curve. Coupled heat and water flow equations were simulated with a modified-&theta model as well as two other existing hydraulic property models: the classical Harlan model, and the modified-<i>K</i> model. The Harlan model overestimated water flow from the unfrozen region to the frozen region. Although the modified-<i>K</i> model well simulated water flow in a frozen soil, calculated pressure heads and liquid water content showed unrealistic changes at the freezing front due to the extremely large decrease in the hydraulic conductivity. The modified-&theta model agreed well with observed liquid water contents and pressure heads in a frozen soil. The unsaturated hydraulic conductivity of a frozen soil based on the modified-&theta model would be more physically feasible than other two hydraulic property models.