Frozen soil thermal conductivity (FSTC, λeff) is a critical thermophysical property that is required for a variety of science and engineering applications. Measurements of λeff in frozen soils are prone to errors because the currently available thermal methods result in phase change (i.e., thawing and refreezing of ice) that affects the estimated λeff, especially near the freezing point of soil water (e.g., −4 to 0 °C) where rapid phase change occurs. In addition, measured λeff data are few compared to other physical properties. Therefore, many FSTC algorithms have been developed and a few of them have been incorporated in numerical simulation programs for calculating λeff. However, large discrepancies between simulated and observed soil thermal regimes have been reported. Previous studies either evaluated the performance of a few FSTC algorithms with limited λeff or simply compared the performance of the algorithms in numerical simulation programs. No study has been performed to systematically assess the performance of the FSTC algorithms included in numerical simulation programs with both observations and model simulations. In this study, 14 FSTC algorithms incorporated in various numerical simulation programs were evaluated with a compiled dataset consisting of 331 λeff measurements on 27 soils from seven studies made at temperatures on or below −4 °C. The Becker 1992 algorithm provided the best estimates of the λeff measurements, but the accuracy of the estimates was not good (i.e., RMSE = 0.46 W m−1 °C−1, Bias = -0.04 W m−1 °C−1 and NSE = 0.51). These FSTC algorithms were also incorporated in the Simultaneous Heat and Water (SHAW) model to compare their effects on the simulating soil temperature and water content at two field sites with contrasting soil textures in USA. The simulation results showed that the average bias of simulated and observed soil temperature for all depths ranged from −2.2 to 2.8 °C and the average differences of liquid water content ranged from −0.08 to 0.1 cm3 cm−3. Generally no FSTC algorithm combined with the SHAW model satisfactorily estimated the dynamic soil thermal regime. Perspectives on future studies are discussed.