The timing, duration, and intensity of the near-surface soil freeze-thaw status is important to hydrology, climate, and agriculture. Many land-surface models have explicitly included the frozen soil processes, but most of these models assumed soil water freezing point at zero degree C. For pure water at atmospheric pressure, this assumption is valid. This may not be valid for the soil water because there may be dissolved salts. This paper takes the community land model (CLM) as an example to provide an in-depth evaluation of the impacts of the freezing point on soil freeze-thaw cycles and runoff. This work is useful to a wide community because CLM has been accepted as a standard land surface model in the coupled climate system model (CCSM) in the National Center for Atmospheric Research (NCAR). CLM is being used also in the NASA's Land Data Assimilation System (LDAS).

The model is evaluated by comparing predicted and observed streamflow for a 20-year period (1979-1998) using data from the Torne/Kalix basins in Sweden. The same dataset has been used in the phase 2(e) of the Project for Intercomparison of Land-surface Parameterization Schemes (PILPS). In the Lansjarv catchment, the snowmelt-induced runoff is too rapid and too large in the original CLM. With the freezing point changing from zero degree C to -2 degree C, guided by observations, the ice fraction in the surface soil layers is reduced significantly, the surface is more permeable, and the streamflow is more accurately simulated. This lowering of the freezing point also has a favorable cooling trend in the soil layers, because the original CLM soil temperature has a warm bias under snow conditions. Larger improvements in the simulations occur in Lansjarv than Abiskojokk because the former has shallower snowpack conditions which correspond to greater freezing depth, larger ice fraction, and colder surface soil layers.