The land-surface parameterization of the Eta model was extended to account for frozen ground processes, and for snow compaction and patchy snow cover. Some constraints were on the new extensions, so that the added physical complexity and soil profile treatment was compatible with general complexity and configuration of the present model. Accordingly, the snow profile is assumed to be uniform, however, the percent snow cover can vary in space (sub- grid variability) as a linear function of the mean areal snow depth. The mean areal snow depth and density are estimated at each time step based on a snow compaction model that was derived as an approximate solution of the Anderson-Yosida's expression of snow density dynamics. Soil moisture and heat fluxes are simulated separately at each time step assuming that there is no significant heat transfer during redistribution of liquid water. Soil moisture phase transitions are calculated under the assumption that the soil water potential and vapor pressure over pure ice are in an equilibrium, and that phase transitions occur in the middle of each soil layer.
Snowmelt/rainfall infiltration is reduced depending on a total amount of ice in the soil profile. Off-line tests of the parameterization were performed using experimental data from the ARS Rosemount site in Minnesota, and the PILPS2d experiment data set from Valdai, Russia. A four-layer version of the model was used to represent a soil profile with the total depth of 2m. The simulation time step was one-half hour in both cases. Simulation results indicate that neglecting cold season processes leads to significant changes in heat and water fluxes of the upper soil layer. As a result, soil temperature during soil freezing/thawing periods can be underestimated or overestimated significantly. It can also lead to the reduction of the total soil moisture content after extensive periods of soil freezing. Although the total amount of an annual runoff from original and modified versions was close, the seasonal distribution of runoff and surface-subsurface runoff partitioning differed significantly.