Frozen soil and snow play an important role in the hydrologic cycle of cold regions. Ice in the soil can severely retard infiltration producing increased surface ponding and runoff. The redistribution of snow through blowing and preferential melt will significantly change the timing and rate of melt. Point models can adequately represent the accumulation and ablation of snow and the freezing and thawing of soils given local surface meteorological data and information about subsurface properties, but for application within large scale hydrologic models a method must be developed to represent the spatial variability of both processes. A field campaign was conducted at the University of Minnesota's Rosemount Agricultural Experiment Station during the winters of 1997-98 through 1999-2000 to collect information about the spatial distribution of snow depth and soil temperatures suitable for parameterizing the effects of spatial variability in these processes within a macroscale hydrologic model. Analysis of the data for the three winters shows that there is considerable variability in soil ice content and snow cover properties both between and within similar vegetation cover types. However, uniform probability distributions provide a good approximation to the probability distributions of both quantities. Parameters derived from the field study were used in the development of parameterizations for the spatial variability of snow cover and frozen soil properties in the Variable Infiltration Capacity (VIC) macroscale hydrologic model. Tests of the algorithms over the upper Mississippi River basin show that spatial variability tends to reduce the effect of soil freezing on large area runoff generation relative to inferences from point considerations alone. The representation of partial snow cover increases the exchange of sensible heat between the atmosphere and land surface during the course of the melt process, simulating a smoother transition between winter and spring than the original algorithm.