Sensitivity of Soil Frost Models to Snow Cover and Density
Brenton Sharratt1
Frozen soils have important implications for the social well-being of those who live in cold regions.
Loss of sediment and nutrients from landscapes during snowmelt, for example, depend upon the
occurrence and depth of soil freezing. Soil frost models, therefore, may be useful in assessing the
impact of land management on soil erosion and water quality in cold regions. There are, however,
few independent evaluations of soil frost models. Four soil frost models were examined for their
ability to simulate snow cover, water content, and frost depth of agricultural soils. Snow cover, soil
water content, and frost depth were measured in plots varying in corn residue cover and height over
two winters in west central Minnesota. These measurements were compared to those simulated by
the SHAW, SOIL, Benoit, and Gusev models. The SHAW and SOIL models use finite difference
equations to predict temperature and water profiles in the soil. The Benoit and Gusev models esti-
mate soil frost depth by balancing heat flow through the snow-residue-soil system.
Snow cover is simulated by the SHAW, SOIL, and Benoit models and only the Benoit model mimics
snowdrift. The SHAW and SOIL models performed better at simulating snow dynamics than the
other two models. The SHAW model was more accurate in estimating soil frost depth than the SOIL
model, partly due to a better simulation of snow cover. The SOIL model, however, mimicked chang-
es in surface soil water content better than the SHAW model. The Benoit model performed poorly in
estimating snow cover due largely to a rapid snowmelt. The Benoit model, using measured and not
simulated snow depths, performed better than the Gusev model in estimating the depth of soil freez-
ing, but was more sensitive to changes in snow density. An increase in snow density from 250 to 350
kg m3 in the Benoit model caused a 150% deeper penetration of soil frost whereas this same in-
crease in the Gusev model resulted in a 20% deeper penetration of frost. This study suggests that
accurately simulating snow cover in soil frost models is critical to their performance in the field.
1USDA-Agricultural
Research Service, 803 Iowa Avenue, Morris, Minnesota 56267 USA
54