To study the variation pattern of snow cover in seasonal frozen soil areas and the effect of related meteorological factors on the snow cover temperature, four land cover patterns (bare land, normal snow, compacted snow, and thickened snow) were created using a field test of a freezing and thawing period. The thermal state of the snow cover and its upper and lower boundary conditions were determined and combined with ground meteorological observation data, and the effect of the thermal state on the upper and lower boundaries of the snow cover was analyzed via statistical analysis and gray correlation analysis. The main meteorological factors affecting the snow cover temperature were selected, and the relationships between the stratification temperature and main meteorological factors were determined. The results show that the differences in snow melting time for different land cover patterns are small and that the snow density and snow depth have only slight effects on the melting rate. The variability of the heat fluxes of the snow layers was above the moderate degree, and the variability of snow from the surface layer to the bottom layer had an increasing-decreasing tendency. The heat exchange at the snow-frozen soil interface during the stable period was weak, and solar radiation could effectively influence the heat flux of 10 to 20 cm of snow. In the freezing period, snow cover treatment compared with bare land can significantly improved the soil temperature and increased the heat preservation effect in the following order: thickened snow > compacted snow > normal snow. When snowfall is certain, compacted snow treatment can increase the net radiation transmission within the snow layer, improve the temperature of the snow layer, and increase the heat preservation effect of the soil. The increase in snow density can increase the internal temperature variation of the snow and significantly change the snow thermal conditions. The regression relationship between the surface of the snow cover and the main meteorological factors was notable, while the model reliability for the middle and bottom layers was poor. [ABSTRACT FROM AUTHOR]