The snow-cover model SNOWPACK was applied to the wet-snow areas of Japan. Simulated variations of snow type, snow depth and weight, profiles of snow density, temperature and liquid-water content were compared with snow-pit measurements. The snow-depth simulation during early winter agreed with the measurements, but the differences between the simulation and the measurements increased during the course of the melt season. These differences were caused by underestimation of the energy balance at the snow surface, mainly that regarding sensible-heat flux during the melt season. The underestimation was caused by the implicit numerical treatment of the heat-transport equation. Consistent with the underestimation of snowmelt, simulated metamorphosis of compacted particles into melt forms was slower than the change shown by the measurements, and faceted snow particles, which constitute a snow type not actually found in the study area, sometimes appeared in the model. The inaccurate melt treatment also influenced simulated densities, which were larger than the measurements at small densities, while they were smaller than the measurements at large densities. Greater accuracy was achieved when an empirical compressive viscosity formulation for wet snows in Japan was introduced. A new version of SNOWPACK, with an accurate treatment of melt processes, is available.