The one-dimensional snow model SNTHERM is validated using field measurements of snow and superimposed ice thickness and surface energy fluxes. These were performed during the spring to summer transition in Svalbard and in the Weddell Sea. Both the seasonal snow thickness decrease and the formation of superimposed ice are well reproduced by the model. During the three observation periods, observed and modeled snow thickness differ only by 8.8 to 19.2 mm on average. In regional studies, the model is forced with atmospheric reanalysis data (ECMWF) and applied to several meridional transects across the Arctic and Southern Ocean. These show fundamental regional differences of the onset, duration, and magnitude of snow thinning in summer. In the central Arctic snow-melt onset occurs within a narrow time range of ±11 days and without significant regional differences. In contrast the snow cover on Antarctic sea ice begins to melt about 25 days earlier and the length of the Antarctic snow thinning season increases with geographic latitude. The importance of melting and evaporation for the modeled snow thickness decrease is very different in both hemispheres. The ratio of evaporated snow mass to melted snow mass per unit area are derived from the model and amount to approximately 4.2 in the Antarctic and is only 0.75 in the Arctic. This agrees with observations and model results of the surface energy balance, and illustrates the dominance of surface cooling by upward turbulent fluxes in the Antarctic.