Using the Arctic regional climate system model (ARCSYM), we investigate the spring seasonal transition and mechanisms controlling snowmelt over a domain covering the northern half of Alaska. Annual simulations for 1992 comparing the Biosphere-Atmosphere Transfer Scheme (BATS) and the land surface model scheme (LSM) show that the BATS experiment enters the spring transition with respect to the large-scale atmospheric regime approximately one month earlier than observed climate and the LSM experiment transitions a month later than observed, even though the air temperature in the LSM experiment is generally warmer than in the BATS experiment. A more detailed examination reveals that each simulation commences and completes the snowmelt period at about the same time but that the LSM snowmelt is more rapid than in the BATS experiment. Controlling the snowmelt is the initial snowpack depth and the surface energy budget, both of which involve a complex series of feedbacks between shortwave and longwave radiation, cloud, surface turbulent fluxes, and vegetation. The snowmelt over tundra regions dominates the more rapid snowmelt seen in the LSM simulation. It is determined that the most crucial differences between the BATS and the LSM schemes are the partitioning of net ground heat flux between patches of snow and bare ground and the formulation of snow albedo. ¿ 1998 American Geophysical Union