The objective of this paper is to investigate the effects of physical parameterization on the simulation of a snowfall event over Korea caused by air-mass transformation by using the PSU/NCAR MM5. A heavy snowfall event over Korea during 3-5 January 2003 is selected. In addition to the control experiments employing simple-ice microphysics scheme, MRF PBL scheme, and original surface layer process, three consequent physics sensitivity experiments are performed. Each experiment exchanges microphysics (Reisner Graupel), boundary layer (YSU PBL) schemes, and revised surface layer process with a reduced thermal roughness length for the control run. The control run reproduces an overall pattern of snowfall over Korea, but with a high bias by a factor of about 2. As revealed in the previous studies, the cloud microphysics and PBL parameterizations do not show a significant sensitivity for the case of snowfall. A more sophisticated cloud processes does not reveal a discernible effect on the simulated snowfall. Further, high bias in snowfall is exaggerated when a more realistic PBL scheme is employed. On the other hand, it is found that the revised surface layer process plays a role in improving the prediction of snowfall by reducing it. Thus, it is found that a realistic design of surface layer physics in mesoscale models is an important factor to the reduction of systematic bias of the snowfall over Korea that is caused by air-mass transformation over the Yellow sea.