A system has been developed for predicting snow distribution in a built-up environment. This system couples the mesoscale meteorological model, which gives precipitation including snowfall in an area, and Computational Fluid Dynamics (CFD) models, which predict snow phenomena in building scale. The system's performance is tested and examined by comparing the predicted data with full-scale measurement. The snowdrift patterns, i.e., erosion around the upwind corners and deposition in front of and behind a building, obtained from the present model shows a good correspondence with those obtained from field observation.