Currently, high resolution mesoscale models, such as MM5, are being tested for real-time operations of short term forecast of snowfall in the terminal area. Previous studies have shown that MM5 has limited skill in forecasting the occurrence of snowstorms. Key issues are the timing, duration, and amount of snowfall predicted by MM5. One way to improve the MM5 forecast of snowfall is to assimilate radar observations into the model initial conditions.

Techniques that can effectively assimilate radar data into MM5 are explored in this work. To be applicable to real-time operations, the less expensive method, Newtonian Relaxation (nudging), is emphasized. The first step of the nudging method is to obtain 3-dimensional analyses of cloud, precipitate and temperature from radar data as well as other available observations. Then the analyses are used to initialize MM5 forecasts in a four-dimensional data assimilation (FDDA) manner (nudging). A case study has been conducted. Preliminary results of the radar analysis and its impact on MM5 forecasts will be presented in the paper. Results from the nudging method will be compared with that from a previous study using MM5-4DVAR.