Precipitation type prediction is one of the more difficult problems facing the NWS and private forecasters during the cold season. An early-season snow and ice storm that occurred in the southern Plains, Southeast, and Mid-Atlantic states from December 3 through 5, 2002 was no exception. While the storm was generally well-predicted by the NCEP NWP models several days in advance, some interesting problems with precipitation type cropped up in areas on the borderline between rain and snow.
Until recently, grid-scale precipitation algorithms have not been sophisticated enough to track precipitation as a forecast variable or 'tracer'. Both precipitation amount and type were diagnosed from the NWP model temperature and moisture fields, with precipitation assumed to occur instantaneously. When the Eta-12 was implemented in late November 2001, a new grid-scale precipitation scheme was implemented that would predict and advect precipitation, and determine precipitation type in each grid box layer, taking into account a number of complex microphysical processes, including interactions among four cloud and precipitation hydrometeor types, fall velocities for differing precipitation types, and so on. More details can be found in the background information in the COMET link found at the end of the menu on the left.
Among the grid-scale forecast variables at each level (including the ground) is a fraction of frozen precipitation falling out of the layer above. It would be potentially useful to examine how well this variable does at predicting precipitation type at the surface. Possible advantages of using a forecast versus diagnosed precipitation type include, but are not limited to:
This short case examines how the predicted fraction of frozen precipitation performed compared to the NCEP precipitation type algorithm, and explains the reasons for the differences in performance. You are encouraged to learn about other precipitation type algorithms (see link at the bottom of the menu, for example) and think about how these other algorithms might have performed in this case as well.
It should be noted here that the NCEP precipitation type algorithm was tuned specifically not to underpredict freezing rain, as this was considered to be the most hazardous winter precipitation type.