Daily snowfall events were identified from six first-order weather station observations: Sault Ste. Marie, MI (ANJ); Muskegon, MI (MKG); South Bend, IN (SBN); Syracuse, NY (SYR); Buffalo, NY (BUF); and Cleveland, OH (CLE). Snowfall events were categorized, using NARR reanalysis maps and radar imagery, as lake-effect, non lake-effect, or a combination of both. Preliminary results show that daily lake-effect snowfall events of greater than two inches make up 32 to 42% of all snowfalls during the seventeen year climatological study period. A small amount, 27 of 413, of events were unclassified.
To generate Cobb Method values for each event, BUFKIT data derived from 1-hr forecast NCEP (13 km) RUC data for the November-March period of 2008-2012 were used in conjunction with the Cobb Method. These data are derived from the nearest grid point to observations, within eight kilometers of one another. Snowfall produced by the Cobb Method generally underestimates daily snowfall observations from each location. The underestimation may be caused in part by the restriction of accumulating snowfall to hours in which greater than a trace of liquid precipitation occurred. The inability of the Cobb Method to produce extreme layer snow ratios of greater than 35 also limits the Cobb Method. The differences may be also be due to the resolution of the RUC NWP model, as the RUC grid spacing is somewhat coarse for depicting the mesoscale circulations of some lake-effect precipitation. The Cobb Method relies on the model's relative humidities, temperatures, and vertical velocities to produce precipitation which may not be parameterized correctly for lake-effect scale. Further research in the convective parameterization of the RUC model (or its predecessor RAP) and its possible effects on the Cobb output is needed.