Sources of uncertainty in precipitation type determination and forecasting

Abstract: Three sources of uncertainty in determining the precipitation type, with particular emphasis on freezing rain, are considered. These are the uncertainty in the choice of 1) algorithm, 2) analysis system, and 3) observations. For part 1, five different precipitation type algorithms are considered using observed soundings. All are rather poor at detecting freezing rain, with hit rates ranging from 0.2 to 60%. Attempts to distinguish freezing rain from other forms of precipitation using metrics gleaned from the observed soundings proved unsuccessful. In part 2, RUC analyses were compared to observed soundings to determine a typical error distribution. In the freezing rain profiles, the error distribution was com- pletely random. Resampling the observed freezing rain profiles 1000 times to reflect this error distribution dramatically reduced the hit rate. In part 3, high-resolution observations of precipitation type are compared. These show that for observation pairs that are 5 km apart (the average radiosonde drift in the lower troposphere, the agreement is only 44%. For a 13-km distance, the agreement is only 35%. Similar statistics have been computed for snow, rain, and ice pellets and will be presented in this talk. Preliminary work addressing these shortcomings will also be presented.