During the winter of 2000-2001, the authors organized an experiment to evaluate a set of precipitation-type algorithms at the Hydrometeorological Prediction Center (HPC) and the Storm Prediction Center (SPC) that forecast either snow, rain, freezing rain, or sleet. The purpose of this study was to evaluate the quality of each algorithm, a consensus forecast, as well as a probabilistic forecast for each type of precipitation. Since previous research indicated that the forecast quality of each of these algorithms was good, it was believed that combining the output from these algorithms would produce high quality probabilistic and consensus forecasts. The consensus forecast was formed from a combination of the output for all the algorithms, and the conditional probability of precipitation type (conditional upon a forecast of precipitation) was determined by averaging the combined algorithm output. For example, the consensus forecast may be rain, whereas the conditional probability of rain is 75% (indicating that three-quarters of the algorithms forecast rain). The consensus forecast indicates the most likely type of precipitation, and the probability indicates the uncertainty associated with that forecast.

Although the authors are still in the midst of a complete evaluation, preliminary results show that the consensus forecast was more skillful than most of the algorithms for each type of precipitation. These results suggest that since one algorithm did not consistently outperform all the other ones for all precipitation types, the consensus forecast may be a better forecast tool than using output from any one algorithm. The probabilistic forecasts, however, exhibited low resolution and were unreliable. Further analysis will be completed to determine the shortcomings of the probabilistic forecasts and to identify any methods of improving these forecasts.