Probabilistic Forecasts of Precipitation Type Based on Global Ensemble Forecasts

Accurate prediction of precipitation type is one of the major weather forecasting challenges during the cool season. Freezing rain is particularly dangerous; it can cause power outages by forming glaze ice and pose significant threats to air and ground traffic. The prediction of this precipitation type typically comes with substantial forecast uncertainty, thus making a case for probabilistic precipitation type forecasting.

We present a statistical post-processing method that uses multi-ensemble (GFS, GEFS, ECMWF deterministic and ensemble, CMC deterministic and ensemble) forecasts of (wetbulb) temperatures at the surface and several pressure levels as predictors for precipitation type probabilities. The different temperature ensemble member forecasts are calibrated, interpolated to full vertical wetbulb temperature profiles, and used as variables for a regularized discriminant analysis algorithm. In a second step, the individual probabilities from the different ensemble members are combined to a single probability forecast. The resulting forecasts are reliable and skillful, suggesting the transition to the test and evaluation team of the National Blend of Models Project for further evaluation and transition to operations.