A Statistical and Approach to Improve Predictability of Cold Season Precipitation Type and Amount for the HRRR-TLE

In this study we are exploring methods to improve the predictability of cold season precipitation type and amount for the operational High-Resolution Rapid Refresh Model (HRRR) Time-Lagged Ensemble (TLE) through application of unique statistical techniques to the model forecast. TLEs are a computationally efficient method to provide a slightly improved probabilistic forecast as the differences between model runs are usually enough to represent initial condition uncertainty. In addition to categorizing each of the HRRR’s grid point locations to “buddy check” all points with nearby points of the same classification, we also apply a decaying average bias correction to all grid points to obtain a forecast PDF. Next, we use a post processing technique known as Bayesian Model Combination (BMC) to optimize the weighting of ensemble model members. Using all of these methods in unison we are able to determine enhanced probabilistic information for both the areal distribution of cold season precipitation and the best timing and location for a transitional line to occur.