Probabilistic Forecasts Based on Statistically-Improved Forecasts

Forecast systems such as the Dynamic Integrated ForeCast system (DICast) produce statistically enhanced forecasts that could be used as the basis for passed to a probabilistic forecast. Probabilistic forecast techniques, such as Bayesian Model Averaging (BMA), typically use bias-corrected raw model data from each ensemble member. However, other calibration techniques, such as statistical post-processing of raw model data, are typicallyoften better than simple bias correction. This talk will present results of a study that compares whether statistically enhanced inputs to a probabilistic forecast process actually result in improved probabilistic forecasts.