Wednesday, 6 June 2018: 4:45 PM
Colorado B (Grand Hyatt Denver)
Due to their limited spatiotemporal resolution and other approximations, Numerical Weather Prediction (NWP) forecasts are fraught with lead-time dependent systematic errors. Typically, users are also presented with disjoint information from a number of NWP systems. In the past decades, numerous methods have been presented for the statistical calibration and combination of NWP forecasts.
In this study, we evaluate a recently developed statistical post-processing (SPP) algorithm called the Bayesian Processor of Ensemble (BPE, Krzysztofowicz et al. 2018). BPE is a theoretically based approach to the calibration and combination of forecast information from various sources. BPE uses the climatological distribution of the predictand as prior information, that it updates with independent information extracted from a user selected set of predictors (NWP high resolution and ensemble forecasts and the latest observation of the predictand).
The calibrated posterior continuous and quantile distribution functions (cdf and qdf), and probability density function (CDF) are expressed in the context of the prior distribution by two posterior moments. Calibrated forecast probabilities, quantiles, and ensemble members are derived from the three calibrated analytical distributions. BPE will be evaluated, and its performance will be compared with the operational Ensemble Kernel Density Model Output Statistics (EKDMOS, Veenhuis 2013) method for predicting 1-10 day lead time 2 m temperature forecasts at 70 sites over the CONUS. Both BPE and EKDMOS use ensemble forecast from the NCEP Global Ensemble Forecast System (GEFS) and the Canadian Meteorological Centre (CMC). The two systems will be compared in terms of their calibration (i.e., statistical reliability) and forecast skill (ie., statistical resolution).
- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner