A Robust Post-processing Algorithm Based on Bayesian Model Averaging

- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner
Wednesday, 7 January 2015
Hong Guan, Systems Research Group Inc./EMC/NCEP/NOAA, College Park, MD; and Y. Zhu and B. Cui

A post-processing algorithm named as Recursive Bayesian Model Process (RBMP) has been developed to apply multi-model (ensemble) forecasts. The method is mainly based on Bayesian Model Averaging (BMA) (Raftery et al. 2005). We adopted the station-based BMA codes developed in MDL to global grid-based codes to calibrate probability distributions for each grid in the global ensemble forecasts. The major task for this method is calculating the weights for each model and a uniform variance based on most recent model performance. These parameters are recursively updated using the decaying averaging technique in order to reduce storage space in operational forecasting.

One of the most important advantages of the BMA is increasing spread and improving under-dispersive of ensemble forecasts. However, this could lead to an over-dispersive problem when the ensemble forecasts were originally not under-dispersive. Therefore we also apply a 2nd moment adjustment process to the BMA-calibrated forecasts.

The RBMP was applied to NUOPC forecasts of 2-m temperature for the summer and fall of 2013. NUOPC combines the three global ensemble forecast systems (GEFS) from the NWS, MSC, and FNMOC. The method is efficient which improves ensemble forecast skill for all lead time with a maximum improvement for short lead-time forecasts. The RBMP is being applied to 10-m wind forecasts and other variables. The result will be also presented.