BMA probabilistic forecasting of the 500hPa geopotential height over Northern Hemisphere using TIGGE multimodel ensemble forecasts

Abstract

Based on the TIGGE multimodel ensemble forecasts taken from ECMWF, NCEP and UKMO, Bayesian Model Averaging (BMA) probabilistic forecasts of the 500hPa geopotential height over Northern Hemisphere with lead times of 1-15 days were conducted. A normal distribution was applied to the geopotential height. Comparison between the original ensemble forecasts and BMA forecasts suggests that the BMA forecasts perform better than the original ensemble forecasts for all lead times. The Brier Skill Score (BSS) of the BMA probabilistic forecasting shows that the BMA probabilistic forecasts are superior to the climatology forecasts. The original ensemble forecasts had a negative deviation, but the BMA prediction PDF was much better calibrated. Generally, the BMA method has significantly improved the quality of 500hPa geopotential height probabilistic forecasts.