A New Scheme of Calibration of Ensemble Forecast Products Based on Bayesian Processor of Output and Its Study Results for Temperature Prediction
Numerical weather prediction (NWP) techniques and ability are developing constantly and the ensemble prediction is a very important part of NWP. An appropriate interpretation process is needed to convey the mass information provided by ensemble prediction to users, thus the interpretation and application of ensemble prediction products are important to realize their utilitarian value. The 00:00 UTC surface temperature at Wuhan Station is selected as the predictand, and its historical observation data as well as NCEP 120 h ensemble prediction data from TIGGE data during January 2008 are used to establish BPO model for each NCEP ensemble member based on a statistical process technique, which is the Bayesian Processor of Output (BPO). The member Bayesian probabilistic forecast is obtained and the performance difference among members is studied. The member Bayesian probabilistic forecasts are integrated into an integrated Bayesian probabilistic forecast which quantifies the ensemble prediction uncertainty according to the weights depending on member Informativeness Scores. The analysis of initial experiment results shows that the performances of ensemble members are different from each other, so are the member Bayesian probabilistic forecast. This new interpretation scheme based on BPO can quantify the forecasting uncertainty of an ensemble prediction, and then a Bayesian integrated probabilistic forecast can be obtained.