Extended Abstract

Table 4 The number of observations inside the 90% confidence intervalsBMA BABJ KWBC RJTDPercentage (%) 0.81 0.16 0.08 0.495. Summary and ConclusionsWe presented a BMA scheme developed by Raftery et al. [2005] and Sloughter etal. [2007] and applied it to generate multimodel ensemble QPFs for Fujian Provinceof the Southeastern China. With 2 month of training data, we estimated the BMAweights and BMA model parameters. Among the three NWPs, the BABJ model seemsto perform better than the other two models based on the training data. The spread ofRJTD model is very small, but poorly calibrated. The BMA ensemble QPFs are moreaccurate and better calibrated than the raw ensemble QPFs generated by the NWPs.This study suggests that BMA is a very promising statistical tool to post-processingthe raw ensemble forecasts.Acknowledgement: The authors acknowledge financial support provided by ChineseMinistry of Science and Technology 973 Research Program [Grant #: S2009041002].References:Duan, Q., N. K. Ajami, X. Gao, and S. Sorooshian (2007), Multi‐model ensemblehydrologic prediction using Bayesian model averaging, Adv. Water Resour., 30,1371–1386, doi:10.1016/j.advwatres.2006.11.014.Raftery, A. E., T. Gneiting, F. Balabdaoui, and M. Polakowski (2005), Using Bayesianmodel averaging to calibrate forecast ensembles, Mon. Weather Rev., 133, 1155–1174, doi:10.1175/MWR2906.1.Sloughter, J.M., A.E. Raftery, T. Gneiting, and C. Fraley (2007), ProbabilisticQuantitative Precipitation Forecasting Using Bayesian Model Averaging, Mon.Wea. Rev., 135, 3209-3220.Weigel, A.P., M.A. Linniger, and C. Appenzeller,(2008), Can multi-modelcombination really enhance the prediction skill of probabilistic ensembleforecasts? Quart. .J. Roy.Meteor. Soc., 134, 241-260.-437-