A coupled pattern projection downscaling technique,using five Global Climate Models (GCMs) data sets from IPCC Fourth Assessment Report (IR4) and station observed precipitation data,was employed to project future precipitation scenario for Haihe River basin.5 model downscaling precipitation predictions and their average,referred as DMME,are first carried out and verified against station observations under cross-verification frame during the 40 year training period from 1960 to 1999.By comparison,a multi-model ensemble of the 5 GCM predicted rainfall anomalies which is called RMME,is also performed.Through coupled pattern downscaling method,multi-predictor optimal selection and multimodel ensemble technique (in Fig.1),DMME finally achieves very good prediction skills with the correlation coefficient significant over 99% confidence level for all stations.RMME,however,fails in predicting the precipitation at all of the stations,(shown in Fig.2)indicating the current GCMs are lack of direct predictability for local precipitation in the Haihe River basin.Moreover,DMME can simulate the change tendency of the observed precipitation in the past 40 years,implying that the current downscaling strategy has the potential ability to project climate change scenario for the Haihe River basin.For each station,using the best predictor selected in the verification procedure,through downscaling from the 5 GCM future 100 year climate scenario simulations of IR4,DMME predicts a decrease precipitation scenario for most stations and a slightly increase for a few stations in the northern part,at the end of this century.(Shown in Fig.3)