Abstract: The prediction skill of four climate models for monthly rainfall over Northeast China was evaluated using three qualitative evaluation methods,i.e.,anomaly correlation coefficient (ACC),trend anomaly inspection evaluation (Ps) and anomaly symbol consistency rate (Pc).Many data were used in this study,including 172 meteorological stations over Northeast China supplied by the National Meteorological Information Center,the hindcast experimental results of rainfall over Northeast China from 1983 to 2010 according to four climate models from China,America,Japan and Europe,and the operational application results over Northeast China from 2011 to 2014.The results indicate that the monthly rainfall prediction skills of EC (European Center for Medium-Range Weather Forecasts) and CFSv 2 (Coupled Forecast System Model Version 2) models are better than those of BCC (Beijing Climate Center) and TCC (Tokyo Climate Center) models.Looking at the spatial distribution,there is a significant difference in the distribution of each monthly Pc for CFSv 2 model,indicating that this model has a big space for its improvement.The CFSv 2 model has some prediction skills in early summer during typical drought and flood years,and the prediction effect in typical flood years is better than that in typical drought years.

Key words: Climate model, Monthly rainfall, Drought/Flood years, Prediction ability, Quantitative evaluation