预见期长达14天的中期降水预报对于旱涝预测与水资源调度管理等具有重要的意义，全球多模式集合预报（TIGGE）资料为区域中期降水预报提供了新基础。基于2008年－2013年江苏省实测降水数据，采用TS 、BS评分、平均偏差和均方根误差等评估指标，重点评估了CMA、CMC、ECMWF、JMA和NCEP这五个TIGGE代表模式在江苏省的中期预报效果。结果表明：整个预见期内各模式对于中小雨的预报评分都较高，小雨的TS评分值最高多在0.5左右，中雨以上等级降水存在较明显的漏报。预见期长达15天的日累积降水量预报只存在三天的衰减期，之后各项误差指标值能维持稳定，3天以上预见期的误差主要为预报值偏大。同时，5天以上累积降水量预报也较实际偏大，且冬季预报效果最好夏季最差，苏北地区好于苏南地区。就各代表模式而言，ECMWF与JMA模式分别在降水分级预报和日累积降水量预报中表现最好。

The 14-day medium-range precipitation forecast is important for drought and flood forecasting and water resources scheduling management. TIGGE (THORPEX Interactive Grand Global Ensemble) provides basic data for regional medium-range ensemble precipitation prediction. Based on the ensemble forecasts of CMA, CMC, ECMWF, JMA and NCEP in the TIGGE datasets in Jiangsu Province, we evaluated their medium-range precipitation forecast results. We mainly used such evaluation methods as the TS, BS, mean absolute deviation, and mean square root error. The results showed that the forecast scores of light and moderate rain during the entire forecast period were all relatively high. The light rain had the highest TS (mostly around 0.5) statistically. However, there were obvious missing reports of precipitations above the moderate level. For the 15-day forecast of daily cumulative rainfall, there was only a 3-day decay period, after which all the error indicators remained stable. The prediction bias in a lead time of over 3 days was mainly overprediction. Meanwhile, the cumulative rainfall forecast of over 5 days was significantly larger than the actual value. The forecast effect was the best in winter and the worst in summer, and the effect in northern Jiangsu was better than that in southern Jiangsu. Across the five representative modes, the ECMWF and JMA modes respectively showed the best performance in rainfall categorical forecast and daily cumulative rainfall forecast.

国家自然科学基金（51579065）；水利部公益性行业科研专项经费项目（201301040；201301070）；全国优秀博士学位论文作者专项资金资助项目(201161)；教育部新世纪优秀人才支持计划（NCET-12-0842）