The average bias of T639 model forecast fields for 1—10 d forecast in 2009—2010 was evaluated. Forecast fields were included 500 hPa geopotential height,850 hPa temperature and 2 m temperature.The result showed that T639 model has obviously forecast system errors for the three fields.The decaying averaging method was trying to correct forecast system errors.Corrected result showed that the decaying averaging method has positive correction skill,but correction skill was decreased with lead time.System errors of East Asia were less than the Northern Hemisphere and the correction skill of decaying averaging method for East Asia was also less than the Northern Hemisphere.Through evaluating correction skill of cold season and warm season respectively it is found that the correction skill has no obvious difference for 500 hPa geopotential height and 850 hPa temperature,but for 2 m temperature,the correction skill of cold season was higher than warm season.In addition,the test on different weight of ＂decaying averaging method＂ showed that the correction skill was best when weight was round about 0.1 for 850 hPa temperature and 2 m temperature and round about 0.06 for 500 hPa geopotential height.%通过统计2009—2010年T639模式500 hPa高度、850 hPa温度和2 m温度的1～10天预报场的平均误差发现,T639模式的这些气象要素预报都存在明显系统误差,且系统误差随着预报时效的增加而增加。利用＂递减平均法＂尝试订正其预报系统误差,订正结果表明：该订正方法总体表现为正的订正技巧,但订正能力随着预报时效的增加而下降;东亚地区的系统误差小于整个北半球,＂递减平均法＂的订正能力总体小于整个北半球。对比夏、冬半年订正效果发现：对于500 hPa位势高度和850 hPa温度的预报场,冬半年和夏半年订正技巧相当;对于地面2 m温度预报场,冬半年订正能力明显高于夏半年。不同权重系数试验表明：对于500 hPa高度场,权重系数约取0.06时,订正效果较好,而对于850 hPa和2 m温度场,权重系数约取0.1时,订正效果最佳。
展开▼
