短期温度、降水的多模式集成预报

Multi-model Consensus Forecast for Temperature and Precipitation of Short-term

通过对T213、NCEP、JMA和德国天气在线4种模式的产品对湖州从2005年1月到2007年6月的温度、降水预报的对比分析,利用前5日和历史权重法建立了对温度、降水的集合预报方法,检验表明:集成的温度、降水预报结果优于大多数模式单独的预报结果,降水24 h、48 h定性预报准确率70%左右。前5日集成模式重要温度预报的平均绝对误差在2℃左右,并且不存在明显的系统误差,预报达到了一定的精度;同时前5日的集成方法比历史集成方法更为优越,可提高温度和降水准确率5%左右,为预报员提供更精细的客观预报参考。

After the comparative analysis of NWP models,including T213,NCEP,JMA and German weather online,for temperature,precipitation forecast from January,2005 to June,2007 of Huzhou,Zhejiang Province,the temperature and precipitation′s ensemble forecast method is suggested.It is showed that the results of the ensemble forecasting model are better than that of most individual models,within 48 h,the forcast accuracies of precipitation are about 70%.The average absolute forecast error of the important temperature of the recent 5 day method is less than 2 ℃,with no clear system error,Futhermore,it is also better than the whole sample method with the 5% increment on the forecast accuracies of temperature and precipitation,and can be used to provide more precise objective prediction to weather men for reference.