多模式集成方法对延伸期降水预报的改进

Multimodal ensemble method improvement for extended range consensus forecast of precipitation

利用TIGGE（THORPEX Interactive Grand Global Ensemble）资料中的CMC、ECMWF、NCEP和UKMO 4个中心全球集合预报模式对2007年10月3日—2008年2月29日逐日累积降水进行多模式集成预报试验。通过集合平均、多模式消除偏差集合平均、加权消除偏差集成3种方法进行试验对比,重点分析各中心模式及多模式集成的240~360h（10~15d）延伸期预报的检验效果。结果表明,多模式集成对逐日累积降水240~360h延伸期预报优于单个中心模式,将逐日降水的预报时效提高了72~168h。3种集成方法对比发现,多模式消除偏差集合平均方法预报效果最好,该方法将晴雨量级的降水预报时效在中短期和延伸期至少提高了1d和5d。

Multimodel consensus forecasts are investigated by using the ensemble mean outcomes of the 24-360 h ensemble fore-casts for the daily accumulated total precipitation from October 3,2007 to February 29,2008 ,which was taken from the CMC, ECMWF,NCEP and UKMO models,focusing on the extended range forecast.The results show that the multimodel consensus that extended the range forecasts of the daily accumulated total precipitation have higher forecast skills than all the single models. The bias-removed ensemble mean has the best performance and the weighted bias-removed ensemble mean technique has even the same performance with Bias-Removed Ensemble Mean （BREM）.