基于频率匹配法的中国降水多模式预报订正研究
投稿时间:2019-01-09  修订日期:2019-03-25  点此下载全文
引用本文:
摘要点击次数: 127
全文下载次数: 0
作者单位E-mail
智协飞 南京信息工程大学 zhi@nuist.edu.cn 
吕游 南京信息工程大学  
基金项目:国家自然科学基金面上项目 北极阁开放研究基金 南京大气科学联合研究中心项目
中文摘要:基于TIGGE资料中欧洲中期天气预报中心(ECMWF)、日本气象厅(JMA)、美国国家环境预报中心(UKMO)及英国气象局(NCEP)1—7天日降水量预报以及中国自动站观测资料与CMORPH降水产品融合的逐时降水量网格数据集,利用频率匹配法(FMM)对中国降水预报进行客观订正。首先利用卡尔曼滤波方法对降水频率进行了改进,并根据不同区域降水强度差异将全国分为7个子区域分别进行频率匹配。结果表明,FMM可以有效减小降水量和雨区面积预报的误差。经过频率匹配法订正后各模式降水预报的MAE大幅减小,且订正后各量级降水的雨区面积更加接近实际观测值。FMM订正特别对小于5mm和大于15mm区间的降水预报技巧改进明显。此外,FMM降低了模式预报的小雨空报率和大雨漏报率,在一定程度上解决了“小雨空报”和“大雨漏报”的问题,并且明显提高了晴雨预报的准确率。FMM明显消除了大范围小雨空报区域,但是对强降水中心仅能从调节降水量大小,改进降水强度和范围,不过强降水区的位置和形状仍然依赖于模式原始预报的预报水平。
中文关键词:多模式预报 频率匹配法 偏差订正 卡尔曼滤波
 
Calibration of the Multimodel Precipitation Forecasts in China Based on Frequency Matching Method
Abstract:Based on the 1-7 day precipitation forecasts from ECMWF, JMA, UKMO and NCEP in TIGGE dataset and the hourly precipitation data merged by China Automatic Station (CAS) and CMORPH precipitation products as the observation data, the precipitation forecast in China is calibrated by using frequency matching method (FMM). Firstly, Kalman filter approach was used to improve the statistics of the precipitation frequency. China was divided into 7 sub-regions for the FMM calibration of the precipitation forecasts in accordance with its different precipitation intensity in different regions. The results show that FMM can significantly reduce the forecast error of precipitation intensity and area. As a whole, FMM can improve the forecast skill of precipitation at different thresholds. After calibration, the ETS score of the precipitation forecast is significantly improved, and the false alarm rate of light rain and the missing rate of heavy rain are considerably reduced, and the forecast skill for "rain or no rain" event is significantly improved. In addition, FMM can make the forecast rainfall area closer to the observed values, especially reducing false alarm rate of light rain in a vast area. However, FMM can only improve the intensity and scope of the precipitation via adjusting the amount of rainfall. The location and shape of the heavy rainfall area depend on the quality of the original model forecast.
keywords:multimodel forecasts frequency matching method bias correction Kalman filter
  查看/发表评论  下载PDF阅读器