龚建东. 2021. 集合预报误差在GRAPES全球四维变分同化的应用研究:II 参数选取与数值试验[J]. 气象学报, (0):-, doi:10.11676/qxxb2021.007
集合预报误差在GRAPES全球四维变分同化的应用研究:II 参数选取与数值试验
GONG Jiandong ZHANG Lin WANG RuichunNational Meteorological Center/Numerical Weather Prediction Center of CMA, Beijing, 100081
投稿时间:2020-04-24  修订日期:2020-09-20
DOI:10.11676/qxxb2021.007
中文关键词:  集合资料同化,集合预报误差,En4D-Var,参数选择,数值试验
英文关键词:Ensemble  data assimilation, ensemble  forecast error,En4D-Var,parameter  setup, numerical  experiment
基金项目:
作者单位E-mail
龚建东 国家气象中心 gongjd@cma.gov.cn 
摘要点击次数: 42
全文下载次数: 27
中文摘要:
      [目的] 将集合预报误差信息在GRAPES全球4D-Var中应用,需要解决大集合样本数高时效生成的计算效率问题,以及与GRAPES全球4D-Var匹配的同化关键参数确定问题。[资料和方法]本文选择基于4D-Var的集合资料同化(EDA)技术生成集合样本,通过重复使用第一个样本极小化迭代过程中产生的预调节信息,并用于其他样本极小化做预调节,将计算效率提高了2倍。通过时间错位扰动(VTSP)方法增加集合样本数,实现集合样本增加2倍。对集合方差进行膨胀,选择水平局地化相关尺度为流函数背景误差水平相关的1.4倍,并通过批量数值试验方法确定背景误差与集合预报误差的权重,当权重取0.7时的预报效果最好。[结果]夏冬两季各自52天的批量试验表明,对分析场的位势高度均方根误差,北半球冷季En4D-Var较4D-Var的改进主要集中在700-30hpa,暖季主要改进在400-150hpa,北半球位势高度的均方根误差变化明显。赤道地区受季节影响较小,对位势高度、风场与温度的改进都较为明显,改善最显著的是赤道地区的南北风场。[结论]本文研发的集合预报误差信息在GRAPES全球4D-Var中应用的技术合理可行。
英文摘要:
      In order to apply the ensemble forecast error information to the GRAPES global 4D-Var, it needs to solve the computational efficiency problem for large ensemble member generation, and to setup the key assimilation parameters to match with the GRAPES global 4D-Var. In this paper we choose 4D-Var based ensemble data assimilation (EDA) technique to generate ensemble members. By saving the pre-condition information generated during first member minimization iteration, and then re-using information to all other members, the computation efficiency is improved by twice times. By using the valid-time-shifting perturbation (VTSP) method, the ensemble member is enlarged by a factor 2 times. The ensemble inflation method is used, and the horizontal localization scale is selected as 1.4 times of the stream-function of background error. The weight of background error and ensemble forecast error is determined by numerical experiment, the weight for ensemble forecast error is 0.7. The 52 days numerical experiment in summer and winter show that the improvement of En4D-Var in cold season is mainly in 700-30hpa, while that in warm season is mainly in 400-150hpa. RMS error of geopotential height in northern hemisphere is decreased obviously. The equatorial region is less affected by seasons, and the improvement of geopotential height, wind field and temperature is more obvious. The most significant improvement is the meridional wind in the equatorial region.
查看全文   查看/发表评论  下载PDF阅读器
分享按钮