刘艳,薛纪善. 2019. 数值天气预报系统GRAPES的新初始化方案[J]. 气象学报, (0):-, doi:10.11676/qxxb2019.029
数值天气预报系统GRAPES的新初始化方案
New Initialization Scheme of GRAPES Numerical Weather Prediction System
投稿时间:2017-11-17  修订日期:2018-04-28
DOI:10.11676/qxxb2019.029
中文关键词:  初始化  数字滤波  四维变分同化  GRAPES
英文关键词:initialization, digital filter, four-dimensional variation assimilation, GRAPES
基金项目:公益性行业(气象)科研专项“面向业务化应用的全球四维变分同化系统研究”(GYHY201506003),中国气象局GRAPES发展专项
作者单位E-mail
刘艳 中国气象局数值预报中心 liuyan@cma.gov.cn 
薛纪善 中国气象科学研究院 jsxue@cma.gov.cn 
摘要点击次数: 35
全文下载次数: 32
中文摘要:
      四维变分同化由于引入预报模式作为一项约束,理论上它的分析场已经具有较好的平衡性,但实际实施时还会有诸多因重力波导致的高频振荡过程,因此,四维变分同化分析仍需要初始化。本文描述了GRAPES全球四维变分同化系统(GRAPES-4DVar)的新初始化方案的科学设计,公式演绎以及试验结果。GRAPES-4DVar的新初始化方案采用数字滤波方案作为代价函数的一项约束控制重力波引发的不平衡结构,约束强加在分析增量上,与极小化迭代过程同步进行。新的初始化方案是变分同化系统的一部分,数字滤波的积分时间与4DVar的同化时间窗一致,不会对4DVar产生额外的计算资源消耗;并能适应长时间窗的同化,不会因为时间窗的延长而削弱慢波过程。新初始化方案中,模式轨迹的光滑程度可在变分同化中通过重力波控制项的权重系数方便地控制。GRAPES全球四维变分同化的理想和循环同化批量试验都表明:在四维变分同化中,重力波的控制依然非常重要,具有初始化的GRAPES试验,无论分析还是预报技巧都较无初始化的分析和预报技巧有明显优势。与以前分析和滤波独立实施的旧初始化方案相比,新方案的分析和预报效果略优,但可以有效地节省循环同化系统的运行时间,这对四维变分同化来说非常重要。
英文摘要:
      Theoretically, the analysis of a four-dimensional variational (4DVar) data assimilation should be balanced due to a forecast model as it’s constrain, but there exist many procedures leading to high-frequency gravity wave in actual implementation. Thus, the process of initialization is still necessary for 4DVar. The scientific design,formulation derivation and experiments of the new initialization scheme of GRAPES(Global and Regional Assimilation Prediction System)numerical weather prediction system are presented in this paper. GRAPES-4DVar adopts the digital filter scheme as initialization. The digital filter is added into the penalty function as a weak constrain term in order to upset the presence of unbalance structure associated with gravity-inertia wave efficiently. The constraint is only imposed on the incremental analysis, and synchronizes with the minimization. The weak constrain term does not add the extra computation cost for the integrated time of digital filter is same as to the 4DVar’s time window. Furthermore, the new initialization scheme is suitable for long time window assimilation, and will not weaken the slow wave process with the extending of the time window. It is easy to control the smooth of model trajectory by tuning digital filter coefficient in the variational assimilation procedure. Results of GRAPES-4DVar assimilation cycling experiments show that the initialization is needed in a 4DVar analysis, the analysis and forecast skill have been improved significantly in the experiments with initialization than those without initialization. Compared to the old initialization scheme in which the analysis and filter are two independent procedures, the new scheme saves the running time of assimilation cycling system efficiently.,although its impact on analysis and forecast is similar to that of the old one.
查看全文   查看/发表评论  下载PDF阅读器
分享按钮