项目名称: 基于集合的四维变分方法用于雷达资料同化的关键技术研究
项目编号: No.41275102
项目类型: 面上项目
立项/批准年度: 2013
项目学科: 天文学、地球科学
项目作者: 邵爱梅
作者单位: 兰州大学
项目金额: 82万元
中文摘要: 基于集合和奇异值分解(SVD)技术的四维变分同化方法(SVD-En4DVar)将集合同化方法和变分同化方法相结合,将由预报集合估计的流依赖的背景误差协方差引入变分同化,利用SVD技术进行降维,使得控制变量显式地出现在目标函数中,避免了伴随模式的运用。本项目拟开展将该方法用于同化多普勒雷达资料的研究,重点研究解决以下3个方面的问题:(1)减少产生预报集合所需计算时间的方法;(2)局地化方法;(3)产生初始扰动样本的方法。通过这一研究,在一定程度上解决将SVD-En4DVar用于雷达资料同化的主要困难,希望其同化效果明显好于现有的3DVar同化效果,并且在计算量上能满足业务运行要求,为推动雷达资料的业务运用作出贡献。
中文关键词: 雷达资料同化;集合同化方法;预报样本集合;局地化方法;四维变分
英文摘要: The 4-D variational method based on ensemble and SVD technique (SVD-En4DVar), as a kind of hybrid method, powfully combines the ensemble method with variational method. In this method, flow-dependent backgournd error covariance estimated from forecast ensemble is introduced into variantional assimilation method, and SVD technique is used to reduce dimensions in the solving process. In this way, the control variables are explicitly expressed in the cost function and the adjoint model is not used. In this project, the SVD-En4Dvar will be carried out the ability of radar data assimilation. The project will focus on solving the following three key problems: (1) the method on reducing the computation cost for producing the forecast ensemble; (2) the localization scheme; (3) the method for pruducing the initial perturbation samples. Through this project, the primary difficulties in using SVD-En4DVar method to radar data assimilation will be solved. It is feasible this method works better that the existed 3DVar system. And the computational time seeks to meet the need for operational run. It will useful and helpful to the further operational application of radar observations.
英文关键词: radar data assimilation;ensemble assimilation method;forecast ensemble;localization scheme;4DVar