集合变分混合同化背景误差协方差流依赖性分析

Flow- dependent characteristics of background error covariance in hybrid variational-ensemble data assimilation

通过单点观测试验的方法,对集合变分混合同化背景误差协方差的流依赖特征、流依赖性影响因子、产生原因,以及集合预报方法对流依赖性的影响进行了研究。结果表明:由于引入了集合信息,集合变分混合同化的分析增量与天气系统的分布有关,具有非均匀、各向异性的特征;这种流依赖特征对混合系数敏感,当集合协方差所占权重很小时,分析增量仍呈现出均匀、各向同性特征;混合同化背景误差协方差的流依赖特征不仅与集合样本有关,还与构造集合协方差的ETKF方法有关,只引入与环流形势密切相关的集合样本并不能使分析增量表现出显著的流依赖性,集合样本和ETKF方法共同作用才能将流依赖信息引入到混合协方差中,使分析增量出现流依赖特征;不同集合预报方法对混合协方差的流依赖特征有显著影响,考虑初值和物理过程的超级集合,以及在超级集合样本上再进行ETKF更新扰动后样本构造的混合协方差流依赖特征更加显著。

A single point observation experiment was carried out to evaluate the flow-dependent characteristics, flow-dependent influencing factors and occurring reasons of the background error covari- anee in hybrid variational-ensemble data assimilation, moreover, the effects of the ensemble forecast method on the flow-dependent characteristics were also analyzed. Results show that due to the introduc- tion of ensemble information, the analysis increments of hybrid variational-ensemble data assimilation are related to weather systems distribution and characterized by the inhomogeneity and anisotropy. The flow- dependent characteristics are sensitive to hybrid coefficients, so when the ensemble-derived covarianee belongs to small weight, the analysis increments are still homogeneous and isotropic. The flow-dependent characteristics of background error covariance in hybrid variational-ensemble data assimilation is not only related to the ensemble samples, but to the Ensemble Transform Kalman Filter（ETKF） method used to generate ensemble error covariance. So the combined effects of both ensemble samples and ETKF method can introduce flow-dependent information into hybrid background error eovarianee and then the analysis increments depend on flow characteristics. Different ensemble forecast methods have different effects on flow-dependent characteristics of hybrid background error covariance. It can be concluded that the flow-dependent characteristics are more significant, when the initial value and multi-physics process in the su- per-ensemble samples are taken into account and the super-ensemble samples are updated by ETKF.