青藏高原区域资料同化的变分质量控制研究：参数优化及其有效同化

Study on variational quality control of data assimilation in Tibetan Plateau: parameters optimization and effective assimilation

中文关键词:  青藏高原  资料同化  变分质量控制  新息向量  观测误差

英文关键词:Tibetan Plateau, Data assimilation, Variational quality control, Innovation vector, Observation error

中文摘要:

      青藏高原(下称高原)区域天气系统变化对其周围及下游区域的数值天气预报质量具有重要影响。然而,高原区域质量偏低且稀少的观测资料严重制约了同化分析效果和预报质量。变分质量控制方案(VarQC)具有合理利用不同质量观测资料改善同化分析性能的良好能力。鉴于此,本文在高原区域观测误差非高斯分布特征分析的基础上,构建了适应高原区域观测误差特征的“高斯分布+均匀分布”的变分质量控制方案(Flat-VarQC),优化了其关键参数,使其能更准确地基于同化新息向量调整观测对同化分析的影响权重,从而提高高原区域稀少观测资料的有效同化率,改善高原区域的同化分析质量。试验结果结果表明：高原区域观测误差具有更明显的“厚尾分布”特征,同化过程中单纯的高斯分布假设将导致观测资料的有效同化率偏低,从而降低了高原区域有限观测资料对同化分析的贡献；适用于高原区域资料特点的变分质量控制关键参数优化是充分发挥高原区域Flat-VarQC方案改善同化分析质量的关键；Flat-VarQC方案能够明显提升高原区域常规资料的有效同化率,吸收更多有益资料信息,剔除有害资料信息,从而提高观测资料对同化分析的正贡献。这对于高原区域中小尺度天气系统的同化分析和强降水预报具有更显著的改善效果。

英文摘要:

      The change of regional weather system in the Tibetan Plateau (hereinafter referred to as the plateau) has an important impact on the numerical weather forecast in its surrounding and downstream regions. However, the low quality and scarce observation data in the plateau seriously restricts the assimilation analysis and prediction quality. The variational quality control scheme (VarQC) has the potential to make full use of scarce observations to improve the quality of assimilation analysis. In view of this, based on the analysis of the non-Gaussian distribution characteristics of the observation error in the plateau region, this paper constructs a variational quality control scheme of "Gaussian distribution plus flat distribution" (Flat-VarQC), which is suitable for the characteristics of the observation error in the plateau region, and optimizes the key parameters, so that it can more accurately adjust the influence weight of the observation on the assimilation analysis based on the innovation vector, so as to improve the effective assimilation rate of the scarce observation data in the plateau region, Improve the quality of assimilation analysis in plateau region. The results show that the observation error in the plateau region has more obvious characteristics of "fat-tail distribution", and the simple assumption of Gaussian distribution in the assimilation process will lead to the low effective assimilation rate of observation data, thus reducing the contribution of limited observation data in the plateau region to the assimilation analysis; The key parameter optimization of variational quality control suitable for the characteristics of plateau regional data is the key to give full play to the Flat-VarQC scheme in plateau region to improve the quality of assimilation analysis; Flat-VarQC scheme can significantly improve the effective assimilation rate of conventional data in the plateau region, absorb more useful data information and eliminate harmful data information, so as to improve the positive contribution of observation data to assimilation analysis. This has a more significant improvement effect on the assimilation analysis of small and medium-scale weather system and heavy rainfall forecast in the plateau region.