时间依赖的多尺度背景误差协方差研究Ⅱ——应用  

STUDY ON TIME-DEPENDENT AND MULTI-SCALE BACKGROUND ERROR COVARIANCE Ⅱ ——Application

作　　者：张旭斌[1,2] 薛纪善[2,3] 万齐林[1,2] 丁伟钰[1,2] 李昊睿[1,2] 

机构地区：[1]广东省区域数值天气预报重点实验室,广东广州510080 [2]中国气象局广州热带海洋气象研究所,广东广州510080 [3]中国气象科学研究院,北京100081

出　　处：《热带气象学报》2015年第2期161-172,共12页Journal of Tropical Meteorology

基　　金：广东省科技计划项目(2012A061400012);十二五科技计划(2012BAC22B00);公益性行业(气象)科研专项(GYHY201406009);广东省气象局科学技术研究项目(2013A04)共同资助

摘　　要：将基于GRAPES 3D-Var资料同化系统构造的具有时间依赖特征与多尺度特征的新型背景误差协方差应用于一个实际天气个例研究,确定出新型协方差中不同尺度标准差合适的缩放系数。该新型协方差在不同尺度的分析过程中产生了量值与尺度差异很大的分析增量,既很好地保留大尺度背景场信息,也使局地观测信息被很好保留并有效提取。不同尺度分析增量的共同作用最终导致分析场调整出更符合天气系统尺度特征的增量,从而更大程度地调整不同尺度天气系统的结构,并进一步改进相应的降水预报。另外,将观测资料按空间分布密度高低分别同化,能更有效提取不同分布密度观测资料中不同尺度的信息。为期一个月的批量试验表明,该协方差对应的前12 h降水预报效果改善明显,尤其是强降水预报。Based on GRAPES 3D-Var data assimilation system, a new-type background error covariance, characterized by time-dependency and multi-scale, was constructed in Part I. This new-type covariance is used in a real-case study in this paper, and the corresponding proper factors for univariate standard deviations at different scales are estimated through a case study. Analysis increments, with large differences in both magnitudes and scales between large-scale and small-scale analysis processes, are generated by the new-type covariance. These increments not only keep the large-scale information from background fields very well, but also primarily retain and effectively abstract the local information from observations. With the combined effect of analysis increments at different scales, the increments, much in accord with the scale characteristics of weather systems at different scales, are produced to adjust the structures of these weather systems to a great degree, and thus to improve the corresponding forecasts for precipitation. Furthermore, assimilating the observations with different spatial distribution density separately can be more effective at abstracting different scale information from observations with different density. The one-month batch experiments show that, application of the new-type covariance is beneficial in terms of forecast performance for precipitation valid at 0-12 h, especially for the heavy rainfall.

关 键 词：三维变分 背景误差协方差 时间依赖 多尺度 

分 类 号：P456.7[天文地球—大气科学及气象学]