| 全球资料同化中误差协方差三维结构的准确估计与应用Ⅰ:观测空间协方差的准确估计 |
| |
| 引用本文： | 龚建东,魏丽,陶士伟,赵刚,万丰.全球资料同化中误差协方差三维结构的准确估计与应用Ⅰ:观测空间协方差的准确估计[J].气象学报,2006,64(6):669-683. |
| |
| 作者姓名： | 龚建东 魏丽 陶士伟 赵刚 万丰 |
| |
| 作者单位： | 1. 国家气象中心,北京,100081
2. 中国科学院寒区旱区环境与工程研究所,兰州,730000 |
| |
| 基金项目： | 国家自然科学基金;国家气象中心资助项目 |
| |
| 摘 要： | 观测误差与背景误差协方差在四维资料同化和业务资料分析系统中起到决定性作用,它决定着观测信息与背景初猜值信息的相对重要性以及这些信息在空间及不同变量间的扩展方式。由于实际大气的真值并不知道,需要发展不同的技巧来估计观测误差与背景误差协方差,其中在观测空间利用观测与背景初猜值之差来分离观测误差与背景误差协方差的方法估计出的结果较为准确,其估计出的观测误差可直接用于资料分析系统中,背景误差可作为标尺来度量其他方法估计结果的可靠性。文章采用国家气象中心T213L31全球中期分析预报系统的6 h预报作为背景初猜场及同时段冬夏两个季节的北半球探空,利用贝塞尔函数拟合方法来分离观测误差与背景误差协方差,并比较了东亚区、北美区、欧洲区3个探空资料均匀密集区的区域与季节变化结果。结果表明,观测空间拟合方法所要求的水平均质、各向同性在欧洲区和北美区成立程度较好,在东亚区略差,使用时需要斟酌。此外均方差区域间差别较大,在冬季明显大于夏季,温度场偏大0.2 K,风场偏大0.9 m/s。温度场在400 hPa以下与150 hPa以上,背景误差略小于观测误差,而在200—300 hPa,背景误差略大一些。风场的特点与温度场比较一致。温度与风场背景误差主要集中在前40波,并在20波左右达到最大,水平相关季节区域差别不大,而温度垂直相关比风场窄,两者相关范围比较大的波数主要集中在前20波。此外利用贝塞尔函数拟合方法获得结果的分析表明,在质量场中不同区域季节间温度误差的稳定性要明显好于高度场,涡度散度的稳定性要明显好于流函数和势函数,特别是对于特征长度更为明显。这表明利用贝塞尔函数拟合方法获得的结果对校准在全球资料同化中采用温度、涡度散度作为资料同化的分析变量具有一定的优势。
|
| 关 键 词： | 观测误差 背景误差协方差 贝塞尔函数拟合 观测空间分离 准确估计。 |
| 收稿时间： | 2005/9/27 0:00:00 |
| 修稿时间： | 2005年9月27日 |
ACCURATE ESTIMATION AND APPLICATION OF 3-D ERROR COVARIANCE STRUCTURES IN GLOBAL DATA ASSIMILATION Part Ⅰ: Accurate Estimation of Error Covariance in Observation Space |
| |
| Gong Jiandong,Wei Li,Tao Shiwei,Zhao Gang,Wan Feng.ACCURATE ESTIMATION AND APPLICATION OF 3-D ERROR COVARIANCE STRUCTURES IN GLOBAL DATA ASSIMILATION Part Ⅰ: Accurate Estimation of Error Covariance in Observation Space[J].Acta Meteorologica Sinica,2006,64(6):669-683. |
| |
| Authors: | Gong Jiandong Wei Li Tao Shiwei Zhao Gang Wan Feng |
| |
| Abstract: | Observation and background error covariances play important roles in the four-dimensional data assimilation and operational data analysis system,and they determine the relative importance of observation and background information,and the spread of these information in the grid space and among various control variables.Because the true atmospheric state is unknown,it is necessary to develop some techniques to estimate observation and background error covariances.Among these techniques,the method which uses innovation vector(observation minus background in the observation space) to partition observation and background error covariances is more accurate than other methods;and its estimated observation errors can be directly used in the data analysis system,and its estimated background error covariance can be used as a benchmark to tune and verify the results from other techniques,such as so called NMC method.Using National Meteorological Centre global medium-range analysis and forecast system T213L31's 6 hour forecasts as a background,and winter/summer two seasonal radiosonde observations in the Northern Hemisphere,the Bessel fitting function is used to partition observation and background error covariances.The analysis and comparison are performed during winter and summer periods,and among the data from three dense radiosonde observations regions of East Asia,North America,and Europe.The results show that the requirements for Bessel fitting method,such as horizontal homogeneity and isotropy,are roughly satisfied in all three regions,relatively better in North America and Europe,but slightly worse in East Asia.The error variances change greatly with different regions and seasons.For example,for temperature and scale-wind in winter those values are about 0.2 K and 0.9 m/s larger than those in summer,respectively.Below 400 hPa and above 150 hPa the temperature background error variances are smaller than the observation error variances,but within 200-300 hPa,the background error is slightly larger.The characteristics of winds are similar with temperatures.In the spectral space,the major temperature and wind background errors are distributed within wave-number 40,with the maximum in about wave-number 20.The horizontal de-correlation length is temporally,spatially stable for all variables,but the vertical de-correlation length of temperature is relative narrower than that of wind,and their main vertical correlations are located within wave-number 20.The analysis results also show that the background error covariance structure for temperature is clearly more stable than that of height,regardless of different seasons or in different regions;and for wind field the covariance structures for vorticity and divergence are clearly more stable than those of stream-function and velocity potential function,especially for horizontal de-correlation length.Those results are useful for tuning and verification of the global data assimilation which adopts temperature,vorticity and divergence as analysis variables. |
| |
| Keywords: | Observation error Background error covariance Bessel function fitting Error separation Accurate estimation |
| 本文献已被 CNKI 万方数据 等数据库收录！ |
| 点击此处可从《气象学报》浏览原始摘要信息 |
| 点击此处可从《气象学报》下载免费的PDF全文 |
|
