【作者】 黄静；

【导师】 邱崇践；

【作者基本信息】 兰州大学 ， 气象学， 2006， 博士

【摘要】 由一般线性反演理论得出结论,在由卫星遥感资料反演大气参数(温度、湿度)时,卫星资料提供的信息中有一部分是“无效信息”,它不会对反演发挥作用,而是造成反演方程病态的重要原因;另一方面,大气温度、湿度的垂直变化特征中有一部份是无法从卫星资料反演的,即存在“不可反演模态”,它既是反演误差的重要来源,也是反演对观测误差极其敏感的根本原因。在解决反演问题的时候应该在参数空间将“可反演模态”和“不可反演模态”分离,在资料空间应该将“有效信息”和“无效信息”分离,这是为获得好的反演结果首先要解决的问题。基于这样的认识,本文首先利用奇异值分解(SVD)方法结合经验正交分解(EOF)技术,导出一种计算大气温度可反演度的方法,并利用NCEP再分析资料首次给出了全球1月和7月分别由HIRS/3和AIRS资料反演大气温度廓线时的可反演度分布,结果表明温度的可反演度在高层和低层较小,400hpa—850hpa之间较大。从地理分布讲,低纬海洋地区和南极洲的部分地区可反演度最小,中高纬和大陆地区较大,反映出和温度变率的关系。比较用AIRS资料和HIRS/3资料得到的可反演度看到,前者比后者高出0.15—0.40,并且前者在低纬地区的反演能力改进更为明显。 论文接着研究了用SVD和EOF技术对传统的一维变分(1DVAR)反演做出改进的方法,基本做法是利用SVD剔除观测中的无效信息,利用EOF技术描写大气参数垂直结构特征。理想资料所作的试验表明,这样做可以提高大气温湿廓线的反演精度,减少对背景廓线的依赖。对比HIRS/3和AIRS资料的反演结果说明这一方法对HIRS/3资料的反演改进作用更为明显。 论文还给出一种新的反演大气温湿廓线的物理统计方法。该方法用一种耦合的SVD技术来实现对大气参数与观测资料的同时分解,让它所产生的基函数不但能表征参数与观测的主要结构特征,还能表现二者之间的联系,进而达到反演的目的。这种物理统计方法由于具有更多的统计特性,因而计算较为简便。利用理想资料所作的试验显示,该方法在反演温度廓线时,大部分高度得到的反演值比背景廓线和1DVAR的反演值更精确,反演水汽廓线时则对中高对流层和近地面层的反演结果有所改进。实际资料计算也说明该方法可以对某些高度的温度廓更多还原

【Abstract】 Some conclusions can be gained from the generalized linear inverse theory. When the atmospheric parameters (temperature, moisture) were retrieved from the remote radiation measurements, there is some invalid information in the data which is ineffective to the retrieval. The invalid information is the important reason for the ill-conditioning of the retrieval equation. On the other hand, some vertical variety characteristics of the temperature and moisture can’t be retrieved from the satellite data, namely the irretrievable mode exists. It’s the important source of the retrieval error as well as the basic reason of the high sensitivity of the retrieval to the observation error. So when we solve the retrieval problem, the parameter space should be separated into the retrievable mode and the irretrievable mode, and the data space should be separated into the effective information space and the ineffective information space. That’s the first problem should be solved for getting a good retrieval. Based on the cognition, a method is developed to assess the retrievability of temperature and the distribution of the global retrievability with HIRS/3 and AIRS data in Jan and Jul are firstly given by using the NCEP reanalysis data. The results show that the retrievability of temperature is low in the upper and lower atmosphere, and high between 400 hPa and 850 hPa. In geographical distribution, the retrievabilities are low in the low latitude marine regions and in some regions in Antarctica, and relatively high in mid-high latitude regions and continental regions. This partly represents the relationship between retrievability and the variability of temperature. In comparing the retrievabilities obtained with the AIRS and HIRS/3 data, the former are 0.15-0.4 higher than the latter and the retrievabilities obtained with the AIRS data are improved more evidently in the low latitude regions.Secondly, we modify the traditional one-dimensional variational (1DVAR) retrieval method by the SVD and EOF technique. The basic method is that utilize the SVD to get rid of the ineffective information and use the EOF to extract the vertical structure of the atmospheric parameters. The tests by the ideal data show the retrieval precision of the temperature and moisture profiles can be improved and the dependingon the background profiles can be reduced. In comparing the retrieval results obtained with the AIRS and HIRS/3 data, the revised method can improve more on the retrieval with HIRS/3 data.A new statistical-physical method for retrieving the atmospheric temperature and moisture profiles is given too in the paper. In the method, we use the coupling SVD technique to decompose the atmospheric parameters and the observation data synchronously. The basic function not only can denote the primary structure characteristics of the parameters and the observation, but also can indicate the relation between them. In virtue of this relation, the retrieval can be implemented. Because the method has more statistical character, the calculation is comparatively simple and convenient. The tests by the model data show that the retrieved temperature is more precise than the background profile and the retrieval by 1DVAR at most height;and the retrieved moisture profile can be improved at mid-high troposphere and the surface layer. The tests by practice data also illuminate that the method can improve the retrieval of temperature at some altitudes and the retrieval precision of moisture profile can be improved more, and the depending on the background profile is relatively small.更多还原