【作者】 马盈盈；

【导师】 龚威；

【作者基本信息】 武汉大学 ， 摄影测量与遥感， 2010， 博士

【摘要】 气溶胶是大气研究的重要参数,它不仅影响着气候变化、水汽循环,同时还因其对电磁波的吸收和散射作用,降低了定量遥感的测量精度,粒径较小的颗粒物还能通过呼吸进入人体,对人类的身体健康构成了一定的危害。为了研究和了解气溶胶的分布及其对气候环境的变化,以及对辐射传输的影响,目前全世界已广泛开展了对气溶胶浓度、分布和光学特性的研究。我国广泛采用的气溶胶观测主要是地基单点的观测,尚未将其观测数据扩展为面数据,即大范围、高密度的多点观测。悬浮在大气中的气溶胶又容易受到大气环流的影响,并不会长期悬浮在气溶胶排放源的上空,而是会产生扩散和漂移。因此,单点的地基观测数据应用范围相对较小。基于此,本文采用了星载卫星观测数据,对我国区域性气溶胶的特性进行分析,不同于以往的被动观测数据,如MODIS (MODerate-resolution Imaging Spectro-radiometer)和MISR (the Multiangle ImagingSpectroRadiometer),本文采用的是主动星载激光雷达的观测数据。从底层数据处理和算法改进、优化等角度入手,在对区域性气溶胶分布特性,光学特性分析的同时,还进一步获得其层次分布特征值。目前,地基激光雷达作为气溶胶观测的重要观测手段已发展的较为成熟,但作为新兴的探测方式,星载激光雷达仍有别于地基激光雷达的传统反演方法。针对NASA官法开发的星载激光雷达气溶胶反演方法,结合其他同步卫星的观测数据和对地观测的分类方法,并提供了可行的改进方案,为大气气溶胶反演的假设前提参数——激光雷达比的正确选择提出了可靠的理论依据。进一步应用改进后的反演结果分析我国部分区域气溶胶的空间分布和光学厚度等特性。首先,本文提出多星数据融合的层次查找方法。主要引入A-Train星系中与CALIPSO (Cloud-Aerosol Lidar and Infrared Pathfinder Satellite Observations)近似同步的观测卫星CloudSat。针对星载激光雷达噪声较大,多层次下气溶胶回波信号与噪声信号易混合的情况,提高层次识别的正确性,为后期的层次分类奠定良好的基础。基于此改善,进一步针对现有的PDF (Probability Density Functions)分类方法中存在的不足提出了相应的改进措施,在训练样本有限的情况下有效的确保层次分类的正确性。为了有效的验证分类的结果,还引入了另一同属于A-Train星系的中Aqua加载的传感器MODIS影像进行分类结果的验证。考虑到我国当前面临的沙漠化程度严重、沙尘暴频发等情况的出现,本文就信噪比较低、背景光噪声较大的白天沙尘数据进行分析。结果显示改进后的分类算法在合理的样本选取情况下,尽管样本数量较少但仍然能获取较好的分类结果。对分类后的结果进行统计分析,获知沙尘气溶胶同冰晶云相比的粒径和形状特征与差别,以及不同高度上的云顶冰晶颗粒的尺度差别。在数据应用部分,参考NASA兰利数据处理中心(Langley Research Center)提出的多分辨率气溶胶参数反演法,对我国局部区域做气溶胶光学参数反演,同时,考虑其他相关的大气要素对气溶胶光学厚度变化的影响给出相应的分析结果。从分析结果中可以获得大气层中,各悬浮层次的气溶胶光学厚度、整层大气的气溶胶光学厚度、区域气溶胶的分布特点。实验结果中可以显示,针对湖北地区的研究发现,受大气运动的影响气溶胶光学厚度的分布特征有别于地表覆盖情况,即非理论上推导的工业排放城市的气溶胶光学厚度相对较大,而农业生产地区的较小,并综合对比了四季的变化特点；此外,在北京地区,气溶胶光学厚度自06年星载激光雷达发射至2008年底气溶胶光学厚度普遍偏低,说明为奥运会的筹办,大气质量已有明显的改善,特别是在2008年7月至9月间的气溶胶光学厚度明显小于2007年同期的观测值：最后融合MODIS和CALIPSO的观测结果对我国华东地区做2008年春季的气溶胶分布特性研究,在3月至5月期间明显的湖北东部,江浙西部气溶胶光学厚度较大,而冬季则显著减小。更多还原

【Abstract】 Aerosol is the important atmospheric ingredient, it influences both the climate change and water cycle, in the same time, by absorbing and scattering of electromagnetic wave, it can change the accuracy of quantitative remote sensing. The particles which have the smaller size can be absorbed into the body through breathing and influence the health. Trying to research the distribution of aerosol, investigate how it change the climate, as well as the radiative transfer, there are many research has been developed in the world wide. At present, the observation in our country is based on the ground based station and does not develop into the regional analysis. The aerosol suspend on the ground does not invariable and always change, commonly, aerosol is usually influence by the atmospheric circulation and the aerosol particles still diffusing and transporting. In the end, the station observation is limited. In this paper, the data is from spaceborne lidar, different from the passive remote sensor, lidar is an active remote sensor. Start from the data processing improvement, then analyzing the aerosol characteristics and the feature of layers.At present, the ground based lidar has been widely used, and it is the mature technology, but as the new method used for atmosphere observation, the algorithm of spaceboren lidar is different from the ground based lidar. Based on the official method, this paper study the pre-processing of spaceborne lidar retrieval, and give the improved scheme, all of these can choose a more appropriate lidar ratio, and support the correct retrieval. After that, used the atmospheric parameter from new processes analyze the distribution and aerosol optical depth in different area, China.At first, in this paper we give the method of data fusion between CALIPSO and CloudSat, and discriminate the layer suspends in the atmosphere which is the aerosol and which is the pseudo layer (cause by noise). Though introduce the CloudSat, which is also a satellite of A-train constellation, it is almost synchronous with CALIPSO. In view of the lower SNR (signal noise rate) of CALIOP observation and the layers signal below thick cloud is difficult to discriminate, the data fusion between them can avoid error layer detection, to lay the foundation for the subsequent analysis. Based on this improvement, consider the deficiency of scene classification about cloud and aerosol, the former algorithm employ the PDF (probability density functions), but it needs many samples and the parameter estimation is difficult. Further, for the effective validation of classification, we also introduce another remote sensor, MODIS. MODIS load on Aqua satellite which is also an A-Train satellite, the interval between them is less than one minute.Because the dust storm is a problem exists in our country, the appearance of dust storm is still serious, this paper analyzes the dust storm which happens in day time and conquer the influence from lower SNR. The result shown that, if the correct samples we choose, then the result is satisfactory, through calculate the particle size and shape, the differences between dust aerosol and ice cloud is known.In the part of data application, we use the Hybrid Extinction Retrieval Algorithms (HERA), and receive the aerosol characteristics of China. From the retrieval result, the aerosol optical depth, and their distribution of each layer is obtained. The result shown, aerosol distribution in Hubei province doesn’t always in according with the ground surface, not the higher aerosol optical depth value above megalopolis. Further, the aerosol optical depth in Beijing is debased obviously, this means for the Olympic Games in 2008, the air quality in the urban areas picked up gradually. In the last part of the paper, the data fusion between MODIS and CALIPSO retrieval data used to analyze the aerosol optical characteristic in the Southeast China, this step can improve the accuracy and give more detailed information. The aerosol concentration in winter is lower than in summer.更多还原