中国科学院遥感与数字地球研究所-高光谱遥感应用技术研究室

【作者】              张浩                                                                                                                                           

【导师】              童庆禧；张兵；郑兰芬

【 学位年度 】    2009

【论文级别】      博士    

【关键词】          气溶胶光学厚度，水汽含量，二类水体，大气纠正，CE318，WHI，CHRIS 

【Key words】   aerosol optical thickness; water vapor column; case 2 waters; atmospheric correction; CE318; WHI; CHRIS  

【中文摘要】

二类水体遥感图像大气纠正一直是内陆水体水色遥感发展面临的首要问题。其困难主要在于两方面：一方面是二类水体尤其是内陆水体成分复杂多变，空间差异大、区域特征明显，下垫面的复杂性难以满足一类水体大气纠正算法的要求；另一方面，二类水体上空气溶胶特性受人为活动和地表要素影响大，气溶胶类型复杂、光学厚度较大，一些算法中对气溶胶光学特性的假定存在问题。因此，本文以我国典型的二类水体－太湖遥感数据的大气纠正为例，针对上述问题尤其是第二个问题进行研究。

        气溶胶和水汽是影响大气纠正精度的主要因素，其中气溶胶散射与吸收更为复杂。本文首先研究了多波段太阳分光光度计CE318反演气溶胶光学厚度与水汽含量的方法，分析光谱响应函数对气溶胶光学厚度反演精度的影响，利用SCIAMACHY精细光谱的气体吸收截面数据以及逐线积分模型LBLRTM分析了H₂O、O₃、NO₂、CO₂、CH₄等气体在非水汽吸收通道(940nm以外通道)吸收特征，改进这些气体吸收光学厚度的计算方法。利用LBLRTM分析940nm水汽吸收透过率与水汽含量、观测角度的关系，提出了一种精度较高的水汽反演模型，能够保证大角度(~63°)观测条件下水汽反演相对误差不超过4%，对应该模型提出了新的水汽通道定标方法，该方法不依赖于临近通道的光学厚度反演结果。

        利用上述改进方法并基于2005～2007年在太湖沿岸获取的CE318数据对该区域的气溶胶光学厚度进行了计算。结合AERONET太湖站的气溶胶产品，系统分析了太湖气溶胶光学特性时空变化规律，构建适合太湖区域季节周期的气溶胶类型，用以提高太湖水体遥感大气纠正精度。

        在这些研究的基础上，分析了气溶胶类型误差、气溶胶光学厚度误差及成像几何误差对二类水体大气纠正精度的影响，确定了大气纠正中查找表建立的原则与方法，分别基于图像自身和基于实测大气数据完成了欧空局PROBA卫星搭载的紧凑型高分辨率成像光谱仪CHRIS数据和航空宽视场高光谱成像仪WHI数据的大气纠正，在涉及辐射传输计算时采用矢量辐射传输模型6SV1，以补偿气体分子散射和气溶胶散射的极化效应。针对基于图像自身的大气纠正，本文首先总结了现有二类水体遥感图像大气纠正算法，提出了高光谱数据的基于查找表的耦合水气辐射传输模型的迭代与非迭代大气纠正算法，同时提取遥感反射率、气溶胶类型、气溶胶光学厚度、水质参数(悬浮物浓度、叶绿素a浓度、黄色物质吸收系数)等，其中大气辐射传输模型采用6SV1，水体辐射传输模型采用简化的生物光学模型。通过对CHRIS数据大气纠正和准同步水面实测数据的比较，发现这两种算法均能够得到较为满意的遥感反射率反演精度。

        本文最后以WHI为例，基于同步实测的靶标光谱、水面光谱、大气观测参数研究了航空高光谱成像仪的替代定标，评价可用于监测二类水体的波段范围，并基于实测大气参数建立查找表完成了WHI的大气纠正。此外，由于航空图像存在严重几何畸变，本文还提出一种基于POS数据的快速航空高光谱数据的几何纠正算法，用于航空数据大气纠正后的进一步处理。

        本文研究区域气溶胶光学特性的方法可推广到其他地区，提出的耦合水气辐射传输模型的大气纠正算法原理适合于包括可见光－近红外谱段的所有高光谱数据，具体实现时需要根据波段设置情况略微改变。

【Abstract】

The atmospheric correction of remote sensing imagery for Case 2 waters is always the main problem to be solved for inland color remote sensing. Two difficulties exist therein: 1) the optical properties of the case 2 waters (especially inland waters) are more complicated and spatially unhomogeneous with more reginal charateritics than those of case 1 waters, which is hard to meet the hypothesis in the atmospheric correction algorithms for case 1 waters; 2) the aerosol optical properties are greatly influenced by human activities and surface factors above the inland case 2 waters. Moreover, the complicated aerosol optical properties and the relative large aerosol optical thickness (AOT) result in the assumptions of some atmospheric correction algorithms questionable. So, the aforementioned probems, especially the second problem was studied during the atmosphere correction of remote sensing imagery for Taihu, the typical case 2 waters.

      Aerosol and water vapor are two main factors affecting the accuraccy of atmospheric correction of remote sensing imagery. The scattering and absorbing effects of the aerosol are more complicated. So, the retrieval algorithm of the AOT and water vapor column (WVC) were studied with the multi-band sunphotometer. The AOT retrieval accuracy was analyzed by the influence of spectral response functions. The absorption features were also analyzed outside the 940nm band for these gases such as H₂O, O₃, NO₂ ,CO₂, CH₄ by Line-by-Line Radiative Transfer Model (LBLRTM) and the gases’ fine spectral cross-sections of SCanning Imaging Absorption sepctroMeter for Atmospheric CartographHY (SCIAMACHY). The calculation algorithms of the optical depths for these gases were improved. Also, the relationship among absorption transmittance, WVC and observation zenith angle in 940nm-band was studied by LBLRTM. As a result, a new algorithm was proprosed with high WVC retrieval accuracy, of which the relative retrieval error was no more than 0.04 even with large zenth angle (~63°). Accordingly, a new calibration method for water absorption band which is independent to the AOTs of the neighbor bands.

      The above mentioned improved methods were applied to CE318’s in-situ data around Taihu from 2005 to 2007 to retrieve AOTs. Combining with the products of Aerosol Robotic NETwork (AERONET) of Taihu, the time-spatial distribtion of aerosol optical properties was analyzed and the seasonal aerosol types of Taihu were also modeled. These results were used to improve the accuarcy of atomospheric correction.

      Based on these studies, it was analyzed the influences of atmospheric correction accuracy for case 2 waters by errors such as aerosol type error, AOT error and the error of geometric imaging parameters( solar zenith, view zenith, et al.). Then a general method was proposed to establish atmospheric correction look-up tables. And the atomopheric correction was accomplished for European Space Agency’s (ESA) Project for On Board Autonomy/Compact High Resolution Iamging Spectrometer (PROBA/CHRIS) and airborne Wide field-of-view Hyperspectral Imager (WHI), respectively by image itself and by in-situ atmospheric data. In order to compensate the polarized effects of gases’s scattering and aerosol’s scattering, the vector radiative tranfer model 6SV1 was used in the radiative tranfer calculation.

      Aimed to image-based atmospheric correction algorithm, present algorthms of case 2 water atmospheric correction were summarized. The iterative and noniterative atmospheric correction algorithms were proposed for hyperspectral remote sensing imagery, which were the coupled ocean-atmosphere correction algorithms with look-up tables. These algorithms could retrive remote sensing reflectance (Rrs), aerosol type, AOT, water quality parameters (suspended matter concentration, chloroyphyII-a concentration and aborption coefficients of colored dissolved organic matter), et al. 6SV1 modle and the simple bio-optical modle were adopted for atmospheric and ocean radiative transfer process, respectively. After atmospheric correction for CHRIS data and compared the results with the in-situ measured Rrs of the quasi-synchronous stations, the satisfactory Rrs retrival accuarcy were abtained by both of the algorithms.

      Take WHI data as a case study: the hypersepctral remote sensing imagery was accomplished for vicarious calibration based on the synchronous measured calibraion tars’ reflectance, water Rrs and atmosphere parameters. Also the valid spectra of WHI were evaluated for case 2 waters. Then the atmospheric correction for WHI was finished based on lookup table, which was estabilshed by measured atmosphere paramters. Besides, the fast geometric rectified algorithm based on POS data was proposed for refctifying the serious geometric distortion of airborne imagery, which was used to futher process the data after atmospheric correction.

      The method for aerosol optical properties in this study can be applied to can extend to other regions. Also the principle of the coupled ocean-atmosphere correction algorithms fit to all the hyperspectral data with the spectra from visable to near infrared ranges; however, small modification may be made to these algorithms according to imagery band settings.