基于非朗伯前向模型的MODIS陆地上空气溶胶遥感反演研究

The uncertainties of aerosol retrieval from satellite observation have been generally attributed mainly to the accurate estimation of surface reflectance and reasonable assumption of aerosol optical properties in the forward model (FM, which is radiative transfer model by coupling earth and atmosphere). However, the accuracy of the FM itself would also introduce uncertainties into aerosol retrieval when an approximated FM is employed to simulate top-of-atmosphere (TOA) reflectance. Focusing on this issue, this study carries out a series of thoroughly research, in order to develop an effective algorithm for aerosol retrieval by using the accurate Non-Lambertian FM instead of the approximated Lambertian FM. Firstly, error in principle between Lambertian FM and Non-Lambertian FM has been analyzed and sensitivity study of various factors to aerosol retrieval has been conducted by using radiative transfer simulations, in order to clarify the mechanism of uncertainties in aerosol retrieval induced by ignoring surface anisotropy. Secondly, this study proposes an effective method to estimate the surface parameters in the Non-Lambertian FM. Thirdly, based on the platform of NASA MODIS land dark-target (DT) algorithm of collection 6, a new algorithm for accurate aerosol retrieval has been developed with precise Non-Lambertian FM. Finally, we conduct aerosol retrieval testing for a long time series of MODIS data to evaluate the new algorithm against ground-based observations. This study aims to provide a theoretical basis and technical support to produce a more accurate aerosol product for climate and air quality research.

在陆地气溶胶卫星遥感中，对前向模型（地-气耦合辐射传输模型）中地表反射率估计和气溶胶类型假设的误差被普遍认为是引起反演不确定性的两种主要来源。然而，前向模型自身的误差也会给反演带来较大的不确定性。本研究以此为切入点，引入精确的非朗伯前向模型替代近似的朗伯前向模型用于反演算法的构建，进而提高陆地气溶胶遥感反演的精度。依次开展以下研究：分析郎伯前向模型的理论误差，开展地表郎伯假设引起气溶胶反演误差关于模型中诸因子的敏感性试验，以阐明忽略地表各向异性引起气溶胶反演的不确定性机理；建立非朗伯前向模型中地表各向异性参数的估计方法，以NASA MODIS陆地气溶胶产品C6版暗目标算法为平台，构建基于非朗伯前向模型的气溶胶反演新算法；最后通过对试验区长时间序列的MODIS观测数据进行反演测试与结果验证，评估新算法的精度及适用性。本研究旨在为气候及空气质量研究更优秀的数据集生产提供理论依据和技术支撑。

忽略地表各向异性而采用朗伯假设的近似前向模型用于气溶胶反演，会导致较大的前向模拟计算表观反射误差，从而在反演过程中给气溶胶光学厚度（AOD）反演带来较大的不确定性。本研究围绕这一科学问题开展以下研究，及主要成果如下：. 1）推导并给出了基于朗伯和非朗伯假设前向模型的理论误差，通过模拟分析，系统地阐述了忽略地表各向异性特性所引起的气溶胶反演的不确定性程度，全面地分析了朗伯假设带来的AOD反演误差大小随地表反射率大小、地表BRDF类型、入射和观测角度以及气溶胶厚度的变化规律。我们发现了一个与前人推断完全相反的现象：地表各向异性引起的AOD反演误差随气溶胶厚度变大而变大（前人推断：气溶胶厚度较低时地表影响较大）。这一点与对MODIS的反演测试验证结果一致。. 2）通过物理机理分析与模拟结果对比，确定了李•非郎伯前向模型用于气溶胶反演的非朗伯前向模型。并通过大量的试验分析，给出了模型中地表参数估计的解决方案，该方案既保证了反演的物理机理完备性又能顾及现阶段参数估计的误差传递。以MODIS陆地气溶胶产品C6版DT算法为平台，构建了一套针对MODIS数据的基于非朗伯前向模型的DT气溶胶反演算法（简称NL_DT）。. 3）选择整个中国区域为试验区，针对三年的MODIS/Aqua数据进行了反演测试，通过与AERONET地基观测的对比验证，结果表明：新算法有效地提高了AOD的反演精度，MODIS C6版基于朗伯假设的DT算法的AOD反演系统性偏高得到了有效地校正。另外，新算法有效地改善了Ångstrom指数、细粒子比FMF的反演结果，展现出MODIS对于气溶胶粒子特性参数的反演具有一定潜力。. 本研究采用非朗伯前向模型并考虑地表的方向性反射特性，能在很大程度上减小MODIS气溶胶反演的不确定性，提高反演的精度，可为气候及空气质量研究更优秀的数据集生产提供理论依据和技术支撑。