【摘要】 由于大气短波红外辐射传输散射的复杂性，目前二氧化碳物理反演算法通常假定为晴空平面平行大气.然而通常的云检测算法难以识别视场内亚像元低层水云存在的场景，直接影响了二氧化碳遥感产品的精度.本文基于三维矢量辐射传输计算方法模拟了三维亚像元水云在短波红外1.6μm波段的散射效应，并估算了忽略此散射效应可能引入的二氧化碳反演偏差.结果表明：1.6μm波段云区表现为强反射特征，阴影区表现为减弱特性.忽略三维云散射后，云区视场辐射模拟结果偏低，且相对偏差量随地表反射率的减小、太阳天顶角的升高和云层光学厚度的增大而增大；而阴影区辐射模拟结果将偏高，相对偏差量随太阳天顶角升高、云层光学厚度的增大而增大；在云区和阴影区吸收通道都表现了相对更大的偏差.估算结果显示在低反射率、高太阳天顶角和高云层光学厚度场景中，晴空平面平行假设可能引入高达10～15 ppmv反演误差，远高于观测需求.如何剔除三维云散射影响的观测，或者在反演算法中修正三维云散射效应对于提高二氧化碳反演精度将非常重要.更多还原

【Abstract】 Current operational carbon dioxide（CO₂） physical retrieval algorithms usually assume a clear-sky plane-parallel atmospheric condition due to the complexity of the multiple scattering problem in short-wave infrared radiative transfer. The existing cloud screening algorithm has difficulty identifying sub-field of view water clouds, which may cause a significant bias in the retrieved CO₂ concentration. In order to deeply understand this issue, the scattering effect of the three-dimensional sub-pixel water cloud in the short-wave infrared 1.6 μm band is simulated based on the three-dimensional vector radiative transfer calculation method. Besides, the introduced carbon dioxide retrieval bias is estimated. The results show that the cloud area shows strong reflection characteristics, while the shadow area shows weakening characteristics. In the cloud area, the radiation error made by the clear-sky plane-parallel approximation increases with the decrease of surface reflectivity and the increase of the solar zenith angle and cloud optical thickness. At the same time, the radiation error in the shadow area increases with the increase of solar zenith angle and the increase of optical depth of the cloud layer. The absorption channels show a relatively more notable deviation than the window channels in the cloud and shadow area. Because the atmospheric CO₂ information comes mainly from the radiation ratio between the absorption and window channels, the radiance simulation error will be translated into errors in CO₂ column retrieval. The estimation results show that in low reflectivity, high solar zenith angle high cloud layer optical thickness scenarios, the assumption of the plane-parallel atmosphere could introduce up to 10～15 ppmv inversion errors, which is much higher than the observation requirement. Therefore, eliminating the observation of the influence of 3 D cloud scattering or directly correcting the 3 D cloud scattering effect in the retrieval algorithm will be very important to improve the accuracy of CO₂ inversion.更多还原