An improved atmospheric correction algorithm for applying MERIS data to very turbid inland waters
An improved atmospheric correction algorithm for applying MERIS data to very turbid inland waters
复制标题
一种改进的大气校正算法，用于将 MERIS 数据应用于非常浑浊的内陆水域
DOI：
10.1016/j.jag.2015.03.004
复制
发表时间：
2015
期刊：
影响因子：
--
通讯作者：
W. Yang and T. Fukushima
中科院分区：
文献类型：
--
作者：
L. M. Jaelani;B. Matsushita;W. Yang and T. Fukushima
大气校正（AC）是利用卫星数据定量监测水质参数的必要过程。然而，对浑浊的沿海和内陆沃茨进行AC仍然是一个重大挑战。在这项研究中，我们提出了一种改进的AC算法N-GWI（新标准的戈登和王的算法与迭代过程和生物光学模型），用于将MERIS数据应用于非常浑浊的内陆沃茨（即，沃茨在864.8 nm处的离水反射率在0.001和0.01之间）。N-GWI算法结合了三个改进，以避免某些无效的假设，限制了现有算法在非常浑浊的内陆沃茨的适用性。首先，N-GWI使用一个固定的气溶胶类型（沿海气溶胶），但允许气溶胶浓度在每个像素不同，这种改进省略了一个复杂的要求，气溶胶模型选择的基础上，只有卫星数据。其次，将参考波段从670 nm移至754 nm，验证了参考波段的总吸收系数可以用纯水的总吸收系数代替的假设，从而可以避免在非常浑浊的沃茨水中对参考波段的总吸收系数进行未经校正的估计。第三，N-GWI产生一个半解析关系，而不是一个经验的粒子后向散射的光谱斜率的估计。我们的分析表明，N-GWI提高了两个非常浑浊的亚洲湖泊（日本霞浦湖和中国滇池）的大气校正的准确性，对于波长大于620 nm的波长，归一化平均绝对误差（NMAE）小于22%。然而，N-GWI在中等浑浊的沃茨中表现出较差的性能（在四个美国沿海沃茨中的NMAE值大于83.6%）。讨论了N-GWI的适用性，包括其优点和局限性。
Atmospheric correction (AC) is a necessary process when quantitatively monitoring water quality parameters from satellite data. However, it is still a major challenge to carry out AC for turbid coastal and inland waters. In this study, we propose an improved AC algorithm named N-GWI (new standard Gordon and Wang’s algorithms with an iterative process and a bio-optical model) for applying MERIS data to very turbid inland waters (i.e., waters with a water-leaving reflectance at 864.8 nm between 0.001 and 0.01). The N-GWI algorithm incorporates three improvements to avoid certain invalid assumptions that limit the applicability of the existing algorithms in very turbid inland waters. First, the N-GWI uses a fixed aerosol type (coastal aerosol) but permits aerosol concentration to vary at each pixel; this improvement omits a complicated requirement for aerosol model selection based only on satellite data. Second, it shifts the reference band from 670 nm to 754 nm to validate the assumption that the total absorption coefficient at the reference band can be replaced by that of pure water, and thus can avoid the uncorrected estimation of the total absorption coefficient at the reference band in very turbid waters. Third, the N-GWI generates a semi-analytical relationship instead of an empirical one for estimation of the spectral slope of particle backscattering. Our analysis showed that the N-GWI improved the accuracy of atmospheric correction in two very turbid Asian lakes (Lake Kasumigaura, Japan and Lake Dianchi, China), with a normalized mean absolute error (NMAE) of less than 22% for wavelengths longer than 620 nm. However, the N-GWI exhibited poor performance in moderately turbid waters (the NMAE values were larger than 83.6% in the four American coastal waters). The applicability of the N-GWI, which includes both advantages and limitations, was discussed.