中国近岸水体泥沙含量较高，且变化梯度大；泥沙浓度可从几千mg／l变化到1mg／l以下。针对上述浑浊二类水体的大气修正一直是水色遥感应用的难点之一。利用水体近红外固有光学特性的Arnone光谱迭代算法在泥沙含量较低时适用，其近红外光谱迭代关系与根据2003年春季黄东海水色试验数据得出的现场光谱关系大致相当。当泥沙含量达到某一程度时，该算法失效，导致蓝光等较短波段的离水辐亮度为负。根据现场数据分析结果，这一分界点大概在泥沙浓度10—20mg／l。因此，本文首次提出将中国近岸浑浊水体进一步区分为中低和中高浑浊水体，并给出初步的划分标准，采用光谱优化方法对中高浑浊水体进行水色大气修正。优化误差函数的选取以现场试验获取的可见光波段光谱关系式为基础。结果表明，优化算法在近岸高浑浊水体可给出满足光谱分布规律的反演结果。与其他大气校正方法一样，优化方法也需要进一步的微调。将Gordon标准算法、Arnone光谱迭代和优化方法结合，对SeaWiFS图像进行处理，分别得出归一化离水辐亮度和总悬浮物（TSM）浓度分布图像，结果令人满意。

The suspendedmatter concentrations are very high in coastalwaters ofChina, and the variations of\nconcentrations gradients are also great. The concentrationsmay vary from several thousandsmg/l to less than\n1mg/.l Atmospheric correction for above turbid waters has been always the unsolved problem in ocean color\nremote sensing. TheArnone iterationmethod using inherent optical properties ofnear-infrared(NIR) bands is\njust applicable in lowly turbidwaters, whose spectral relationsbetween red andNIR bands aremostly smi ilar to\nthatderived from ocean color expermi ent data in Yellow Sea and East China sea, spring, 2003. But the\niterationmethod failed, and thewater-leaving radiances atshorterbands such asblue bands are negative, when\nthe suspendedmatter concentrations exceed 10—20mg/,l which is the result of in situ datamentioned above.\nTherefore, in this articlewe distinguish turbidwaters into low and high turbid waters and give the prelmi inary\ncriterion. Optmi izationmethod has been applied to atmospheric correction of highly turbid waterswhose error\nfunction is determined by spectral relations between various visible bands from above in situ data. The results\nshow that the optmi izationmethod could get reasonablewater-leaving radiances, although itneeds further fine-\ntuning. The normalizedwater-leaving radiances and totalsuspendedmatterconcentrationsderived from SeaW iFS\naremostly satisfied when Gordon standard algorithm, Arnone iteration method and optmi ization are used in atmospheric aorrection.