The modeling of underwater light field is essential for the understanding of biogeochemical processes, such as photosynthesis, carbon fluxes, and sediment transports in inland waters. Water-column light attenuation can be quantified by the diffuse attenuation coefficient of the downwelling irradiance (K_d) using semi-analytical algorithms (SAA). However, the accuracy of these algorithms is currently limited in highly turbid environments, such as Amazon Floodplains, due to the SAA parametrization steps. In this study, we assessed an SAA approach for K_d retrieval using a sizeable (n = 239) and diverse dataset (e.g., K_d (490) ranging from almost 0 to up to 30 m⁻¹ with mean values of 5.75 ± 3.94 m⁻¹) in Amazon freshwater ecosystem. The main framework of this study consists of i) re-parametrization of a quasi-analytical algorithm using regional in-situ inherent optical properties (IOPs) and ii) application and validation of SAA for K_d retrieval using in-situ and Sentinel-2/MSI (n = 49) derived from Remote Sensing Reflectance (R_rs). Overall, the performance of the calibrated SAA was satisfactory for both in-situ and satellite R_rs. The validation results with in-situ data achieved a Mean Absolute Percentage Error (MAPE) lower than 22%, Correlation Coefficient (R) > 0.80, Root Mean Square Error (RMSE) lower than 1.7 m^−1, and bias between 0.73 and 1.34 for simulated visible bands of Sentinel-2/MSI (490, 560 and 660 nm) (VIS). The results using MSI imagery were similar to those of in-situ, with R > 0.9, MAPE < 20%, RMSE < 1.25 m⁻¹, and bias between 0.98 and 1.10 for VIS bands, which illustrate the viability of this methodology for K_d mapping in Amazon Floodplain Lakes. Therefore, this study demonstrates a successful application of satellite remote sensing data for the spatialization of the K_d in the optically complex waters of Amazon Basin, which is essential for the ecological management of the Amazon Floodplain Lakes.

水下光场的建模对于理解生物地球化学过程（例如光合作用，碳通量和内陆水域的沉积物运输）至关重要。水柱光衰减可通过使用半解析算法（SAA）的下流辐照度的扩散衰减系数（K _d）进行量化。但是，由于SAA参数化步骤，目前在高度浑浊的环境（例如Amazon Floodplains）中，这些算法的准确性受到限制。在这项研究中，我们评估K中的SAA的方法_d检索用一个相当大的（N = 239）和多样的数据集（例如，K _d（490）为几乎为0至高达30米^-1与5.75±3.94平均值m ^-1）在亚马逊淡水生态系统中。这项研究的主要框架包括：i）使用区域原位固有光学特性（IOPs）对准分析算法进行重新参数化，以及ii）应用原位和Sentinel-2进行K _d检索的SAA的应用和验证/ MSI（n = 49）来自遥感反射率（R _rs）。总体而言，校准后的SAA的性能对于现场和卫星R _rs都令人满意_。现场数据验证结果得出的平均绝对百分比误差（MAPE）低于22％，相关系数（R）> 0.80，均方根误差（RMSE）低于1.7 m ^-1，对于Sentinel-2 / MSI（490、560和660 nm）（可见光）的模拟可见光带，其偏差在0.73和1.34之间。使用MSI图像得到的结果与原位相似， R> 0.9，MAPE <20％，RMSE <1.25 m ^-1和VIS波段偏差在0.98和1.10之间，这说明了该方法在K波段上的可行性。_d在Amazon Floodplain Lakes中映射。因此，这项研究证明了卫星遥感数据在亚马逊盆地光学复杂水域中K _d的空间化成功应用，这对于亚马逊洪泛区湖泊的生态管理至关重要。