Research highlights

• Compared performance of absorption decomposition algorithms in optically complex waters.

• Good performance of absorption decomposition algorithms in inherent optical property retrieval from the measured total absorption coefficient.

• Limited applicability of absorption decomposition algorithms in longer wavelengths.

• Higher errors in total absorption coefficient derived from semi-analytical algorithms.

• Need for better semi-analytical algorithms to derive accurate total absorption and absorption due to subcomponents.

Abstract

The present study evaluates the applicability of two absorption decomposition algorithms (ADA) (Zhang and Lin’s models) in the retrieval of subcomponent Inherent Optical Properties (IOPs) from coastal waters. These ADAs use measured or model-derived total non-water absorption coefficient \({a}_{nw}(\lambda )\) (total absorption coefficient subtracted water absorption coefficient) as input and provide absorption subcomponents – absorption due to phytoplankton \({a}_{ph}(\lambda )\) and coloured detrital matter, \({a}_{dg}(\lambda )\) as outputs. Coast Colour Round Robin match-up dataset, NASA’s bio-Optical Marine Algorithm Dataset (NOMAD) and Kochi (Indian coastal waters) datasets that consist of coincident measurements of remote sensing reflectance, \({R}_{rs}\left(\lambda \right)(s{r}^{-1})\) and IOPs were used for the evaluation of various models. With measured \({a}_{nw}(\lambda )\) as input, both Zhang and Lin’s models demonstrated good performance with an average spectral mean absolute percentage error (MAPE) in the range of 19–44% for the derived absorption subcomponents for all three datasets. Quasi-Analytical Algorithm (QAA) and LS2 are two semi-analytical algorithms (SAAs) that use \({R}_{rs}(\lambda )\) as input and provide \({a}_{nw}(\lambda )\) as output. The QAA-Zhang, QAA-Lin, LS2-Zhang and LS2-Lin models (combination of SAA and ADA) resulted in higher average spectral MAPE (33–59%) values for the derived absorption subcomponents as compared to other existing SAAs, owing to errors present in both SAAs and ADAs. These results indicate that improved SAAs are required to derive accurate \({a}_{nw}(\lambda )\) to improve the applicability of ADAs in remote sensing applications.

中文翻译：

---

从沿海水域测量的总吸收系数和遥感反射率反演子组件固有光学特性

研究亮点

• 光学复杂水中吸收分解算法的性能比较。

• 吸收分解算法在从测量的总吸收系数中检索固有光学特性方面的良好性能。

• 吸收分解算法在较长波长下的适用性有限。

• 从半分析算法得出的总吸收系数的更高误差。

• 需要更好的半分析算法来推导出准确的总吸收和子成分吸收。

抽象的

本研究评估了两种吸收分解算法 (ADA)（Zhang 和 Lin 的模型）在从沿海水域检索子组件固有光学特性 (IOP) 中的适用性。这些 ADA 使用测量的或模型导出的总非吸水系数\({a}_{nw}(\lambda )\)（总吸水系数减去吸水系数）作为输入，并提供吸收子组件——由于浮游植物的吸收\ ({a}_{ph}(\lambda )\)和有色碎屑，\({a}_{dg}(\lambda )\)作为输出。Coast Color Round Robin 匹配数据集、NASA 的生物光学海洋算法数据集 (NOMAD) 和 Kochi（印度沿海水域）数据集，包括遥感反射率的重合测量，\({R}_{rs}\left(\lambda \right)(s{r}^{-1})\)和 IOP 用于评估各种模型。以测量的\({a}_{nw}(\lambda )\)作为输入，Zhang 和 Lin 的模型都表现出良好的性能，导出的平均光谱平均绝对百分比误差 (MAPE) 在 19-44% 的范围内所有三个数据集的吸收子组件。准分析算法 (QAA) 和 LS2 是两种半分析算法 (SAA)，它们使用\({R}_{rs}(\lambda )\)作为输入并提供\({a}_{nw}(\拉姆达 )\)作为输出。与其他现有 SAA 相比，QAA-Zhang、QAA-Lin、LS2-Zhang 和 LS2-Lin 模型（SAA 和 ADA 的组合）导致衍生吸收子组件的平均光谱 MAPE（33-59%）值更高，原因是SAA 和 ADA 中存在的错误。这些结果表明，需要改进的 SAA 才能获得准确的\({a}_{nw}(\lambda )\)，以提高 ADA 在遥感应用中的适用性。