Zhou Wen, Lin Junfang, Ma Ronghua
Appl Opt. 2019 May 1;58(13):3509-3527. doi: 10.1364/AO.58.003509.
Inherent optical properties (IOPs) play a key role in modulating an aquatic light field; they are the core link for remotely sensing water constituents based on ocean color remote sensing. Many semi-analytical algorithms (SAAs) have been developed to obtain IOPs from remote sensing reflectance (R) data; these algorithms require a forward model (FM) to link the IOPs to R. Most currently available SAAs use the FM presented by Gordon et al.[J. Geophys. Res.93, 10909 (1988)JGREA20148-022710.1029/JD093iD09p10909] (G88 hereafter) without knowledge of how other models would impact the retrieval of IOPs from R. This study evaluates the effects of two popular SAAs, namely, the quasi-analytical algorithm (QAA) and the generalized IOP algorithm (GIOP), combined with six different FMs on the retrieval of IOPs from a synthetic data set generated with Hydrolight software. The results indicated that different FMs can have quite different effects on the computed R(λ), and the effects were not uniform across the R spectrum. Of the six FMs tested, G88 and P05 [Appl. Opt.44, 1236 (2005)APOPAI0003-693510.1364/AO.44.001236] produced the best estimates of R(λ) at 350, 440, and 550 nm in both oceanic and coastal sub-datasets; they also were less impacted by changes in the particle phase function. M02 also produced a good estimation of R but only at 440 nm, and L04 performed well only in the oceanic condition. When the two SAAs were combined with the six FMs, in the oceanic condition, QAA and GIOP combined with M02 (QM02 and GM02) provided better quality for the absorption coefficient [a(λ)] at 350, 440, and 550 nm when compared with the SAAs combined with the other models. However, for the retrieval of the particle backscattering coefficient [b(λ)] in the oceanic condition, QAA and GIOP combined with L04 (QL04 and GL04) performed better than the others, and GL04 always provided a better estimation of b(λ) than QL04. In the coastal condition, QAA and GIOP combined with G88 or P05 produced slightly better quality of IOPs compared with the other four FMs. Compared with GIOP in the coastal condition, QAA combined with G88 or P05 always showed better quality of retrieval of a(λ) but weaker quality of retrieval of b(λ).
固有光学特性（IOPs）在调节水体光场中起着关键作用；它们是基于海洋颜色遥感反演水体成分的核心环节。人们已经开发了许多半分析算法（SAAs）来从遥感反射率（R）数据中获取IOPs；这些算法需要一个正向模型（FM）来将IOPs与R联系起来。目前大多数可用的SAAs使用Gordon等人[《地球物理研究杂志》93, 10909 (1988)JGREA20148 - 022710.1029/JD093iD09p10909]（以下简称G88）提出的FM，而不清楚其他模型对从R反演IOPs会有怎样的影响。本研究评估了两种常用的SAAs，即准分析算法（QAA）和广义IOP算法（GIOP），与六种不同的FM相结合对从Hydrolight软件生成的合成数据集反演IOPs的影响。结果表明，不同的FM对计算得到的R(λ)可能有 quite different effects on the computed R(λ), and the effects were not uniform across the R spectrum. 在测试的六种FM中，G88和P05 [《应用光学》44, 1236 (2005)APOPAI0003 - 693510.1364/AO.44.001236]在海洋和海岸子数据集中，在350、440和550 nm处对R(λ)的估计最佳；它们受粒子相位函数变化的影响也较小。M02也能对R进行良好估计，但仅在440 nm处，而L04仅在海洋条件下表现良好。当将这两种SAAs与六种FM相结合时，在海洋条件下，与其他模型相结合的SAAs相比，QAA和GIOP与M02相结合（QM02和GM02）在350、440和550 nm处对吸收系数[a(λ)]的反演质量更好。然而，对于海洋条件下粒子后向散射系数[b(λ)]的反演，QAA和GIOP与L04相结合（QL04和GL04）的表现优于其他组合，并且GL04对b(λ)的估计始终优于QL04。在海岸条件下，与其他四种FM相比，QAA和GIOP与G88或P05相结合产生的IOPs质量略好。与海岸条件下的GIOP相比，QAA与G88或P05相结合在a(λ)的反演质量上始终表现更好，但在b(λ)的反演质量上较弱。
