Evaluating semi-analytical algorithms for estimating inherent optical properties in the South China Sea.

Using large amounts of bio-optical data collected in the South China Sea (SCS) from 2003 to 2016, this study checks the consistency between well-known semi-analytical algorithms (SAAs)-the quasi-analytical algorithm (QAA) and the default generalized inherent optical property (GIOP-DC)-in retrieving the non-water absorption coefficient (a(λ)), phytoplankton absorption coefficient (a(λ)) and particulate backscattering coefficient (b(λ)) from remote-sensing reflectance (R(λ)) data at 412, 443, 490, 531, and 555 nm. The samples from the SCS are further separated into oligotrophic and mesotrophic water types for the comparison of the SAAs. Several findings are made: First, the values of an(λ) derived from the two SAAs deliver similar performance, with R values ranging from 0.74 to 0.85 and 0.74 to 0.87, implying absolute percent error differences (APDs) from 37.93% to 74.88% and from 32.32% to 71.75% for the QAA and GIOP-DC, respectively. The QAA shows a value of R between 0.64 and 0.91 and APDs between 43.57% to 83.53%, while the GIOP-DC yields R between 0.76 to 0.89 and APDs between 44.65% to 79.46% when estimating a(λ). The values of b(λ) derived from the QAA are closer to the in-situ b(λ) values, as indicated by the low values of the normalized centered root-mean-square deviation and normalized standard deviation, which are close to one. Second, a regionally tuned estimation of a(λ) is proposed and recommended for the SCS. This consistency check of inherent optical properties products from SAAs can serve as reference for algorithm selection for further applications, including primary production, carbon, and biogeochemical models of the SCS, and can provide guidance for improving a(λ) estimation.