Improving satellite data products for open oceans with a scheme to correct the residual errors in remote sensing reflectance

An approach to semianalytically derive waters' inherent optical properties (IOPs) from remote sensing reflectance (R-rs) and at the same time to take into account the residual errors in satellite R-rs is developed for open-ocean clear waters where aerosols are likely of marine origin. This approach has two components: (1) a scheme of combining a neural network and an algebraic solution for the derivation of IOPs, and (2) relationships between R-rs residual errors at 670 nm and other spectral bands. This approach is evaluated with both synthetic and Sea-viewing Wide Field-of-view Sensor (SeaWiFS) data, and the results show that it can significantly reduce the effects of residual errors in R-rs on the retrieval of IOPs, and at the same time remove partially the R-rs residual errors for low-quality'' and high-quality'' data defined in this study. Furthermore, more consistent estimation of chlorophyll concentrations between the empirical blue-green ratio and band-difference algorithms can be derived from the corrected low-quality'' and high-quality'' R-rs. These results suggest that it is possible to improve both data quality and quantity of satellite-retrieved R-rs over clear open-ocean waters with a step considering the spectral relationships of the residual errors in R-rs after the default atmospheric correction procedure and without fixing R-rs at 670 nm to one value for clear waters in a small region such as 3 x 3 box.