The Ocean Color Estimation by principal component ANalysis (OCEAN) algorithm performs atmospheric correction of
satellite ocean-color imagery in the presence of various aerosol contents and types, including absorbing mixtures, and for
the full range of water properties (Case 1 and Case 2 waters), retrieving diffuse water reflectance with good theoretical
accuracy. It is easy to implement and has several advantages for operational processing lines: (1) It has de-noising
abilities, for it is based on principal component analysis and neural networks, (2) it is able to perform atmospheric
correction through cirrus and thin clouds, (3) it is able to retrieve water reflectance in the presence of Sun glint until a
glint reflectance of 0.2, and more importantly, (4) it is less sensitive to absolute radiometric calibration and directionality
than classical ocean-color algorithms. This allows multi-sensor merging (denoted hereafter Level 4 synthesis). These
abilities may improve dramatically the daily spatial coverage of ocean color products. In the companion paper (Part I),
the theoretical performance of OCEAN in situations of both Case 1 and Case 2 waters is presented for various multispectral
radiometers (i.e., POLDER, SeaWiFS, MODIS, MERIS). In this paper (Part II), the focus is made on OCEAN
de-noising and merging properties. The ability of the algorithm to work in situations of Sun glint and cirrus/thin clouds is
illustrated using MERIS imagery. Multi-directional merging is demonstrated using POLDER imagery (daily and
temporal merging), and multi-sensor merging using SeaWiFS and MODIS imagery (daily merging). The resulting
products do not show directional artifacts.
|