Atmospheric correction of ocean color imagery: test of the spectral optimization algorithm with the Sea-viewing Wide Field-of-View Sensor

We implemented the spectral optimization algorithm [SOA; Appl. Opt. 37, 556 0 (1998)] in an image-processing environment and tested it with Sea-viewing Wide Field-of-View Sensor (SeaWiFS) imagery from the Middle Atlantic Eight and the Sargasso Sea. We compared the SOA and the standard SeaWiFS algorit hm on two days that had significantly different atmospheric turbidities but , because of the location and time of the year, nearly the same water prope rties. The SOA-derived pigment concentration showed excellent continuity ov er the two days, with the relative difference in pigments exceeding 10% onl y in regions that are characteristic of high advection. The continuity in t he derived water-leaving radiances at 443 and 555 nm was also within simila r to 10%. There was no obvious correlation between the relative differences in pigments and the aerosol concentration. In contrast, standard processin g showed poor continuity in derived pigments over the two days, with the re lative differences correlating strongly with atmospheric turbidity. SOA-der ived atmospheric parameters suggested that the retrieved ocean and atmosphe ric reflectances were decoupled on the more turbid day. On the clearer day, for which the aerosol concentration was so low that relatively large chang es in aerosol properties resulted in only small changes in aerosol reflecta nce, water patterns were evident in the aerosol properties. This result imp lies that SOA-derived atmospheric parameters cannot be accurate in extremel y clear atmospheres. (C) 2001 Optical Society of America.