How precise are SeaWiFS ocean color estimates? Implications of digitization-noise errors

Various subtle but important digitization round-off and noise errors are fo und in SeaWiFS imagery. These errors often cause large relative errors at a pixel and cause pixelization or "speckling" across the image, which is par ticularly obvious in the SeaWiFS chlorophyll standard product. Using simula tions and current SeaWiFS algorithms, we show the effect of digitization-no ise errors on the SeaWiFS ocean-color data products. It is found that the S eaWiFS mission goals, namely to estimate water-leaving radiances to within +/-5% and chlorophyll-a concentrations to within +/- 35% for Case I waters, cannot always be met. For maritime aerosol conditions the errors in the es timated chlorophyll concentrations over oligotrophic waters can easily reac h +/- 60% and the absolute values over adjacent pixels can vary 2-3 fold. S everal schemes can be used to reduce this type of error, among which the sp atial smoothing of the atmospheric-correction bands is recommended. This is because the errors in these bands are propagated and exaggerated to the vi sible bands and to the chlorophyll product through the atmospheric-correcti on process. The smoothing scheme, however, will cause more pixels to be dis carded over cloud edges. Ultimately, a significant increase in both digitiz ation bits and sensitivity is essential to achieve the stated mission goals . Such improved specifications are available on MODIS. (C) 2001 Elsevier Sc ience Inc. All rights reserved.