Deriving Ocean Color From Satellite
(Correcting For Atmospheric Aerosols)

As water changes from clean mid-ocean conditions to coastal areas (with soil particles) or highly productive waters (with phytoplankton), the ocean color changes from blue to green. By observing this change in color it is possible to estimate phytoplankton concentrations (in cases where soil particle amounts are known to be small). In both blue and green waters, the ocean appears fairly dark when viewed from satellite and light scattering from atmospheric aerosols contributes a significant portion of the light seen by the satellite. In order to derive phytopankton concentrations from satellite, the aerosol contribution to the satellite radiance must be removed. This is traditionally accomplished by measuring the satellite radiance at two longer wavelengths (~760 and 870 nm) ,where the ocean is assumed to be dark (due to increased absorption by water). , Next the slope of the aerosol radiance at these two longer wavelengths is used to select an aerosol model from some relatively small set of aerosol possibilities. Using the selected aerosol model, the aerosol radiance at shorter wavelengths is calculated  allowing one to obtain the water leaving radiance in regions where the water is less absorbing (~450-550 nm). Atmospheric correction reamins a challenge in current NASA products. This is evident in SeaWifs images processed near Hawaii. When the Hawaii volcano plume is present in some area, then the water leaving radiances predicted by the ocean color algorithm are significantly higher.   This is illustrated in the figures below. The top figure shows two regions of large aerosol loading (downwind of Hawaii Island and north of Oahu). The bottom figure shows the water leaving radiance calculated from the SeaWifs algorithm (after removing the aerosol contribution). The regions with large aerosol loading are also the regions with large water leaving radiance. This problem is discussed in Porter et al., (2002) explaining the problem which occurs due the selection of poor aerosol models. A likely solution is also discussed.