Predict the ocean's biogeochemical response to and its influence on climate change;
Predict the variability in the structure of the phytoplankton community and its links with higher trophic levels as well as ocean biogeochemistry; and
Develop the scientific basis necessary to manage the sustainable resources of the coastal marine ecosystem effectively.
To address these three objectives, the remote sensing community has developed an evolving suite of satellite sensors to collect measurements of ocean color in the visible portion of the electromagnetic spectrum. The basic measurement of phytoplankton biomass relies on the strong absorption of visible light by chlorophyll (the primary light-harvesting pigment), which peaks near 443 nm (Gordon and Morel, 1983; Kirk, 1994). This absorption characteristic is a robust feature across a broad range of productivity levels in the world's oceans.
The challenge for spaceborne sensors is that 80 to 90 percent of a satellite-sensed signal originates in the atmosphere (Gordon and Morel, 1983). Much of this atmospheric signal is Rayleigh (or molecular) scattering, primarily from stratospheric ozone. After accounting for satellite and solar geometry for a particular scene, it is relatively straightforward to make corrections based on knowledge (or estimates) of extraterrestrial solar radiation, ozone concentration, and atmospheric pressure. However, aerosol scattering, primarily from hydrophilic particles in the marine boundary layer, is a much more complex process. It varies strongly as a function of time and location. Because it is not yet possible to make direct measurements of these aerosols and their contribution to atmospheric optical properties, the remote sensing community has relied on an indirect approach. Because the ocean is largely “black” in the red and near-infrared portion of the spectrum, it can be assumed that any radiance measured at these wavelengths originated in the atmosphere and was not backscattered out of the ocean. Relying on ratios between the remaining wavelengths, the spectral dependence of aerosol scattering can be propagated down to the short wavelengths in the blue portion of the visible light spectrum.
The atmospheric correction schemes have matured considerably over the past 20 years since the launch of the first ocean color sensor, the Coastal Zone Color Scanner (CZCS), on Nimbus-7. The original atmospheric schemes relied on locating a “clear-water pixel” where chlorophyll concentrations were low and the spectrum of water-leaving radiance therefore well known. The atmospheric correction for this clear-water portion of the image was then extrapolated across the entire scene. Obviously there are serious limitations to this approach; for example, there may not be a low-chlorophyll pixel in the image, or atmospheric properties may change significantly within an image that covers nearly a million square kilometers. The next step was to enable pixel-by-pixel correction, thus eliminating the need for an imagewide correction and a low-chlorophyll region.
As analysis and processing of CZCS data continued, it became apparent that the atmospheric correction schemes had to accommodate multiple scattering by molecules. The first-generation algorithms assumed that a photon would be scattered only once. However, at low Sun angles or at the edge of the sensor swath, the probability of multiple Rayleigh scattering increased substantially. Moreover, post-CZCS sensors— e.g., the Sea-viewing Wide Field of View Sensor (SeaWiFS)—have substantially higher signal-to-noise ratios (SNRs), which means processes such as Rayleigh-aerosol scattering become significant. Atmospheric correction algorithms for the Moderate-resolution Imaging Spectroradiometer (MODIS) incorporate an approach to these issues, and researchers are using algorithms to explore the effects of absorbing aerosols in the stratosphere, especially sulfate aerosols associated with large volcanic eruptions.
Based on these processes, a minimal band set for atmospheric correction can be defined. First, bands should be positioned to avoid specific absorption features in the atmosphere such as water vapor and oxygen. Second, at least two bands with some minimum spectral separation are necessary to characterize the spectral trends with sufficient accuracy. Lastly, bands should be placed in the near infrared (NIR) as noted above. A recent report by