There are several possible methods or schemes to detect clouds.

  1. Maximum Temperature -- all observations of a small surface area over a relatively short period of time are compared. The highest temperature is retained as the best estimate of temperature in that area. This method is based on
    a. ocean surface features are more persistent than clouds
    b. clouds are colder than the surface.
    Caveat: This method works poorly for persistent, thin clouds.
  2. Two Wavelength Infrared -- compare temperatures from 3.7 µm and 10.5 µm. If the temperatures are the same, then we can assume the measured signal came from
    a. the sea surface, OR
    b. uniform clouds, which will probably be detected in a visual image of the area of interest.
    If the temperatures at the two wavelengths are different, then there are scattered, undetected clouds in the scene.
  3. Infrared Variability -- temperatures of clouds tend to be much more variable in space than temperature of the sea surface. Therefore, all areas having a small deviation from a mean brightness temperature close to that expected of the sea in the region are accepted as good values.
  4. Two Wavelength Visible-Infrared -- uses reflected sunlight to detect clouds on the assumption that the sea is much darker in visible wavelengths than clouds (because clouds are a better reflector of visible light than the sea surface!).

Once cloud-free pixels have been identified, the infrared radiance of the remaining pixels must be corrected for the influence of water vapor and aerosols in the atmosphere in order to obtain accurate values for SST.