Standard Algorithm (as implemented and tuned for reprocessing #4):
albedo = 100*Lt(865)*t*t0/F0/mu0
if (albedo > 1.045) then cloud
Proposed Algorithm:
rhos(865) = pi/(F0*mu0*t*t0) * (Lt(865)/t_oz - Lr)
if (rhos(865) > 0.027) then cloud
where:
As an example, Figure 1 shows a full GAC orbit swath which appears to contain clear ocean pixels at the highest solar zenith angles. The second panel of Figure 1 shows the distribution of solar zenith angle, which exceeds 83-deg at the northern extreme. This image was processed to obtain chlorophyll retrievals using the standard software and settings proposed for reprocessing #4, but with the cloud masking disabled. The retrieved chlorophyll is shown in the 3rd panel. The final two panels show cloud albedo as described above, and rhos(865). The lack of retrievals above 80-deg solar zenith is due to a limitation of the atmospheric correction tables.
In Figure 2, the average albedo and average reflectance per scanline are plotted against latitude and solar zenith angle. Only pixels for which straylight-free chlorophyll was retrieved were included in the averages. The plots illustrate the fact that the cloud "albedo" of relatively clear ocean observations generally increases with solar zenith angle, with the rate of increase rising rapidly above 65-deg. In contrast, the surface reflectance distribution remains relatively flat up to 75-deg, primarily due to the fact that the change in Rayleigh path radiance has been accounted for. The high reflectance at 10-20 deg latitude is due to a dust plume. The solid horizontal lines show the threshold value above which pixels will be flagged as cloud. The threshold on rhos(865) was chosen to give similar results to the current albedo threshold at moderate solar and viewing angles, but the surface reflectance test will allow more observations at higher solar and viewing zenith angles to pass through unmasked. This is further illustrated by the scene-specific examples below.
The effect of the proposed cloud masking change on the Level-3 binned products is also shown below. Application of the proposed cloud mask has been found to increase overall coverage (in terms of number of filled 9-km bins) by 1% to 2%, with smaller gains seen in the monthly composite and larger gains seen in the daily composite. As expected, the increased coverage is seen primarily at high latitudes, where the solar zenith angle is large. The average number of samples per bin also generally increased by 1% to 2%. The most significant impact to the derived products is a general increase in mean aerosol optical thickness, especially at higher solar angles. This is not suprising, as the operational cloud mask will tend to mask even moderate aerosol loads when the solar and viewing angle is large, thus biasing the average toward lower aerosol optical thicknesses.
