When we use the BBSO or EFM methods for removing scattered light we end up with a number for the DS flux – usually in a box on the DS.
When we apply the FFM we end up with a number for the terrestrial albedo – based on all pixels in the image.
The precisions of the two approaches is not the same. In a test on synthetic images with added, realistic, noise, and realistic variability in the offset and PSF-parameter alfa we find that the
S.D. of the determined albedo is 0.12%, while the
mean error on DS intensity is 1.2 %.
This difference was seen in the Figures of the paper we submitted. The present results are a re-calculation of those simulations – it looked too good to be true in the paper, but I can conform that it is indeed the case. I speculate that it is due chiefly to more pixels being used to determine the albedo in FFM than the DS flux in the EFM.
Further calculation of the albedo from DS intensity would carry the 10-times larger error forward to the albedo!
The box is about 50×50, (a bit bigger).
Woah! that’s a huge difference! FFM is obviously the method of choice. I am
wondering if there could be more to it than just the difference in the number
of analysed pixels, because that would have to be a difference of a factor of
about 100. Is the box in the DS ~ 1% of the total frame i.e. ~ 2500 pixels, or
of order 50×50 pixels?