In our fits of model-images to observed images we fit a model with several parameters. One of them is a ‘sky pedestal’ which is a constant that is added to the model image before fitting in order to accomodate sky brightness that is not due to the scattered-light halo.
We include this term because we figure that the sky might be uniformly bright for several reasons:
1) moonlight could be reflected from the ground up to the sky and back into the telescope
2) airglow could brighten the sky
3) unresolved stars and zodiacal lights shine from the sky behind the Moon
4) light could be scattered inside the telescope and contribute a more or less uniform ‘haze’
We show here the magnitude plot of that pedestal term against lunar phase for our 5 filters:
Error: axis labels have been switched!
In order to interpret this plot know that the pedestal term is ADDED to the model image, so a positive value of the pedestal implies that the model is representing a sky brightness, and that a negative value of the pedestal implies that the model needs to subtract something before it can fit the observation well.
Notice that the pedestal term is negative for all but the VE2 filter. That is, only in VE2 is a well fitting model one that has a sky brightness contribution. For the remaning filters the model needs a subtracted term for the fit to be good.
Notice also that the VE2 pedestal is about 4-6 times larger than the other filter’s, in absolute terms.
Notice also that the term goes to zero for large phases (i.e. near New Moon).
So – turning to speculation now – what is going on?
Only VE2 behaves as expected: As the Moon gets brighter more light has to be removed. This could be Moonlight scattered from the ground, onto the sky, and back to the telescope [See the paper by Bernstein, R. a. R. ∼A. “The Optical Extragalactic Background Light: Revisions and Further Comments.” The Astrophysical Journal 666, 663–673 (2007). They model the moonlight reflected from the ground near a telescope.]. Or light scattered inside the telescope – but not airglow (why would it depend on lunar phase?). Nor can it be zodiacal light because although it depends on lunar phase when you observe the Moon in particular it has the wrong phase-dependence – towards Full Moon the ZL near the Moon is low; towards New Moon the ZL near the Moon is stronger (because the Moon is nearer the SUn and the ZL is strongest near the Moon).
So – the pedestal for VE2 behaves as we expect widely scattered light from the Moon should behave.
By the way, let us call this light ‘ADL’ from now on – ambient diffuse light, to tell it apart from the light scattered in the halo, which has a strong profile with distance from the Moon.
But for the other filters the pedestal behaviour is opposite!
I think this could be a sign of the fitting process having a problem – the method has to specify negative pedestals for something else to work. That could perhaps also be the case for VE2 but to a smaller extent than the conventionally expected scattering: if it is bigger for VE2 a contribution could still be taken out leaving a positive pedestal for that filter.
Do we have any reasons to believe in stronger ADL for the VE2 wavelengths than the rest? The VE2 filter is of a different nature than the rest – perhaps it scatters more light? This would match the phase-behaviour.
What could the role of airglow be? While the zodiacal light and the galactic background and stars come from behind the Moon and are blocked by this, the airglow is all over the image. What happens when we fit such images with just one pedestal term? Can the procedure go wrong in the observed way?
What do we know about airglow? Does it depend on wavelength?
Let us assume that: In the VE2 filter there is a lot of scattering, and the method of fitting deals with this by specifying a large positive pedestal. For reqasons not yet understood, there is a tendency to make the pedestal a little too small – butthis is not seen in VE2 because the scattering is so large. In the other filters the scattering is less and the error comes through as a small negative pedestal.
Work on understanding why the pedestal needs to be a small negative number for the other filters. Perhaps it is a consequence of some of the choices of frozen parameters in the model. We do freeze ‘core factor’ and ‘rlimit’, for instance. Optimally, the factors that need to be frozen should be done so at values that do not induce strange problems such as the present one.
Secondarily, find out if the small negative pedestal influences albedo determinations significantly. With the DS intensity being proportional to albedo we can see that the small negative pedestal values directly correspond to albedo biases in the 0.2% range.