We took data (e.g. visit 72045) with the moon near the edge of the SuNSS imaging field.  Use these data to estimate the spectrum of the moon as observed on that night.  

We can probably assume that the emission line component is approximately constant, so let's start there.

This spectrum is modified by atmospheric scattering and absorption, and this needs to be corrected.  We need to correct the measured spectrum for these effects, and probably this is best done using theoretical models of the atmosphere as provided by MODTRANS or libradtrans;  I recommend the later as it's open source (Emde).

Note that we don't actually care about this absorption, we just need to correct it to the same conditions as are present on our "science" night;  this means that we could use two libradtrans runs to correct from one to the other — but it'd be easier to clean this spectrum.

In the longer term we also need to correct for the change of the lunar spectrum with phase angle and the fractions of mare/highlands visible on the lunar disk.  I am hoping that this can be down with a low-order smooth function (e.g. a spline).