In some of our images sky conditions were such that the halo obviously was 'too large' for good results to be expected. We seek to eliminate these images now.
One simple measure of a halo that is "too large" is by looking on the BS and finding those images where the halo drops slower than in other same-phase images. We calculate and index of halo-narrowness based on the ratio of brightness 20 pixels beyond the BS edge, and the maximum brightness in the image. We call this the Q20 index.
On nights with 'large halos' we also often see a bright sky quite far away from the lunar disc, on the DS part of the sky. We quantify this by the mean brightness of the 10x10 box in the lower DS corner of the bias-subtracted image.
We plot now the phase vs Q20 and the sky brightness vs Q20:
Each row is for one of the 5 filters, left column is Q20 against the illuminated fraction of the lunar disc (i.e. phase). Right column shows sky brightness on the DS vs Q20. Q20 is the ratio of the BS intensity 20 pixels from BS edge on the sky and maximum brightness of the image.
We see that Q20 is quite constant against lunar phase, implying that we are correctly normalizing the halo. We see only a small drop in Q20 for (perhaps) V and VE2 filters as phase becomes very small.
We see some outliers where Q20 is high compared to most of the data at same phase.
We see (right column) that sky brightness and Q20 are related - for large values of both, but otherwise Q20 is a constant.
In fitting images to models - discussed abundantly elsewhere in this blog - we realize the need to concentrate mainly on images with as little halo as possible, so we wish to eliminate the images with a large Q20 value. Where shall we set the cutoff?
On the plots the dashed line shows 1.3 times the median Q20. We (arbitrarily) choose to get rid of all images with Q20 above this line.
We also note that Q20 is largest for VE2 and smallest for B - light is scattered further the longer the wavelength is. Does this tell us anything about the scattering mechanism? I'd venture that if the scattering was basedon the Rayleigh mechanism we should see more scattering in th eblue - wonder how Mie scattering goes? At least some of the scattering is in the optics and is fixed and unrelated to the atmsopheric effects - perhaps that can be confirmed by considering some optics?
What of the sky brightness?
We see that VE2 and V have higher sky brightness values than the other filters - VE2 so more than V. Most B, VE1 and IRCUT images have sky brightness below 1 (count), while VE2 has mos between 1 and 10. V has most between 0.,1 and 1 butr a fair fraction above 1.
For VE2 we spculate that the sky brightness is realted to long-wavelength flux from the sky - this could be from water vapour - so we suggest that VE2 sky brightness is compared to MLO weather data for humidity. Since the H2O may be high up in the atmsophere and not a ground level where the hygrometer is, this may not be easy. Satellite imagery for the nights in question could be brought to play on this issue. [Student Project!].
We are uncertain why V should have sky brightness issue if nearby B and VE1/IRCUT do not.
We arbitrarily suggest to also eliminate all images with sky brightness above 2 counts (10 for VE2).
Removing such images from 'Chris list of best images' we are left with 494 in number, and the list itself is available on request.