Determine model parameters

[STSpencer]

NSB requires numerous parameters, a spreadsheet containing which can be found here.

These need to be changed for use on the Cerro Paranal site with SSTCAM. In particular:

• [ ] Determine Krisciunas A and B sky brightness parameters using pyASB (https://github.com/mireianievas/PyASB) and all-sky camera data. Need all sky cam data to do this.
• [x] Determine correct degree of gaussian blurring to use (probably using pyASB with stellar data, any other alternatives?) (Almost certainly 0).
• [x] Decide appropriate gaia limiting magnitude to use in final investigations (15.0)
• [x] Decide appropriate healpix level for SSTCAM final investigations (11 seems about right, needs to produce healpix maps smaller than 1 CHEC pixel).
• [x] Determine appropriate SkyAperture shape. Currently using rectangular but circular also an option.
• [x] Determine appropriate height and width for pixel sizes on sky (currently using 0.19 degrees), this isn't stored in the ctapipe optics description as far as I can see, and is a function of SiPM selection decision.

[STSpencer]

Added photutils definable params

[STSpencer]

Part of the issue here is computational cost, NSB is a slow, computationally costly tool at the best of times. Given the current parameter setup, each fov map for the camera takes around 8 hours to run on a single core (and it's RAM heavy, so this probably rules out using the grid). The three parameters that influence the amount of time taken to run a camera pixel map the most are:

Min gaia magnitude: There's a cutoff on the minimum magnitude of gaia stars to include in the analysis. The H.E.S.S. recommended setting for science analysis is 15, but we can cut this back if we think we're not sensitive to mag 15 stars.

Healpix level: I.e. how fine a binning to use on the skymap data. At the minute I'm setting this to 11 as otherwise one healpix pixel can be significantly larger than the equivalent size of a pixel on the sky (and I'm not sure how well the photutils pixel extraction code would work if this was much lower). Basically we don't want this sort of scenario (healpix level 8) where the pixels in the camera are much smaller than the bins of the skymap:

Size of fov map This represents the number of pixels over which to calculate the summed intensity in the fov map (this is the slowest part of the calculation since you have to sum the fluxes of thousands of stars per pixel). This is currently set to 10000. We could turn this down, but the problem is that you wind up generating nan pixel intensity values otherwise.

[STSpencer]

Also, I checked and zodiacal light isn't explicitly included in the Krisciunas model, but I suspect it would be a small factor. More importantly, is there a pixel-level model of telescope albedo that ~exists? There's no way to directly include it in NSB itself, but it could be included in the ctapipe level code.

[STSpencer]

The only remaining free parameters that we have to determine are the Krisciunas A and B coefficients, and these have their own issue (#9), so I think we can safely shut this issue.