This README isn't a proper documentation, will more be an overview of the code, and will explain which bits are hard-coded, that you probably want to change if you're going to use it. I'll chuck TODOs throughout for anything I think is dodgy that might need checking as well

As a brief overview, I wrote this to code to have a large and deep as possible sky model for SKA_LOW at 150MHz, that could be processed with as little GPU time as possible. To do that, the code takes the the large-sky FITS files linked in the original TRECS paper (https://doi.org/10.1093/mnras/sty2603) and does the following:

1. Does a flux cut to reduce the number of sources for less compute time in OSKAR sims
2. Moves the tables from DEC centre $\delta_0=0^\circ$ to $\delta_0=-26.7^\circ$ so it's directly over the MRO
3. Converts the TRECS sources into OSKAR sky models. SFG galaxies are either a single point source or a single gaussian, AGN are a combination of 4 gaussians to try to create hot-spots and lobes, and a point source for the central core.

As an aside, this code was written to work with python2.7, and makes OSKAR (v2.7) sky models.

(Included this in case it's useful). You can use this bash script to obtain the data, via the wget command:

./wget_the_data.sh

(The total data set is > 30GB so download what you need/want if stuck for space)

The following command:

python flux_cut_and_reproject.py --fitsfile=catalogue_AGNs_complete_wide.fits \
    --flux_cut=0.5

will produce the catalogue catalogue_AGNs_complete_wide_reduced.fits. The argument --flux_cut=0.5 cuts anything below 0.5mJy (which were the settings I ran with for my talk in Manchester). flux_cut_and_reproject.py also has the option to do a flux cut by confusion noise for a given maximum baseline length.

It also takes the current declination field centre of $\delta_0=0^\circ$ and moves it to $\delta_0=-26.7^\circ$, attempting to scale the RA appropriately depending on the new declination. TODO Remove/check the trigo of the co-ordinate transform.

Finally, the script on takes a certain amount of columns from the original cats, just enough to calculate a single SI for each source, and some size parameters. TODO this is all hard-coded to 150MHz so would need to be adapted.

Running the command

python create_osm.py --TRECS_tables=files_to_convert.txt \
    --osm=TRECS_sky.osm --plot_AGN --plot_SFG

will convert the *reduced.fits listed in files_to_convert.txt into an OSKAR sky model. The SFG are either represented by a single gaussian or just a point source, depending on how big they are (TODO you could make that a user supplied argument - I set it to 30 arcsec).

The AGN are built using one point source and 4 gaussians; I simplify an AGN into consisting of a point source core, 2 gaussian hotspots, and 2 gaussian lobes (to try and get some kind of diffuse emission from the AGN). I use the TRECS $R_s$ and size params to set the location and size of the hoptspot and lobes. I've heavily commented the code to try and explain how I do that.

The optional arguments --plot_AGN and --plot_SFG show you examples of the sources being included. Example AGN look like this:

TODO The amount of flux split between the core, hoptspots and lobes is hard-coded and the same for all AGN - you probably want to randomise that, and maybe give options to the user for acceptable ranges?

As an added bonus, I've included my attempt at taking an .osm sky model and convolving it with a restoring beam, to anticipate what a perfect CLEANed image of the sky model would look like. It's all hard coded and I haven't had time to comment it, but I've included it in case it's useful for Fred. Currently you just do

python plot_osm.py

which produces the raw sky model TRECS_sky_raw.fits and the instrument convolved TRECS_sky_convolved.fits (I haven't included these FITS files because they are >100MB each, but they are quick to generate). The convolved FITS looks like this: