Construct the depth map with stars or with the SNR given the background?

[fjaviersanchez]

Right now I am constructing the depth map using the dimmest star detected at 5-sigma. In principle, this should be enough to map the changes in number density (since the distribution of stars is quite homogeneous across the field of view) but, should we try also to construct the depth map using the depth computed from the noise in a PSF-like aperture?

def get_depth_map(ra,dec,mags,nside=2048):
    good = np.logical_or(np.logical_not(np.isnan(ra)),np.logical_not(np.isnan(dec)))
    pix_nums = hp.ang2pix(nside,np.pi/2.-dec[good]*np.pi/180,ra[good]*np.pi/180)
    map_out = np.zeros(12*nside**2)
    for px in np.unique(pix_nums):
        mask = px==pix_nums
        if np.count_nonzero(mask)>0:
            map_out[px]=np.max(mags[mask])
        else:
            map_out[px]=0.
    return map_out```

[egawiser]

You can create a map of background rms in PSF-sized apertures. That can then be turned into a depth map by multiplying by 5 (to get 5 sigma instead of 1 sigma) and including an aperture correction. But if what you really want is the map of 5 sigma point source detection depths (not background rms), using the magnitude of stars detected with 5 sigma significance is more direct. I worry about using the dimmest star in each HEALPixel, since there will be depth fluctuations; instead you could use the reported S/N for each star to determine what magnitude star at that location would have been detected at 5 sigma and then take a median to more robustly represent the depth. Does that seem reasonable?

[fjaviersanchez]

@egawiser, that sounds reasonable! I did something similar to the method that you suggest for a different study. I was just wondering if it would make sense to also try the other approach using the background fluctuations. Given your answer I’d say that it is probably not worth it.

[fjaviersanchez]

This is the depth using the dimmest stars in imSim dithered imsim_dithered_depth_dimmest_stars.pdf

And this is using the median magnitude that would have been detected at 5-sigma

imsim_dithered_depth_snr_stars.pdf

The results look very similar (as expected)