ivezic
commented
9 years ago I have image simulations (based on gaussians, but good enough for this context) that I used to measure performance of various star-galaxy separators (including the SDSS method and the optimal Bayesian method). I also have a decent recipe for calculating galaxy size distribution vs. magnitude. I will cook up a metric that will give what you asked for: what fraction of galaxies are resolved at X% confidence level as a function of magnitude and achieved SNR at that magnitude. This morphological result can then be combined with color-based and other schemes. If we did this together, we'd reach the end result much faster, let's talk!
ctslater
dnidever
drphilmarshall
willclarkson
akvivas
Much of the Galactic structure science (with non-variable stars) will be limited in depth by our ability to separate stars from small unresolved galaxies. This is mostly dependent on seeing for morphological s/g separation and photometric precision for color-based s/g separation. We would like to be able to translate the conditions of numerous epochs into a limiting depth at which we experience e.g. 10% or 50% contamination by galaxies.
Suggestions for how to perform this exercise are welcome. My first thought is that this will likely require modeling how we plan to perform s/g separation (psf minus aperture mag is probably not the Theoretically Optimal solution), including a realistic background population of galaxies. We will probably want individual metrics for morphology and color separation.