Open rongpu opened 4 years ago
I guess it's just a coincidence that the masking radius for a G=16 star is 7.8" (G=17 -> 5.6").
This is very curious. I guess the sweeps don't have the "iterative" column in them... any chance you could go back to the tractor files and plot these maps, splitting on iterative=True/False?
Here are the plots with the split of iterative=True/False:
So it's pretty clear that all the spurious sources have iterative==True.
Is iterative detection disabled in the MEDIUM mask (and maybe that's why we don't see this problem in brighter stars)?
Yes, iterative detection is turned off for MEDIUM (and BRIGHT and GALAXY and CLUSTER)
On Tue, Jun 2, 2020 at 8:07 PM Rongpu Zhou notifications@github.com wrote:
Is iterative detection disabled in the MEDIUM mask (and maybe that's why we don't see this problem in brighter stars)?
— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/legacysurvey/legacypipe/issues/579#issuecomment-637878489, or unsubscribe https://github.com/notifications/unsubscribe-auth/AAIEH7JQFS7HWDIDQU22IJTRUWH5XANCNFSM4NQDE3MQ .
Thanks for making those plots -- that's quite striking!
Here are the same plots but for the south (all sources, iterative==True, iterative==False):
The iterative==True source show similar clustering around stars, but the overall source density looks fine. So I guess the clustering of iterative==True sources around stars does not necessarily mean they are spurious sources.
Here are separate plots for 16<G<16.5 and 16.5<G<17.0, with distance normalized to the MEDIUM mask radius, and using the dr9f (north) catalog to get better statistics:
The overdensity is still visible in 16.5<G<17.0, but the much of it is now beyond the mask radius.
Things look much better in the south, although there is still slight overdensity along diffraction spikes and flares (mostly in 16<G<16.5):
So if we want a really clean sample, we might want to extend the MEDIUM star by half a magnitude in the south as well.
The increase in masked area from extending MEDIUM star magnitude limit is very small: extending to G=16.5 increases the MEDIUM mask area by ~4%, and it's ~7% increase for G<17.0. The plot below shows the cumulative masked area (relative to the current G<16 limit):
There is an excess of sources in dr9i/j north at ~7 arcsec from 16<G<17 stars.
Examples: http://legacysurvey.org/viewer-dev/?ra=225.0404&dec=41.2355&layer=dr9i-north&zoom=14&dr9i-north http://legacysurvey.org/viewer-dev/?ra=222.6617&dec=41.9391&layer=dr9i-north&zoom=14&dr9i-north http://legacysurvey.org/viewer-dev/?ra=225.5813&dec=40.8953&layer=dr9j-north&zoom=14&dr9i-north
The plot includes all sources in the sweep catalog (no quality cuts or masks). The extra sources only appear around stars with 16<G<17, and not around brighter or fainter stars; see below for 15<G<16 (showing dr9j, although dr9i is similar).
DR8 does not have this problem.
My guess is that the iterative detection is more sensitive to the unmodeled flux near the edge of the PSFEx profile (which is ~6.8 arcsec in radius). The fact that the spurious sources are preferentially found along the diffraction spikes seems to support this. But I don't know why it does not affect G<16 stars, since as I understand the MEDIUM mask for G<16 have no special behavior in dr9j.