-
Notifications
You must be signed in to change notification settings - Fork 2
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
missing imaging photometry in metadata table #121
Comments
I began looking into this issue and determined that it's not a bug. The vast majority of these sources genuinely lack DR9 photometry and the stellar mass estimate comes from the spectrum alone. However, there are a handful of cases where In essence, the code checks if there exists a However, consider this (secondary) target:
There is a source in the But in the (Recall that the matching radius is 1 arcsec.) However, since the declination ( So the options are to:
Below is the code currently in
|
@weaverba137 @stephjuneau @sbailey this issue would need be addressed in And note that whatever we decide I do not think we should regenerate the Fuji LS/DR9 VAC, so this would only impact the DR1 (Guadalupe + Iron) VACs and database loads. |
Hi @moustakas et al., the example source indeed lacks a Tractor entry within 1" when querying against the combined tractor table. Not sure how it was selected as a secondary but it seems to be a small HII region in a spiral or tidal arm (and in good company given all the sources outside the SGA ellipse there - wow!). I agree with "do nothing" for Fuji. For others tables, I think what you proposed in terms of searching in both N and S and then picking the best one could be fine. The caveats are (1) we create inconsistencies with the formal "resolve" for joining the S+N footprints; (2) we might risk creating a cross-match to the wrong targets depending on which secondary program it is (on this topic, we might be able to make decisions on a program per program basis rather than an object per object basis -- for example the PV secondary program should be handled with care as cross-matching positions meant to be along the major or minor axis of a large galaxy to assign them their own photometry is questionable). |
Expanding on Stephanie's caveat "we create inconsistencies with the formal "resolve" for joining the S+N footprints": IIUC, elsewhere we treat the +32.375 deg boundary as the strict and only definition of whether sources should be taken from the N or the S footprints, i.e. as if there was no overlap, with the actual overlap being a expert-level debugging feature for data cross checks. In that spirit, I think you should "do nothing" and not do anything clever with checking both footprints in the overlap region. Strictly following the boundary as if the overlap didn't exist would be more consistent with how final target selection resolve catalogs are generated and the benefits to consistency outweigh the recovery of photometry for a small number of targets, especially given that a match in one but not the other is already a sign of something special/problematic about that target. Mentioning @geordie666 in case there are any caveats to how strict the +32.375 boundary is. |
To clarify: primary targets will always get the correct photometry because I'm only talking about secondary targets here for which I think it's not crazy to try to check both north and south catalogs in the very rare case of an object like |
Even for secondary targets, I advocate that you should use the same resolve algorithm as primary target selection, i.e. the +32.375 N/S boundary, without further refinements/complications. i.e. we do not do anything tricky like "if a dec>32.375 target fails target selection in the N but passes in the S, keep it", so by extension I think you also should not do anything like "if an ra,dec>32.375 location doesn't have a positional match in the N but does in the S, keep it". Perhaps the tie breaker for @geordie666 to comment on: target selection has the option for secondary targets to be position-matched to primary targets and inherit the primary TARGETID if there is a match. If a dec>32.375 secondary does not have a N match, does the code check if there is a match in the S, or is the S data ignored and this is treated as a non-match full stop? i.e. I think desispec.io.photo should resolve to the same N/S positional matching algorithm that target selection would have, if matching had been requested at the time of target selection. But perhaps that ship has already sailed and you are already "trying harder" to find a match in other ways... |
Regarding the tiebreaker: The target referenced above by @moustakas is also in the Main Survey, during which
Note that this is an Philosophically, I'm not opposed to breaching the N/S boundary when matching secondaries to primaries. But, if the question is asked purely on the basis of "what would Note that I haven't checked all possible corner cases in the code, so I can't promise One note on the N/S resolve: The boundary isn't purely at dec=32.375o. Sources are resolved based on Galactic latitude, too. Sources have to be both north of the Galactic plane (in Galactic coordinates) and north of Dec ≥ 32.375o |
From @sbailey:
I strongly disagree and I'll note that this is not what the code currently does. This algorithm also doesn't extend to other imaging datasets like DR10 where there is no north-south split and you'd be throwing away data if you did a resolve. From @geordie666:
For secondary targets, In the end, after looking at many other of these targets, I don't think we should do anything different than what we're already doing. I'm going to close this ticket since it's on the critical path for the next set of VACs but if anyone else has thoughts or comments please feel free to add them. |
It looks like in some cases the output metadata table is not populated with the input (broadband) imaging photometry.
I thought this issue had to do with not being able to measure the aperture correction, but it appears to be something else.
Unfortunately, the bug affects the
Iron/v1.0
,Fuji/v2.0
, andGuadalupe/v2.0
catalogs; fortunately, it affects only a relatively small fraction of objects.Iron/v1.0
: 43,607 (0.242%)Fuji/v2.0
: 12,942 (0.926%)Guadalupe/v2.0
: 580 (0.026%)E.g., for Iron:
I'm not sure what the resolution is, but need to make sure to create a unit test to prevent this issue from future productions.
The text was updated successfully, but these errors were encountered: