Agenda And Minutes for 10/14 10am CDT¶
1. Dial Toll-Free Number: 866-740-1260 (U.S. & Canada)
2. International participants dial:
Toll Number: 303-248-0285
Or International Toll-Free Number: http://www.readytalk.com/intl
3. Enter your 7-digit access code, followed by “#”
Your access code: 7691719
- Scope of the telecon
- Plans for star/galaxy separation
- How to use (and plans for using) the mangle mask
- Photometric systematics: photometric calibration?
Attending: Eli, Steve K., Ricardo, Felipe, Angelo, Alex, Marissa, Eric S, Javier, Anne B, Huan, Enrico, Douglas, Brian Y, Bob A., Aurelien, peter M., Wayne B., Nacho, Joseph C., Joe Z., Robert G., Keith B., William W., others...
1. Star/Galaxy Separation
Nacho discussed his tests on SPREAD_MODEL and CLASS_STAR. Tests have been performed on sims and data, specifically COSMOS (deeper than full depth). SPREAD_MODEL behaves similar to the sims, with a bit more scatter. A SPREAD_MODEL cut of "0.002-0.003" gives a 98% pure galaxy selection but only 60-70% complete (mostly due to unresolved faint galaxies). This is okay for LSS, maybe not everybody.
Maayane's multiclass should help at the faint end by using all available data. This is running and she should have results soon. Questions remain on how to incorporate these data: into the database or on the portal...?
Some concerns in LSS about correlations of star/galaxy separation and seeing.
Nacho also noted that in some fields stars live on the slightly negative side of the SPREAD_MODEL histogram. This is due to PSF modeling issues.
Long-term, it would be nice to have a true probabilistic star/galaxy classifier.
Action items: Huan will produce full-depth equivalent coadds to better test star/galaxy separation with the proper depth. Eli will look at where the stars live in SPREAD_MODEL space as a function of position in SVA1.
2. Mangle Mask
The mangle mask in the database tells the local depth where galaxies are, but do not tell where galaxies aren't, which is used to create random masks. The LSS team has used the full set of mangle flat files to create random points which is certainly an option though it would be nice to have some sort of data compression.
Aurelien has generated a healpixelization of the maps with NSIDE=2048 (1.72' side) and NSIDE=4096 (0.85' side). The healpix maps can be browsed here. Note that you will need the standard des username and password to view the files. The current SVA1 maps are NESTED ordering scheme in GALACTIC coordinates. The resolution (NSIDE) is given in the file name. Holymolys are the maps with star holes, bleed trails, etc. The value 0 is used to represent areas which have been observed, but are masked, versus the HEALPix UNSEEN value (-1.6375e+30) for regions which are not observed.
There is a question of whether we need higher resolution pixelized maps (or whether at some point we have to go to the raw flat files) for certain applications.
We had a discussion of mask issues; some bright stars and bleed trails are not properly masked. There are also "edge bleeds" which show up as problems perpendicular to the bleed trails in the coadds which are not properly masked. Many of these issues have been fixed for Y1A1, but there are no plans (or reasonable ability) to backport these fixes to SVA1 (in an, eg, SVA2).
Action items: Aurelien will produce both RA/DEC and Galactic healpix maps, clearly labeled, so that we don't have to independently mess up our coordinate conversions. Aurelien and Anne will each continue to do tests of converting the mangle limits (2" aperture) to MAG_AUTO, MAG_MODEL, and MAG_DETMODEL limits.
Eli pointed out that the "red sequence uniformity":
https://cdcvs.fnal.gov/redmine/projects/des-sci-verification/wiki/SV-A1_Color_Uniformity_From_Clusters is best with Bob's SLR, and better than Huan's Galaxy Locus. Huan will do some more tests, but right now SLR looks like the best bet for SVA1 analysis.
Right now Bob's analysis is limited to tile-by-tile fits, but he has plans to do some bilinear (or other) interpolation to remove edge transitions. Keith is set up to do SLR on smaller regions with interpolation, but hasn't run this yet. SLR fits on a smaller scale (14' x 14') with interpolation examples from SV cluster fields
Huan pointed out that more tests need to be done on the disk only MAG_MODEL magnitudes that were run on SVA1, particularly with red galaxies (no disk!).
Anne will continue her systematic tests correlating the different magnitudes with various quantities, such as sky brightness. That is, does the sky brightness, seeing, etc, impart (unwanted) structure into the galaxy catalog using the different magnitudes?
Action items: Keith will run SLR. Somebody please volunteer to do disk only MAG_MODEL tests...?