Agenda And Minutes for 10/28 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
- Star/Galaxy Separation Updates: uniformity...
- Bad Photometry in SVA1: Eli: Regions of Bad Photometry
- SLR Updates: Bob / Keith / Huan
- MAG_MODEL tests?
- Mangle -> depth?
- Rough guide to flags: RichardM/SophieR(Cambridge) https://docs.google.com/presentation/d/1ddqRmZ-geXzTKimzpnzkHw-ZyYYFYQ7dSF0jodKbhcA/edit?usp=sharing
- Some tests on depth with Mangle Mask (Diego)
Attending: Eli, Douglas, Angelo, Ofer, Tom, Anne, Eric, Joe C., Yuanyuan, Keith, Alex D-W, Ross C, Aurelien, Wayne, Bob, Robert G., Huan, William W., Diego, others...
1. Updates on Star/Galaxy separation
Eli reviewed the spread model locus systematic shifts, especially in the SN fields, at Spread Model Maps. These biases are mostly caused by imperfect star selection before running psfex to generate the psf model. In an email thread, Emmanuel noted that in the future a "two pass" psfex run would fix the problem: one run using the default star selection, and a second run after doing a better job at selecting stars. In the meantime, this is what we have. Nacho suggested a local threshold would be sufficient for most studies, but LSS could need a map such as this. Ofer reminded us that Maayane's neural network output will be available soon, and because it incorporates all available data it may not be as susceptible to biases. William (I think) also pointed out that the spread model error contains useful information (though Nacho noted that this was not the case in the DC6B tests). Though exactly how best to incorporate spread model error and all that we know is still TBD (Note from Nacho: in DC6B, these tests consisted in thresholds of the kind of SPREAD_MODEL+n*SPREAD_ERR_MODEL following Emmanuel's suggestions).
Meanwhile, Huan will soon have available COSMOS field coadds with no psf homogenization with a limited number of images so that it matches DES full depth as well as possible. This will allow better tests of how spread model behaves with different depth imaging.
2. Bad Photometry
Eli reviewed the bad photometry regions detected here: Regions of Bad Photometry. Unfortunately, these regions are not captured by the mangle mask which is why they show up as clusters of bad objects.
There are actually 2 types of bad regions. One is unmasked bleed masks. Robert pointed out that mkbleedmask missed about 1/100 bleed trails in SVA1. Unfortunately, it appeared that this was correlated in that if a bleed was missed once it would be missed multiple times. It may be possible to mask these easily simply by removing objects with crazy colors (exact cuts TBD), since the area covered is small enough that it may not impact mask/random point generation. The second type is scattered light. Eli will generate a pixelized map of SVA1 to see what regions need to be eyeballed. Nacho has offered to help with the hand scanning. Meanwhile, mitigation of this problem -- to ensure these don't go unmasked in the coadds -- will be discussed at the DESDM telecon.
Keith has SLR updates (linked above). Zeropoint errors appear to be 1-3% via bootstrap resampling on small 30' pixels. That seems pretty good!
Yuanyuan reports Chris Miller will be investigating the biases input by using a disk only model in MAG_MODEL.
5. Mangle depth
Diego has started using the mangle masks to select a uniform depth region for galaxy evolution studies. One issue is that the mangle mask uses 2" aperture magnitudes, while the total depth (and completeness!) depend on seeing. Diego will next be checking how uniform the MAG_AUTO/MAG_MODEL depths are given a uniform selection via the 2" apertures. The goal is to be able to know not just the depth but the completeness everywhere in this selection.
In general, I hope we can converge on some areas to use, with the understanding that some science needs want more area while some want more uniform depth...Eli will start an email thread to discuss these issues.
Diego also pointed out that some molygons in the database have depth 0 -- these should be regions that are masked by bright stars. There are also molygons that have areas tagged with > 6 Sr. Aurelien will investigate.
Richard M. and Sophie R. have started looking at the various flags that are in the catalog. The goal is to further document these so we actually know what we want to select from the database for different use cases (and which flags tell us something is universally bad).