Agenda And Minutes for 7/21 10am CDT¶
1. Dial Toll-Free Number: 866-740-1260 (U.S. & Canada)
2. International participants dial:
Toll Number: 303-248-0285
UK Toll-Free: 08004960576
Or International Toll-Free Number: http://www.readytalk.com/intl
3. Enter your 7-digit access code, followed by “#”
Your access code: 7691719
Attending: Rick K., Peter M., Sahar, William W., Robert G., Anne B., Mike J., Gary B., Alex D-W, Rich M., Erin S., Nacho, others...
The majority of the telecon was spent discussing the astrometric offsets (see linked page) and what to do about them for SVA1 and Y1+. While there will be a relative astrometry step in Y1A1, until now there was no plan in distributing the new .head files. However, these are essential for the multiepoch fits being done by the WL folks, and Robert will make sure these files are saved and distributed. There was some discussion about the possible problems of images that show up in multiple coadd tiles may have different astrometric solutions in each tile and this may cause some issues. It is agreed that the correct solution is to do a full global astrometric solution, and Gary is working on this (including tree rings, etc). However, this will not be ready for Y1. In the meantime, it was noted that (a) the Y1 algorithm fits the astrometry on all overlapping exposures (which are much bigger than tiles) so there is considerable overlap in the fits of neighboring tiles thus lessening this problem; and (b) the goal of course is that the residual offsets are small enough to not matter. This will have to be checked.
The impact of these astrometric offsets on the coadd photometry is less certain. It is expected that these problems should be worse in psf/model based magnitudes rather than auto-based magnitudes. However, Anne B. has checked the 2pt correlation for galaxies selected with i_detmodel<22.5 and i_auto<22.5 and found that they are consistent on scales of 0.05-6 degrees. However, this is a broader test.
Eli is making new coadds on the two test tiles with new astrometry (not the same algorithm as will be used in Y1A1, but it does remove the offsets and greatly reduce the rms of the residuals.) This should allow us to get a sample of the photometric offsets in magnitudes and colors caused by the less-than-ideal astrometry. (Erin, meanwhile, is testing the multiepoch fitting on these two fields).
Another issue that was brought up is: what is the end game? Eli guesses that at the very least we'll have to re-solve the astrometry for all of SVA1 to get shapes that meet our standards. But will we have to redo the coadds as well? This is an operational challenge, of course, but we shall see what the tests turn up.
Next, Nacho gave a review of the star/galaxy separation; see linked page for summary of the star/gal chat. In the end we will add new columns to SVA1 Gold (and beyond!) with improved star/galaxy separation. However, at the moment the current classifiers work well on the COSMOS training and validation sets, and not as well on the full SPT-E region with different depths and observing conditions. The next round will concentrate on fixing these problems and making sure the classifiers stick to the best physically motivated object features.
Finally, Jim discussed some offsets between his YaCal zeropoints using i-band on SPT-E and the Gold photometry. There is a strong gradient which is primarily due to Galactic reddening which is taken into account by SLR (which uses stars above the dust layer). However, when using SFD or Planck dust maps there are still significant residuals (~3.8%). Eli asked Jim if this is also seen in the color calibration -- the SLR primarily fixes the colors, and relies on 2mass for the absolute (single-band) scale. While the color uniformity has been tested with red galaxies, no such independent test has been done on the absolute calibration.