Verification Datasets Meeting 2016-04-06 Minutes

Data Management
#1

Verification Datasets meeting:
Attendees: Simon, Colin, David, Frossie, Hsin-Fang

Simon:

  • astrometric offsets between the different bands, not much headway
  • template quality coadds for differencing, defaults don’t work, not-PSF-matched
  • Kaiser coadds, adding in likelihood space, similar to “optimal coadd” method, paper came out recently
  • no movement on MOPS at the moment
  • Jess Ford, gave talk on new MOPS method, looking at clustering of points phased onto a test orbit
    would have to do billions of test orbits, but might be feasible

Colin:

Hsin-Fang:

  • nothing to update

Angelo:

  • QA database, queries to select samples
  • SQR-008 tech note
  • ideas for improving the current baseline schema
  • schema reflects what’s in the DPDD
  • need way of selecting point sources, in coadds have better method, but need one at single-frame-level
    at least to compute the image quality to see if it’s good enough to go on to coadd processing
  • Simon mentioned that there will be QA metrics from Level 1 that could be used in Level 2, these
    QA metrics are not well defined and not clear if they will be persisted
  • Colin mentioned that there will be lots of quanities that will be measured and could be persisted
    but are not specified in DPDD, this would be a good exercise to flesh out what we want to save

David:

  • COSMOS re-reduction with SDSS as reference
    https://confluence.lsstcorp.org/display/SQRE/COSMOS+DECam+data+reduction
  • still trying to figure out why the photometric scatter is larger than expected
  • more failures were because there was no reference data at that position in the sky, i.e. no SDSS coverage for the southern standard star fields
  • for calexps where there were reference data 99.8% succeeded in proccessCcdDecam
  • Angelo suggested using pixel quality flags to select good sources
  • Simon, photcalib spits out RMS per chip, at bright end, compare to scatter from multiple measurements
  • are the chip-to-chip fluxmag0 variations constant over time, would explain ~7% variations in fluxmag0
    variations in z-band, but low scatter from bright stars from multiple measurements
#2

SDSS is not very good in Cosmos.

I offered previously to run your DECam data against the PS1 catalog and provide you a DECam-based calibrated catalog, but I was crazy busy at the time. If you make the necessary science frames and calibs available for me to download, I might be able to do it now.

#3

That would be great. All of the data are on lsst-dev:/lsst8/decam/redux/cp/cosmos/.

#4

What visits should I use?

#5

Even better, what command-lines should I use?

#6

Here are the four groups of visits (the rest are of standard star fields, most at low DEC):
176837-176854
177071-177122
177337-177367
177726-177756

These are the config overrides that I use, but note that I haven’t updated to the new processCcd yet.
config.calibrate.doPhotoCal=True
config.calibrate.doAstrometry=True
config.calibrate.measurePsf.starSelector.name="secondMoment"
config.doWriteCalibrate=True
config.doDeblend=True

#7

There’s no need to reduce the entire dataset to construct a reference catalog. Please choose a subset of visits that will cover the area in each filter.

You should use the objectSize star selector instead of secondMoment.