SV Efforts by the Large Scale Structure SWG

A preliminary plan outlining verification and validation tests for the large scale structure group was presented at the Stanford (!!!) collaboration meeting (docDB #3945). The latest version (though outdated too) describing DC6B tests and some ideas for SV is in (docDB #4497).

The following are is a first shot for SV activities concerning LSS. After first round of discussion, this could find its way into a CWG-like doc.

Broadly speaking, LSS needs to cover the following goals, in 4 rough categories:

Mask: no specific operational requirement

  • does its depth represent the real depth of the images? Plot SExtractor error in magnitude for objects against residual of magnitude of object-magnitude_limit: should be 0.11 at zero residual.
  • is it accurately mapping the area? Use the polyid function over the catalog and the mask (without star holes): <1% of objects should be marked as not in the mask (careful with noise though).
  • is it correctly marking stars, bad columns?: will need visual inspection for this of a few images with saturating stars, known CCD defects.

Contaminants: ideally, full-depth observations including spectroscopically identified stars and galaxies (S/G separation); darks of varying exposure(?) for artifacts; full-depth observations on fields with photometric data to i>24 (object detection completeness).

  • (detection completeness): build detection completeness vs S/N plot (photometric field). Could be enhanced separating stars and galaxies. In this case, using the expected counts from well known luminosity functions for specific populations of galaxies could be used.
  • (s/g separation): purity of sample (using spectroscopic field and matched objects) > 99% up to magnitude limit
  • (artifacts): using darks and photometric field, evaluate potential contamination of artifacts passing cuts. Purity > 99% up to magnitude limit

Homogeneity: as large area as possible, not necessarily to full-depth; known star field up to i<21 in this area

  • (detection completeness): use known star field to verify detection completeness has not a large scale variation (only up to intermediate magnitudes) < 2%.
  • (s/g separation):
  • (artifacts): (does this make sense? It will when tiles are made... test TBD)
  • (calibration errors): use known star field to check median errors. Plot median_rms vs scale of sub-areas in which we divide the area. Median_rms < 2%.

Photoz: full-depth observations including galaxies with external spectra, at least LRGs, ideally all types.

  • Perform (photo-z challenge) battery of tests (at least for low redshifts). How much statistics we need vs what will be available at first may be first constraint.