Project

General

Profile

Mask-G2

Test a)

In this test we check if the footprint mask is underestimating the true extent, or star-masks are being wrongly applied.

Prerequisites

Any single-epoch or coadded catalogs with associated footprint mask produced.

This test can be updated incrementally as the SV proceeds.

Procedure

1 - Check visually if the catalog scatter plot (using topcat for example) and footprint-mask (using XXX) have a roughly similar footprint.
2 - Filter an object catalog with bright objects (g,r,i,z < XXX,XXX,XXX,XXX) and other quality cuts to avoid junk from source extraction -> catalog I
3 - Produce an object catalog with objects within the footprint mask.
Tool: polyid (Mangle software tool).
Input: catalog I, footprint mask (has {1,0} in each polygon, depending on whether it is inside the footprint, bright-star mask included).
Output: catalog O with the columns RA, DEC with objects within the footprint mask

Verdict

size(catalog O) > 0.99*size(catalog I). With this we make sure that the footprint is not underestimating the true spatial extent of the catalog.

To check if footprint is larger than the catalog, see Test b).

Consequences

There are several reasons for the pass-fail criterion not being met.
  • too many spurious bright objects (check features of those objects, compatibility with artifacts, autocorrelation, dN/dmag) --> better quality cuts, tweak SExtractor parameters
  • mask eliminating true objects (check star-mask automated generation, as of July 2012, this doesn't exist yet and it's done via a standard star listing) --> revisit mask pipeline
  • mask not evaluating properly the borders of survey (check distribution of objects not in catalog O) --> revisit mask pipeline

Test b)

In this test we check whether the footprint is overestimating the extent of the catalog (borders, missed holes in real data)

Prerequisites

Any single-epoch or coadded catalogs with associated footprint mask produced.

Procedure

1 - Create a dense (TBD) catalog of random objects.
Tool: ransack (Mangle software tool)
Input: footprint mask, number of random objects (TBD).
Output: catalog (R) of random objects (ra, dec) covering the same area as the true catalog (T).
2 - Match objects in catalogs R and T using a python, C script (already available from DC analyses) -> catalog O.

Verdict

Check the spatial pattern of catalog O to look for obvious errors in the footprint (structures following the borders of the catalog, large connected areas inside footprint).
TBD: numerical check. There will be real 'holes' in the object distribution, how to know if they are real, not fake from a bad footprint?

Consequences

If the unmatched patterns are around catalog T, review footprint creation , how images are being fed into the Mangle pipeline.
If the unmatched patterns are inside the connected area of the patch, the footprint has missed or is not accurate enough to follow small areas that have not been observed. Check against depth mask. If this mask DOES NOT include these regions either, revisit mask generation.

Test c)

In this test we check the depth mask accuracy in terms of magnitude limit.

Prerequisites

Any single-epoch or coadded catalogs with associated depth mask produced. (To be checked with DESDM where in the pipeline is Mangle mask computed). Catalogs must have MAG_APER and MAGERR_APER computed for 2" diameter (dubbed MAG_APER2 and MAGERR_APER2 here). The 2" is the aperture value used by Mangle mask in past Data Challenges.

This test can be updated incrementally as the SV proceeds.

Procedure

1 - Produce a catalog with a residual with respect to the magnitude limit for each entry:
Tool: polyid (Mangle software tool, has to be modified for required output, already done for LSS tests).
Input: catalog with quality cuts applied, depth mask (has a magnitude limit value in each polygon)
Output: catalog with the columns RA, DEC, MAG_APER2 - magnitude limit, MAGERR_APER2
2 - Make a histogram of MAGERR_APER2 for objects with |MAG_APER2 - magnitude limit| < 0.1

Verdict

Check if <MAGERR_APER2> = 0.11 within 2-sigma.

Consequences

Keep gathering statistics throughout SV and all bands. If no convergence, create mask by hand (to eliminate possibility of pipeline messing up). If still wrong, check with Molly!