MAPCOR Purge Algorithms for Release Cleanup

T. Evans - IPAC

This document will explain the MAPCOR tuning algorithms and parameters to other 2MASS team members who will be helping clean up artifacts in the Sampler Release Point Source Catalog prior to its public release. This clean-up is necessary for several reasons:
  1. Not all bright stars are actually detected and extracted within the point source processing pipeline -- a prime example of this is beta Peg.  Of course, if a star isn't found in the source lists, MAPCOR can't find the artifacts it creates.
  2. Even when a very bright star is found in the source lists, it often has a poor-quality mag, which means MAPCOR might not have correctly found all of its artifacts.
  3. Since MAPCOR works only within a scan, it will not find any artifacts in a scan that are caused by a bright star in a nearby scan (for example, beta Peg's diffraction spikes).
  4. Any reasonably bright star located right on the edge of a scan may be extracted in one scan but not the other, but can still cause artifacts in both scans.
Thus, the clean-up will start with 2 (or more) lists:  one listing very bright stars, for items 1, 2, and 3 above, and one listing moderately bright stars found near the edges of scans, for item 4.  The faint limits for possible parent source mags are: Anything fainter than these mags should not cause artifacts.  The faint limits for all sources included in the point source catalog are: Using the mags found in the possible parent lists, the algorithms and parameters below will determine what areas of the sky, and thus what coadd images, should be examined to find possible missed artifacts.   However,  one important caveat is that we really don't know yet if these parameters, tuned using parent sources with ~ 3 or 4 < m < ~ 8, are valid for stars brighter than ~ 4th mag.  So, for those very bright stars, the sky areas determined by these calculations could very well be too small or too large, and thus must be used only as a rough guide.  When you look at the coadd images, you'll be able to tell if you need to expand the sky area to be examined.

A rough outline of the clean-up process is the following:  the person examining the coadds will, for each parent, determine  the coadd images needed and get them, overlay the catalog source positions (we don't want to look at sources already excluded from the catalog), determine which sources are actually artifacts, and add them to a list that includes the scan #, ID #, some kind of parent source ID, and the artifact type.

The positional parameters below are all given in units of arcsec (which is the size of the coadd pixels).
 

Persistence

There can be many persistence artifacts following very bright stars, but not all of them are extracted as point sources.   My advice is to start at the parent source and move outward (in the scan's trailing direction), looking for persistence sources at the proper positions, until the persistence images are no longer detectable by eye.  Note that if, for example, the 3rd persistence artifact isn't extracted as a point source, that doesn't mean the 4th or 5th or even 6th persistence artifact also won't be extracted.

The ith persistence position is approximately:

RA(i) = RA + i(xoffs)
Dec(i) = Dec + i(yoffs)
where xoffs is the average frame-to-frame x coordinate offset, and yoffs is the average frame-to-frame y coordinate offset.  The signs of the offsets depend on the scan direction.  The values to use are:
For north-going scans:
xoffs = 0.5
yoffs = 83

For south-going scans:
xoffs = -0.5
yoffs = -83
 

How quickly the persistence fades out, or in other words how large i can get, depends on the parent source's mag.  The mag of the ith persistence is approximately given by:
mag(i) = mag_parent + pers_dm(i),
where the pers_dm(i) values are:
 
i  1 2 3 4 5 6 7
 pers_dm(i) 5.8 7.6 8.4 8.8 9.1 9.4 9.8

pers_dm(i) is not well-determined for values of i above about 7.  Nominally, you must search for persistence for values of i up to the point at which the persistence mag + its sigma (which is ~ 1 mag) is fainter than the catalog mag limits given above (16.5 mag, 16.0 mag, 15.5 mag for J, H, K).  Because of the lack of good  pers_dm(i) values beyond i=7, you can only guess you'll be searching for at least 12 or 15 or more persistence sources from very bright parent stars (brighter than 3rd mag).
 
 

Filter Glints

In the north data, there are two glints per parent source in each band.  For very bright stars, however, all of the glints tend to be lost within the confusion mess in the center of the parent source.  The position of glint j, where j = 1 or 2, is:
RA(j) = RA + dRA(j)
Dec(j) = Dec + dDec(j)
where dRA(j) is the offset in RA between glint j and its parent in that band, and dDec(j) is the offset in Dec between glint j and its parent in that band.  The values of dRA(j) and dDec(j) are:
 
dRA(1) dDec(1) dRA(2) dDec(2)
J band 2 -14 4 -28
H band 1 -14 5 -6
K band 1 -15 -4 16
with an uncertainty of ~ 2 arcsec.

To be marked as a glint, a source within ~2 arcsec of the correct position must have a mag within +/- 3.0 mag of mag(j), the predicted mag of the jth glint:

mag(j) = mag_parent + glint_dm(j),
The glint_dm(j) values are:
 
glint_dm(1) glint_dm(2)
J band 6.7 10.0
H band 6.6 8.2
K band 7.7 7.2

Of course, the glint source mags will be contaminated by the "halo" if the glint is located in the central confusion region of a very bright star (which is not unusual).  In that case, you can either ignore the mag constraint of a source located at the glint position, or even ignore the glint search entirely if you decide that all of the glints are WELL within the confusion radius (but be careful here!).
 

Diffraction Spikes

The diffraction spike width is nominally 4-6 arcsec (full width), but we know that it tends to spread out as it extends farther from the parent source, and thus you should take that value as a minimum only.  The predicted length of each spike in arcsec is:
l = l_0 { 10^[ a ( m_0 - mag_parent ) ] }
where the l_0, a, and m_0 parameters depend on the band:
 
l_0 a m_0
J band 66 0.15 6.0
H band 58 0.18 6.0
K band 54 0.18 6.0

Remember, though, that the length calculation has not been well-tested for very bright (< 4 mag) parent stars, so it should only be taken as a rough guide to the length to examine.  If you find a spike is no longer visible in the image, you don't need to go farther in that direction whether or not you've reached the nominal limit; of course, if you still see the spike beyond the limit, you do need to go farther.

Any sources extracted within the diffraction spike area are either part of the spike itself or are contaminated by the spike.  In MAPCOR, if a spike source is brighter than mag_parent + diff_dm, it is considered contaminated by the spike; if it is fainter, it is considered part of the spike.  While examining the images, you can override this mag constraint when you can see that a source, though fainter than the limit, is a real source.  The values of diff_dm for each band are:

 

NICMOS Stripes

The NICMOS horizontal stripes caused by very bright stars are located at the star's Dec and at Dec +/-  256 arcsec, and extend across the entire coadd image.  They are about 8 arcsec wide (full width, north to south), and often have ghosts at the locations (RA +/- 0, 256"), (Dec +/- 0, 256").  In fact, for the brightest stars, you might actually see vertical stripes located at the star's RA and at RA +/- 256 arcsec; the ghosts' locations are the intersections of the vertical and horizontal stripes.  (I've seen hints of these vertical stripes on some images, but never plainly enough to be sure they're real.)

These stripes are much fainter than diffraction spikes; MAPCOR does not even search for them if parent star's mag is fainter than 6.7 in the J band, 6.2 in the H band, and 5.7 in the K band.  For stars brighter than those limits, anything found within the stripe areas is marked as contaminated by the stripe.
 
 

Confusion

There are two basic types of confusion for bright stars: The confusion radius, unfortunately, is one of the more difficult MAPCOR parameters to tune using a large sample of parents, essentially because it's hard to decide at what distances the halos no longer contaminate neighboring source extractions or cause false extractions.  Chas has looked at a few bright stars and OK'd the parameters that way.  MAPCOR's predicted size of the confusion radius in arcsec is:
r = r_0 { 10^[ b ( m_0 - mag_parent ) ] }
where the r_0, b, and m_0 parameters depend on the band:
 
r_0 b m_0
J band 12.3 0.197 6.2
H band 9.4 0.216 6.2
K band 7.1 0.265 6.2

Once again, this calculation has not been well-tested for very bright (< 4 mag) parent stars, so it should only be taken as a rough guide.   As for the diffraction spikes, it is important to determine if the confused source is real or an artifact (one of many R2-1 extractions in the center of a bright star), because the former are still allowed in the catalog, but the latter are not.
 


e-mail: T. Evans
Last modified: 24 Sep 1998