Multiple sources close to the resolution of 2MASS may not be resolved or split in all bands depending on the seeing or color and relative brightness of the components. If a source component that is resolved, or deblended, in one band is merged with another band in which it was not split, the resulting colors of the bandmerged source can be erroneous. This is illustrated in Figure 1 which shows a J-H-Ks color-color diagram for 3-band sources in the IDR2 PSC with b>+70o and a blend flag (bl_flg) value of 112 (red points), 121 (green points), or 211 (blue points). These inconsistently deblended sources were deblended in only band but not the two others in which they were detected. For comparison, see Figure 2, taken from Section VI.5 of the IDR2 Explanatory Supplement, which shows the color-color diagram for all 3-band IDR2 PSC sources in a 3.1 deg2 high galactic latitude region.
Figure 1 - Color-color diagram of
3-band IDR2 PSC sources in the northern galactic cap that are deblended
in J-only (blue), H-only
green), and Ks-only
(red).
Figure 2 - Color-color diagram for IDR2 PSC sources in
3.1 deg2 circular region at l=301.3o,
b=+77.9o. Sources shown in blue have SNR(Ks)>10.
The colors of a random field sample, such as is shown in Figure 2, cluster tightly around the dwarf and giant stellar color tracks. The inconsistently deblended sources shown in Figure 1 scatter over a much larger range of color space than do the field sources. The sense of the scatter is that the objects tend to be too faint in the deblended bands. For example, the sources deblended in J but not H and Ks (bl_flg = 211 - blue points) extend to much redder J-H color, but cover a relative normal range of H-Ks stellar colors.
The inconsistent deblending problem exists now in the 2MAPPS 2.x pipeline output where only passive deblending is used. Passive deblending refers to the condition when multiple sources detected within ~5" or each other are fit to PSFs simultaneously in PROPHOT to extract photometry. It will continue to be a problem even if active deblending is implemented in PROPHOT in 2MAPPS v3.0, because the chi-squared and/or magnitude thresholds used to trigger active deblending will not always trigger consistently between bands. Thus, deciding how bandmerge must handle inconsistent deblends is independent of whether or not active deblending is implemented.
Several proposals have been made on how to deal with the problems caused by inconsistent deblends, and they are outlined below. We must come to a consensus on the proper course as quickly as possible if changes are required in BANDMERGE for 2MAPPS v3.0. This remains one of the last major development liens in 2MAPPS v3.0.
Let bandmerge/bandfill operate as they have in preliminary processing. The bl_flg in the PSC indicates sources that have abberant colors, so users can avoid them if they bother to look.
Advantages:
Disadvantages:
Tom Chester has suggested an algorithm similar to that used in the IRAS bandmerge processing. In this scheme, BANDMERGE tests to determine if a deblend is consistent over all candidate bands in which a source is detected. If each band in the merge has a consistent deblend, then the merge is made between the relevant components of the blends.
If BANDMERGE finds that the candidate band sources being considered for a merge have inconsistent deblends, then the merge is carried out only in the bands in which the deblend is consistent. For the bands in which the deblend is inconsistent, a 97% confidence upper-limit is formed from measurements in the non-deblended band(s). For both sources, the magnitudes are replaced by the upper limit(s) in the non-deblended band(s).
Examples:
Advantages:
Disadvantages:
The first part of this algorithm is similar to the Chester proposal. BANDMERGE tests to determine if a deblend is consistent over all candidate bands in which a source is detected. If each band in the merge has a consistent deblend, then the merge is made between the relevant components of the blends.
If BANDMERGE discovers that the deblends are inconsistent between all bands in the candidate merger, then BANDMERGE re-blends sources in the deblended band(s). A single, unresolved (non-deblended) source results.
Examples:
Advantages:
Disavantages:
The 2MAPPS software engineers held a preliminary implementation design session to discuss the proposed plans. The "Chester" proposal was found to be far preferable in its design and coding since it treats both possible types of deblends in a consistent fashion. There was no clear idea on how to obtain the reblended source parameters other than brightness and position for the re-merge proposal.
