2MASS Final Catalog Source Selection
2MASS Final Catalog Source Selection Criteria


I. Useful Links Concerning Catalog Source Selection

II. Catalog Generation Working Databases

Catalog Generation Working Point and Extended Source Databases will be drawn from all scans scans meet the release scan selection criteria. The Catalog Generation Databases will consist of all sources in the selected scans that meet the following criteria:

Table 1 - Point Source Catalog Generation DB Selection Criteria

Requirement ExplanationSQL
1. Edge-Boundary Limits Source must lie >10.0 arcsec from any 3-band tile edgedist_edge_ew>10.0 AND dist_edge_ns>10.0
2. Not an artifactNo sources identified as an artifact in MAPCOR scan processingcc_flg NOT MATCHES '*[C-P]*'
3. Point Source Valid PhotometryValid measurement in at least one bandrd_flg MATCHES '*[123]*'
4. SNR LimitsScan SNR >7 OR Measurement SNR>7 (at least one band) ((rd_flg[1] != '0' AND (j_snr ≥ 7.0 OR j_msig ≤ 0.1551))
OR (rd_flg[2] != '0' AND (h_snr ≥ 7.0 OR h_msig ≤ 0.1551))
OR (rd_flg[3] != '0' AND (k_snr ≥ 7.0 OR k_msig ≤ 0.1551)))
OR
( ( (rd_flg MATCHES '[1-3]??' AND rd_flg MATCHES '?[1-3]?' AND rd_flg MATCHES '?? [1-3]') AND ( (j_snr ≥ 5.0 OR j_msig ≤ 0.2171) OR (h_snr ≥ 5.0 OR h_msig ≤ 0.2171) OR (k_snr ≥ 5.0 OR k_msig ≤ 0.2171) ) ) ) )
5. Ptsrc - Band DetectionGood detections in 3-bands AND Scan SNR >5 OR Measurement SNR>5 in at least one band (OR'd with SNR threshold)
6. Ptsrc - chi-squared limit chi-squared < 10 (in at least one band), to select against cosmic rays and meteor trailsj_psfchi<10.0 OR h_psfchi<10.0 OR k_psfchi<10.0
7. Ptsrc - N/M limitsPoint source has saturated R1 detection in at least one band, or for bright sources (J<14.5, H<14.0, Ks<13.5), a source must have been measureable on at least 3 frames (M) and must have been detected at the >3-sigma level (in frame aperture photometry) on at least 41% of the frames on which it was measured (N). Or, if a source was measured on only 2 frames, it must be detected on both, and must be a 3-band detection. The so-called N/M statistic. rd_flg MATCHES '*[3]*'
OR (rd_flg[1] = '2' AND j_m > 14.5)
OR (rd_flg[2] = '2' AND h_m > 14.0)
OR (rd_flg[3] = '2' AND k_m > 13.5)
OR ((ndet[2] BETWEEN 3 AND 4 AND ndet[1] ≥ 2) OR (ndet[2] > 4 AND ndet[1] ≥ 3))
OR ((ndet[4] BETWEEN 3 AND 4 AND ndet[3] ≥ 2) OR (ndet[4] > 4 AND ndet[3] ≥ 3))
OR ((ndet[6] BETWEEN 3 AND 4 AND ndet[5] ≥ 2) OR (ndet[6] > 4 AND ndet[5] ≥ 3))
OR (rd_flg = '222' AND ((ndet[1] = 2 AND ndet[2] = 2) OR (ndet[3] = 2 AND nde t[4] = 2) OR (ndet[5] = 2 AND ndet[6] = 2)))


Table 2 - Extended Source Catalog Generation DB Selection Criteria

Requirement ExplanationSQL
1. Edge-Boundary Limits Source must lie >15.0 arcsec from any 3-band tile edge dist_edge_ew>15.0 AND dist_edge_ns>15.0
2. SNR LimitsScan SNR >7 OR Measurement SNR>7 (at least one band) ((j_msig_k20fe ≤ 0.1551) OR (j_msig_7 ≤ 0.1551) OR (h_msig_k20fe ≤ 0.1551) OR (h_msig_7 ≤ 0.1551) OR (k_msig_k20fe ≤ 0.1551) OR (k_msig_7 ≤ 0.1551))
3. E_Score and G_score0 < e_score ≤ 1.4 or 0 < g_score ≤ 1.4 (e_score>0 AND e_score≤1.4) OR (g_score>0 AND g_score≤1.4)


Database select statments for drawning the Catalog Generation Databases:


1. Edge Boundary

Safety boundaries are used around each scan to avoid problems with non-three-band coverage due to array mis-alignment, to provide some buffer for scan "wander" when using great-circle interpolation between reconstructed scan corners, and for avoiding edge-effects in point and extended source photometry routines (i.e. source apertures or sky annuli intersecting a scan edge).

Boundaries of 10" for point sources and 15" for extended sources were used for the Second Incremental Release. No changes are proposed.

2. Artifact Flags

Sources identified by MAPCOR with artifacts during scan processing are eliminated. Sources which are considered "real", but which may be affected by a nearby artifact are passed into the Catalog Generation DB. Further artifact cleanup takes place in Final Product Generation with the DBMAPCOR process and Extended Source Queries.

3. Point Source Valid Detections

Analysis of point sources that have rd_flg values of only 4,6 or 9 show that all are spurious detections, or are detections on edges of scans.

4. SNR Thresholds

The Point and Extended Source Catalogs (i.e. PSC and XSC) will contain sources having at least one band with Scan SNR AND Measurement SNR >10:1. This conservative cut is used to insure very high reliability in the Catalogs, and compliance with the letter of the Level 1 Science Requirements for the Survey. This threshold is more conservative than that used for the IDR2.

The PSC and XSC will be generated simultaneously with fainter source lists that will still have very high reliability (>96% TBD), but that will contain sources with lower SNR's. The nominal SNR threshold for these source lists is at least one band with SNR >7:1 (TBD). This would be consistent with the Catalog threshold used for IDR2. Discussion is ongoing about using an even lower SNR threshold together with other reliability indicators for the faint source lists (see below).

5. Band-Detections

Reliability of SNR 5-7 Sources

A powerful 2MASS source reliability indicator is multi-band detection. Most spurious detection triggers such as bright star confusion halos and diffraction spikes, meteor trails, cosmic rays, hot pixels, etc. produce "sources" that do not band merge. Bright star persistence and dichroic glint detections do tend to bandmerge, but are also the most reliably identified in scan processing. Thus, three-band sources are generally highly relible, two-band sources are slightly less reliable, and single-band sources are the least reliable.

Therefore, multi-bandedness can be used together with a lower SNR threshold to produce fainter source lists that augment the PSC and XSC that are still highly reliable. Reliability, in this context, refers to a detection that corresponds to a real source on the sky, but that may suffer from some flux-overestimation bias.

The table below provides band-detection statistics drawn from the final processing (V3) Working Point Source Database. Each row gives the total and fractional number of sources in the DB that have the specified band-detection combinations. The first two columns give the counts for all sources in the DB, while the next five pairs give the band-detection counts for the subsets of the DB that have edge-sources and known artifacts filtered out (dist_edge_ew > 10 AND dist_edge_ns > 10 AND cc_flg[?] != C,D,G,P), and have at least one detected band with Scan SNR greater than the indicated limit.

These statistics were assembled while the DB is being loaded, so the total number of sources available was changing. The size of the DB increased from 45 to 48% of the predicted total size while the queries were run. The last row in the table indicates the approximate fraction of all sources that were in the DB at the time of each query. The fractional band-detection statistics are the most meaningful. The total counts must be adjusted for the DB size change.

Table 2 - Band-detection stats with 45-48% of processing complete
No edge srcs or artifacts
  All sources snr>3 snr>5 snr>6 snr>7 snr>10
Bands Det.TotalFrac.TotalFrac.TotalFrac.TotalFrac.TotalFrac.TotalFrac.
Total 566873540 1.0000 512732365 1.0000 363027227 1.0000 328037032 1.0000 289926231 1.0000 233428930 1.0000
J-only 112373878 0.1982 97590866 0.1903 40307566 0.1110 24443896 0.0745 14345630 0.0495 4346527 0.0186
H-only 72984245 0.1287 56712025 0.1106 9238229 0.0254 4394271 0.0134 2432227 0.0084 948635 0.0041
K-only 69729971 0.1230 52323885 0.1020 5400307 0.0149 3223416 0.0098 2428185 0.0084 1707846 0.0073
JH 61112022 0.1078 59505482 0.1161 57033879 0.1571 49119579 0.1497 38310661 0.1321 15716883 0.0673
JK 7439712 0.0131 6802531 0.0133 5528228 0.0152 3945704 0.0120 2511492 0.0087 641608 0.0027
HK 8245564 0.0145 7617382 0.0149 7070994 0.0195 6330905 0.0193 5378690 0.0186 4484334 0.0192
JHK 234988148 0.4145 232180194 0.4528 238448024 0.6568 236579261 0.7212 224519346 0.7744 205583097 0.8807
Fraction of Total0.45 0.46 0.467 0.48 0.46 0.455 

A consideration in how much to augment the PSC and XSC content with fainter objects is the size of the catalog generation databases that much undergo final product generation (FPG). FPG planning scoped the point source catalog generation DB at ~5.5e8 sources, with a margin of +20%. 70,744 scans are being run through 2MASS final processing and the full-sky release is covered by ~59,700 tiles. The catalog generation database for the PSC (SNR>10) will contain approximately 4.3e8 candidates. Including candidates down to SNR>7 in any band will yield a catalog generation DB containing ~5.3e8 candidates. If 3-band sources are included down to SNR>6 or SNR>5, the catalog generation DB will contain 2.2e6 or 1.0e7 more sources than the SNR>7 DB, respectively. Thus, inclusion of lower SNR 3-band sources does not introduce massive numbers of additional sources, and the resulting catalog generation DB is within the scope of planned FPG.

6. Chi-Squared Threshold

Thresholds on the chi-squared value from psf-fitting photometry are used to help filter out spurious detections of transients such as cosmic rays, hot pixels and meteor trails. They may also have an impact on marginally resolved sources. The limits currently used (<10 in at least one detected band) are those used in the previously releases.

Chi-squared limits are mostly useful for eliminating brighter spurious sources. At low SNR, photon statistics dominate the chi-squared evaluation, so chi-squared approached unity for even relatively not-point-like faint sources.

7. N/M Threshold

N-out-of-M Statistics in Final 2MASS Point Source Catalog Selection

The N/M threshold is also designed to help filter out spurious detections of transients such as cosmic rays, hot pixels and meteor trails. The lower limit on M, the number of frames on which a source could be measured, is designed to minimize noise detections in low-coverage areas.

The IDR2 N/M limits, if applied to the final processing databases, have the unfortunate effect of filtering out 3-band saturated R1 sources. Therefore, a provision is added to the N/M criteria to pass to the CatGen DB any source that has a saturated R1 measurement in any band and otherwise satisfies the preceding selection criteria. This will also pass a very small number of possibly unreliable sources that have saturated R1 measurements in one or two bands and non-detections in the remaining band(s). There will be <200 of these in the final database. They will all be examined by hand and removed if necessary.

8. E_Score and G_Score Thresolds

E_score and G_score thresholds are designed to eliminate "sources" that are not truly extended. These scores are set using the oblique decision tree logic in GALWORKS during scan processing. The limits given in Table 2 are those used in the IDR's.

Last Updated: 14 March 2002
R. Cutri - IPAC