Appendix 4. 2MASS Calibration Scan Working Databases and Atlas Images

2. General Properties of the Calibration Scan Products

a. Scope of the Calibration Scan Atlas and Working Databases

b. Calibration Field Properties

Table 1 - Calibration Field Scan and Data Properties
Tile1Center2Coverage Limits3Scan Footprints4All-Sky PSC Comparison
RA_J2000 (deg)Dec_J2000 (deg)RAmin(deg)RAmax(deg)Decmin(deg)Decmax(deg)RA_bias5Dec_bias5Photom.6

Notes to Table 1

  1. Click on the Tile number to see a 3-color J,H,Ks composite image of the field from one representative scan. An asterisk denotes LMC/SMC calibration tiles.
  2. Average central coordinates of all scans of a field. This may not coincide with the mean geometric area because some fields have a few outlying scans.
  3. Approximate coordinate boundaries of the smallest simple rectangle that encloses all scans of each field. These do not define precisely the coverage area of each field which can be irregular.
  4. Sky maps showing the outlines of all scans covering each calibration field. The maps show a 1°x2° area in cartesian projection with the RA scale exaggerated to emphasize the scan cross-stepping.
  5. RA and Dec offsets between average Cal_PSWDB positions and All-Sky PSC positions, plotted as a function of RA and Dec within each calibration field. Average Cal-PSWDB positions are taken from the Merged Calibration PSWDB
  6. Photometric residuals between average Cal-PSWDB point sources and All-Sky PSC, plotted as a function of source magnitude. The average Cal-PSWDB magnitudes are taken from the Merged Calibration PSWDB

i. Calibration Tiles and Sky Coverage

2MASS calibration observations cover approximately 6 deg2, distributed in 40 separate fields, or tiles.

The calibration field tile number listed in the first column of Table 1 is the unique identifier given to each field on the sky. For the main survey calibration tiles, the tile number is derived from the primary calibration star(s) in each field (see III.2.d). Magellanic Cloud calibration tile numbers (A3.4.c), denoted with "*" in Table 1, were set arbitrarily to 90298 and 90299 for the SMC fields, and 90400-90402 for the LMC fields. The tile number is contained in the Calibration Scan Information Table, to provide a simple way to search metadata for individual calibration fields.

Each calibration scan swept over an area 8.5' wide in RA and 1° long in declination. The first scan of the six comprising a calibration observation was centered on the nominal tile position. The 2MASS telescopes were cross-stepped 5" east in RA between each subsequent scan of the observation set to minimize pixelization effects. So an area of approximately 0.15 deg2 was covered by each observation.

The footprint of all scans covering each calibration field can be seen by clicking on the Scan Footprint links in Table 1. In most fields, the observations were well-registered, resulting in consistent coverage patterns such as those seen in 90182 or 90234. A small fraction (<5%) of the scans in 10 fields were displaced from the nominal position due to occasional telescope commanding errors or small pointing variations. This produced more irregular coverage patterns such as those seen in 90813 and 92202.

The Center RA and Dec listed in Table 1 are the average of the reconstructed center positions of all scans of each field. This may not correspond to the exact geometric center of coverage because of irregular coverage in some fields (e.g. 90273). The Coverage Limits columns in Table 1 give the approximate coordinates of the smallest rectangular region that contains all the scans of each field. These limits can be used to select source extractions from a individual fields when querying the calibration Working Databases. Because of the irregular coverage, the depth of coverage within these limits can range from zero at the extreme boundaries, to hundreds or thousands near the field centers.

ii. Photometric Properties

The 2MASS calibration observations were taken with the same exposure time and scanning method as the main survey observations, and photometry was extracted using the same reduction software. As a result, the photometric accuracy and precision achieved by the calibration scan data are very similar to the levels achieved for the All-Sky PSC and XSC. Because of this similarity, much of the characterization and validation of the main survey photometric performance described in sections VI.1, VI.2 and VI.3 are based on analyses of the calibration scan point source photometry.

The average SNR=10 sensitivity level achieved for point source photometry in the 2MASS calibration scans is J=16.3, H=15.4, Ks=14.8 mag. However, the sensitivity achieved in all scans of a given field may span a range of up to 1 mag because of varying atmospheric conditions.

Direct measurements of the photometric sensitivity were obtained for every 2MASS calibration observation using the repeatability of point souce photometry within the six scans comprising the observation. Figure 1 shows the measured root variance in brightness plotted as a function of the mean point source magnitude of each source detected in at least five of the six scans in one observation of the 90249 calibration field. The magnitude at which the root variance distributions intersect the horizontal line at =0.1086 mag is the SNR=10 sensitivity level for the observation. This metric was measured for every 2MASS calibration observation taken under a wide variety of observing conditions, which made possible the derivation of the empirical relationship between achieved photometric sensitivity and atmospheric transparency, seeing and background levels described in VI.2.

Figure 2 shows the distributions of the SNR=10 magnitudes for all northern and southern observatory calibration scans computed using the relationship in VI.2. The peaks of the distributions occur near J=16.4, H=15.4 and Ks=14.7 mag. The distributions in each band have an extent of nearly one magnitude, so the calibration observations of any one field can span a large range of sensitivity. Ks-band images from the two scans of the 90013 calibration field that have the worst and best estimated sensitivities are shown in Figure 3. Note the differences in image quality due to seeing and background structure. The SNR=10 levels of the two images differ by ~0.8 mag. The distributions in Figure 2 are qualitatively similar to those for the main survey. Both have comparable average and maximum sensitivity levels. However, the calibration data include scans with sensitivities that extend to lower levels than the survey scans because the calibration data were not required to meet the survey's sensitivity specifications. 11.5% of all calibration scans achieved SNR=10 levels at magnitudes brighter than the survey's required levels (J=15.8, H=15.1, Ks=14.3 mag). The low sensitivities most commonly occur in the H-band (10.3% of the total) because of atmospheric OH airglow emission.

Figure 1 - Point source photometric repeatability as a function of mean magnitude measured in 6 scans of the 92409 field on the night of 11/16/1997 UT. Black crosses are the root variances for individual stars detected multiple times. The green points are the average RMS levels measured in 0.5 mag wide bins. The horizontal dashed line marks the SNR=10 (=0.1086 mag) levels. The vertical dashed lines represent the SNR=10 levels required for the main survey. Figure 2 - Distribution of SNR=10 point source sensitivity levels achieved in all 2MASS calibration scans, estimated using atmospheric transparency, seeing and background levels. These sensitivity levels do not take into account the limitations of confusion.
Figure 3 - Ks images showing the same 7'x7' region in the 90013 calibration field from two scans with significantly different achieved sensitivities. (left) Scan 37 from 12/07/1998 UT, SNR=10 @ 14.23 mag. (right) Scan 95 from 11/29/1999 UT, SNR=10 @ 15.05 mag.

There is no measurable bias between Calibration and Survey observation photometry.

The links in the "All-Sky PSC Comparison/Photom." column in Table 1 lead to plots that compare point source photometry from the Cal-PSWDB and the All-Sky PSC for each calibration field. The average Cal-PSWDB point source brightness is used for these plots, and is taken from the subset of Merged Cal-PSWDB sources detected at least 500 times. Figures 4 shows the Cal-PSWDB/All-Sky PSC residuals for all of the calibration fields on a single plot. The Cal-XSWDB/All-Sky XSC residuals for extended sources detected at least 300 times in all calibration fields are shown in Figure 5. Extended source residual plots for the individual fields are not provided because there are so few galaxies in the small areas covered by each.

There are no systematic biases between photometry derived from the calibration observations and the survey scans. This agreement is more of a confirmation of the survey calibration effectiveness rather than the accuracy of the calibration scan photometry, though. Calibration scan photometry has slightly better calibration accuracy than scans in the main survey because the calibration scans were photometrically calibrated using the in situ measurements of standard stars in each field (A4.4.a). As a result, calibration scan zeropoint adjustments are less affected by short timescale atmospheric transparency variations and errors in fitting the nightly zeropoint solutions.

Figure 4 - Differences between the average 2MASS Cal-PSWDB magnitude of sources detected at least 500 times and magnitudes of the same sources in the All-Sky PSC, plotted as a function of the average Cal-PSWDB magnitude. Individual points represent the differences between the average source magnitudes from the Merged Cal-PSWDB and All-Sky PSC. Contours trace the density of points. Figure 5 - Differences between the average 2MASS Cal-PSWDB 7" circular aperture magnitudes of sources detected at least 300 times and the magnitudes of the same sources in the All-Sky XSC, plotted as a function of the average Cal-XSWDB magnitude. Solid points represent the photometric differences for non-confused merged groups. Open circles are confused merges. The bright Ks-band only confused sources are detections of very red nebulosity in the 90009 calibration field in rho Ophiuchus.

Photometric measurements of bright, non-saturated sources in the 2MASS Cal-PSWDBs and Cal-XSWDBs have an internal precision of 1.5-2.0%. The intrinsic measurement scatter increases monotonically with decreasing brightness for J>13, H>12.5 and Ks>12 mag.

The repeatability of Cal-PSWDB and Cal-XSWDB photometry is demonstrated in Figures 6-8 and 9-11, respectively. These diagrams show the root variance flux (RMS) of the large number of independent measurements of sources in the two WDBs plotted as a function of the mean source brightness. The RMS values are computed using statistics accumulated during the generation of the Merged Calibration PSWDBs and XSWDBs. For point sources, the combined default magnitude is used: [JHKs] RMS = [jhk]_mstdev * sqrt([jhk]_n-1), where [jhk]_mstdev is the standard deviation of the mean flux and [jhk]_n is the number of detections going into the calculation of the mean in each band. The combined fluxes in 7" circular apertures are used for extended sources: [JHK]7 RMS = [jhk]_mstdev_7 * sqrt([jhk]_n_7-1).

The distributions each show a well-behaved locus of points that has an approximately constant RMS value of 0.015-0.020 mag for bright but non-saturated sources, and that rises systematically towards fainter flux levels. The photometric dispersions are larger for the flux estimates of fully saturated sources (J<5.5, H<5 and Ks<4.5). There is also small population of sources at all non-satured flux levels that extend to higher RMS values than the main loci.

The constant RMS value for bright sources corresponds to the fundamental precision limit to individual 2MASS measurements in the non-photon noise-limited regime dictated by PSF undersampling effects due to the large 2MASS detector pixels. The dispersion increases for fainter sources approximately as the inverse square root of the source brightness, as expected from photon statistics. Although it is not apparent from these diagrams, the growth of the point source RMS values flattens slightly faintward of J>16.2, H>15.6 and Ks>14.8. This slope change occurs because the merged source statistics include only detections in each band. Low SNR sources fainter than the completeness limits are detected preferentially when noise drives up their apparent brightness, resulting in an overestimate in flux and underestimate in measurement dispersion.

Figure 6 - J-BandFigure 7 - H-BandFigure 8 - Ks-band
Root variance (RMS) flux for all sources in the Cal-PSWDB detected at least 500 times, plotted as a function of the average brightness. The average magnitudes and RMS values are taken from the Merged Cal-PSWDB: [JHKs] RMS = [jhk]_mstdev*sqrt([jhk]_n-1). The small red points represent individual merged sources. Black contours trace the surface density of the points.
Figure 9 - J-BandFigure 10 - H-BandFigure 11 - Ks
Root variance (RMS) flux in 7" circular apertures for all sources in the Cal-XSWDB detected at least 300 times, plotted as a function of the average brightness. The average magnitudes and RMS values are taken from the Merged Cal-XSWDB: [JHKs] RMS = [jhk]_mstdev_7*sqrt([jhk]_n_7-1). The solid red points are from non-confused merged groups. Open circles are from confused groups.

The elevated measurement dispersion seen for some objects in Figures 6-11 may be caused by a number of effects, the most interesting of which is true flux variability. However, large measurement scatter is most frequently the result of measurement contamination due to confusion with one or more nearby objects or nearby image artifacts. PSF-fit measurements in the Cal-PSWDB of objects that are actually extended or marginally resolved will also exhibit artificially large measurement dispersions because of centroiding uncertainties and the poor match between the PSF and true source profile. Similarly, measurements in the Cal-XSWDB of extended "sources" that are actually detections of galactic nebulosity or compact groups of point sources will not repeat consistently because of centroiding variations and source characterization difficulties. For example, most of the confused extended sources with large photometric scatter seen only in the Ks-band are detections of red nebulosity in the 90009 calibration field in rho Ophiuchus. The few non-confused sources with larger photometric dispersion in the Merged Cal-XSWDB are actually detections of a compact, embedded clusters in the low latitude 90312 calibration field, not galaxies.

The Merged Calibration Point and Extended Source WDBs contain several cautionary flags that indicate if any of the conditions are present that may lead to artificially large photometric dispersions. These "Merge Quality" flags and their use are described in A6.2.v. To identify measurements of objects in the Cal-PSWDB and Cal-XSWDB that have the least likelihood of contamination, select entries that are associated with groups in the Merged Cal-WDBs that have:

Use of the merge photometry caution flag (ce_flg) constraint is very conservative, and may result in the rejection of uncontaminated groups. This is particularly true for the calibration WDBs because with thousands of independent measurements the probability is relatively high that one or more measurements may be affected by the conditions tracked by ce_flg.

Figure 12 shows the flux chi-squared statistic distributions for all objects in the Merged Cal-PSWDB that satisfy the above criteria and that were detected at least 80% of the time they were observed (to insure reliability). Objects with large flux chi-squared values in this clean sample have the highest probability of being true flux-variable point sources. J-band light curves of three representative objects in the 90067 (M67) calibration field drawn from the individual measurements in the Cal-PSWDB are shown in Figure 13. The measurement distribution of the first of these objects, 2MASS J08512240+1151291, shown in the top panel, has a small RMS and low chi-squared value consistent with the majority of objects in Figures 6 and 13. Its lightcurve shows no evidence for variability. The second and third objects, 2MASS J08510483+1145568 and 2MASS J08512530_1202563, have larger RMS and chi-squared values than the main locus of points in Figures 6 and 13, and the spread in their lightcurves is larger than that of 2MASS J08512240+1151291. Both of these are previously known variable stars. The former is a W UMa variable contact binary identified in the M67 by Stassun et al. 2002, and the latter is the peculiar spectroscopic binary AG Cnc (e.g. van den Berg et al. 2002).

Figure 12 - Flux chi-squared distributions for sources in the Merged Cal-PSWDB that have gcnf=0, ce_flg=0, and n_galcontam=0, >80% detection rate, and >100 detections per band. Lightcurves for the three objects denoted by the magenta circles in top panel are shown in Figure 13. Figure 13 - Lightcurves of three representative sources selected from the "clean" subset of the Merged Cal-PSWDB in the 90067 calibration field. 2MASS source designations are from the All-Sky PSC. Shown in the top panel is an example of a source with a low RMS, low chi-squared flux distribution. The middle and bottom panels are objects selected to have distributions with larger RMS and chi-squared values. These are both previously known variable stars: (center) P=0.36d W UMa variable, and (bottom) AG Cnc - a P=2.8d peculiar spectroscopic binary.

iii. Astrometric Properties

Source positions in the 2MASS Cal-PSWDBs and Cal-XSWDBs are reconstructed with respect to the USNO-A2.0 catalog. Because the Tycho 2 catalog was used as the astrometric reference for the main survey, positions in the Calibration scan products exhibit systematic offsets of up to approximately 0.6" with respect the the All-Sky PSC and XSC.

Position reconstruction for sources detected in the 2MASS Calibration observations was carried out using the USNO-A2.0 catalog as the primary astrometric reference rather than Tycho 2 which was used for the main survey (A4.4.b). Because of small, systematic offsets between these two reference catalogs, similar biases exist between the Cal-WDB positions and those in the All-Sky PSC and XSC. The "RA_bias" and "Dec_bias" columns in Table 1 contain links to diagrams that show the RA and Declination offsets between the mean position of sources detected multiple times in the Cal-PSWDB and the position of those sources in the All-Sky PSC for the individual calibration fields. The astrometric biases are position-dependent within the calibration scans, and they differ in amplitude and structure between calibration fields. However, they are consistent among all scans of a given calibration field

Figures 14 and 15 show the RA and Dec Cal-PSWDB/All-Sky PSC position residuals for all calibration fields on the same diagram. The grouping of points in RA and Dec corresponds to the location of individual calibration fields. The vertical scatter among the points for each field can be as large as 0.2-0.3", and is indicative of the systematic behavior of the astrometric bias within the fields.

Because of the astrometric biases in the Cal-PSWDB and Cal-XSWDB, users should defer to the absolute position of objects given in the All-Sky PSC and XSC whenever possible.

Figure 14 - Average declination offset between stars in the Cal-PSWDB and All-Sky PSC, plotted as a function of declination. The discrete groups of points correspond to the individual calibration fields. Average Cal-PSWDB positions taken from the Merged Cal-PSWDB. Figure 15 - Average right ascension offset between stars in the Cal-PSWDB and All-Sky PSC, plotted as a function of right ascension. The discrete groups of points correspond to the individual calibration fields. Average Cal-PSWDB positions taken from the Merged Cal-PSWDB.

2MASS Cal-PSWDB source positions have an average internal radial repeatability of 60 mas over the range 9<Ks<13.5. The scatter in position measurements for fainter sources increases monotonically with decreasing source brightness.

The consistency of position reconstruction in the Cal-PSWDB is illustrated in Figure 16 which shows the average radial separations between individual positions and the mean position of objects detected >100 times in the Cal-PSWDB plotted as a function of the mean source Ks brightness. This diagram is constructed using the sep_avg parameter in the Merged Cal-PSWDB Information Table. The positions of sources in the 9<Ks<13.5 mag range are repeatable to ~60 mas radially. Brighter sources that are measured in the 51 ms READ1 exposures exhibit slightly larger position residuals because of the effects of seeing on the short exposures. Position residuals increase with decreasing source brightness for objects fainter than Ks>13-14 mag. This positional repeatability is comparable to that of the All-Sky PSC (II.2.f).

The green circles in Figure 16 correspond to several nearby M-dwarf stars that are the primary photometric standards in several of the calibration fields: BRI0021-0214 in 90021, LHS191 in 90191, LHS2026 in 92026, LHS2397a in 92397, TVLM 868-53850 in 90868, and BRI2202-1119 in 92202). These objects all have significant proper motions (200-1000 mas/yr), and each stands out clearly as having large average radial separation parameters for their brightness. The astrometric dispersion for objects in the Cal-PSWDB and Cal-XSWDB may be artificially elevated by the same confusion and resolution effects discussed above that result in degraded photometry. As with the photometric precision, the sources least likely to have contaminated position reconstruction will be those that correspond to unconfused groups in the Merged Cal WDBs (gcnf=0). An important caveat to this is that high proper motion objects that move farther than 2", the correlation radius used to merge the Cal-PSWDB, during the time covered by the 2MASS observations, may be split into multiple, confused groups. For example, the individual detections of LHS191 (mu = 1049 mas/yr) is split into three groups, and would be missed in a selection of sources limited to groups with gcntr=0.

Figure 16 - Average radial separation (sep_avg) from the mean position of objects detected >100 times in the Cal-PSWDB, plotted as a function of average source brightness. The red points represent individual objects and the black contours trace the density of points. The average separation is taken from the in the Merged Cal-PSWDB Information table. Only unconfused groups not identified with extended objects are shown. The green circles indicate the position of nearby M-dwarf stars in several calibration fields with large proper motions.

iv. Completeness and Reliability

The 99% completeness limits of the full Cal-PSWDBs are in the range 15.5<J<16.5, 14.5<H<15.5 and 14.0<Ks<15.0. The spread in completeness levels is a result of different source surface densities in the calibration fields, and the variation in atmospheric conditions in which each field was observed.

Average J, H and Ks completeness curves for the 2MASS Cal-PSWDBs are shown in Figure 17. In these figures, completeness is computed as the ratio of the number of times a source was detected in a band to the total number of times it was observed, [jhk]_n/spos, using statistics accumulated in the Merged Cal-PSWDB Information Table. Although the objective of the 2MASS calibration observations was to measure the relatively bright standard stars in each field, source detection thresholds were set to the same low level (SNR~3.5) that used for the main survey scans. As a result, the completeness of detections in the calibration scans is comparable to that achieved in the main survey observations.

The completeness curves in Figure 17 exhibit nearly one magnitude of spread in the brightness at which they begin to turn over. This spread is a consequence of the very different source density regions sampled by the 2MASS calibration observations, and the differing atmospheric conditions in which each field was measured. An approximately constant SNR threshold was used for source detection during all 2MASS data processing, but the noise estimate included the contribution of source confusion. Therefore, high source density fields in which confusion noise is significant have a brighter detection threshold than sparse fields at high galactic latitude. As described in A4.2.i, the achieved sensitivity can vary by nearly one magnitude among observations of each individual calibration field because of differing atmospheric seeing, background and transparency conditions.

Source extractions in the Cal-PSWDBs that are not identified as artifacts are highly reliable for J<16, H<15.5 and Ks<15 mag. However, the total fractional reliability of the Cal-PSWDBs is ~20% for the fields with the lowest source surface density and ~90% for the highest source density fields.

The differential reliability of extractions in the 2MASS Cal-PSWDBs as a function of magnitude is illustrated in Figure 18. Reliability here is defined as the ratio of the number of reliable source detections to the total number of extractions in all observations of a calibration field. Cal-PSWDB extractions that are identified as image artifact detections (cc_flg=[P,C,D or G]) are excluded. Reliable source detections are identified using the same repeatability analysis described in A5.2 that is used to define the reliability scoring criteria for the Survey, 6x and Calibration PSWDBs.

In addition to the reliable detections of astrophysical sources, the Cal-PSWDB contains a large number of spurious extractions of faint noise excursion near and below the reliability limits shown in Figure 18. Figure 19 shows the total fractional reliability of each calibration field plotted as a function of the surface density of all extractions. The total fractional reliability is the ratio of the total number of reliable source detections, as defined by the reliability criteria discussed in A5.2, to the total number of extractions in the Cal-PSWDB. The fractional reliability of the Cal-PSWDB is highest in the most dense calibration fields, and lowest in the sparsest calibration fields. This seemingly counterintuitive dependence, and the approximately constant number of extractions for the lower density fields is a consequence of how source detection thresholds were allowed to adjust dynamically in response to increasing confusion noise during 2MASS data processing. Detection thresholds were set at intentionally low SNR levels to maximize completeness in sparse fields where confusion is least likely to corrupt measurements. Detection thresholds adjusted to conservative levels in dense fields because source confusion ultimately limits measurement accuracy.

Because of the large number of spurious extractions that are present in the Cal-PSWDB and Cal-XSWDB, all entries are assigned a reliability score (rel) that is related to the probability that the extraction is the detection of a real astrophysical source at the time of the 2MASS observation. Select WDB extractions that have a reliability flag value of rel="A" to minimize the number of spurious extractions. Caution should be exercised when using any Cal-PSWDB or Cal_XSWDB extractions with a lower probability of reliability.

Figure 17 - Completeness curves for the 2MASS Cal-PSWDB. Completeness computed as the ratio of the number of times a source is detected to the number of times it was observed, [jhk]_n/spos, using statistics from the Merged Cal-PSWDB. Only merged sources detected >100 times in at least one band are included. Figure 18 - Differential reliability as a function of J, H and Ks magnitude for the extractions in the 2MASS Cal-PSWDB that are not flagged as artifacts. Derived using the repeatability analysis described in A5.2. Figure 19 - Total fractionally reliability of extractions in each 2MASS calibration field plotted as a function of the surface density of all extractions. The fractional reliability is the ratio of the number of reliable source detections, as defined in A5.2, to the total number of extractions in the multiple observations of the field.

[Last Updated: 2008 February 18; by R. Cutri]

Previous page.   Next page.
Return to Explanatory Supplement TOC Page.