II. User's Guide to the 2MASS All-Sky Data Release

---

3. Extended Source Catalog (XSC)

b. How to Use the 2MASS XSC

The 2MASS Extended Source Catalog (XSC) is comprised of near-infrared sources that are clearly resolved and extended beyond the beam/PSF. Sources range in angular size from 10´´ to 2°: from small, faint galaxies to the largest Local Group galaxies. By their very nature, galaxies are complex and, therefore, require characterization measurements which are much more detailed than that for point sources. As such, the XSC includes many different kinds of measurements and various flavors of, e.g., position, orientation, size, symmetry, surface brightness and integrated flux. The user should take care to understand these parameters and their limitations. Here we answer some "Frequently Asked Questions" and give some brief guidelines on how to use the XSC to maximum effect.

Fundamental: The XSC is an extended source catalog. Although galaxies dominate the total numbers (and we frequently refer to the XSC as a galaxy catalog), the XSC is also comprised of Milky Way entities, such as globular and open clusters, planetary nebulae, HII regions, ISM emission and nebulosity, young stellar objects and compact star-formation regions, and even Solar System comets. These Galactic sources tend to be confined to the plane of the Milky Way. Refer to the "All-Sky" maps in II.3c2 to see how these sources are distributed across the sky.

Fast links to: IRSA   astrometry   photometry   large galaxies   Milky Way objects   completeness/reliability   outliers   references/links

Parameters: What is available in the XSC?

See Table 1 of IV.5a for a detailed description of all the XSC parameters.

Data Mining: Where do I get the data? How do I use IRSA?

The catalog tabular and imaging data is archived within the IPAC science archive IRSA (Infrared Science Archive). Use the GATOR catalog query service to retrieve data. Also each source has a corresponding image cube, which is a small FITS image incorporating the object. The FITS and jpeg previews of the object can be obtained with the IRSA XSC image retrieval service.

IRSA includes many services to retrieve data and to compare across data sets. The two most important services relevant to the XSC are:

• GATOR, the catalog query service to retrieve catalog data
  
• XSC Postage-Stamp Image Retrieval Service

Where else is the data archived?

The NASA Extragalactic Database (NED) contains the names, positions and basic photometric information from the XSC. Images, including links to the IRSA Image Retrieval Service and jpeg previews are also available in NED. For example, compare the 2MASS images with other (mainly optical) images of the M51 (Whirlpool) galaxy.

Astrometry: Which Position Do I Use and How Accurate Is It?

Each XSC source is tagged with two different sets of photometry. (1) The first set is based on the J-band peak pixel of the source. Since each pixel is 1" in size (while the PSF beam is closer to 2.5"), the relative accuracy of this measurement is between 0.3 and 0.5". The source name, or designation, is derived from this position. (2) The second set is based on the intensity-weighted centroid of the combined J+H+K_s image. The relative accuracy of this measure is nearly twice as good as the peak-pixel method, due to the improved S/N and centroiding method.

The absolute astrometric accuracy of 2MASS is better than 100 milli-arcsec for point sources. For extended sources, you can expect 0.5" accuracy for the peak-pixel coordinates and 0.3" accuracy for the centroid coordinates. For a comparison of the 2MASS peak-pixel coordinates with the FIRST radio survey, see II.3d5.

Astrometry: Why is the 2MASS Designation Different from the Centroid Positions?

The source name or designation is derived from the J-band peak pixel of the source. These coordinates can be slightly different from those derived from the intensity-weighted centroid of the combined J+H+K_s image. You can expect differences ranging from 0 to 0.5".

Photometry: Which Should I Use?

Because extended sources are complex by nature, a variety of apertures are used to compute the integrated flux. Choose the aperture or method that best matches your science goals. Since there is no "one size fits all" aperture, the XSC does not provide "default" mags (unlike the PSC). See note below. We do have some guidelines.

For most applications, the elliptical isophotal aperture is a good choice, both in terms of capturing most of the integrated flux (~80-90%) and providing accurate colors for galaxies of all sizes. Choose the "total" apertures (e.g., Kron or the Extrapolation SB Profile) if you need integrated fluxes that reflect the total flux of the source. But beware, these apertures are vulnerable to stellar contamination and surface brightness irregularities. The photometry based on the extrapolation of the surface brightness profile seems to be the more robust of the two methods (primarily because stellar contamination is minimized by averaging over the azimuthal isophote(s) used to derived the median surface brightness profile.)

The most robust aperture in the XSC is the circular, 7" radial aperture. This is a good choice if you are focused on faint or small compact sources. But beware, for larger galaxies the small apertures are not a good choice -- they are subject to a "bulge" color bias. Conversely, for small galaxies the circularizing effect of the PSF alters the shape (orientation) and radial sizes -- beware.

Below are links to explanations of the aperture photometry available in the XSC:

• XSC Aperture Photometry
  
• Isophotal Photometry
  
• "Total" Photometry
  
• Effective or Half-Light Aperture Photometry
• Photometric Repeatibility
• The 2MASS PSF/Beam

Photometry: Why Are There No Default Magnitudes for the XSC?

Although "default" mags were provided in the incremental XSC releases, they are not provided in the final XSC release (which supercedes all incremental releases). It was decided that since extended sources are complex entities and no one aperture (or method) satisfies all science requirements, the notion of "default" in the XSC is mis-leading at best. The user should decide which aperture best matches the desired science goals. We do offer some guidelines as to what photometry works best for common cases. See the above "Photometry: Which Should I Use?"

That said, if the user so desires to re-capture the "default" mags that were used in the incremental releases, then target the circular (K-band fiducial) isophotal mags, which are available in the XSC.

Photometry: What About Total Magnitudes?

"Total" mags refer to apertures in conjunction with corrections that account for the "total" flux of a source. We have determined that between 10 and 20% of the flux (galaxy morphology dependent) of an object is lost in the formidable background noise. This flux is recovered using the median (elliptical) surface brightness profile. But beware, "total" mags are vulnerable to stellar contamination and to irregularities in the surface profile (e.g., asymmetries). "Total" mags are not a good choice when colors are desired (since one is effectively RSS-combining two large uncertainties to compute a color). For colors, use a smaller, more robust aperture (e.g., isophotal mags; See "Photometry: Which Should I Use?" above.)
• XSC Aperture Photometry
  
• Isophotal Photometry
  
• "Total" Photometry
  
• Effective or Half-Light Aperture Photometry
• Photometric Repeatibility

Large Galaxies: Why Are Large Galaxies Special Objects in the XSC?

The XSC is complete for all galaxies larger than ~10-15" in diameter, including the largest galaxies in the sky. However, due to their proximity to a survey "scan" edge, galaxies larger than ~1' or 2' will have photometry that is systematically incomplete. We have rectified this situation.

The 2MASS survey acquired images of the sky using relatively small arrays. A "tile" or scan is 8.5' in angular width, and the typical overlap between scans is 50". Galaxies that are smaller than this overlap are guaranteed to be fully sampled in at least one survey scan. Larger galaxies may be truncated based on their proximity to a scan edge. Therefore it was necessary to construct an atlas of large objects made from "pieces" of adjoining scans. The net outcome is that we will fully recover galaxies that are currently "lost" or misrepresented in the 2MASS extended source pipeline.

The resultant set of mosaics and corresponding source characterizations are collectively called the 2MASS Large Galaxy Atlas (LGA). This LGA information has been incorporated into the XSC for the largest 550 galaxies in the sky. LGA galaxies are easily identified n the XSC with the cc_flg parameter; they are tagged with: cc_flg = "Z". We have also identified sources that are in close proximity to large galaxies, whose photometry has been eliminated from the catalog due to their unreliable nature: cc_flg = "z".

The 2MASS Large Galaxy Atlas is described in detail in Jarrett et al. (2003, AJ, 125, 525).

Large Galaxies: Why are some "large" galaxies in fact globular clusters or barely resolved galaxies?

The Large Galaxy Atlas includes objects that are not traditionally thought of as galaxies. For example, the Tarantula Nebula Of the LMC (a piece of A Local Group galaxy). Globulars are another interesting category, as they are sometimes thought of as Local Group objects (e.g., globular Omega Cen is often referred to as a dwarf spheroidal LG galaxy). In Jarrett et al. (2003, AJ, 125, 525), the globular 47 Tucanae (NGC 104) is used to represent a (very nearby) compact dwarf elliptical galaxy.

The Atlas also contains a handful of objects that are not very large, including low surface brightness galaxies, Seyferts and AGN. In addition to reconstruction of large galaxies, the spirit of the LGA is to recover galaxies that are either lost from the 2MASS automated pipeline because of (1) confusion, or (2) because of surface brightness. Confused objects (including pairs, groups and cluster cores) require special processing to deblend the individual component flux. Seyfert/AGN might require special processing to account for the extreme color difference between 2MASS bands. Finally, even though the galaxy may be large at optical wavelengths, in the near-infrared the object might be quite small or invisible (e.g., extreme late-type galaxies). An example of a very small galaxy reconstructed in the Large Galaxy Atlas is IRAS 07598+6508 (Figure 1).

Figure 1

Star-Galaxy Separation: How do the e_score and g_score work?

To distinguish extended (resolved) sources from point source we use a variety of measures to separate these two general classes of objects. Effective discriminators include the size, central surface brightness and color. In all, we use about 10 different measures (or dimensions), whose values are correlated to one degree or another. Read more about this parameter set in Star - Galaxy Discrimination.

We combine and exploit this multi-dimensional parameter space using a "decision tree" method. We build the decision tree training sets to do one thing only: separate the extended (resolved) sources from the point-like sources. It does nothing else. This method cannot (and will not) tell you (1) how extended an object is, or said another way, it will not tell you the probability of extendedness or otherwise, and (2) it will not tell you the nature of the sources tested against the decision tree. It can only tell you if the source is extended or if it is not.

That said, we do attempt something a bit more useful to our needs. The decision tree method is applied to the 2MASS parametric data for each band, independently. Although the band-to-band information might be correlated to some degree, by combining the decision tree results for each band we create a "pseudo-probability". We can "assign" a pseudo-probability by using a weighted average of the decision tree classifications for each band. This probability has a value between 1 and 2, with 1 being "extended" and 2 being "point-like". But owing to the correlated nature of source characterization between 2MASS bands, one should not read too much into small differences (e.g., there is virtually no difference between a value of 1.0 and 1.2). Two different probability flavors are generated: (1) g_score and the (2) e_score. These are employed as the final arbiter for star-galaxy separation. It turns out that a value of around 1.4 gives a satisfactory separation, while maintaining satisfactory completeness. Summary: extended objects have scores between 1 and 1.4, while point-like objects have values between 1.4 and 2.0.

g_score
The "g_score" is geared towards recognizing galaxies. Not only does it include the standard star-galaxy parameters (e.g., shape, size & SB, ... etc) but it also includes the "color attribute". Galaxies are almost always redder than Milky Way stars (beyond ±2° or 3° from the Plane) and so a powerful discriminent is the J-K_s color itself. A g_score less than 1.4 indicates that it is probably an extended source.

e_score
The "e_score" is a simpler version of the g_score: it has only three parameters that are tested in the decision trees, including surface brightness, shape and the double/triple star discriminator. It does not include a color attribute. In this sense, it is less biased than the g_score. It was conceived to capture Milky Way objects that might have unusually blue colors, or shapes distinctly different from galaxies. The XSC is drawn from a database sample that has an e_score or g_score value less than 1.4.


It should be understood that: (1) the e_score is less reliable than the g_score, and (2) although it may be less biased to color, the e_score is not necessarily better at finding Milky Way objects (the g_score is more than happy to find HII regions, nebula, YSOs, etc).

Note that the Level-1 Requirements are only relevant to galaxies and the g_score (which is the most reliable discriminator anyway). See VI.5b.ii.

To summarize:

• the XSC is drawn from the 2MASS extended source database, thresholded at g_score or e_score < 1.4;
  
• g_score is a more reliable indicator that a source is extended (particularly for normal galaxies);
  
• one cannot use the e_score or g_score to distinguish one galaxy from the next; these scores simply tell you whether the object is extended or not.
  

Milky Way: How Do I Distinguish Milky Way Fuzz from Background Galaxies?

• See Galactic Extended Sources: Distribution & Colors

  The 2MASS XSC includes sources that belong to the Milky Way: H II regions, stellar clusters, planetary nebulae, young stellar objects, emission-line nebulae, reflection nebulae and solar system comets. Nearly all Milky Way objects are tightly confined to the Plane of the Galaxy: |glat| < 5°. The exception to this rule are the giant molecular clouds that float above the Plane (e.g., Orion, Taurus, Rho Ophiuchus), and nebulosity in the LMC and SMC. You can see the Milky Way objects quite easily in the XSC Allsky image, lining the Plane in color "red". These objects are both intrinsically red (e.g., HII regions are dominated by emission bands in the 2 micron window) and dust-reddened. Although it is not possible uniquely to identify Milky Way objects in the XSC, most exhibit the following characteristics:

  • Milky Way objects are mostly confined to the Plane: |glat| < 5 deg
    
  • Milky Way objects tend to be very red (J-K_s > 1.5)
    
  • Milky Way objects are often large in size (radius > 20")
    
  • Milky Way objects sometimes have corrupted mags (due to their oddball shape and confusion)
    

  We have identified some 8,000 sources in the XSC that are probably Milky Way in origin, a smaller subset of 3712 sources are highly probably Galactic in nature. The smaller table of Milky Way sources is given in Table 1.

Completeness: Why Is My Favorite Galaxy Missing from the XSC?

The Level-1 Science Requirements demand a completeness level of >90% for K_s brighter than 13.5 mag and |glat| > 30° (to avoid Milky Way confusion). The actual completeness that the XSC realizes is probably greater than 95%; see the analysis of the Virgo Cluster Completeness.

What is missing from the XSC? In the near-infrared, particular kinds of galaxies are nearly invisible, including late-types (Sd, Sdm, Im), blue compacts and LSBs; see the 2MASS near-infrared galaxy morphology sequence. The reason is that the near-infrared window is sensitive to giant stars and the older stellar populations that comprise the "backbone" of galaxies. It is not sensitive to massive star formation regions and hot blue stars, which might dominate the emitted light for late-type disk galaxies. A comparison between the optical and near-infrared windows is given in 2MASS Large Galaxy Atlas. In particular, see the discussion in sections 6.4, 6.5, 6.6 and 6.7. The end result is that catalog comparisons between the optical and NIR will show significant differences in completeness, surface brightness and color.

Galaxies are also missed due to confusion from nearby stars. The automated pipeline might peak up on the star, thus reducing the "extendedness" of the source, thereby excluding it from the XSC. Galaxies might also be missed due to bright star masking, where the confusion radius, diffraction spike or horizontal strip masking either eliminates the source completely or disrupts it to the point that it is quality-rejected from the XSC. Stellar confusion can also render the position of the "star+galaxy" to be inaccurate (the position may in fact correspond to the star itself, or the centroid of the blended object).

Another thing to consider is the position of your source. Occasionally the published position of a galaxy is quite inaccurate (tens of arcsec). The source may yet be in the XSC, but the user needs to search a larger radius to properly catch it.

Your object may belong to the 2MASS Large Galaxy Atlas, where oddball galaxies and invisible galaxies are recovered (in addition to large galaxies).

If you decide that your missing galaxy should have been detected by 2MASS (considering the arguments given above), then send the coordinates of the object to the 2MASS group (re: T. Jarrett), and we'll attempt recovery of your object.

Reliability: What Is a Star Doing in the XSC?

The Level-1 Science Requirements demand a reliability level of >98% for K_s brighter than 13.5 mag and |glat| > 20°. Sources of unreliability include stars, double stars, triple stars, artifacts and disrupted objects. Double stars are the most common false XSC object for |glat| > 20°. The cause of this unreliability is that double stars under some conditions closely mimic galaxies (including their color). With atmospheric "seeing" variations, the PSF is occasionally smeared to the point that stars may also mimic real extended objects. We have gone to great pains to minimize this deleterious effect.

In the Plane where source confusion reigns, triple stars are a major source of unreliability, particularly for faint XSC sources. The Level-1 Science Requirements do not apply to the Galactic Plane (|glat| < 10°); however, the reliability appears to be at least 80-90%, depending on the source confusion (as measured with the density metric.) The XSC stellar contamination is summarized in VI.5b.ii.

Duplicity: Why Do Some Galaxies Appear to be Duplicates?

On rare occasions you may find a "duplicate" in the XSC. These arise from scan-to-scan differences in position for each source. In one scan the source might be a few arcsec different from the same source in a different scan. Most of these cases are rectified and corrected. Please send us the name and coordinates of the "duplicate" and we'll place the object in our anomaly file.

Contamination: Why Do Some Galaxies Appear to have Unphysical Colors or Brightnesses?

Large aperture photometry (i.e., anything larger than the PSF) is subject to contamination from stars and background gradients. The pipeline attempts to identify stars and remove them from the galaxy before characterization measurements are carried out (see e.g., before (Figure 2) and after (Figure 3) stars are removed from a dense field). The pipeline is not perfect, there are cases in which the star removal process is either incomplete or not done at all (e.g., the stars are not properly identified due to confusion). In such cases, the photometry will be affected accordingly, possibly resulting in very strange characterization measurements (e.g., isophotal shape, surface brightness, size) and photometry (e.g., colors). If you encounter an extended object with suspect integrated fluxes or colors, it is best to consult the postage-stamp image or the Atlas image containing the source.

Other potential problems include (1) corruption from nearby bright stars (e.g., Beta Pegasi and extended source artifacts) and (2) airglow gradients in the background. Both of these phenomena can result in corrupted (or contaminated or modified) fluxes in one or more bands.

Figure 2	Figure 3

What is the Spatial Coverage of the XSC?

Alternatively, how does one compute the areal coverage for a specified region (e.g., to normalize the number counts) ?

2MASS fully covered the sky, providing a uniform census of galaxies in the local Universe. The XSC is only limited by (1) masking of bright stars, and (2) confusion noise in the Plane. Bright star masking is fully tracked and tabulated for the XSC on scales set by the 2MASS Atlas images (8.5 X 17 arcmin). The bottom line is that the spatial coverage is typically better than 98% for most of the sky. See the coverage maps here: Extended Source Spatial Coverage. A set of tan-projection FITS images are available for download at this link: II.6.f.

PSC: Do galaxies appear in the PSC?

Yes. There are both XSC objects with PSC detections and PSC-only detected galaxies. The PSC-only galaxies tend to be faint and point-like. But there are many of these galaxies in the PSC. Take a look at PSC Contains Many (Faint) Virgo Galaxies, and the Evaluation of Extragalactic Content of the PSC/WSDB.

What Do the Allsky Maps Show? What are the Streaks in the Allsky Maps?

The XSC allsky maps can be found here: Allsky Images of the XSC. They show the XSC distributed across the sky, both in Equatorial and SuperGalactic projections. The "structure" seen in the maps is real and corresponds to clusters and supercluster filaments. It remains to be seen just how well these projections trace the large scale structure of the Universe.

A consequence of squeezing the allsky into equal-area aitoff projections is that the map edges take on distortions that may look unphysical. For example, notice the streak-like structure in the lower right-hand portion of the equatorial map, corresponding to a coordinate: RA=20h 04m, Dec = -55d 56m. Here is a zoom (Figure 4) of this map region. Re-projecting the area using a standard "flat" tangent projection reveals the following (Figure 5): the field includes several known (catalogued) galaxy clusters, many of which are aligned along a diagonal to the equatorial plane. Note the prominent galaxy clusters at 19h 50m, -55d (Abell 3651) and 20h 12m, -56.8d (Abell 3667), both located at a redshift corresponding to 0.05 to 0.06.

Figure 4	Figure 5

How Do You Estimate the FWHM of the PSF?

The 2MASS point spread function (PSF) is characterized using an exponential function, although it is also roughly gaussian in shape. The relation between the 2MASS PSF metric and the corresponding FWHM is described here: The 2MASS BEAM. This is another opportunity to caution the user that small galaxies are circularized by the 2MASS beam, and interpretation of axial ratios, diameters and half-light radii should be tempered accordingly.

Helpful References

• 2MASS Extended Source Catalog: Overview and Algorithms (Jarrett et al. 2000)
  
• 2MASS Large Galaxy Atlas; (Jarrett et al. 2003)
  
• 2MASS All-Sky Data Release Frequently Asked Questions Page (FAQ)
  
• IRSA
  
• NED

[Last Updated: 2015 Oct 21; by T. Jarrett and R. Cutri]

---

Return to Section II.3.