previous page   back to table of contents

V. AllWISE Data Processing

V.4. Catalog Generation


V.4.a. Tile Edge Boundaries
V.4.b. Duplicate Source Resolution
V.4.b.i. Identifying and flagging SSST duplicates
V.4.b.ii. Pre-filtering Detections
V.4.b.iii. Duplicate Resolution Results
V.4.c. Source Reliability Criteria
V.4.c.i. Detection and SNR Limits
V.4.c.ii. Artifact Flagging
V.4.c.iii. Frame Depth-of-Coverage

The Catalog Generation process for AllWISE separates the deep-source extractions in the full Working Database (WDB) produced by the Multiframe Pipeline processing into the distinct AllWISE Source Catalog and Reject Table.

Multiframe processing for AllWISE generated a WDB that contains all 1,176,421,279 extractions that are detected on the coadded Atlas Images with SNR>2.4, and measured on both the individual Single-exposure images and Atlas Images. In addition to extractions of real astrophysical sources, the WDB contains:

The AllWISE Source Catalog is intended to be a highly reliable and complete set of single, unique detections for compact objects on the sky. To be included in the Source Catalog, WDB entries must satisfy all of the following criteria:

747,634,026 extractions in the WDB meet these criteria, and therefore constitute the AllWISE Source Catalog. The 428,787,253 WDB entries that do not meet these criteria are contained in the AllWISE Reject Table.

V.4.a. Tile Edge Boundaries

The position of a WDB entry must be >50 arcsec from the boundary of the Atlas Image from which it was extracted to be included in the AllWISE Source Catalog. This safety boundary is designed to eliminate entries that may have incorrect aperture photometry in the largest measurement aperture, which has a 50 arcsec radius, because of truncation by the image edge.

The distance to an Atlas Image boundary is derived from the pixel coordinates of a source (wx, wy). Atlas Images are 4095x4095 pixels in size and the coadd pixel scale is 1.375"/pix. Thus, the edge boundary criterion requires a source to have pixel coordinates in the range

36.4≤wx≤4058.6 pix and 36.4≤wy≤4058.6 pix.

Because Atlas Tiles, overlap each other by at least 180 arcsec, there will always be a location that lies more than 50 arcsec from the edge of at least one Tile.

The application of the Tile edge boundary criterion is integrated into the Duplicate Source Resolution process which is described below.

V.4.b. Duplicate Source Resolution

The AllWISE Source Catalog is intended to contain one entry for each compact source of mid-infrared radiation on the sky. Because the AllWISE WDB contains duplicate, redundant extractions of the some sources, the duplicate resolution process identifies groups of duplicate extractions, and select one apparition in each group that will be eligible for the Catalog.

There are two types of duplicate source extractions in the AllWISE WDB:

Identification and resolution of the Tile overlap duplicate extractions for AllWISE uses the same process and algorithm used in production of the WISE All-Sky Release Source Catalog that are described in section V.3.c of the All-Sky Release Explanatory Supplement. Source Catalog. SSST duplicates are identified and classified in a pre-duplicate resolution analysis step and filtered out of the list of WDB extractions that are input to the Tile overlap duplicate processing.

The results of both the Tile edge and duplicate resolution processing are encoded in the WDB use_src column. Extractions that satisfy the Tile edge limits, are either not part of a duplicate source group (i.e. a "singleton") or they are part of a duplicate group and have been selected by the duplicate resolution process have use_src=1, and are eligible to be included in the Source Catalog. Extractions that do not meet these conditions have use_src=0 and will be in the Reject Table.

Sources in the AllWISE Catalog must have use_src=1, so that column is not carried explicitly in the Catalog entries. However, WDB extractions that have use_src=1 but fail the other Catalog reliability criteria will wind up in the Reject Table. Therefore, the use_src column is included in the Reject Table so that it is possible to investigate duplicate resolved sets of fainter source detections.

V.4.b.i. Identifying and flagging SSST duplicates

SSST duplicate groups were identified prior to Tile overlap duplicate resolution processing by positionally autocorrelating the extractions in each individual Tile. The correlation radius for the SSST groups was 3.9 arcsec, smaller than the resolution possible with WISE. The members of each SSST group were classified according to the order in which they were extracted within their Tile. The first source extracted in each SSST is usually the proper extraction because it was the brighter detection and was fit first. Later extractions in the Tile were typically the fainter detections that were pulled towards the brighter neighbor. The SSST classification process also attempts to recover valid extractions of very bright sources that were incorrectly cross-flagged as spurious artifact detections.

There are a very small number of SSST groups whose members are unflagged and all persist into the AllWISE Catalog because their separations are slightly larger than the 3.9" correlation radius that was used. These rare occurrences can be recognized as groups of two or more sources that have slightly larger than 3.9" separations, and where only one of the group clearly corresponds to the position of a source when viewed on the AllWISE Atlas Images.

SSST group member classification is encoded in the rel flag that is carried in the AllWISE Catalog and Reject Table. The rel flag was used to pre-filter extractions for duplicate resolution processing, as well as in the Catalog Generation reliability criteria. The values of the rel flag, their meanings, and the number of each in the AllWISE Source Catalog and Reject Table are listed in Table 1. Approximately 3.8% of all WDB entries are in SSST groups, and 2.6% and 5.9% of Source Catalog and Reject Table have non-null rel values.

Table 1 - rel flag values, definitions and the number of occurrences in the AllWISE Source Catalog and Reject Table
relDefinitionNo. in CatalogNo. in Reject Table
nullThe extraction is not a member of a SSST group, and is eligible for inclusion in the Catalog if it satisfies other criteria. 728,064,747403,503,726
sThe extraction is a member of an SSST group, and is classified as the correct source measurement that is eligible for inclusion in the Catalog if it satisfies other criteria. 19,560,7362,339,345
cThe extraction is a member of an SSST group, is classified as the correct source measurement, is saturated and flagged as a spurious detection of an artifact in all bands. Such extractions are eligible for inclusion in the Catalog and can override the artifact flagging requirements in Catalog Generation. 8,543657
rThe extraction is a member of an SSST group, and is classified as a spurious low SNR detection that was shifted onto the position of a brighter source. It is not eligible for the Catalog and will be found only in the Reject Table. 022,943,525

V.4.b.ii. Pre-Filtering Detections

Not all extractions associated within a duplicate group are eligible to be selected in the duplicate resolution process. Extractions are pre-filtered using the criteria listed in Table 2. These criteria have been updated from those used in the preparation of the All-Sky Release Source Catalog (c.f. section V.3.c.iii.1 in the All-Sky Release Explanatory Supplement.

The changes to the pre-filtering criteria for AllWISE are:

Table 2- ALLWISE Duplicate Resolution Pre-Filter Criteria
MagnitudeArtifact Flagrel flagDistance to Tile EdgeGroup detectonSingleton
detected1 in at least one band - r - RejectReject
upper limits in all 4 bands - not 'r' ≥50 arcsec RejectAccept
- --minimum distance to edge <50 arcsecRejectReject

Notes to Table 1
1 magnitude is NOT an upper limit

V.4.b.iii. Duplicate Resolution Results

Figures 1a and 1b show detections from overlapping tiles (color coded) over approximately a 3.5°x 3° region of sky. Figure 1a shows detections prior to duplicate resolution and Figure 1b shows detections after duplicate source resolution. Figure 1c shows detections after duplicate source resolution in grey tones, so there is no Tile differentiation.

Figure 1a - All WDB extractions before duplicate resolution processing. Tiles differentiated by color. Figure 1b- Duplicate resolved WDB extractions. Tiles differentiated by color. Figure 1c- All duplicate resolved extractions without the color differentiation.

Figures 2 shows all duplicate resolved WDB extraction in a 1°x1° region that contains a bright star and its artifacts. Extractions from different Tiles are coded by having the same color. Artifacts are not filtered out in the ALLWISE duplicate resolution processing, so they appear in this image. Figure 3 shows the duplicate-resolved extractions shown in Figure 2 that also satisfy the reliability criteria for the AllWISE Source Catalog.

Figure 2 All duplicate resolved detections from overlapping tiles. Tiles differentiated by color. Figure 3 - Duplicate resolved detections shown in Figure 2 that also satisfy the Catalog source reliability criteria. Tiles differentiated by color.

V.4.c. Source Reliability Criteria

To be included in the AllWISE Source Catalog, a WDB entry must be selected in the duplicate resolution processing described above (i.e. have use_src=1), and it must have a high-quality, reliable detection in at least one WISE band. A high quality, reliable detection in a band is defined as meeting all of the following criteria in that band:

Source extractions that meet the above criteria in one band will be in the Catalog even if they do not satisfy them in other bands. For example, it is common for a Catalog source to have a flux measurement that is SNR≥5 in W1, but SNR<5 in all other bands.

The first two criteria listed above have special exceptions that are described below to accommodate the presence of W1 saturation encountered during early 3-Band Cryo observations and the SSST duplicates.

The SQL statement that was used to select the AllWISE Source Catalog from the full WDB is:

use_src = 1 and 
( (w1mpro is not null and w1m > 4 and (cc_flags[1] not matches '[DHOP]' or (rel = 'c' and cc_flags[1] matches '[DH]')) and ((w1sigmpro is not null and w1snr >= 5) or (ph_qual[1] = 'Z' and w1sigp2 is not null and w1sigp2 <= 0.2171 and (cc_flags[2] not matches '[DH]' or (rel = 'c' and cc_flags[2] matches '[DH]'))))) or
  (w2mpro is not null and w2sigmpro is not null and w2snr >= 5 and w2m > 4 and (cc_flags[2] not matches '[DHOP]' or (rel = 'c' and cc_flags[2] matches '[DH]'))) or
  (w3mpro is not null and w3sigmpro is not null and w3snr >= 5 and w3m > 4 and (cc_flags[3] not matches '[DHOP]' or (rel = 'c' and cc_flags[3] matches '[DH]'))) or 
  (w4mpro is not null and w4sigmpro is not null and w4snr >= 5 and w4m > 4 and (cc_flags[4] not matches '[DHOP]' or (rel = 'c' and cc_flags[4] matches '[DH]'))) ) 

The components of this selection statement are described in the following sections.

V.4.c.i. Detection and SNR Limits

A reliable band-detection in the AllWISE WDB must have a valid detection in the band (e.g not a 2σ upper limit) and that detection must have have SNR≥5. The detection validity and SNR measurement is normally required to be from the profile-fit photometry. For W2, W3 and W4, this criterion translates to the following SQL statement:

w?mpro is not null and w?sigmpro is not null and w?snr≥5,

where "?" = 2, 3 or 4.

The profile-fit W1 flux uncertainties of a small number of valid source detections covered by saturated W1 frames during the early 3-Band Cryo survey phase were were incorrectly set to zero because of the presence of large numbers of hard saturated pixels. These detections have valid W1 flux measurements, but the missing W1 flux uncertainties mean there is no valid W1 SNR value to test for Catalog inclusion. These extractions are identified in the Catalog and Reject Table as having ph_qual[1]='Z'. For these cases, the standard deviation of the individual W1 frame flux measurements, w1sigp2, is used as a proxy for the profile fit SNR value, and the W1 portion of the Catalog detection and SNR requirement becomes:

w1mpro is not null and ((w1sigmpro is not null and w1snr≥5) or (ph_qual[1]='Z' and w1sigp2 is not null and w1sigp2≤0.2171))

This exception for W1 extractions affected by the saturated 3-Band Cryo images "rescues" 3,449 sources that would have otherwise been excluded from the Source Catalog.

V.4.c.ii. Artifact Flagging

To filter out spurious detections of image artifacts caused by bright sources, a reliable source band-detection is one that is not flagged as a spurious detection of an image artifact in the band. Spurious detections of artifacts are nominally flagged by having a capital letter value in the cc_flags WDB. However, some valid extractions of saturated sources were incorrectly flagged as spurious artifact detections because of the source flux subtraction error that produced the SSST duplicates. Many of these incorrectly flagged source entries were identified as part of the during the SSST identification and classification process, and are denoted by having rel='c'. The artifact selection criterion is:

cc_flags[?] not matches '[DHOP]' or (rel='c' and cc_flags[?] matches '[DH]')

where "?" = 1, 2, 3 or 4. Note that the exception for the rel='c' sources is used only if they are flagged as diffraction spike or scatter light artifacts because those will be produced for objects in close proximity, unlike optical ghosts or latent images.

V.4.c.iii. Frame Depth-of-Coverage

A reliable band-detection in the WDB must have been extracted from a region with a minimum depth-of-coverage of at least five independent frames. The SQL expression of this requirement is:

w?m > 4

where "?" is the band, 1, 2, 3 or 4.

This requirement is designed to eliminate spurious detections in low coverage areas within Atlas Tiles. Because they do not persist on the sky, transient events such as cosmic ray strikes, residual satellite trails, hot pixel events and scattered light from very bright moving objects such as the moon and planets are usually suppressed by pixel outlier rejection during the image coaddition process. However, outlier rejection does not operate when fewer than five frames are available for coaddition, and single frame transients can persist into the coadds and induce spurious detections.

The Catalog may contain sources that are measured on fewer than five frames on one or may bands as long as there is a reliable detection in another band that meets all other selection criteria.

Last update: 13 May 2014

previous page   back to table of contents