previous page   back to table of contents   next page

V. Atlas and Catalog Generation

2. Scan, Tile and Frameset Selection


  1. Scan Selection
  2. Tile Selection
  3. Frameset and Frame Selection
  4. Polar Tile Coverage Attenuation

The first step in the Atlas and Catalog Generation for the WISE All-Sky Data Release was to select the input data set for Multiframe Pipeline processing, and to define the set of Atlas Tiles for which single-exposure images would be coadded and sources extracted.

a. Scan Selection

The WISE All-Sky Data Release includes survey data acquired during the full cryogenic mission, 7 January 2010 to 6 August 2010 UTC, while both inner and outer cryogen tanks held solid hydrogen. During this time, WISE conducted 6,331 survey scans, 00712a to 07101a, and collected 1,491,686 framesets of image data. WISE obtained multiple, independent exposures covering each point on the entire sky during the full cryogenic period. The nominal frame depth-of-coverage accumulated during the survey is encoded by the different colors in Figure 1. There are two strips in ecliptic longitude, 26°<λ<40° and 199°<λ<226°, in which the sky was covered a second after the first six months of surveying, and thus has on average two times deeper coverage than the single-epoch coverage areas.

The Single-exposure Images and Source Database contain the calibrated image sets and extracted source lists from all 1,491,686 of the framesets contained in the 6,331scans included in the scope of the All-Sky Release.

Figure 1 - Ecliptic sky plot showing the average raw frameset depth-of-coverage for the WISE full cryogenic survey period in 23' spatial bins. The key on the left side of the image indicates the depth of coverage encoded by each color, and in parenthesis, the cumulative area (in square degrees) having that coverage or greater. This is not the All-Sky Release area coverage.

b. Tile Selection

The All-Sky Release uses all 18,240 Atlas Tiles that tessellate the sky for the purpose of coadding Single-exposure images and extracting sources in Multiframe Pipeline processing.

The names (coadd_id) and position information for the Atlas Tiles that comprise the All-Sky Data Release Atlas and Catalog are listed in Table 1 in IV.4.f.

c. Frameset and Frame Selection

Not all of the 1,491,686 individual framesets acquired during the full cryogenic survey were used in the All-Sky Release Multiframe processing. Framesets, and in some cases individual band-frames were used for image coaddition and source extraction only if they satisfied minimum quality requirements for the survey. Frameset and frame exclusion was done both as a static filtering of the inputs to the Multiframe pipeline, and as dynamic frame filtering in the coaddition process.

Framesets and/or band-frames were filtered out of the inputs to the Multiframe pipeline (i.e. static filtering) in the following cases:

Dynamic frame rejection was carried out in the image coaddition subsystem ( during Multiframe processing. This filtering identified and excluded image frames that contain a large number of aberrant pixels as tagged in the Single-exposure Bit Masks. Frames were rejected if they are heavily contaminated by scattered moonlight, large numbers of saturated pixels, or large numbers of glitches and cosmic rays.

The total number of frames that contributed to the full set of coadded Atlas Images for the All-Sky Data Release in each band is listed in Table 2. The frame count differs between bands because some of the static and all of the dynamic frame filtering was band-dependent. Approximately 96%, 95%, 90% and 91% of the frames acquired during the WISE full cryogenic survey were included in the construction of the Image Atlas and Source Catalog, in W1, W2, W3 and W4, respectively.

Table 2 - Number of Frames Used in Mulitframe Processing for the All-Sky Data Release
BandNumber of Frames

In addition to the frameset and frame-level filtering, pixel-level outlier rejection was performed during the Multiframe pipeline coaddition process. Pixel outlier rejection affects the coverage on smaller scales in the resulting Atlas Images that can be tracked by their corresponding Depth-of-Coverage Maps. The coverage that was available for the measurement of each Catalog source is tabulated in the Catalog w1m, w2m, w3m and w4m columns. A detailed discussion of survey and data release sky coverage is given in VI.2.

The achieved depth-of-coverage after static and dynamic frameset and frame filters, and pixel-level outlier rejection in the Multiframe processing is shown in Figure 2. Comparison with the nominal survey depth-of-coverage (click on the Figure 2 thumbnail to do this interactively), illustrates several differences of note.Most significant are the horizontal bands of low coverage at λ,β= 100°,+45° and 290°,-45°. Early in the survey, the spacecrafts' magnetic torque rods were enabled to dump accumulated momentum when scans approached within 45° of the ecliptic poles. Activating the torque rods resulted in a small jump in the telescope pointing and smearing of the resulting images. Because the smearing occurred near the same point on each orbit, and the smeared images were flagged as having degraded image quality in the QA process, low-coverage "holes" developed at those locations. Later in the survey (2010 May 02 UTC), torque rod enabling was staggered between 45, 57.5 and 70° latitude on alternating orbits so that any image smearing would not occur at the same point on the sky on each orbit.

Framesets that contribute to each Atlas Image are tracked in the All-Sky Release Atlas Image Frame Cross-Reference Table. This table contains a listing of frameset identifiers that correspond to each Atlas Tile in each band. This table can be used to determine which framesets were used to construct a particular Atlas Image, or to determine to which Atlas Images a particular frameset or frame contributed.

Figure 2 - Net Coverage in W1 for the WISE All-Sky Data Release. Colors encode the average frame depth-of-coverage in 15x15 arcmin spatial bins. Click on this image to view a larger version that allows an interactive comparison between the achieved coverage and raw scan coverage prior to Tile selection and frameset and frame filtering.

d. Polar Tile Coverage Attenuation

The WISE survey scanning strategy produced frameset depth-of-coverage that increased with increasing ecliptic latitude. Typically, 12 independent Single-exposures exposures were accumulated at each point on the sky near the Ecliptic, but over 3000 exposures were obtained directly at the north and south Ecliptic poles. The extremely large number of framesets involved with Tiles near the Poles exceeded the processing memory and runtime capabilities for source extraction, so coverage was artificially limited to an average depth of approximately 165 framesets for 41 Tiles near the Ecliptic poles during the Multiframe Pipeline processing that produced the All-Sky Image Atlas and Source Catalog (20 in the north and 21 in the south).

Although source extraction using all available Single-exposure framesets was not feasible for these 41 Tiles, the full-depth coadded image products were constructed using all framesets that satisfied the basic selection criteria described above. The full-depth coadded Ecliptic Polar Tile image products can be accessed from II.3.h.

Table 1 - List of All-Sky Atlas Tiles Near the Ecliptic Poles with Attenuated Coverage

i. Attenuation Algorithm

Attenuation of high coverage frames was done in a manner meant to produce a roughly uniform coverage at a desired level over the parts of Tiles requiring attenuation, while preserving temporal sampling as fully as possible. To achieve these goals, an iterative, weighted, pseudo-random frame elimination algorithm was employed that followed these steps:

  1. A coarse 8x8 coverage depth matrix was computed for each Tile. Each cell in the depth matrix contained references to all frames which added coverage to that part of the Tile.
  2. The deepest coverage cell in the coverage matrix is identified.
  3. The frames in the deepest coverage cell are inverse weighted by the number of cells each frame impinges upon.
  4. So weighted, a frame in the deepest coverage cell is drawn at random and removed from all cells in which it falls. Weighting in this manner favors removal of frames that minimizes consequences far from the removal cell. Note that the area covered by a frame is approximately one quarter of the size of an Atlas Tile.
  5. A new maximum depth cell is identified and the process is repeated.
  6. When the coverage in the maximum depth cell drops below the target depth, the Tile attenuation is complete.

Coverage Depth and Uniformity

The nominal target depth of an attenuated area on the high coverage Tiles was 150 frames. To account for the approximate nature of the algorithm, some pad was added and the actual target attenuation depth was set to 165. The attenuation process is not easily controlled at fine scale because of the large frame area. This resulted in some areas that were less than the desired minimum. However, the non-uniformity caused by the attenuation is small compared to the natural coverage depth variation that results from the survey strategy.

Temporal Sampling

The temporal power spectra for the frames in several attenuated tiles were examined and all sampling intervals were preserved thanks to the random removal process and the highly repetitive nature of the survey strategy. No additional aliasing was introduced by the attenuation over and above those produced by the orbital period of WISE and its harmonics. Thus, the utility of the variable source characterization is preserved.

ii. Example - The North Equatorial Pole

Three-color composite images of the All-Sky Atlas Image and Full-depth Atlas Image for Tile 2709p666, that covers the north Ecliptic pole, are shown in Figures 3 and 4. The corresponding W1 pixel depth-of-coverage maps for these two images are shown in Figures 5 and 6.

Figure 3 - W1/W3/W4 composite color image showing the attenuated coverage All-Sky Atlas Image Tile 2709p666_ab41 covering the NEP. Figure 4 - W1/W3/W4 composite color image showing the Full-depth Atlas Image Tile 2709p666_ab42 covering the NEP.
Figure 5 - Pixel depth-of-coverage map for the W1 All-Sky Atlas Image Tile 2709p666_ab41 covering the NEP. The color scale at the bottom of the image describes the depth-of-coverage. Figure 6 - Pixel depth-of-coverage map for the Full-depth Atlas Image Tile 2709p666_ab42 covering the NEP. The color scale at the bottom of the image describes the depth-of-coverage.

The coverage attenuation results in All-Sky Atlas Images and Source Catalog in those areas that do not reach the depth possible if all available Single-exposure data were used. However, the sensitivity decrease is not significant in W1 and W2 because the confusion noise limit is reached even in W1 and W2 even in the attenuated coverage, as illustrated in Figures 7 and 8. The coverage attenuation does result in significant sensitivity loss in W3 and W4, though, as shown in Figures 9 and 10.

Figure 7 - W1 Figure 8 - W2
Figure 9 - W3 Figure 10 - W4
The figures above compare the All-Sky Atlas Image (left) with ~165 frame depth-of-coverage with the same area in the Full-depth Atlas Image (right) where the coverage is 1500-2000 framesets, for a zoomed-in region in Atlas Tile 2709p666.

Last update: 2012 March 16

previous page   back to table of contents   next page