Global and local variations in the depth-of-coverage are expected from the nominal survey design (section I.4.a.3). However, there are regions where the depth-of-coverage varies considerably due to the exclusion of "bad" input frames and individual pixels when generating the Image Atlas. This could be due to contamination from the moon, cosmic rays, and/or other outliers (e.g., induced by fast moving objects), or bad quality frames in general. The latter are usually associated with elevated noise or degraded image quality caused by excursions in spacecraft tracking. For details on the frame selection process, see section V.2.c. A general description of the construction of the depth-of-coverage maps and the physical meaning of pixel values therein is given in section II.3.d.
Examples of Atlas Images with low depth-of-coverage due to the above-mentioned effects are shown in section VI.2 (Figures 16 and 17). An immediate consequence for regions with low depth-of-coverage (covered effectively <~ 6 times) is that the temporal pixel-outlier rejection process becomes unreliable. Even though robust estimators were used in this method, they are still relatively "noisy" when the number of samples in the stack was low. Strictly speaking, no temporal pixel-outlier rejection is performed for depths-of-coverage ≤ 4. This criterion avoided throwing away too many good pixels during the coaddition process, with the caveat of increasing the number of unreliable or spurious detections (cosmic rays, fast moving object trails, and other transients). Note that the source catalog selection process explicitly excluded extractions from regions with a depth-of-coverage ≤ 4.
If one were to measure the equatorial (J2000) positions of sources off Atlas Images, e.g., via a flux-weighted centroid or peak-finding algorithm, there will be a systematic difference with the astrometrically calibrated positions from the WISE source catalog or any other astrometric reference catalog. For details, see section II.3.g. The absolute (radial) differences between Atlas Image-derived source-peaks and "true" astrometric positions anywhere on the sky are up to 0.5 arcsec in bands 1, 2, 3, and 1.4 arcsec in band 4. A procedure to correct the Atlas Image derived source positions and image WCS in general is given in section II.3.g.
Pixel units in the Atlas Image (and Single-Exposure Intensity images) are not calibrated in terms of absolute surface brightness. Their units are digital numbers (DN). The images are designed for photometric measurements relative to a local background value using the photometric Zero Point Magnitude (MAGZP) value provided in the Atlas Image FITS headers. Methods on how to convert photometric measurements made off the image products to absolute flux units are outlined in section II.3.f.
One of the first steps in the Atlas Image generation pipeline is background level matching of the single exposures. Given the WISE mission had no requirements on the background, we adopted a simple, fast, locally self-consistent method. Only the single exposures touching an individual Atlas Image tile were matched with respect to each other. Three caveats have been identified from this method:
 |
 |
| Figure 1 - Section of a band 3 (left) and band 4 (right) Atlas Image measuring ~54 x 46 arcmin2 containing the M83 galaxy at far left. Note the dark halo (depressed background) surrounding M83. The image stretch was set to exacerbate this depression. The depression here is ~0.5-0.6% relative to the median background per pixel on larger scales. | |
The Anomaly Gallery contains examples of the artifacts commonly encountered in Atlas Images. Overall, these fall in two broad categories, with much overlap between the two:
Saturated pixels are masked in the single exposure images and omitted when constructing the Atlas Images. For bright regions and sources exceeding the saturation limits, pixels in the Atlas Image intensity and uncertainty maps will be associated with NaN (Not-a-Number) values. The corresponding depth-of-coverage map will have zero coverage in these regions. These "holes" are typically seen at the location of the peak-emission of sources. An example is shown here.
The Atlas Image generation pipeline includes a step to tag frames containing scattered moon-light for further dissemination downstream (see section IV.5.a.iv for details). The initial moon-tagging step is done using a moon-centric mask that indicates regions most contaminated by the moon. A frame that hits any one these masked regions is tagged as containing scattered moon-light. However, following generation of the preliminary-release Image Atlas, a bug was discovered when computing the position angle of a frame in the moon's reference frame as defined by the mask. This error is expected to have mis-tagged only band 3 & 4 frames that would otherwise have landed on one of the moon diffraction-spike arms or outer ring in the mask. Frames landing in the central symmetric core would have been correctly tagged. Bands 1 & 2 are not affected since the moon masks are circularly symmetric. Nonetheless, we do not expect this error to have had a significant impact since the moon masks only served as a prior tagging step. It is likely that the "missed" frames would have been excluded by the downstream frame-outlier rejection step.
Last update: 2011 May 23