Please familiarize yourself with the cautionary notes on using the Single-exposure Image products before reading this section.
The NEOWISE survey consists of repeated observations of each point on the sky using overlapping exposures that are offset from each other according to the survey-scan strategy. This results in typically 12 or more independent measurements of each point on the sky per observation epoch, therefore sampling different regions of the W1 and W2 arrays. Therefore, one can take advantage of this redundancy to mitigate the impact from bad pixels, charged-particle hits, and other detector glitches at the Single-exposure level by co-adding the Single-exposures. One can use the accompanying bit-mask images and/or perform outlier rejection within stacks of repeated pixel measurements to omit bad/outlying pixels. The corresponding uncertainty images can also be used to weight the pixel measurements. This will also enable one to go deeper, in particular by including Single-exposures from the earlier Cryogenic and/or Post-cryogenic phases of the WISE mission. Therefore, the benefits of co-addition are enormous. Co-addition can be performed in either the inertial equatorial reference system (using the native WCS of each Single-exposure), or, with additional tweaking of the WCS metadata, in the reference frame of a moving object.
Unlike the earlier 4-band (All-Sky), 3-band Cryo, and AllWISE Data Releases, there are no co-added image products (i.e. an Image Atlas) accompanying the NEOWISE Data Releases. An interactive image co-addition service that is optimized for WISE and NEOWISE Single-exposure images is available through IRSA. This service enables users to create custom co-adds with specifiable subsets of WISE and NEOWISE Single-exposure images at either stationary or moving positions and with optional resolution enhancement.
The Single-exposure images are contaminated by large numbers of bad (unusable) pixels, charged particle hits, and saturated pixels. Concentric or elliptical aperture photometry will be impacted by these bad and "outlying" pixels, leading to biased and erroneous measurements. To obtain the photometry of compact (point-like) sources of interest on the Single-exposure images, your best option is to check the Single-exposure Source Database and use the PSF-fit photometry measurements therein, if available. The PSF-fitting process in the WSDS pipeline automatically omits bad and discrepant pixels using the bit-mask images. The fitting process also uses the uncertainty images to assign relative pixel weights. Please also read the cautionary notes associated with measurements in the Single-exposure Source Database.
Furthermore, in light of the benefits of image co-addition noted in section III.3.a above, the photometry for stationary and non-variable sources, including large-aperture and elliptical-aperture photometry for extended objects, can be obtained from the deeper Source Catalogs included in the All-Sky and AllWISE Data Releases. These Catalogs were generated by combining all good-quality Single-exposures acquired in earlier phases of the WISE mission.
If you plan to perform your own aperture photometry on the NEOWISE Single-exposure images directly, it is recommended you examine the intensity images for bad and discrepant pixels that could affect your estimation process (including background estimation). A method on how to use the intensity and uncertainty images for performing simple aperture photometry is outlined in section IV.3.a of the AllWISE Explanatory Supplement. It's important to note that this methodology is optimized for co-added image products (e.g., Atlas Images from the previous WISE releases), where image-stacking would have eliminated most of the bad and outlying pixels. It is mentioned here only as a guide, and care should be exercised if used on the Single-exposure images. Similar approaches are used in some freely available source-extraction tools, e.g., SExtractor and MOPEX. These tools can be configured to accept the NEOWISE Single-exposure image products.
The pixel values in the Single-exposure (intensity and uncertainty) images are in units of digital numbers (DN). Integrated (pixel-summed) estimates in DN can be converted to calibrated Vega-based magnitudes using the Magnitude Zero Point "MAGZP" in the FITS header of a Single-exposure image. This effectively represents the magnitude corresponding to 1 DN of signal. Equation 3 in section IV.3.a. of the AllWISE Explanatory Supplement describes this conversion. Note that uncertainties in the time-dependent instrumental MAGZP estimates are generally small compared to uncertainties contributed by photon-counting, read-noise, and other detector effects at the pixel level. Figure 1 in section IV.2.d could be used to infer a crude uncertainty.
The Vega-based W1 and W2 magnitudes can be converted to absolute flux density units (e.g., Jy) using the formulae and conversion factors presented in section IV.4.h.1 of the All-Sky Release Explanatory Supplement. Beware that these conversion factors depend on the SED of your sources or emission of interest. Suggested color corrections are also given in this section.
The Single-exposure Bit-Mask Images provide information about the status of each pixel in the Single-exposure intensity images. The pixel values therein are stored as 32-bit unsigned (long) integers with each bit assigned to a specific condition that may have been set either onboard the spacecraft (e.g., saturated-pixel tagging), during the instrumental calibration pipeline, and/or during continued detector characterization, trending, or the generation of instrumental calibrations. As described in section IV.2.a, not all conditions are "fatal", i.e., rendering a pixel unusable. For your convenience, these "fatal" pixels have been set to NaN in the intensity and uncertainty images. However, see cautionary note i in section III.2 for possibly missed bad pixels.
Any pixels with non-zero values in the bit-mask images that are not NaN'd in the intensity images should be examined for possible exclusion from your analyses. Depending on your use case, these pixels may or may not be detrimental. If you plan to use an image co-addition tool that accepts the NEOWISE Single-exposure bit-mask images (in 32-bit integer FITS format), for conservatism we recommend masking pixels associated with any of the following bits: b = 1, 3, 4, 6, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 21, 30, or equivalently in decimal: d = ∑2b = 1076887130. This masking is accomplished in software by performing a "logical AND" between d and all the pixels of a specific bit-mask image, then omitting the pixels where this AND'ing results in a non-zero value.
If the bad pixels reside in well isolated, relatively uniform background regions of an intensity image, i.e., not associated with the cores of (possibly saturated) sources or with larger contiguous bad-pixel clumps, you can attempt to replace them with some local estimate of the pixel signal. This is a dangerous operation if your goal is compact-source photometry. It can also be challenging for extended-source photometry. Any bad-pixel replacement is at your own discretion. For compact (point) source photometry, your best option is to check the Single-exposure Source Database for your sources of interest and if available, use the PSF-fit photometry measurements therein. Please also see the cautionary notes associated with these measurements.
The Single-exposure uncertainty images store the 1-σ uncertainty estimates for each pixel in the corresponding intensity images. Their computation and validation against robust measures of the local pixel-RMS noise in the intensity images are outlined in section IV.2.a.iv. These uncertainty images account for all instrumental calibration steps and can be used to compute uncertainties in photometric measurements made on the intensity images. Beware that these uncertainties do not include any component of confusion noise, which may be significant in regions with a high source density. Note that source extraction in the WSDS pipeline accounts for possible confusion noise when deriving the photometric uncertainties of extractions in the Single-exposure Source Database.
A methodology on how to use the pixel-uncertainty values to estimate uncertainties in integrated (pixel-summed) fluxes is outlined in section IV.3.a of the AllWISE Explanatory Supplement. This section shows the formalism on how these products would be used, in case you are interested in implementing your own software. However, some freely available source-extraction tools (e.g., SExtractor and MOPEX) can also be configured to accept pixel-uncertainty images. However, please be aware of the caveats on performing aperture photometry directly on the Single-exposure intensity images (section III.3.b above).
The Single-exposure pixel uncertainties are generally accurate to <~ 4% in W1 and W2. This is based on global trending analyses and comparisons to the pixel-RMS noise in the intensity images (section IV.2.a.iv). However, we cannot guarantee this accuracy applies to all products when used on an individual image basis. We recommend validating an uncertainty image before using it for photometry. This entails checking if the uncertainty values are over- or under-estimated relative to some robust measure of the pixel-RMS in the accompanying intensity image. Metrics to perform a coarse check are provided in the Image Metadata Table. These metrics are w1unclocscal2 and w2unclocscal2 (for W1 and W2 respectively), and represent the ratio of a robust measure of the intensity-pixel RMS to the modal pixel uncertainty in a specially selected region of the Single-exposure, i.e., with minimal contamination by sources and complex background emission. Values of w?unclocscal2 ~ 1 (analogous to the traditional reduced χ2 metric) indicate plausible uncertainties. Values of w?unclocscal2 that deviate by >~ 10% from unity, however, (i.e., with |1 - w?unclocscal2| >~ 0.1 ) may require a manual rescaling of the uncertainty image, i.e., a multiplication of the uncertainty image with w?unclocscal2. We say "may" since these metrics are still not totally immune to the presence of complex backgrounds or confused-source emission. Nonetheless, they can be used as "conservative" scaling factors, leading to perhaps slightly overestimated uncertainties. Note that this rescaling is an approximation since it only tests whether the local background RMS is more or less consistent with the overall pixel-uncertainty estimates, and not whether the Poisson-noise component associated with source emission over the full dynamic range is correct. The Poisson-noise component assumes the detector electronic gains are correct.
Last update: 18 March 2024