Forced photometry of the unWISE images given SDSS objects
Forced photometry of the WISE images given SDSS detections
The data described here are available at:
or for SDSS folks, riemann:/clusterfs/riemann/raid000/dstn/wise-sdss-forced-phot
I took all SDSS "v5b" survey_primary objects, and found all unWISE coadd tiles touching them.
The "unWISE coadds" are new coadds of the WISE imaging, described in [Lang 2014, AJ, 147, 108].
The unWISE tiles have the same centers as the official WISE Atlas Images; they are on equal-Dec rings, roughly 1.5 x 1.5 degrees in size, with a bit of overlap.
For each tile, I gathered all the SDSS objects within and near the tile, and then cross-matched to the WISE catalog ("AllWISE release") to find WISE-only sources. I treat the WISE-only sources as point sources, since no profile-fit measurements in the form I need are available in the WISE catalog. For SDSS sources, I apply some tests for sources that do not have reliably measured sizes, and treat those as point sources in the forced photometry; there is a flag for that in the output files.
With that list of sources, I apply the WISE PSF model to predict the appearance of each object in the WISE image, and fit for a linear combination of those bases that best match the WISE image. The best-fit linear weights are the measured fluxes. The unWISE coadds already have a scalar sky background level subtracted, so I do not allow the sky level to vary during fitting.
After fitting each tile, I need to split the sources back into files that match the SDSS field-by-field file layout. SDSS fields that are entirely contained within one WISE tile are written in one phase, and then the remaining ones that span multiple tiles are resolved (only the measurement with the closest tile center is kept) and written out in a second phase. SDSS fields that contain no primary objects will not have a corresponding photoWiseForced file.
The results are row-by-row parallel to the SDSS "v5b" "photoObj" files; the files
each have 368 rows, and describe row-by-row the same objects.
The "photoWiseForced" files have the following columns:
True for SDSS sources that were photometered in WISE. The object must be PRIMARY in SDSS for this to be set. When this column is False, all other columns have value zero.
- ra, dec
Copied straight from the SDSS photoObj file.
Zero-indexed pixel position on the unWISE 2048x2048 image tile.
The SDSS source is a galaxy (objc_type == 3) but was treated as a point source for the purposes of forced photometry. If you want an optical/WISE color, it would be best to use the SDSS PSF mags, not the model mags, for these objects.
The SDSS source is a point source (objc_type == 6).
The unWISE coadd tile name.
And then for each WISE band:
The WISE flux and formal error as an inverse-variance. Take the formal errors with a grain of salt, of course. NOTE that these are in the native WISE photometric system: Vega, NOT AB, so some would say these should not be called "nanomaggies". A source with magnitude 22.5 in the VEGA system would have a w1_nanomaggies flux of 1.
The WISE team's suggested conversions to AB are available here: http://wise2.ipac.caltech.edu/docs/release/allsky/expsup/sec4_4h.html#conv2ab
And are (or were, when this was written),
mag_AB = mag_Vega + dm,
Band dm W1 2.699 W2 3.339 W3 5.174 W4 6.620
Vega magnitude and formal error in the forced photometry. These are simple conversions from the "nanomaggies" columns.
Profile-weighted chi-squared and number-of-pixels values. "Profile-weighted" means that these are weighted according to the brightness of the source in the WISE images; they are supposed to mean "how bad is the fit, where I am?" The "pronpix" is probably not useful in this dataset -- it effectively counts what fraction of the source was inside the image, and should be nearly 1. everywhere.
"proflux": profile-weighted, how much flux comes from other sources in the model? "profracflux" = proflux divided by the flux of this source. This is an indication of blending or confusion; a source with a large profracflux has a significant contribution from a nearby source.
How many pixels were included in the fit.