UKIRT Data Products
For A963 we obtained JHK mapping, and for A2192 we obtained HK. More details can be found below.
ASCII catalogs of NIR detections in each field
a963_ukirt_catalog_v1.dat.tar.gz
a2192_ukirt_catalog_v1.dat.tar.gz
The catalogs above include SExtractor flags and stellarity indices (0-1, where 1 is very star-like) from each band.
Maps
The maps linked here have been compressed using FPACK and they've been compressed with tar. After they've been uncompressed using tar -zxvf *.fits.fz.tar.gz, you can use funpack *.fits.fz to get the full mosaic. There are also some IDL tools, like READFITS, which can read in the FPACK-compressed mosaics.
If you wish to do your own source extraction and photometry on these mosaics, you'll want to use the following:
- GAIN = 13.5
- MAG ZERO PT = 30.0 [resulting in mags on the AB system!]
Each mapping consists of four pointings, and each pointing has four frames, so that the final mosaic consists of 16 frames and nearly even coverage. However, since each field was mapped twice I needed to coadd the two mappings in each band. Prior to coadding, I multiplied each frame by a factor to correct for slight differences in the magnitude zeropoint (calculated by the UKIRT pipeline) and put them all on the same mag zeropt. Then I used SWARP to coadd, and measured photometry using SExtractor. After applying an aperture correction, and bandmerging the catalogs for each field, I checked the mosaics and source positions by hand to remove a small number (roughly 1%) of bright stars and spurious sources associated with image artefacts. Note that I only excluded the sources which are obviously stars by eye. I haven't yet excluded fainter stars by making a stellarity cut, or a NIR color cut.
After refining the bandmerged catalogs for A963 and A2192, I ran some sanity checks to be sure that the photometry and astrometry are all right. I did a crude matching of sources to the 2MASS point source catalog, and below I plot a photometry comparison for A963:
And for A2192:
And here is a plot of the positional offset for all UKIRT-2MASS matches in A963:
And in A2192:
Follow-up checks up on the photometry
The photometry comparison plots above show an over-estimation of the brightness in the UKIRT bands (or under-estimation in 2MASS?) for fainter sources, and my first assumption is that I did something wrong in my reduction or photometry in our UKIRT maps. To test this assumption, I downloaded a large chunk of the UKIDSS Deep Survey (UDS), specifically in the Large Area Survey portion. The Data Release 8 for UDS is available here, and the Large Area Survey contains tens of millions of UKIRT detections in J, H, and K. I downloaded a large enough subset of the sample to obtain just over 50000 UDS-2MASS matches. I use UDS magnitudes obtained with the same circular aperture size (7.5 pixel radius) that I used to measure photometry in A963 and A2192. Below I plot the photometry comparison for those UDS-2MASS matches:
There's definitely a similar offset that becomes especially apparent at magnitudes of 17 or fainter. Examining the photometry for stars and galaxies separately could be useful, so I split the sample of UDS-2MASS matches by their star-galaxy parameter in a simple, crude way (if prob(star) is greater than or equal to 0.5 classify it as a star, and if prob(galaxy) is greater than or equal to 0.5 then classify it as a galaxy) and plotted the two populations separately below. First are the likely stars:
While the bulk of the brighter star-like sources follow the 1:1 line pretty closely, there's a sizeable number of sources showing a systematic offset at fainter magnitudes. Below I plot the photometry comparison for galaxies only.
What's going on?
The good news is that the offset that we're seeing when comparing photometry between UKIRT and 2MASS in A963 and A2192 is also seen in other fields mapped by UKIRT. It was suggested by Min, and I agree with this, that the primary reason for the over-prediction of the brightness in 2MASS at the faint end is the Eddington Bias. You can imagine that the scatter of points above and below the red line at y=0, if the only factor affecting the magnitudes was intrinsic photometric measurement errors, should be perfectly symmetrical at all magnitudes. However, when you're looking at objects near the detection limit of a survey (as AB magnitudes around 16-18 are for 2MASS), and sources have inherently larger photometric errors as they get fainter, those sources whose flux uncertainty happens to amplify their brightness will be detected more readily than those which happen to diminish their brightness, because those in the latter group will likely not be detected above the noise at magnitudes of around 16-18. This was mentioned briefly in Hodgkin et al. (2009), which is the standard UKIDSS photometric calibration paper. In section 5 they acknowledge an observed over-prediction of the 2MASS flux at the faint end, and attribute it to the Eddington Bias (see figure 13 of the paper for a magnitude difference plot using 2MASS calibrator sources).
However, I also noticed that this doesn't look like it entirely explains what we're seeing in the UDS-2MASS comparison plots above. Specifically, when I plot just the objects classified as galaxies, they clearly tend to lie systematically above the red line at all magnitude ranges, and not just at the faint end. Furthermore, the photometric comparison of galaxies, particularly in J-band, between UDS and 2MASS galaxies seems to show a bimodality at the bright end, with a heavily-populated branch lying just above the y=0 line but a separate branch at around J(UDS)-J(2MASS)~0.5 or so. Nevertheless, I'm not too concerned with this because a careful inspection of our photometry comparison plots from A963 and A2192 doesn't really show such a bimodality. I'm inclined to conclude that we're only seeing the effects of Eddington Bias in our data.