T. Jarrett, IPAC
(980715)
As part of the 2MASS survey strategy, several fields are targeted for periodic repeated observations to serve as validation test fields. The first 2MASS validation field is located at high glat, partly encorporating the Corona Borealis galaxy cluster, Abell 2065 (z=0.072). A nice pic of the core can be seen below.
The most recent data taken of this validation field is from 980620n, scans 7 to 14 (abell 2065 is located at the very top of the tile in declination). We can compare this data set with scans taken last year, including two scans from 970608n, two scans from 970615n, and four scans from 970617n. (note: there are other scans from different nights as well, but the photometric quality is poor).
The 980620n scans were taken with special tile positions so that the inscan and cross-scan positions are offset about 4' from the original (1997) tiles. This feature is necessary to delineate any systematics in the photometry due to sky position (on the 2mass arrays).
The following plots show the difference in position, photometry and completeness between the 1998 data and the 1997 data. In general, due to the incomplete overlap between the data sets, there is only one set of points per source (one from 980620n, the other from either 960617n or 970615n or 970608n). So the "delta" mag and position values refer to one 1997 value versus one 1998 value.
Some numbers:
About 210 galaxies have corresponding matches (i.e., overlap) with data from 1997 (970617n has the best overlap with 980620n, including the abell 2065 cluster itself).
Coordinate Position Comparison
GALWORKS determines the coordinate position for galaxies in two different ways. One uses the peak pixel position of the galaxy as seen in the J-band image (J is used instead of Ks due to the superior sensitivety). This method, therefore, has a built in uncertainty due to the "fat" 2MASS pixels (on order of 1/3 to 1/2 of a pixel) and due to sky noise pushing around the peak flux of the galaxy between pixels. The second method uses the "super" coadd (J+H+Ks) with an intensity-weighted centroiding. The latter method is generally superior to the former method (as the plots below demonstrate). The only problem with the "super" coadd is that the galaxy is slightly smeared due to band to band registration and extreme colored galaxies will have intrinsically lower SNR when all three bands are added. The latter problem is rare, however.
Conclusion: there is no systematic position difference between 1997 and 1998 data sets. The scatter is on order of 0.3 to 0.4".
Photometry Comparison
The most robust flux measurement used by GALWORKS is that obtained from fixed circular radius = 7" aperture photometry. Any flux difference observed between the 1997 and 1998 data sets is confined to: (1) real photometric scatter, (2) systematic in the zero point calibration and (3) background determination differences (driven by sky noise).
Conclusion: there is no or, at most 1 to 2%, systematic photometric difference between 1997 and 1998 data sets. The scatter ranges from a few % (for the highst snr points) to 15 to 20%. The magnitude in the scatter roughly corresponds to the expected value (essentially sqrt(2) times the individual photometric uncertainty).
Photometric Difference as a function of the coadd cross-scan position
There is some worry that a cross-scan systematic in the detector Q.E. can introduce a few % bias in the photometry as a function of the cross-scan position. We can test this by plotting the cross-scan position (essentially the coadd "x" position) as a function of the photometric difference between the 1997 and 1998 data sets (note again that the 1998 data is offset in cross & in-scan position from that of 1997).
Conclusion: there does not appear to be any systematic in the photometry (within the scatter) across the tile (cross-scan).
Repeatibility
A test of the internal repeatibility and reliability between the '97 and '98 data sets is conducted. We have applied a simple (and mostly harmless to C) color-criteria to the data to cull out obvious false sources (essentially blue double stars).
Conclusion: the repeatibility is generally 95- 100% all across the board, with the rate of false detections something less than 1% (one or two false things for 200 gals or so).