Position differences for the four catalog test fields are analyzed. These results provide some insight into expected catalog quality of positions, position uncertainties as well as their consistency. The quality of the astrometry in the Polar1 test field, although well within the requirement, is significantly less than in the other three fields. This inconsistency points up the need to characterize the 2MASS astrometry as a function of sky position.
Figure 1 plots mean radial difference (Ave_dr) as a function of Ks magnitude for each on the four test fields. The Equator1, Equator2 and South1 test fields are quite consistent with a gradual increase in Ave_dr with magnitude, starting around 13. On the bright end a steeper increase with decreasing magnitude starts near the Read1:Read2 transition at about 8.5. The Polar1 Ave_dr values are worse, particularly for sources fainter than about 10th magnitude. Figure 2 repeats the previous figure for chi-square values, with mean x-scan chi-squares in the upper panel and in-scan in the lower. Results for the Equator1, Equator2 and South1 test fields are reasonably consistent, with chi-squares near one over most of the magnitude range. For these fields uncertainties generally look good, but with a tendency to be a little underestimated at the bright end (Read1). The Polar1 uncertainties, on the other hand, are significantly underestimated for magnitudes fainter than 10.
Figure 3 plots mean radial difference as a function of declination for the Polar1 test field. Note that the radial differences are elevated over the entire polar cap Dec range with a sharp spike upward very near the pole and a lessor trend upward near the south end. This behavior is puzzling. The spike very near the pole suggests an accuracy problem. The checking was originally done using vector differences executed in double precision. It was then repeated using quad precision with no significant change in the results. Thus, the spike appears to be real. It's source is unclear since the original transformations performed during pipeline processing used double precision vectors. The good news is that it affects a very small sky area. It also should be noted that the error spike very near the pole has a disproportionate effect on the overall Polar1 statistics because the number of overlap pairs increases dramatically near the pole (see Figure 4). The elevated errors over the entire polar Dec band, while not a total surprise (nightly QA checking gave some indication), are not well understood. A first look at Tycho-2 differences over the entire sky indicates that increased errors in the polar Dec band are part of a global variation in errors with a gradual increase toward the poles. This can be seen in Figure 5 which plots mean radial differences between 2MASS and Tycho-2 stars also found in Tycho-1. Note that the errors near the South pole are larger than in the North.
Figure 6 plots mean chi-square values as a function of Declination
for Polar1, with x-scan values in the upper panel and in-scan in the lower. In both cases uncertainties
are underestimated within a degree of the pole. The in-scan uncertainties are slightly underestimated
at the south end as well.
A strong case can be made for including the pole scans in the list of scans to receive
uncertainty adjustments. This would not be too difficult since relatively few scans are involved.
