Problem-Scan Positions vs Tycho-2 for 2nd Incremental Release

Two sets of scans from the 2nd Incremental Release were selected as having position reconstruction problems, independent of any comparison to the Tycho-2. The first set of 421 scans was simply composed of those which, because of ACT counts or distribution, did not meet the criteria for an ACT reconstruction. These scans were reconstructed using the lower-accuracy higher-density USNOA as the reference catalog. The pipeline switched from USNO-A1.0 to USNO-A2.0 on the processing date of Sept. 16, 1998 and all but a handful of the 421 "non-ACT recon" scans were processed with version 2.0. The second set of 184 problem scans was obtained by selecting all scans where the martinizing algorithm wanted an adjustment in excess of 0.75 asec in either direction, on any one of the 12 scan segments. This resulted in 208 scans, which then dropped to 184 when "non-ACT-recon" scans were removed. The resulting group of 184 scans will be referred to as the "prob-ACT-recon" set.

II. Comparison of the Two Types:

At first it might seem odd that there is so little overlap between the two lists. This can be explained by the fact that the "non-ACT-recon" scans were reconstructed with the much higher density USNOA. Although the USNOA is far less accurate than the the ACT, it's higher density generally prevents the solution from random-walking off to really large errors. This point is illustrated by histograms of the 2MASS/Tycho-2 residuals. Although residuals for all 421 "non-ACT-recon" scans have been provided, residuals for only 102 of the 184 "prob-ACT-recon" scans are as yet available. Dave was given the second set some time after the first.

Figure 1 and Figure 2 histogram cross-scan differences for the "non-ACT-recon" and "prob-ACT-recon" sets, respectively. Note that although the cross-scan sigma is sightly higher for the "prob-ACT-recon" set (0.54 vs 0.47 asec) it is driven by the long tails. The FWHM of the "prob-ACT-recon" distribution is in fact half as wide as that of the "non-ACT-recon" distribution. Figure 3 and Figure 4 repeat the histograms for the in-scan differences. Once more the "prob-ACT-recon" sigma is slightly higher (0.45 vs 0.37 asec) but it's FWHM is only half as wide. Note also that the "non-ACT-recon" set has a quarter arc-second in-scan bias which is not present in the "prob-ACT-recon" set.

In closing, it should be pointed out that problem scans discussed here are very much in the tails of the overall distribution of residuals. Dave Monet reports overall 2nd Incremental Release sigmas for the 2MASS/Tycho-2 differences of 0.18 and 0.17 arc-seconds for cross-scan and in-scan respectively. It remains to be demonstrated that the 2MASS position uncertainties conservatively reflect the errors associated with these problem scans. Once the associated position uncertainty data are made available to me, I will amend this URL to include a chi-squared analysis.