Problem-Scan Positions vs UCAC for 2nd Incremental Release

Identification of possible problem scans was described previously in an analysis of 2MASS/Tycho-2 difference statistics. As described there, the potential problem scans were selected in two ways. The first set was composed of all 421 scans which did not meet the criteria for an ACT reconstruction. These will be refered to as "non-ACT-recon" scans. A second set of 184 "prob-ACT-recon" scans was obtained by selecting ACT-reconstructed scans for which the Martinizing algorithm wanted to make unusually large adjustments. From these two data sets Dave Monet returned position differences for 199 scans with UCAC overlaps. For this analysis, the 2MASS/UCAC comparison set was further reduced to 188 scans for which 2MASS uncertainties were readily available. It remains a sizable comparison set with over 70,000 source matches.

II. Comparison of Position Errors with Quoted Uncertainties:

A number of sample scans comparing position error with the 2-sigma uncertainty envelope over the length of each scan will be presented. The quoted uncertainties apply to clean sources with minimum extraction errors. The plotted 2MASS/UCAC differences are for 2MASS sources selected via the same criteria. Each figure is a four-panel plot with RA differences/uncertainties in the upper-left panel and Dec differences/uncertainties in the lower left. The corresponding chi-squared histograms can be found in the panels on the right.

Typical problems associated with "non-ACT-recon" scans are illustrated by scans 980727s119, 981007s075, 981023s060, 981213s031, 980710s147 and 980710s135. Typical problems associated with "prob-ACT-recon" scans are illustrated by scans 980430s083, 980517s069, 980731s043, 981112s014, 980904s036 and 981202s013. From these examples it can be seen that, although the reconstruction errors can become quite large, the quoted uncertainties are conservative. This further demonstrated by the Chi-Square histograms for the 2MASS/UCAC matches from all 188 scans which show mean values of 0.72 x-scan and 0.81 inscan.

III. Improvements Possible with Martinizing:

This set of identified 2MASS/UCAC matches provides a unique opportunity to evaluate the potential effectiveness of the Martinizing algorithm. Note that for the results presented here the distortion model has not yet been implemented. It should also be mentioned that no UCAC or Tycho-2 information was used in the creation of the Martinized positions. The UCAC is merely used as a truth table after the fact. To demonstrate the effectiveness of the Martinizing algorithm, the same sample scans presented previously are revisited. This time each four-panel figure plots mean 2MASS-UCAC differences as a function of Declination in the left panels. Original differences are plotted with solid black lines and differences with Martinized corrections with dashed red lines. Histograms of position differences are presented in panels to the right with the same color coding. Cross-scan differences are plotted in the upper panels and in-scan differences in the lower panels.

Martinized results for the "non-ACT-recon" sample scans are presented for 980727s119, 981007s075, 981023s060, 981213s031, 980710s147 and 980710s135. Martinized results for the "prob-ACT-recon" sample scans are presented for 980430s083, 980517s069, 980731s043, 981112s014, 980904s036 and 981202s013. From these examples it can be seen that the Martinizing algorithm provides significant improvement to problem scans. This is further demonstrated by the pre/post correction position difference histograms for the 2MASS/UCAC matches from all 188 scans. Note that the improvement for problem scans would be expected to be more dramatic than for normal scans since the need for improvement is so much greater. On the other hand, the imminent public release of Tycho-2 and implementation of the distortion model should further improve the Martinized results.