11: stack astrometry correction with Gaia/WISE
This pipeline stage is new in CSC 2.1.
In CSC 2.1, the astrometric reference frame for the majority of stacks is tied to the Gaia DR3 celestial reference frame (Gaia-CRF3). The astrometry update is based on a translational transformation of the mapping between SKY (X,Y) coordinates and WCS RA and Dec coordinates that was constructed by matching CSC sources to Gaia counterparts, and then performing a weighted least square transformation of the coordinates. For stacks with insufficient CSC/Gaia matches, we match the CSC sources to ALLWISE catalog sources, and then use the Gaia archive-provided best neighbor pairings between Gaia and ALLWISE to perform a two-step transformation. For a given stack, the transformation is applied to all stack detections, and to the per-observation detections of the contribution observations. This reflects in the final coordinates of the master sources. The transformation also applies to the associated full-field and region data products both at the stack and per-observation levels, such as background maps, sensitivity maps, event files, etc. The resulting coordinate shift for each detection is recorded in the deltax and deltay columns in both the stacked observation detections table and per-observation detections table. The man_astrom_flags column is set if the stack to which the detection belongs has seen its astrometry updated. A detailed description of the algorithm used can be found in the Stack Astrometry Correction Specification Document.
Stack astrometry correction and absolute position error
Anchoring the CSC 2.1 astrometry to Gaia-CRF3 has resulted in a significant improvement in the absolute position error in the new version.
[Version: full-size]
The vast majority of sources (~90%) have a coordinate shift of 0.4 arcseconds or less in the new version of the catalog, with the distribution peaking at 0.2 arcseconds.
[Version: full-size]
[Version: full-size]
The stack astrometry correction has reduced the systematic component of source position uncertainties in CSC 2.1. This systematic error term in the position error has gone from 0.7 arcseconds in CSC 2.0 to 0.29 arcseconds in CSC 21, per axis, at the 95% confidence level. The determination of the absolute astrometric error was done using the same procedure as in Rots & Budavari 2011 (ApJS 192, 8), based on the cross-match between CSC and Gaia sources.
Note on Quality Assurance
To ensure that the automatic stack astrometry correction does not introduce incorrect shifts due to low number statistics or confusion in very crowded regions, we perform a manual QA step for those stacks where the density of sources is above a certain threshold, when there are too few matches, or when the resulting automatic shift is larger than 1 arcsecond. In that case we manually select the matches, or opt for the identity transformation shift if that is more appropriate.