The atrometric precision in some sub-images, which are at the corner of the full image, depends on the position in the sub-image. The plot below shows the astrometric precision within sub-sub-images of a sub-image, i.e., a sub-image divided into further sub-images. The sub-sub-images are created in order and thus will show gradients within the sub-image. A clear trend is shown in the plot.
on the image: /last10e/data1/archive/LAST.01.10.01/2023/11/08/proc/164333v0/LAST.01.10.01_20231108.164633.263_clear_field1213_000_001_001_sci_coadd_Image_1.fits
Is the back and variance variable across the sub image?
Another possibility is that the extrapolation of the WCS from the central sub image was not accurate enough and some stars were missed in the astrometric refinmenet that in turn led into a bias in the astrometry - if this is the case, then this should be an intermitent problem (is it?) and can be maybe avoided by increasing the search radius in the astrometryRefine code.
The atrometric precision in some sub-images, which are at the corner of the full image, depends on the position in the sub-image. The plot below shows the astrometric precision within sub-sub-images of a sub-image, i.e., a sub-image divided into further sub-images. The sub-sub-images are created in order and thus will show gradients within the sub-image. A clear trend is shown in the plot.
The plot was produced with the script: script.zip
on the image: /last10e/data1/archive/LAST.01.10.01/2023/11/08/proc/164333v0/LAST.01.10.01_20231108.164633.263_clear_field1213_000_001_001_sci_coadd_Image_1.fits