The BANZAI pipeline uses SEP for extracting sources from images. The brightest 200 stars extracted by SEP are submitted to astrometry.net for astrometric calibration. For nearly all images, i.e. unless the image is obscured by clouds, an astrometric solution is found, and the WCSERR header variable is set to 0. LCO does not guarantee that unobscured images will yield WCS solutions, merely that our procedure is consistent, and that our source extraction parameters (detection threshold, minimum source size, etc.) were selected following thorough testing.
A few times every semester, a few LCO users have reported images, acquired under good conditions, that do not have astrometric (WCS) solutions. The images with WCSERR=4 typically share the following characteristics: (i) longer (>120s) exposure times; (ii) redder filters (rp, ip, or R); (iii) several bright stars, with saturated pixels and diffraction spikes, are in the field-of-view. We’ve determined that this final characteristic, pixel saturation, is critically important for determining whether an astrometric solution is fit to the image. When a bright star saturates the detector, then the saturated pixels are masked and therefore not available to SEP to attempt extraction. Consequently, the bright star is omitted from the source catalog. The more saturated stars in an image, the more sources are “missing” from the catalog, and the more difficult it becomes for astrometry.net to determine an astrometric solution.
In the image below, the red points are the 200 sources from the catalog that are sent to astrometry.net. Note that the ~15 brightest stars in the field-of-view are not included; they are identified by diffraction spikes but not by red points. Even some moderately bright stars without diffraction spikes near the center of the field (green circles) are not extracted by SEP. Note also that a few SEP “sources“ are false positives located within the diffraction spikes of saturated stars. The astrometric service failed to determine a solution for this image.
If you have unobstructed images for which no astrometric solution was found (WCSERR=4), check whether the images have the characteristics listed above. Simply reprocessing the images through the BANZAI pipeline will have no effect; the same SEP parameters will yield the same result. However, you could install SEP on your own, specify your own extraction parameters, and submit your BANZAI-processed images to it. The result, a different source catalog, could then be submitted to astrometry.net and may yield an astrometric solution. To see how BANZAI calls SEP, and the default extraction parameters, see photometry.py in the BANZAI Github repository.
For future observations of the same field, we recommend two strategies: