Known Issues

General Issues

The most up-to-date issues and in-progress fixes are available on the GitHub issues pages for the Legacy Surveys website or the legacypipe pipeline.

Issues with CCDs that may have affected the quality of DECam observations are recorded on the DECam CCDs page.

Header inconsistencies

Headers may be inconsistent with catalog values for about 2% of bricks in the Tractor catalogs. This is due to some files being altered while rsync processes were running.

Duplicate ref_id values

A very small number of ref_id values for Gaia sources appear twice in the Tractor catalogs. One of the duplicates corresponds to the "correct" Gaia source and the second duplicate has all of the correct information for a Gaia source but with an incorrect ref_id. The ref_id that would be correct for the second duplicate is missing entirely from the Tractor catalogs.

An example from DR9 is ref_id 303293444934132224, which appears in Tractor catalog tractor-0238p302.fits with correct information (RA, Dec, etc.) and in Tractor catalog tractor-0238p305.fits with information that actually corresponds to the Gaia source with ref_id 303315950562852864.

The cause of this problem has yet to be fully diagnosed, although the discrepant Gaia sources tend to lie very close to the boundaries of a brick.

This issue will also propagate to the sweep files.

Bricks where source-fitting failed

A few dozen bricks were intractable on a reasonable timescale as they overlapped regions of high source density. These bricks were "bailed out" of before they finished processing. Sources in such bricks can be identified as they have the BAILOUT flag set in the MASKBITS bitmask.

In addition, certain high-density regions are missing completely from the Legacy Surveys. A good example is the Large Magellanic Cloud.

Bricks for which SUB_BLOB was set

To facilitate rapid processing of sources in high-density regions, some DR10 bricks were split into smaller "sub-blobs." These bricks can be identified as they have SUB_BLOB set in the MASKBITS bitmask for some pixels in the image stacks version of the maskbits product. Due to a bug in processing these sub-blobs, some genuine sources were not initially extracted and included in DR10. The locations of the 598 affected bricks are plotted below


These bricks were re-processed in the summer of 2023, effectively fixing the bug, and the DR10 release directory was updated with the relevant files on the weekend of September 9, 2023. This bug fix can be thought of as "DR10.1." Sources that were updated for DR10.1 have a RELEASE integer of 10002 (compared to the standard RELEASE integer of 10000).

Files that were updated to fix the SUB_BLOB bug include the coadd stacks, tractor and tractor-i files, random catalogs, sweep files, external-match files, and metric and log files. The full list of the updated files is available here. The random catalogs, and the tractor, tractor-i and coadd files were completely replaced, and we recommend always using the latest versions. The original and updated versions of the sweep and external-match files were both retained. The original versions of the sweeps are in the directories 10.0, 10.0-extra and 10.0-lightcurves and the updated versions are in the directories 10.1, 10.1-extra and 10.1-lightcurves. The original and updated versions of the external-match files are both in the external directory, but the original files contain the string dr10 and the updated files contain the string dr10.1.

Large galaxies missing from the Siena Galaxy Atlas

Fifty-two galaxies were overlooked in the version of the Siena Galaxy Atlas (SGA) used to process DR10. The upshot of this oversight is that these galaxies were not flagged as GALAXY in the MASKBITS bitmask. Further details are provided in legacypipe issue #680. Note that these fifty-two galaxies were included in the publicly released version of the SGA.

Updates to the brick-level summary statistics files

The initial brick-level summary file released with DR10 contained an incorrect accounting of the number of sources in some bricks. This was likely caused by processing results before all relevant files in the release were finalized and rsynced. Multiple columns in the summary file were inaccurate to varying degrees, including nexphist_*, nobjs, npsf, nrex, nexp, ndev, nser, psfsize_*, psfdepth_*, galdepth_*, ebv, trans_*, cosky_*, ext_* and wise_nobs. In December of 2023, we completely replaced this file. The table of the number of sources of each morphological type on the description page was also updated to reflect the new information.

The dr10-south-depth.fits.gz and dr10-south-depth-summary.fits.gz files were also updated in December of 2023, although all of the changes in these summary files were caused by fixes to the 598 bricks affected by the SUB_BLOB bug (described above).