Old Photometric Redshift Sweeps

The DR9 photometric redshift sweeps were updated in July of 2023, and we recommend that users adopt the newer versions.

Below, for posterity, we retain the information for the earlier versions of the photometric redshift sweeps, which are stored at 9.0-photo-z/sweep-<brickmin>-<brickmax>-pz.fits.

The Photometric Redshifts for the Legacy Surveys (PRLS, e.g., see Zhou et al. 2021) catalog is row-by-row-matched to the DR9 sweep catalogs as described previously for the other types of sweeps files.

The photometric redshifts are computed using the random forest algorithm. Details of the photo-z training and performance can be found in Zhou et al. (2021). For computing the photo-z's, we require at least one exposure in \(g\), \(r\) and \(z\) bands (NOBS_G,R,Z>1). For objects that do not meet the NOBS cut, the photo-z values are filled with -99. Although we provide photo-z's for all objects that meet the NOBS cut, only relatively bright objects have reliable photo-z's. As a rule of thumb, objects brighter than \(z\)-band magnitude of 21 are mostly reliable, whereas fainter objects are increasingly unreliable with large systematic offsets.

The photo-z catalogs do not provide information on star-galaxy separation. Stars are excluded from the photo-z training data, and we do not attempt to identify stars. To perform star-galaxy separation, one can use the morphological "TYPE" and/or the photometry (e.g., the optical-WISE color cut, as applied in Zhou et al. 2021, can be very effective for selecting redshift ≳ 0.3 galaxies) in the sweep catalogs.

Work which uses this photometric redshift catalog should cite Zhou et al. (2021) and include the additional acknowledgment for photometric redshifts.