DR9 will be the final release of the Legacy Surveys that is used for DESI target selection. As such, a major push was made to update the DR8 code and data model to prepare for DESI observations. This page documents those updates, in brief.

## Algorithmic changes for optical data

• All raw images from Legacy Surveys observations are initially reduced using the NOIRLab Community Pipeline (CP). For DR9, the CP has been rerun in full on each of the raw images:

• The CP now imposes no interpolation.

• The CP has improved fit parameters for good seeing images.

• The CP WCS now consistently uses Gaia DR1.

• The CP now doesn't remove pixels in masks, rather, it just flags those pixels as having been masked

• The CP now has improved star flats fit across all of the imaging.

• Changes in the CP (in particular the switch from dark sky flats to DES star flats) produced residual sky patterns that have been corrected.

• New fringe templates (and scaling) were applied instead of the CP fringe correction for $z\hbox{-}\mathrm{band}$ DECaLS imaging.

• The Tractor pipeline now includes iterative fitting:

• After a first round of fits, the data-model residuals are calculated and a second round of fits is conducted.

• A modified, extended PSF model is now used to subtract the extended wings of bright stars from DECam images.

• This updated PSF model leads to different sky levels.

• The "COMP" (composite) morphological type from DR8 has been replaced by a Sersic profile (type=SER).

• The criterion used to force a Gaia point source was changed for DR9:

• Previously, Gaia sources were fixed to type=PSF if:

• (G $\leq 19$ AND ASTROMETRIC_EXCESS_NOISE < $10^{0.5}$) OR (G $\geq 19$ AND ASTROMETRIC_EXCESS_NOISE < $10^{0.5 + 0.2(\mathtt{G} - 19)}$

• For DR9, Gaia sources were fixed to type=PSF if:

• (G $\leq 18$ AND ASTROMETRIC_EXCESS_NOISE < $10^{0.5}$) OR (G $\leq 13$)

• Sources in the SGA and globular cluster and planetary nebula catalogs are "pre-burned" [1] for DR9, and these catalogs have also been extensively improved:

• The globular cluster and planetary nebula reference catalog has been updated, and the angular sizes of many sources in this catalog have now been set by visual inspection (see legacypipe PR 469 and PR 504).

• The properties of Local Group dwarf galaxies in the SGA have been updated based on Vizier records, measurements in McConnachie (2012), and visual inspection. In particular

• Fornax and Sculptor have been moved from the SGA large galaxy catalog to the globular cluster and planetary nebula catalog.

• A total of 26 galaxies that, on visual inspection, were deemed to not benefit from "local" Tractor fitting have been removed from the SGA.

• Some additional galaxies, beyond those that were included for DR8, have been added to the SGA large galaxy catalog for DR9.

## Algorithmic changes for WISE

• WISE imaging has been updated to include observations through the 6th year of the NEOWISE W1/W2 data.

• The unWISE astrometry has been re-generated, and proper motions of stars have been correctly accounted for when conducting time-resolved forced photometry.

• unWISE epochs have been resampled to have more uniform coverage for the WISE light curves, particularly near the ecliptic poles (e.g. for columns such as lc_flux_w1 on the catalogs page).

• The unWISE epoch index number is now recorded in the Tractor catalogs.

• Bit-flags for pixels that lie in some of the brightest WISE bright star masks have been updated.

## Other algorithmic changes

• The DR9 external match files now include matches to Data Release 16 of the SDSS.

• Starting with DR9, we recover any duplicates in the SDSS spectroscopic files. The result is that all sources from SDSS spectroscopy should be in the external match files, even if they are duplicates from repeat spectroscopy.

## Data model changes

• All WISE light-curve columns from the Tractor catalogs are now packaged with the sweep files in their own set of light curve sweep files, which are row-by-row-parallel to the standard sweep files.

• Any remaining columns from the Tractor catalogs that are not in the standard sweep files or the light curve sweep files are now packaged in their own set of extra sweep files, which are row-by-row-parallel to the standard sweep files.

• The shape parameters in the Tractor catalogs (and downstream) have changed, as there are now no sources of type=COMP (which corresponded to a "composite" galaxy):

• The DR8 catalog columns:

• fracdev, fracdev_ivar

• shapedev_e1, shapedev_e1_ivar, shapedev_e2, shapedev_e2_ivar, shapedev_r, shapedev_r_ivar

• shapeexp_e1, shapeexp_e1_ivar, shapeexp_e2, shapeexp_e2_ivar, shapeexp_r, shapeexp_r_ivar

• have been replaced, in DR9, by:

• shape_e1, shape_e1_ivar, shape_e2, shape_e2_ivar, shape_r, shape_r_ivar

• The logic behind this change is that, in DR8, the composite type comprised some fraction (fracdev) of a de Vaucouleurs profile, with the remaining fraction being an exponential profile:

• the shapedev_ and shapeexp_ parameters, in DR8, defined the appropriate parameters for the de Vaucouleurs and exponential profiles. In the event that something was fit with type=DEV or type=EXP, only the columns that corresponded to the relevant profile would be populated.

• the "COMP" profile has been replaced by a Sersic profile in DR9, so there is no need to have "mixes" of "DEV" and "EXP".

• In DR9, every profile is defined by its type, and shape_r, shape_e1 etc. are the appropriate parameters for that type (so, i.e., shapedev_r and shapeexp_r can now be collapsed to just the single shape_r column).

• As there are no sources of type=COMP, the type column is now a 3-string ('S3') rather than a 4-string ('S4').

• Two new columns related to the Sersic profiles now exist in the Tractor catalogs for DR9:

• sersic and sersic_ivar

• The wise_x and wise_y coordinates in the unWISE coadds (e.g. files like legacysurvey-<brick>-image-W1.fits.fz) have been added to the Tractor catalogs for DR9. The transformations between these locations and the bx and by coordinates in the optical coadds are:

• wise_x = bx$~\times~0.262/2.75 - 0.94327$

• wise_y = by$~\times~0.262/2.75 - 0.94327$

• The transformations between, e.g., wise_x and bx arise from the fact that the WCS headers of the optical and WISE stacks share the same (brick) center, which is pinned to the center of each image. The WISE image size is chosen by rounding down from 3600 pixels scaled by the relative central pixel scales of the tangential projections (which are 0.262 and 2.75 arcsec/pix for the optical and WISE images respectively).

• Two versions of the noise equivalent area are now supplied in the Tractor catalogs for DR9:

• nea_g, nea_r and nea_z

• blob_nea_g, blob_nea_r and blob_nea_z

• A number of additional columns have been added to the Tractor catalogs for DR9:

• lc_epoch_index_w1 and lc_epoch_index_w2

• apflux_blobresid_g, apflux_blobresid_r and apflux_blobresid_z

• apflux_w1, apflux_w1, apflux_w3 and apflux_w4

• apflux_resid_w1, apflux_resid_w2, apflux_resid_w3 and apflux_resid_w4

• apflux_ivar_w1, apflux_ivar_w2, apflux_ivar_w3 and apflux_ivar_w4

• A new bitwise mask fitbits has been added to the sweep files and Tractor catalogs. This new bit-mask details properties of how each source was fit and is documented on the DR9 bitmasks page. This includes a bit for iterative detection and fitting.

• The gaia_pointsource column has been removed from the Tractor catalogs for DR9 (as compared to DR8). This column has been absorbed in the new fitbits bit-mask.

• Three additional columns exist in the region-specific survey bricks files.

• cosky_g, cosky_r and cosky_z

• The large galaxy (Siena Galaxy Atlas; SGA) catalog has been updated extensively and is "pre-burned." This produces many additional files of data from fitting in SGA regions.

• DR9 uses version 3 of the SGA, so the ref_cat (or REF_CAT) column in the Tractor catalogs and sweep files is populated with L3 for SGA sources in DR9 (this column was populated with L2 for SGA sources in DR8).

• The DR9 coadds include a new set of files legacysurvey-<brick>-blobmodel-<filter>.fits.fz which record a new version of the Tractor's model prediction of the images. The blobmodel information is similar to the information in the legacysurvey-<brick>-model-<filter>.fits.fz files (which have already been made available in prior Data Releaseses). In the blobmodel files, though, the models are clipped to the blobs within which they are measured. In other words, the models used to derive the maps in the blobmodel files are not extrapolated beyond the pixels in the blobs that are used to fit the models.

Footnotes