Siena Galaxy Atlas 2020

• SGA-2020 is a comprehensive, multi-wavelength imaging atlas of nearby galaxies covering 19,721 deg² and cataloging 383,620 sources for detailed spatial studies.
• It integrates high-quality optical (DESI Legacy Surveys DR9) and infrared (unWISE) data to provide robust photometry, geometry, and mosaics for galaxy evolution and DESI targeting.
• The atlas employs advanced processing with tools like The Tractor to deliver precise measurements of galaxy sizes, profiles, and orientations, supporting studies from peculiar velocities to multi-messenger astronomy.

The Siena Galaxy Atlas 2020 (SGA-2020) is a uniform, multi-wavelength optical and infrared imaging atlas of nearby, large angular-diameter galaxies, built from DESI Legacy Imaging Surveys Data Release 9 $grz$ imaging and six-year unWISE coadds in four infrared bands. It contains $N = 383{,}620$ galaxies over $19{,}721\ \mathrm{deg}^2$ of the extragalactic sky, and was designed as a consistently processed reference data set for spatially resolved studies of the local galaxy population, for high-fidelity targeting and photometry in the DESI Bright Galaxy Survey, and for downstream applications including peculiar-velocity work, transient host-galaxy identification, and multi-messenger counterpart searches (Moustakas et al., 2023).

1. Scope, design goals, and sample construction

SGA-2020 was designed for precision studies of nearby galaxies in the regime where resolved structure is observationally accessible. Its stated purposes are to enable detailed studies of star formation and mass assembly in the local universe; provide high-quality photometry, geometry, and mosaics for large galaxies so as to mitigate shredding and bright-end incompleteness in DESI’s Bright Galaxy Survey; deliver legacy-value reference images and metadata for time-domain and multi-messenger astrophysics; and support local-velocity-field measurements via the Tully–Fisher and Fundamental Plane relations (Moustakas et al., 2023).

The atlas covers approximately $20{,}000\ \mathrm{deg}^2$ in the Legacy Surveys DR9 footprint, with final angular coverage of the cataloged sample equal to $19{,}721\ \mathrm{deg}^2$, corresponding to about $50\%$ of the sky and $74\%$ of the available $|b|>20^\circ$ extragalactic sky. The selection targets nearby, large angular-diameter systems and imposes no strict redshift cut, although comparisons to HECATE $(z<0.047)$ and the DESI Peculiar Velocity program $(z<0.15)$ make the low-redshift focus explicit. Within the DESI footprint of about $N = 383{,}620$0, DESI is expected to obtain spectra for more than $N = 383{,}620$1 SGA galaxies by survey completion; $N = 383{,}620$2 spectra were already available in the DESI EDR (Moustakas et al., 2023).

The parent sample was built from HyperLeda and supplemented by RC3/OpenNGC, a subset of Local Group dwarfs, and Legacy Surveys DR8 large sources. An initial HyperLeda query returned $N = 383{,}620$3 sources, which were curated to $N = 383{,}620$4 parents. The sample is largely limited to $N = 383{,}620$5, where $N = 383{,}620$6 is the major-axis diameter at the $N = 383{,}620$7 optical isophote, with a tail extending down to approximately $N = 383{,}620$8–$N = 383{,}620$9; an upper limit $19{,}721\ \mathrm{deg}^2$0 excludes the LMC, SMC, and Sgr dSph. This construction emphasizes large, nearby systems while retaining enough breadth to support wide-area statistical work (Moustakas et al., 2023).

2. Imaging basis and processing pipeline

SGA-2020 combines optical imaging from the DESI Legacy Imaging Surveys DR9 with infrared imaging from unWISE six-year coadds. The optical component uses $19{,}721\ \mathrm{deg}^2$1, $19{,}721\ \mathrm{deg}^2$2, and $19{,}721\ \mathrm{deg}^2$3 imaging from DECaLS, BASS, and MzLS; the infrared component uses W1, W2, W3, and W4, spanning $19{,}721\ \mathrm{deg}^2$4–$19{,}721\ \mathrm{deg}^2$5. The optical data typically have PSF FWHM of about $19{,}721\ \mathrm{deg}^2$6–$19{,}721\ \mathrm{deg}^2$7, photometric calibration from Pan-STARRS1 PSF photometry with precision better than $19{,}721\ \mathrm{deg}^2$8 in $19{,}721\ \mathrm{deg}^2$9 for bright stars, and astrometry anchored to Gaia DR2 at approximately $20{,}000\ \mathrm{deg}^2$0 for DECam/Mosaic-3 and $20{,}000\ \mathrm{deg}^2$1 for 90Prime (Moustakas et al., 2023).

Component	Source	Core characteristics
Optical	Legacy Surveys DR9	$20{,}000\ \mathrm{deg}^2$2, PSF FWHM $20{,}000\ \mathrm{deg}^2$3–$20{,}000\ \mathrm{deg}^2$4, $20{,}000\ \mathrm{deg}^2$5/pixel
Infrared	unWISE six-year coadds	W1–W4, $20{,}000\ \mathrm{deg}^2$6/pixel
Catalog products	SGA-2020.fits	ELLIPSE and TRACTOR HDUs

Image reduction used the NOIRLab Community Pipeline with instrument-specific processing for DECam, Mosaic-3, and 90Prime. Large-scale sky was modeled per CCD via spline fits after masking sources, with removal of reflection or pupil-ghost patterns for DECam and Mosaic-3 and fringe corrections for MzLS $20{,}000\ \mathrm{deg}^2$7-band and 90Prime $20{,}000\ \mathrm{deg}^2$8. Residual median-scaled sky patterns were also removed in DECam $20{,}000\ \mathrm{deg}^2$9, $19{,}721\ \mathrm{deg}^2$0, and $19{,}721\ \mathrm{deg}^2$1. Small-scale sky estimation used $19{,}721\ \mathrm{deg}^2$2-pixel boxes and spline interpolation, with galaxies and Gaia stars masked during background estimation (Moustakas et al., 2023).

The mosaics were built as inverse-variance weighted sums, with optical coadds in tangent-plane projections using Lanczos-3 resampling. Per-group tiling scaled the mosaic diameter to group size, typically $19{,}721\ \mathrm{deg}^2$3–$19{,}721\ \mathrm{deg}^2$4 the group diameter. In the north $19{,}721\ \mathrm{deg}^2$5 the atlas uses BASS+MzLS, while elsewhere it uses DECam imaging from DECaLS+DES. Source modeling employed The Tractor, fitting $19{,}721\ \mathrm{deg}^2$6 jointly with PSF, REX, EXP, DEV, or Sérsic models. Deblending was tuned to reduce shredding in resolved galactic structure by adopting elevated saddle-fraction and saddle-min settings, and the coadd PSF was built from the inverse-variance weighted average of the contributing pixelized PSFs (Moustakas et al., 2023).

3. Measurement definitions and released data products

The principal release file is SGA-2020.fits, organized into row-matched ELLIPSE and TRACTOR HDUs. The ELLIPSE HDU contains per-galaxy geometry, photometry, and surface-brightness-profile results, including RA and DEC, PGC, $19{,}721\ \mathrm{deg}^2$7, PA, BA, $19{,}721\ \mathrm{deg}^2$8, $19{,}721\ \mathrm{deg}^2$9, average $50\%$0 within $50\%$1, moment-based positions $50\%$2, semi-major axes at specified isophotes from $50\%$3 to $50\%$4, curve-of-growth parameters per band $50\%$5, and quality or failure flags in ELLIPSEBIT. The photometric system is AB, and fluxes are reported in nanomaggies, where $50\%$6 nanomaggie is the flux density of AB $50\%$7 mag. The TRACTOR HDU contains model-fitting outputs, including model family, shapes, fluxes, and depths (Moustakas et al., 2023).

SGA-2020 defines $50\%$8 as the semi-major axis length at the $50\%$9-band surface-brightness isophote $74\%$0, and $74\%$1 as the corresponding major-axis diameter. Azimuthally averaged optical surface-brightness profiles were measured with photutils on elliptical isophotes using fixed geometry from ellipse moments, sigma-clipping with nclip=2 and sclip=3, and median-area integration. The quantity $74\%$2 is reported in $74\%$3 along semi-major axes sampled at $74\%$4 pixel $74\%$5 intervals out to about $74\%$6 the estimated semi-major axis. Radii and integrated magnitudes at $74\%$7 are recorded, and the half-light semi-major axis $74\%$8 is derived analytically from the best-fit curve-of-growth parameters (Moustakas et al., 2023).

Beyond the tabular catalog, the release includes multi-wavelength mosaics, Tractor model stacks and residual images, PSF stamps, $74\%$9 depth maps, maskbit images, and a group catalog based on friends-of-friends linking with a $|b|>20^\circ$0 linking length. Ancillary metadata include imaging depth, PSF FWHM per mosaic, and Galactic transmission per band. Public access is provided through the SGA web portal, the NOIRLab Data Lab, and an interactive viewer; file organization is per group under RASLICE/GROUP_NAME, and example usage includes validating masking through model and residual images, extracting $|b|>20^\circ$1 profiles from the ELLIPSE HDU, deriving $|b|>20^\circ$2 and $|b|>20^\circ$3, and constructing Tully–Fisher samples using SGA geometric parameters and DESI rotation measures (Moustakas et al., 2023).

4. Completeness, catalog fidelity, and known limitations

The atlas is reported to be at least $|b|>20^\circ$4 complete for galaxies with $|b|>20^\circ$5 and $|b|>20^\circ$6, measured at $|b|>20^\circ$7, and more than $|b|>20^\circ$8 complete for galaxies larger than $|b|>20^\circ$9 and brighter than $(z<0.047)$0 at the same isophotal limit. After careful vetting against HECATE and correction of catalog problems including spurious sources, shreds, overestimated diameters, and coverage gaps, only $(z<0.047)$1 of HECATE galaxies with $(z<0.047)$2 are reported as genuinely missing. At the same time, the release notes emphasize that surface-brightness completeness varies with the heterogeneity of the input catalogs, and some sky areas without uniform three-band coverage are excluded (Moustakas et al., 2023).

Several systematic effects are explicitly documented. Sky-subtraction systematics can bias low-surface-brightness structure, especially for early-type outer envelopes. In BASS+MzLS regions north of $(z<0.047)$3, Mosaic-3 pattern-noise subtraction distorts $(z<0.047)$4-band profiles and colors, making galaxies appear too green in $(z<0.047)$5 composites. In the infrared, W1/W2 backgrounds are sometimes over-subtracted because unWISE median background modeling uses $(z<0.047)$6 grids. Crowded fields and bright stars can also degrade masking, centroiding, and profile recovery, with the Coma cluster cited as an example regime where such failures occur (Moustakas et al., 2023).

The release further enumerates pipeline-level failure modes. Ellipse fits were skipped for $(z<0.047)$7 galaxies with small Tractor sizes; $(z<0.047)$8 galaxies in very large groups did not complete processing; $(z<0.047)$9 fits were rejected after visual inspection; and $(z<0.15)$0 galaxies lack profiles despite ELLIPSEBIT=0. A catastrophic bug affected aperture photometry in the per-galaxy ellipse files, although the curves of growth were recovered from the profiles and reported in the merged catalog, so the recommended source for profiles and curve-of-growth information is the ELLIPSE HDU of SGA-2020.fits. A few of the largest galaxies, larger than $(z<0.15)$1, also exhibit centroid offsets of a few arcseconds relative to independent infrared catalogs such as WXSC-100. Practical use therefore requires checking ELLIPSEBIT, maskbit images, and, in difficult fields, direct validation of centroiding and masking (Moustakas et al., 2023).

5. DESI coupling and principal scientific use cases

A central motivation for SGA-2020 is its integration with DESI spectroscopy. The atlas was built to provide robust sizes, geometries, photometry, and mosaics for large galaxies in order to improve the completeness and fidelity of the DESI Bright Galaxy Survey and to serve as an imaging framework for follow-up analyses. DESI spectra covering $(z<0.15)$2–$(z<0.15)$3 at $(z<0.15)$4–$(z<0.15)$5, with spectrophotometric precision of about $(z<0.15)$6, are expected for more than $(z<0.15)$7 SGA galaxies. These data support stellar-population and ISM diagnostics in the central regions of nearby galaxies when combined with the atlas’s resolved photometric information (Moustakas et al., 2023).

The atlas was also designed for peculiar-velocity and local-flow studies. For spirals, the intended scaling relation is the Tully–Fisher relation,

$(z<0.15)$8

with the equivalent luminosity form $(z<0.15)$9. For early-type galaxies, the relevant relation is the Fundamental Plane,

$N = 383{,}620$00

The DESI Peculiar Velocity program targets SGA galaxies, including off-nuclear positions along major axes, to constrain distances and the $N = 383{,}620$01 velocity field (Moustakas et al., 2023).

Beyond DESI’s core cosmology program, SGA-2020 functions as a reference sample for time-domain and multi-messenger astronomy. Its accurate positions, sizes, and multi-band reference images are intended to aid host-galaxy prioritization for transients and gravitational-wave events over large localization regions, including error ellipses exceeding $N = 383{,}620$02. Relative to earlier catalogs such as RC3 and HyperLeda, it updates positions, sizes, ellipticities, and position angles using modern deep imaging over a much larger footprint; positions generally agree within about $N = 383{,}620$03 median scatter, while $N = 383{,}620$04 values are on average about $N = 383{,}620$05 larger than HyperLeda’s $N = 383{,}620$06, reflecting deeper imaging. Relative to NSA/SDSS, the atlas is specifically intended to avoid the bright-end incompleteness and shredding that affect SDSS catalogs at $N = 383{,}620$07–$N = 383{,}620$08 (Moustakas et al., 2023).

6. Later interpretations, derivative analyses, and planned extensions

Subsequent work has used SGA-2020 as a benchmark for galaxy-size methodology. The TNG50-SKIRT Atlas study compared simulated and observed half-light radii using an explicitly matched nonparametric curve-of-growth framework: concentric elliptical isophotes, semi-major-axis half-light radii, and band-by-band comparison in $N = 383{,}620$09, $N = 383{,}620$10, and $N = 383{,}620$11. In all $N = 383{,}620$12 matched bins in $N = 383{,}620$13-band absolute magnitude and $N = 383{,}620$14 color, median effective radii decrease from $N = 383{,}620$15 to $N = 383{,}620$16 to $N = 383{,}620$17. For the reddest galaxies $N = 383{,}620$18, TNG50-SKIRT and SGA medians agree closely across luminosity, whereas for bluer galaxies $N = 383{,}620$19 systematic offsets appear, reaching an approximately $N = 383{,}620$20 excess in the simulated median $N = 383{,}620$21-band $N = 383{,}620$22 relative to SGA near $N = 383{,}620$23. The same study found median ratios $N = 383{,}620$24, $N = 383{,}620$25, and $N = 383{,}620$26, and decomposed the wavelength dependence of size into an approximately $N = 383{,}620$27 contribution from stellar population gradients and an approximately $N = 383{,}620$28 contribution from dust attenuation. This suggests a quantitative framework for translating SGA light-weighted sizes into stellar-mass size proxies, especially in $N = 383{,}620$29 band, while also indicating that blue, luminous systems require greater caution (Baes et al., 2024).

SGA-2020 has also been used in large-scale-structure alignment studies. One analysis combined SGA morphologies, axis ratios, and position angles with filament spines derived from the Bisous process on SDSS DR12 galaxies at $N = 383{,}620$30. The final sample comprised $N = 383{,}620$31 spirals and $N = 383{,}620$32 ellipticals within $N = 383{,}620$33 of filament spines. Elliptical galaxies showed a strong perpendicular alignment of spin normals relative to filament orientation, inconsistent with random by up to approximately $N = 383{,}620$34, while spiral galaxies showed a weaker but nonzero alignment signal of approximately $N = 383{,}620$35 in the full within-$N = 383{,}620$36 sample. The maximum significance occurs at $N = 383{,}620$37–$N = 383{,}620$38 for spirals and $N = 383{,}620$39–$N = 383{,}620$40 for ellipticals. A plausible implication is that SGA’s uniform shape and morphology measurements are sufficiently stable for environment-dependent orientation statistics on samples exceeding $N = 383{,}620$41 galaxies (Muralichandran et al., 2 Sep 2025).

The primary release also defines a clear extension path. Planned directions include redefining the parent sample by direct detection of large galaxies in Legacy Surveys imaging, including deep-learning approaches; quantifying completeness through injection tests across size, flux, and surface-brightness limits; improving centroiding and masking in mergers, around bright stars, and in crowded cluster cores; and developing sky-subtraction methods that better preserve low-surface-brightness envelopes. Proposed footprint and wavelength extensions include additional DECam imaging from DR10 and beyond, inclusion of DECam $N = 383{,}620$42-band, GALEX FUV $N = 383{,}620$43 and NUV $N = 383{,}620$44 coadds where available, and full UV–IR surface-brightness profiles and integrated photometry over $N = 383{,}620$45–$N = 383{,}620$46. Future releases are also expected to incorporate DESI spectroscopic redshifts, spectrophotometry, and derived SED-based quantities such as stellar mass and star-formation rate (Moustakas et al., 2023).