UNIONS: Northern Optical Legacy Survey

• UNIONS is a wide-field, deep, multi-band imaging survey that combines data from CFHT, Pan-STARRS, and Subaru to deliver high-quality ugriz imaging over approximately 6250 deg².
• It employs specialized pipelines and precise calibrations, using Gaia and Pan-STARRS references to optimize photometric redshifts and weak-lensing measurements.
• The survey underpins diverse astrophysical research—from dark matter and galaxy evolution to low-surface-brightness studies—establishing a robust northern legacy complementing LSST and Euclid.

The Ultraviolet Near-Infrared Optical Northern Survey (UNIONS) is a wide-field, deep, multi-band imaging program of the northern extragalactic sky, explicitly described as a “collaboration of collaborations”. It combines observations from the Canada-France-Hawai'i Telescope, the Pan-STARRS telescopes, and the Subaru Observatory to obtain $ugriz$ imaging over a core survey region of $6250\,\mathrm{deg}^2$, with $10\sigma$ point-source depths in a $2^{\prime\prime}$ aperture of $[u,g,r,i,z]=[23.7,24.5,24.2,23.8,23.3]$ AB. UNIONS was designed both as a stand-alone northern legacy survey and as northern optical support for Euclid, particularly for calibration of the wavelength dependence of the Euclid VIS PSF and for photometric-redshift estimation in the North Galactic Cap (Gwyn et al., 18 Mar 2025).

1. Survey conception, footprint, and strategic role

UNIONS was assembled from previously distinct survey efforts with overlapping scientific and strategic goals. Its four principal components are CFIS on CFHT/MegaCam, Pan-STARRS, WISHES on Subaru/HSC, and WHIGS on Subaru/HSC. The survey paper emphasizes two drivers. The first is scientific breadth: dark matter, weak lensing, structure formation, Milky Way assembly, dwarf galaxies, stellar streams, low-surface-brightness structure, galaxy evolution, and strong lensing. The second is hemispheric strategy: the south will be exceptionally well served by Rubin/LSST, whereas the north would otherwise lack an equivalent deep optical legacy survey. In that sense, UNIONS is intended to become the major ground-based legacy survey for the northern hemisphere for the next decade and an essential northern complement to the static-sky science of LSST (Gwyn et al., 18 Mar 2025).

The core UNIONS footprint is primarily defined at $|b|\ge 25^\circ$ and is subdivided into three regions: $4606\,\mathrm{deg}^2$ in the North Galactic Cap with $\delta\ge 30^\circ$, $1381\,\mathrm{deg}^2$ in the North Galactic Cap with $15\le\delta\le30^\circ$, and $6250\,\mathrm{deg}^2$0 in the South Galactic Cap. The North Galactic Cap is singled out as the highest-quality Euclid sky in terms of low backgrounds from zodiacal light, stellar density, extinction, and Galactic cirrus. In the Euclid support model described in the survey paper, KiDS and DES provide southern optical support, whereas north of Declination $6250\,\mathrm{deg}^2$1 the required $6250\,\mathrm{deg}^2$2 data are provided by UNIONS (Gwyn et al., 18 Mar 2025).

A frequent source of confusion is that UNIONS area can be quoted in slightly different ways depending on the product definition. The survey overview defines the core survey region as $6250\,\mathrm{deg}^2$3, while a later MOC-based implementation of the released $6250\,\mathrm{deg}^2$4 footprint reports $6250\,\mathrm{deg}^2$5; the latter is explicitly noted to be MOC-definition dependent. This distinction reflects product realization rather than a scientific redefinition of the survey (Ahad et al., 11 May 2026).

2. Facilities, passbands, and observing system

UNIONS uses CFHT/MegaCam for deep $6250\,\mathrm{deg}^2$6 and $6250\,\mathrm{deg}^2$7, Pan-STARRS for deep $6250\,\mathrm{deg}^2$8 and part of the $6250\,\mathrm{deg}^2$9 coverage, Subaru/HSC WISHES for deep $10\sigma$0, and Subaru/HSC WHIGS for deep $10\sigma$1. Because $10\sigma$2 is supplied by two distinct systems, the survey effectively works with six passbands: CFIS $10\sigma$3, WHIGS $10\sigma$4, CFIS $10\sigma$5, Pan-STARRS $10\sigma$6, Pan-STARRS $10\sigma$7, and WISHES $10\sigma$8. All magnitudes are on the AB system (Gwyn et al., 18 Mar 2025).

Band	Median depth ($10\sigma$9, $2^{\prime\prime}$0)	Median image quality
$2^{\prime\prime}$1	23.69	$2^{\prime\prime}$2
$2^{\prime\prime}$3	24.54	$2^{\prime\prime}$4
$2^{\prime\prime}$5	24.21	$2^{\prime\prime}$6
$2^{\prime\prime}$7	23.79	$2^{\prime\prime}$8
$2^{\prime\prime}$9	23.31	$[u,g,r,i,z]=[23.7,24.5,24.2,23.8,23.3]$0
$[u,g,r,i,z]=[23.7,24.5,24.2,23.8,23.3]$1	23.41	$[u,g,r,i,z]=[23.7,24.5,24.2,23.8,23.3]$2

The $[u,g,r,i,z]=[23.7,24.5,24.2,23.8,23.3]$3-band is operationally central. CFHT queue scheduling observes in $[u,g,r,i,z]=[23.7,24.5,24.2,23.8,23.3]$4 only when seeing is below $[u,g,r,i,z]=[23.7,24.5,24.2,23.8,23.3]$5; the average $[u,g,r,i,z]=[23.7,24.5,24.2,23.8,23.3]$6-band exposure time is $[u,g,r,i,z]=[23.7,24.5,24.2,23.8,23.3]$7, with an actual range of $[u,g,r,i,z]=[23.7,24.5,24.2,23.8,23.3]$8 to $[u,g,r,i,z]=[23.7,24.5,24.2,23.8,23.3]$9. CFIS uses three images per pointing per band with large dithers. The $|b|\ge 25^\circ$0-band exposure time is fixed at $|b|\ge 25^\circ$1. Pan-STARRS contributes approximately $|b|\ge 25^\circ$2 exposures and a required total exposure time of $|b|\ge 25^\circ$3 per pixel, equivalent to about 450 overlapping raw images. WISHES uses $|b|\ge 25^\circ$4 per HSC pointing split into three dithered sub-exposures, and WHIGS uses a nominal $|b|\ge 25^\circ$5 split into three $|b|\ge 25^\circ$6 sub-exposures (Gwyn et al., 18 Mar 2025).

The technical asymmetry across bands matters scientifically. CFIS $|b|\ge 25^\circ$7 provides the best image quality in the survey and is the natural lensing band. The $|b|\ge 25^\circ$8 band is unusually deep for a survey of this scale and is important for Galactic archaeology and high-redshift color selection. Subaru/HSC fills the $|b|\ge 25^\circ$9 and deep $4606\,\mathrm{deg}^2$0 gaps needed for full $4606\,\mathrm{deg}^2$1, while Pan-STARRS provides wide, uniform $4606\,\mathrm{deg}^2$2-band coverage and part of the $4606\,\mathrm{deg}^2$3 baseline (Gwyn et al., 18 Mar 2025).

3. Calibration, coadds, and catalog products

UNIONS is not a single pipeline but a federation of instrument-specific processing streams followed by homogenized products. CFIS data are detrended with Pitcairn and processed with MegaPipe. Astrometry is tied to Gaia DR3, with a typical quoted astrometric uncertainty of $4606\,\mathrm{deg}^2$4. CFIS $4606\,\mathrm{deg}^2$5-band photometric calibration is transferred from Pan-STARRS with typical residuals of $4606\,\mathrm{deg}^2$6, while $4606\,\mathrm{deg}^2$7-band calibration is derived from Gaia DR3 BP/RP spectra with a typical precision of $4606\,\mathrm{deg}^2$8. Run-level photometric “super-flats” are usually accurate to $4606\,\mathrm{deg}^2$9. CFIS coadds are generated with SWarp on $\delta\ge 30^\circ$0 pixel tiles at $\delta\ge 30^\circ$1, with about $\delta\ge 30^\circ$2 overlap between adjacent tiles (Gwyn et al., 18 Mar 2025).

The non-CFIS components preserve their own processing identities. Pan-STARRS products come through the IPP and are noted to have highly textured PSFs in the stacks because of the large number of contributing images. WISHES uses hscPipe with joint calibration and typically achieves $\delta\ge 30^\circ$3 zeropoint uncertainty. WHIGS uses the LSST Science Pipeline, calibrated against Gaia DR2 and Pan-STARRS DR1, with a typical zeropoint uncertainty of $\delta\ge 30^\circ$4 (Gwyn et al., 18 Mar 2025).

The first homogeneous five-band catalog is based on GAaP photometry. In that framework, PSFs are modeled with shapelets, each image is convolved to a common spatially constant isotropic Gaussian PSF, and forced photometry is performed in Gaussian-weighted elliptical apertures defined from CFIS $\delta\ge 30^\circ$5-band detections. The product is optimized for accurate colors rather than total fluxes, and the survey paper is explicit that GAaP can bias colors toward central regions in extended galaxies with color gradients. Photometric redshifts are estimated with BPZ, using the standard normalized residual $\delta\ge 30^\circ$6, and the survey states that $\delta\ge 30^\circ$7-only performance already approaches the Euclid requirement $\delta\ge 30^\circ$8 even before adding Euclid $\delta\ge 30^\circ$9 data (Gwyn et al., 18 Mar 2025).

UNIONS also produces specialized low-surface-brightness products. A separate CFIS $1381\,\mathrm{deg}^2$0-band stack set uses Elixir-LSB methods to preserve faint extended emission. This product class is scientifically consequential: the standard non-LSB CFIS stacks remove the large-scale optical fluctuations needed for background-fluctuation work, whereas the LSB products enabled a Herschel–UNIONS cross-correlation detected at $1381\,\mathrm{deg}^2$1, or $1381\,\mathrm{deg}^2$2 after masking individually detected galaxies (Lim et al., 2022).

4. Weak-lensing infrastructure and cosmological use

Weak lensing is one of the survey’s defining technical pillars. The UNIONS overview describes a dedicated lensing pipeline, ShapePipe, based on ngmix and Metacalibration, and notes an internal lensing catalog spanning $1381\,\mathrm{deg}^2$3 with an effective background source density of $1381\,\mathrm{deg}^2$4 to $1381\,\mathrm{deg}^2$5 galaxies per arcmin$1381\,\mathrm{deg}^2$6 (Gwyn et al., 18 Mar 2025). The first cosmology-grade release, UNIONS-3500 Weak Lensing: I, makes the pipeline more concrete: it analyzes approximately $1381\,\mathrm{deg}^2$7 of $1381\,\mathrm{deg}^2$8-band MegaCam imaging, more precisely $1381\,\mathrm{deg}^2$9 of imaging with an effective footprint of $15\le\delta\le30^\circ$0 after spatial masking, and produces a final catalog of 62 million galaxies with $15\le\delta\le30^\circ$1 and $15\le\delta\le30^\circ$2 (Hervas-Peters et al., 13 May 2026).

That release is notable for its explicit treatment of systematic error. It documents end-to-end masking, conservative source selection, Metacalibration response estimation, additive-bias correction, and an externally calibrated multiplicative-bias prior $15\le\delta\le30^\circ$3. PSF leakage is modeled both at the object level and through two-point diagnostics using $15\le\delta\le30^\circ$4- and $15\le\delta\le30^\circ$5-statistics; after correction, the induced contamination to $15\le\delta\le30^\circ$6 is brought below a $15\le\delta\le30^\circ$7 threshold on the scales retained for cosmology, while larger angular scales are removed or modeled (Hervas-Peters et al., 13 May 2026).

The first configuration-space cosmic-shear analysis from this catalog is intentionally conservative: a single-bin, 2D analysis in flat $15\le\delta\le30^\circ$8CDM using $15\le\delta\le30^\circ$9. It yields

$6250\,\mathrm{deg}^2$00

consistent within $6250\,\mathrm{deg}^2$01 with Planck 2018 and with precedent cosmic-shear results. The same paper emphasizes that the dominant present limitations are the non-tomographic nature of the source sample, residual shear-calibration uncertainty, and nuisance sensitivity to intrinsic alignment and redshift calibration rather than raw shape noise (Goh et al., 13 May 2026).

UNIONS lensing is also being used beyond cosmic shear. A direct intrinsic-alignment analysis in overlap with BOSS/eBOSS over $6250\,\mathrm{deg}^2$02 reports a $6250\,\mathrm{deg}^2$03 detection with CMASS galaxies, a $6250\,\mathrm{deg}^2$04 detection with LRGs, and a result compatible with the null hypothesis for ELGs; it further finds PSF contamination to $6250\,\mathrm{deg}^2$05 below $6250\,\mathrm{deg}^2$06 on the relevant scales (Peters et al., 2024). In cluster outskirts, a preliminary UNIONS shape catalog was used to derive excess surface mass density profiles and infer $6250\,\mathrm{deg}^2$07 and $6250\,\mathrm{deg}^2$08 from the infall region of DESI Legacy clusters, while measuring mean splashback radii for three public cluster samples (Mpetha et al., 15 Jan 2025).

5. Scientific reach across astrophysics

UNIONS was designed for broad astrophysical use, and the published record already reflects that breadth. In Galactic archaeology, deep CFIS $6250\,\mathrm{deg}^2$09-band photometry combined with PS1 colors and Gaia astrometry enabled identification of the NGC 5466 stream across $6250\,\mathrm{deg}^2$10 of sky, with a strong heliocentric distance gradient $6250\,\mathrm{deg}^2$11 (Jensen et al., 2021). The same survey infrastructure then supported the discovery of Ursa Major III / UNIONS 1, described as the least luminous known satellite of the Milky Way, with $6250\,\mathrm{deg}^2$12, total stellar mass $6250\,\mathrm{deg}^2$13, and a heliocentric distance of $6250\,\mathrm{deg}^2$14 (Smith et al., 2023).

Low-surface-brightness galaxy science is another domain where survey design is decisive. Using UNIONS $6250\,\mathrm{deg}^2$15 data over $6250\,\mathrm{deg}^2$16, the GOBLIN pipeline produced about $6250\,\mathrm{deg}^2$17 million low-surface-brightness candidates and identified 42,965 dwarf candidates with probability $6250\,\mathrm{deg}^2$18, including 23,072 with probability $6250\,\mathrm{deg}^2$19 (Heesters et al., 23 May 2025). The survey’s LSB-optimized $6250\,\mathrm{deg}^2$20-band products also enabled the cross-correlation of the cosmic far-infrared background with optical-imaging fluctuations, with a detection at $6250\,\mathrm{deg}^2$21 and an inferred correlated optical fluctuation level of about $6250\,\mathrm{deg}^2$22 in $6250\,\mathrm{deg}^2$23 (Lim et al., 2022).

Strong-lensing work illustrates a different use of the same depth–seeing–area combination. An early CFIS-based search over $6250\,\mathrm{deg}^2$24 confirmed five new doubly imaged lensed quasars, with separations from $6250\,\mathrm{deg}^2$25 to $6250\,\mathrm{deg}^2$26, demonstrating the utility of deep $6250\,\mathrm{deg}^2$27 color information and $6250\,\mathrm{deg}^2$28-class $6250\,\mathrm{deg}^2$29-band image quality for small-separation systems (Chan et al., 2021). A later dedicated search for edge-on late-type galaxy lenses over $6250\,\mathrm{deg}^2$30 of CFIS $6250\,\mathrm{deg}^2$31-band imaging found 4 grade A, 20 grade B, and 58 grade C new edge-on lens candidates, effectively doubling the known sample of such systems (Barroso et al., 13 Mar 2025).

UNIONS has also become a survey for obscured populations and radio-selected AGN. In a $6250\,\mathrm{deg}^2$32+UKIDSS+WISE study over $6250\,\mathrm{deg}^2$33, UNIONS imaging enabled the identification of 382 infrared-bright dust-obscured galaxies, including 120 broken-power-law DOGs among 244 power-law DOGs; these were interpreted as more heavily obscured AGN than ordinary power-law DOGs (Yoshida et al., 21 Apr 2025). In the UNVEIL radio-identification program, cross-matching VLASS Epoch 2 to UNIONS over $6250\,\mathrm{deg}^2$34 yielded 146,212 radio galaxies down to $6250\,\mathrm{deg}^2$35, with 63,019 valid photometric redshifts, 8,692 sources at $6250\,\mathrm{deg}^2$36, 1,171 at $6250\,\mathrm{deg}^2$37, and about 49,000 radio-loud AGN with $6250\,\mathrm{deg}^2$38 (Zhong et al., 11 Jul 2025).

Galaxy-evolution studies have exploited the survey in both morphology and lensing modes. A CNN-based post-merger analysis on CFIS $6250\,\mathrm{deg}^2$39-band imaging produced a visually confirmed sample of 699 post-mergers and found that the visually confirmed star-forming subset has a star-formation rate about a factor of two above matched controls (Bickley et al., 2022). A later galaxy–galaxy lensing study then used UNIONS both to define 1,623 post-mergers and to measure their excess surface density profiles against $6250\,\mathrm{deg}^2$40 matched controls, finding no significant difference with current data and placing a $6250\,\mathrm{deg}^2$41 confidence upper limit excluding extreme merger-induced starbursts in which more than $6250\,\mathrm{deg}^2$42 of the final stellar mass formed in the burst (Cheng et al., 1 Feb 2025).

6. Status, caveats, and legacy position

As of January 2025, UNIONS was advanced but still incomplete in several bands and regions. Coverage north of $6250\,\mathrm{deg}^2$43 is substantially more mature than the extension to $6250\,\mathrm{deg}^2$44; for example, the $6250\,\mathrm{deg}^2$45-band extension in that declination range was only $6250\,\mathrm{deg}^2$46 complete, whereas $6250\,\mathrm{deg}^2$47-band coverage north of $6250\,\mathrm{deg}^2$48 was complete in area but not yet in uniform depth. No future $6250\,\mathrm{deg}^2$49 data were expected, while $6250\,\mathrm{deg}^2$50-band depth and uniformity were still improving. Extension programs to $6250\,\mathrm{deg}^2$51 had been approved and were to begin in 2025 (Gwyn et al., 18 Mar 2025).

Several caveats recur across the survey documentation. UNIONS is heterogeneous by construction: different facilities, different PSFs, different stack geometries, and different processing histories. Pan-STARRS stacks have highly textured PSFs. Stack-based GAaP photometry is optimized for colors rather than total fluxes. Weak-lensing cosmology in the first release is deliberately non-tomographic because full color information is not yet homogeneous across the whole source sample. These are not incidental details; they define the present boundary between what is already robustly mature and what remains under active refinement (Gwyn et al., 18 Mar 2025).

At the same time, the survey’s legacy role is already well established. The overview paper states that UNIONS data are only slightly shallower than Year 1 LSST, that the survey will remain unsurpassed from the ground in the northern hemisphere, and that it had already produced more than 30 peer-reviewed publications by early 2025 (Gwyn et al., 18 Mar 2025). In practical terms, UNIONS has become the northern optical substrate for Euclid support, a cosmology-grade $6250\,\mathrm{deg}^2$52-band lensing survey, a resolved-star discovery machine for Milky Way halo structure, a low-surface-brightness survey for dwarfs and diffuse light, and a wide-area discovery survey for rare strong lenses, obscured galaxies, and radio-selected AGN. Its defining characteristic is not any single result, but the fact that one coordinated $6250\,\mathrm{deg}^2$53 survey architecture has proven useful across all of those regimes.