Rubin DP1 Data Preview

• Data Preview 1 (DP1) is a science-ready Rubin Observatory dataset derived from commissioning observations, providing calibrated images, coadds, and catalogs across seven fields in the ugrizy system.
• It comprises 1792 high-quality exposures over 48 nights, covering approximately 15 deg² with median image quality near 1.13 arcseconds and deep 5σ point-source limits.
• DP1 supports diverse studies—from time-domain variability and Galactic structure to low-surface-brightness and multiwavelength analyses—while serving as a robust pipeline testbed.

Rubin Observatory Data Preview 1 (DP1) is the first science-ready Rubin data release derived from real on-sky commissioning observations with LSSTComCam on the Simonyi Survey Telescope in late 2024. It comprises calibrated single-epoch images, coadds, template coadds, difference images, detection catalogs, and ancillary products across seven non-contiguous fields in the $ugrizy$ system. DP1 is therefore both an early-science dataset and a practical preview of Rubin data formats, calibration products, catalogs, images, and access services ahead of full Legacy Survey of Space and Time operations (Team et al., 24 Mar 2026).

1. Release architecture and survey footprint

DP1 was assembled from 1792 science-grade exposures acquired over 48 distinct nights between 2024-10-24 and 2024-12-11. The release covers approximately 15 deg$^2$ across seven deliberately chosen fields: 47 Tucanae Globular Cluster, Extended Chandra Deep Field South (ECDFS), Rubin SV Euclid Deep Field South (EDFS_comcam), Fornax Dwarf Spheroidal Galaxy, Rubin SV Low Galactic Latitude Field, Rubin SV Low Ecliptic Latitude Field, and Seagull Nebula. The field sampling is intentionally uneven: ECDFS is the deepest and most heavily observed field with 855 exposures, while the other fields probe crowded stellar systems, Galactic structure, and low-latitude environments (Team et al., 24 Mar 2026).

Across the release, the median delivered image quality is 1.13 arcseconds FWHM for the PSF, and the best images reach about 0.82 arcseconds FWHM. The deepest coadds are in ECDFS, where the reported 5$\sigma$ point-source limits are $u=24.55$, $g=26.18$, $r=25.96$, $i=25.71$, $z=25.07$, and $y=23.10$. The other fields are generally within about 2.2 mag shallower than ECDFS in bands with nonzero coverage (Team et al., 24 Mar 2026).

Quantity	Value	Source
Science-grade exposures	1792	(Team et al., 24 Mar 2026)
Distinct nights	48	(Team et al., 24 Mar 2026)
Total area	$\sim 15$ deg$^2$0	(Team et al., 24 Mar 2026)
Fields	7	(Team et al., 24 Mar 2026)
Bands	$^2$1	(Team et al., 24 Mar 2026)
Distinct astrophysical objects	$^2$2 million	(Team et al., 24 Mar 2026)
Solar system objects	431	(Team et al., 24 Mar 2026)
New solar system discoveries	93	(Team et al., 24 Mar 2026)

The release is explicitly smaller than future LSST data products, but it already spans a wide range of astrophysical environments and observing conditions. This makes DP1 scientifically useful in its own right while also serving as a shared-risk preview of Rubin data management, calibration, and analysis workflows (Team et al., 24 Mar 2026).

2. Products, catalogs, and access patterns

DP1 includes several major product classes. The image products comprise raw images, visit_image calibrated single-epoch images, deep_coadd images, template_coadd images, difference_image products, and associated background images. The catalog layer includes the Source, Object, ForcedSource, DiaSource, DiaObject, ForcedSourceOnDiaObject, SSObject, SSSource, CcdVisit, and Calibration tables. The release also provides 14 survey property maps, HiPS visualizations, task configuration files, processing logs, calibration products, and bandpass transmission curves (Team et al., 24 Mar 2026).

Access is provided through the Rubin Science Platform (RSP), organized around the Portal, Notebook, and API interfaces. Catalog access is exposed through IVOA services, including TAP/ADQL backed by Qserv, while the Rubin Data Butler provides direct access to images, catalogs, maps, and ancillary products. The DP1 repository is hosted in Google Cloud, and the release is available to Rubin data-rights holders during the proprietary period (Team et al., 24 Mar 2026).

A parallel computational theme in DP1 work has been the repackaging of Rubin catalogs into scalable analysis formats. An LSDB/HATS “afterburner” pipeline converted the Rubin-provided catalogs into tiled parquet catalogs with nested time-series columns. In that representation, the object table contains 2,299,726 rows and 268,796,943 forced sources, while the diaObject table contains 1,089,818 rows and 196,911,566 forced sources. This work treated DP1 as a realistic testbed for Rubin-scale catalog analysis rather than merely a small static data release (Malanchev et al., 30 Jun 2025).

3. Time-domain and variability science

DP1 has a commissioning cadence that differs from the planned LSST main survey cadence. Fields were observed multiple times per night, often in up to three filters, rather than every few days. This denser short-baseline sampling is particularly effective for variability with periods shorter than a day and, in practice, even for periods of order an hour (Malanchev et al., 30 Jun 2025).

One early demonstration is the discovery of the SX Phoenicis star LSST-DP1-O-614435753623041404, abbreviated LSST-C25. The object has 217 flux measurements distributed across $^2$3, a best-fit period of $^2$4 days, mean magnitudes $^2$5 and $^2$6, and pulsation amplitudes $^2$7 mag and $^2$8 mag. Using an SX Phe period–luminosity relation, the authors estimate a distance of 16.6 kpc, placing the star in the outer Galactic disk at a Galactocentric distance of roughly 22 kpc; its position and proper motion are interpreted as consistent with the Monoceros Ring (Carlin et al., 30 Jun 2025).

Extragalactic transient work likewise showed that DP1 difference-imaging products are already scientifically usable. A search over 369,644 DIA objects in three extragalactic fields was reduced through automated cuts and visual vetting to 11 final extragalactic transients. Photometric typing with Superphot+ yielded a subtype mix of 6 SN Ia, 2 SN II, 2 SN Ibc, and 1 SN IIn. The observed counts are in slight $^2$9 tension with the literature-based expectations of $\sigma$0 SNe Ia and $\sigma$1 core-collapse SNe, a discrepancy attributed to commissioning-era template construction in which transients can be partially self-subtracted because they are present in both search and template images (Freeburn et al., 30 Jul 2025).

RR Lyrae analysis provides a complementary view of DP1’s current time-domain limits. A cross-match against public RR Lyrae catalogs produced approximately 600 stars with adequate light-curve sampling across five DP1 fields. Metallicities inferred from the Rubin $\sigma$2 color relation are reasonable for well-sampled stars but unstable for sparse light curves. Distances derived from the $\sigma$3 period–Wesenheit–metallicity relation agree well with literature values, with a mean offset of $\sigma$4 mag, whereas multiband PLZ distances are systematically too large. This suggests that DP1 time-series photometry is already useful for RR Lyrae astrophysical inference, but that cadence and phase coverage remain the dominant limitations (Ngeow et al., 1 May 2026).

At the pipeline level, LSDB-based variability searches recovered RR Lyrae, eclipsing binaries, cataclysmic variables, quasars, AT2024ahyy, and a likely new M-dwarf flare candidate. The periodic pipeline targeted periods from 5 minutes to 12 hours, while the transient pipeline used a constant-model versus Bazin-function comparison to flag flare-like light curves (Malanchev et al., 30 Jun 2025).

4. Extragalactic structure and circumgalactic studies

A major early DP1 result is the detection of very low-surface-brightness structure around external galaxies. Using DP1 $\sigma$5 coadds in ECDFS, a long stellar stream was identified around LEDA 751050 at spectroscopic redshift $\sigma$6. The feature extends along the minor axis of the host, has an angular size of about $\sigma$7, corresponding to roughly 130 kpc by 15 kpc, and is described as a “polar” stream and an analog of the M31 Giant Stellar Stream. Its mean surface brightnesses are $\sigma$8, $\sigma$9, and $u=24.55$0 mag arcsec$u=24.55$1, with the faintest parts reaching $u=24.55$2 mag arcsec$u=24.55$3. The inferred stellar mass is $u=24.55$4, compared with $u=24.55$5 for the host, implying a merger mass ratio of approximately 1:250 (Johnson et al., 6 Jan 2026).

DP1 has also been used to measure diffuse circumgalactic dust. Using only 4.6 deg$u=24.55$6 of ComCam imaging, a stacked background-galaxy analysis detected a chromatic reddening signal from projected separations of about 10 kpc to 1 Mpc around foreground galaxies. Interpreting the color excess with a Milky Way extinction curve gives

$u=24.55$7

within 120 kpc. The analysis used 18,573 foreground galaxies and 19,710 background galaxies, and the amplitude and radial dependence agree with earlier SDSS, KiDS, and DES measurements despite the much smaller area (Crenshaw et al., 23 Jun 2026).

Cross-domain association studies are another important DP1 use case. A crossmatch between X-ray catalogs and DP1 identified 2314 optical counterparts to 3830 X-ray sources in the final footprint using the Object catalog alone, corresponding to about 60.4%. Including DiaObjects raises this to 2566 of 3830, roughly 67%. Match quality is strongly field dependent: E-CDF-S provides the best performance, whereas crowded fields such as 47 Tuc and Fornax dSph expose incompleteness in the coadd-based object catalog and the importance of difference-imaging detections in crowded regions (Yuankun et al., 18 Jul 2025).

These studies collectively show that DP1 is already useful for low-surface-brightness photometry, projected galaxy–dust correlations, and multiwavelength source association, even though the data were not yet optimized for all low-surface-brightness or crowded-field use cases.

5. Stellar populations, crowded fields, and Galactic structure

Crowded-field performance is one of the clearest areas where DP1 is simultaneously capable and limited. In the 47 Tucanae field, the coadd-based Object catalog yields a remarkably clean color–magnitude diagram, recovers both the 47 Tuc main sequence and the background Small Magellanic Cloud, and supports the selection of 4506 candidate cluster members. However, the standard coadd pipeline shows a sharp completeness decline inside approximately 28 pc of the cluster center. A forced-photometry approach using diaForcedSource measurements at DIAObject positions increases the working sample to 116,963 objects after quality cuts, recovers 14,744 possible cluster members, and extends the usable radial reach to about 14 pc, albeit with strong contamination and systematics, including a spurious blue plume near $u=24.55$8 (Wainer et al., 4 Jul 2025).

The outer 47 Tuc field has also been used for unresolved-binary work. Using DP1 coadd photometry and an $u=24.55$9 versus $g=26.18$0 CMD, unresolved main-sequence binaries with mass ratio $g=26.18$1 were detected between roughly 18 arcmin and 40 arcmin from the cluster center, outside the half-light radius and nearly to the tidal radius. The measured binary fraction is

$g=26.18$2

derived from 1308 main-sequence stars and 25 binary stars in the selected mass range. This is larger than the inner-region HST estimate quoted in the paper and suggests a radial dependence consistent with stronger dynamical processing in the cluster core (Cordoni et al., 4 Sep 2025).

For Galactic structure, DP1 broadband photometry has already been used for photometric distances and metallicities of blue main-sequence stars. In three southern-latitude fields—ECDFS, EDFS, and Rubin SV 95$g=26.18$325—a photoD-based analysis found a clear deficit of faint blue main-sequence turn-off stars at $g=26.18$4 relative to TRILEGAL simulations. The authors interpret this as evidence that the stellar halo density profile over Galactocentric distances of about 10–50 kpc is steeper than the canonical $g=26.18$5 form assumed in default TRILEGAL models, with fitted slopes roughly in the range $g=26.18$6–$g=26.18$7, while noting the strong degeneracy between halo flattening and slope (Palaversa et al., 30 Dec 2025).

A closely related methodological problem is star–galaxy separation at the faint end. A Random Forest classifier trained on DP1 ECDFS data showed that multi-band Rubin photometry alone substantially outperforms morphology-only classification at faint magnitudes; colors involving the $g=26.18$8 band are especially important, and adding photometric uncertainties yields the best performance. In the best configuration, the independent validation set reached Accuracy = 98.3%, and the study reports that galaxy contamination remains negligible almost the whole magnitude range probed, namely $g=26.18$9 mag (Gatto et al., 26 Mar 2026).

6. Analysis infrastructure, calibration experiments, and limitations

DP1 has also been used to validate Rubin-adjacent analysis frameworks. The RAIL photo-$r=25.96$0 study ran eight algorithms on DP1 photometry, using six-band ECDFS data and four-band Rubin_SV_38_7 data. Across methods, the six-band sample achieved per-galaxy scatter of $r=25.96$1 with outlier fractions around 10%, degrading at faint magnitudes and at $r=25.96$2. Among the tested methods, FlexZBoost performed best on the ECDFS test set, with $r=25.96$3, $r=25.96$4, and $r=25.96$5. The paper concludes that the machine-learning methods satisfy the LSST Year 1 bias requirement on average, while also noting that redshift-dependent bias still motivates external calibration and that Euclid near-infrared photometry improves performance at $r=25.96$6 (Zhang et al., 8 Oct 2025).

Several recurring limitations emerge across DP1 studies. Because LSSTComCam can hold only three filters at once, band coverage is incomplete in some fields; this matters for both stellar-parameter inference and photo-$r=25.96$7 work (Team et al., 24 Mar 2026). Commissioning-era transient templates were built from science images, leading to reduced sensitivity through self-subtraction for long-lived transients (Freeburn et al., 30 Jul 2025). In crowded regions such as 47 Tuc and Fornax dSph, coadd deblending failures limit the completeness of the Object catalog and can bias counterpart reliability estimates or stellar selection (Yuankun et al., 18 Jul 2025). For RR Lyrae and other variables, sparse phase coverage in DP1 can destabilize template fitting and propagate directly into metallicity and distance estimates (Ngeow et al., 1 May 2026).

The general pattern is that DP1 already supports scientifically meaningful measurements in low-surface-brightness astronomy, Galactic structure, crowded-field stellar populations, photo-$r=25.96$8 calibration, time-domain classification, and multiwavelength crossmatching, but often as a proof-of-concept constrained by commissioning-era cadence, footprint, templates, and deblending. That combination of immediate scientific utility and explicit pipeline stress-testing is central to DP1’s role within the Rubin program (Team et al., 24 Mar 2026).