ATLAS Refcat2 Catalog: Stellar Calibration

• ATLAS Refcat2 is a comprehensive stellar reference featuring 991 million stars with sub-1% photometric uniformity and Gaia-based astrometry.
• It integrates data from multiple surveys like Gaia DR2, Pan-STARRS, and SkyMapper to achieve uniform, high-precision calibration across the full sky.
• The catalog supports time-domain surveys by providing robust astrometric matching and photometric zeropoint determination for accurate, wide-field calibrations.

The ATLAS Refcat2 catalog is a comprehensive, all-sky astrometric and photometric stellar reference comprising approximately 991 million stars with $g, r, i, z$ magnitudes to $m \sim 19$. Designed to undergird the Asteroid Terrestrial-impact Last Alert System (ATLAS) and serve as a photometric and astrometric calibration resource for any wide-field optical survey, Refcat2 achieves sub-1% photometric uniformity and leverages Gaia DR2 astrometry for high positional accuracy. Its broad input base and PS1 bandpass homogenization make it pivotal for precise, wide-area photometric work in time-domain and static-sky astronomy (Tonry et al., 2018).

1. Objectives, Scope, and Scientific Motivation

ATLAS Refcat2 was developed in response to the calibration requirements of ATLAS, which necessitates sub-0.03 mag photometry nightly over tens of square degrees in order to detect faint asteroids, produce light curves of variable objects, and perform image differencing at the photon noise limit. No pre-existing catalog offered the requisite combination of full-sky coverage, $\sim$19th mag depth, $\lesssim 0.01$ mag photometric uniformity, and sufficiently dense stellar sampling for instruments with 6″ pixels. Refcat2 thus set the following main design objectives:

• $\geq 99\%$ completeness for $g$, $r$, or $i < 19$ over $-90^\circ <$ Dec $< +90^\circ$
• Inclusion of the brightest stars ($m \sim 19$0), augmenting with Tycho-2 and the Yale Bright Star Catalog
• Photometry on a uniform Pan-STARRS bandpass system ($m \sim 19$1)
• Gaia-quality astrometry (positional uncertainty $m \sim 19$2 mas), including proper motions and parallaxes

These criteria ensured Refcat2 would uniquely support wide-field surveys requiring high-precision, spatially uniform calibrations over multi-degree fields (Tonry et al., 2018).

2. Data Sources and Astrometric Reference System

The construction of Refcat2 involved the merger of eight principal data sources, with Gaia DR2 serving as the fundamental astrometric backbone:

• Gaia DR2: Positions, proper motions, parallaxes, and $m \sim 19$3, $m \sim 19$4, $m \sim 19$5 photometry
• Pan-STARRS1 DR1: $m \sim 19$6 photometry ($m \sim 19$7)
• SkyMapper DR1.1: $m \sim 19$8 ($m \sim 19$9)
• ATLAS Pathfinder: $\sim$0 (dedicated 5° diameter astrograph, Mauna Loa)
• Re-processed APASS: $\sim$1 ($\sim$2)
• Tycho-2 and Yale Bright Star Catalog: inclusion of very bright stars ($\sim$3)
• 2MASS: $\sim$4 bands for extinction and color correction

Each Refcat2 entry with Gaia DR2 association reports epoch 2015.5 ICRS coordinates with $\sim$5 mas uncertainty and proper motions accurate to $\sim$60.2 mas yr$\sim$7. The inherently uniform Gaia astrometry forms the absolute spatial reference for the catalog (Tonry et al., 2018).

3. Photometric Calibration Methodology

Refcat2 enforces a homogeneous photometric system based on Pan-STARRS bandpasses through a three-stage protocol:

1. Photoflat construction: For each frame, images are pre-flattened with twilight sky flats. Residual, star-wise magnitude differences (relative to Pan-STARRS in overlap regions) are fitted with an $\sim$8 spline and applied multiplicatively to correct spatial variations to $\sim$9 mag.
2. Zeropoint determination: Exposure zeropoints are derived via (i) direct comparison to established standards (Pan-STARRS, Gaia+2MASS regression), (ii) airmass regression on photometric nights, and (iii) inter-pointing overlap (“ubercal”) constraints.
3. Homogeneous transformation: Empirical color-term transformations are computed to cast all contributing surveys onto the $\lesssim 0.01$0 system. For example, for re-reduced APASS (AP):

$\lesssim 0.01$1

$\lesssim 0.01$2

$\lesssim 0.01$3

For southern sky ($\lesssim 0.01$4) or saturated-bright stars, the “GMP” subset (Gaia + 2MASS + low-reddening, well-behaved color stars) enables polynomial regressions from Gaia+2MASS to PS1 magnitudes. The regressions include crowding bias suppression terms. Equivalent relations exist for $\lesssim 0.01$5, $\lesssim 0.01$6, and $\lesssim 0.01$7 (Tonry et al., 2018).

4. Catalog Content, Data Structure, and Distribution

Refcat2 encodes 991 million stars, each with up to 44 catalog fields. Key quantities include:

• RA, Dec (epoch 2015.5, in $\lesssim 0.01$8 deg units)
• Gaia astrometry ($\lesssim 0.01$9, $\geq 99\%$0, $\geq 99\%$1, with uncertainties)
• Gaia photometry ($\geq 99\%$2, $\geq 99\%$3, $\geq 99\%$4, uncertainties in mmag)
• Effective temperature ($\geq 99\%$5), Gaia extinction ($\geq 99\%$6), variability, duplicity flags
• SFD $\geq 99\%$7-band extinction ($\geq 99\%$8 in mmag)
• Proximity radii where neighbor flux exceeds set thresholds
• Final PS1 $\geq 99\%$9 magnitudes/uncertainties/$g$0/contribution bitmaps
• 2MASS $g$1 with uncertainties

Entries are organized into $g$2 one-degree-square CSV files (e.g., "270−20.rc2"), compressed to $g$345 GB total ($g$450 bytes per star). Access is via MAST, by bulk download or cone-search SQL interfaces (Tonry et al., 2018).

5. Accuracy, Systematics, and Completeness

Astrometric uncertainties are inherited from Gaia DR2: $g$5 mas ($g$6), proper motions to $g$70.2 mas yr$g$8, and parallaxes to $g$90.1 mas. Photometric systematics show degree-scale RMS of $r$0 mmag for $r$1 and $r$23 mmag systematics (GMP-based) in the south. Localized errors, up to $r$320 mmag, are present in crowded or high-reddening fields (e.g., Galactic center, LMC), affected by crowding and unresolved stellar populations.

Random uncertainty per star is estimated based on input catalog performance and a floor (APASS: 1 mmag + floor, Tycho/BSC: 10 mmag, Gaia-only conversions: 20 mmag). Completeness exceeds 99% for $r$4 globally. Very bright stars ($r$5) are present, though with photometric errors at a few percent. Non-Gaia objects constitute $r$6 of the catalog, with effective artifact rejection possible via $r$7 filtering (Tonry et al., 2018).

6. Operational Use and Data Access

ATLAS Refcat2 is optimized for pipeline calibration in time-domain surveys. A standard nightly reduction workflow includes:

1. Extraction and PSF fitting on new exposures ($r$819 mag depth, 30 s exposure)
2. Star-matching ($r$9 per field) to Refcat2 for astrometric alignment to 0.07″ RMS, leveraging Gaia positions
3. Determination of photometric zeropoints via reference offset calculation and cloud-mapping ($i < 19$0 mag field corrections)
4. Final field-wide photometric precision of $i < 19$1 mag

SQL query access via MAST CasJobs or TAP (e.g., to retrieve Refcat2 stars for a sky region and magnitude interval) is supported. Several optimized query templates are provided for both Refcat2 and Pan-STARRS sources (Tonry et al., 2018).

7. Limitations and Flagging Protocols

Systematics are most severe in crowded regions and for very bright sources. Flags are embedded to indicate:

• Gaia variability
• Duplicity (potential blending)
• Large $i < 19$2 in $i < 19$3 band fits

This enables downstream filtering for high-reliability photometric and astrometric subsets. Non-Gaia objects are rare, and those at fainter magnitudes are often spurious or extragalactic contaminants. The catalog's built-in flags and recommended filtering parameters (e.g., $i < 19$4) facilitate exclusion of such entries in scientific analysis (Tonry et al., 2018).