Rapid ASKAP Continuum Survey (RACS) Overview

• RACS is a wide-area, multi-frequency radio continuum survey using ASKAP’s 36-antenna array with phased-array feeds, providing detailed continuum and polarization data.
• Its observing strategy employs 15-minute snapshot integrations across hundreds to thousands of pointings, covering up to 90% of the sky in multiple frequency bands.
• RACS serves as both a calibration backbone for deeper ASKAP studies and a public resource for investigating radio source morphology, spectral indices, and polarimetric properties.

Rapid ASKAP Continuum Survey (RACS) is a wide-area, shallow, multi-band radio continuum survey conducted with the full 36-antenna Australian Square Kilometre Array Pathfinder (ASKAP). It was conceived as a real-sky reference or “Global Sky Model” for calibration of future deep ASKAP surveys, as a vehicle for commissioning the full ASKAP system, and as a public survey resource bridging metre-wavelength and decimetric radio surveys. RACS combines 15 min snapshot observations, phased-array-feed beamforming over roughly 30–31 deg² per pointing, and repeated releases in low-, mid-, and high-frequency bands, ultimately sampling the ASKAP range from 700 to 1800 MHz with total-intensity and polarimetric products (McConnell et al., 2020, Koribalski, 2022).

1. Conception and instrumental basis

RACS is inseparable from the ASKAP instrument architecture. ASKAP comprises 36 × 12 m antennas, each equipped with a dual-polarisation phased-array feed (PAF) that forms 36 simultaneous beams over an instantaneous field of view of about 30–31 deg². The array is optimized for surface-brightness sensitivity on angular scales of 10″–3′ at 1.4 GHz, and the PAF system is the central reason that ASKAP can execute rapid, all-southern-sky continuum mapping at moderate resolution (McConnell et al., 2020, Koribalski, 2022).

The survey’s initial scientific motivation had three linked components: to provide a shallow all-sky continuum model for calibration of deeper ASKAP surveys, to exercise the full telescope and software stack before the longer survey program, and to deliver a community dataset deeper and sharper than the principal legacy southern-sky comparators, especially NVSS and SUMSS. In that original framing, RACS was not only a science survey but also a calibration infrastructure layer for the broader ASKAP portfolio (McConnell et al., 2020).

This calibration role remained central as the project evolved. Later descriptions characterize RACS as a multi-band all-sky program intended to build a shallow global sky model for ASKAP calibration, while also supporting spectral-index, polarization, time-domain, and morphology-driven science. A plausible implication is that RACS should be understood less as a single catalogue release than as a survey framework with repeated observing epochs and derivative pipelines (Duchesne et al., 2023, Thomson et al., 16 May 2026).

2. Observing strategy, tiling, and sky coverage

The first low-band epoch observed the sky south of declination $+41^\circ$ at a central frequency of 887.5 MHz using 903 individual pointings, each integrated for 15 min. In the first-epoch design, the tile centres were arranged in constant-declination rows, with a square_6×6 beam footprint and 1.05° pitch, yielding a surveyed area of 34 240 deg² (McConnell et al., 2020). A closely related summary describes RACS-low as an ASKAP observatory project covering the entire sky south of $\delta=+41^\circ$, amounting to $\sim 30\,000$ deg² (Koribalski, 2022).

Subsequent epochs expanded both frequency coverage and northern declination reach. RACS-mid at 1367.5 MHz used 1493 tiles over 36 449 deg², with images extending to $\delta_{\rm J2000}\lesssim +49^\circ$, and employed the closepack36 PAF footprint with 36 overlapping beams spaced by 0.9° (Duchesne et al., 2023). RACS-high at 1655.5 MHz likewise used 1493 snapshot pointings, each of $\sim 15$ min and 288 MHz bandwidth, to span the sky from the South Celestial Pole to about $+48^\circ$, corresponding to $\approx 90\%$ of the celestial sphere or $\sim 37\,000$ deg² (Duchesne et al., 9 Jan 2025). The second low-band epoch, RACS-low2, used 947 unique fields to extend low-band coverage to $\delta_{\rm J2000}\approx +48^\circ$, largely matching the mid- and high-band footprints (Duchesne et al., 15 Jun 2026).

Across these releases, the core observing logic remained stable: each field is a single ASKAP pointing tracked on the sky; the 36 PAF beams are mosaicked to flatten sensitivity; and adjacent pointing centres are spaced to sample the PAF primary-beam pattern with near-uniform coverage (Koribalski, 2022). This common structure made RACS suitable for direct cross-band comparison even as the beam footprint, declination range, and beam models evolved.

3. Calibration, imaging, and catalogue construction

RACS data are processed with the ASKAPsoft science-processing pipeline. Common steps include radio-frequency-interference detection and flagging, bandpass calibration using bright calibrator scans, gain calibration per beam and per antenna, primary-beam correction using PAF beam models, and mosaicking of individual beams or fields (Koribalski, 2022). In the first low-band release, ASKAPsoft performed flux-scale and bandpass calibration using daily scans of PKS B1934−638, phase and amplitude self-calibration, Briggs-robust imaging on a 2.5″ pixel grid, primary-beam correction via linmos, and mosaicking with SWARP after convolution to a common beam where required (Hale et al., 2021).

Later releases added more aggressive direction-dependent control of artefacts. RACS-mid used wide-field $w$-projection gridding, multi-scale CLEAN, two-term multi-frequency synthesis, custom primary-beam models, selective peeling of bright off-axis sources, and wide-field Stokes $\delta=+41^\circ$0 leakage correction based on Zernike-surface models (Duchesne et al., 2023). RACS-high introduced a universal baseline cut of $\delta=+41^\circ$1 m to suppress large-scale ripples and Sun-induced artefacts, self-calibration and CLEAN imaging of Stokes $\delta=+41^\circ$2 and $\delta=+41^\circ$3, peeling of 81 bright off-axis sources with the PotatoPeel script, and holography-derived primary-beam models for three beam-forming epochs (Duchesne et al., 9 Jan 2025).

Source finding is catalogue-specific but generally PyBDSF-based. In the first RACS-low Stokes $\delta=+41^\circ$4 catalogue, PyBDSF operated on 25″ mosaics with local-noise tracking, $\delta=+41^\circ$5 island thresholds, $\delta=+41^\circ$6 pixel thresholds, and atrous wavelet decomposition for extended emission; duplicates in overlaps were removed by tile-centre proximity (Hale et al., 2021). In RACS-mid DR2, full-sensitivity tiles were convolved to the lowest common resolution of neighbouring images before source finding, and de-duplication between overlapping tiles used a separation threshold based on the major-axis PSF of the two entries (Duchesne et al., 2023). RACS-high identified sources on full-sensitivity mosaics with a $\delta=+41^\circ$7 threshold, then grouped Gaussian components into sources and merged overlapping tiles by retaining the detection closest to the tile centre (Duchesne et al., 9 Jan 2025).

The usual thermal-noise estimate for RACS is

$\delta=+41^\circ$8

with $\delta=+41^\circ$9. For ASKAP, typical SEFD values of $\sim 30\,000$0–$\sim 30\,000$1 Jy and $\sim 30\,000$2 MHz over $\sim 30\,000$3 s imply $\sim 30\,000$4–$\sim 30\,000$5 mJy beam$\sim 30\,000$6, matching the measured survey depths. The survey-speed figure of merit is commonly written $\sim 30\,000$7, emphasizing why the large PAF field of view is fundamental to RACS (Koribalski, 2022).

4. Major data releases and measured survey performance

The sequence of public releases moved from a common-resolution low-band catalogue toward higher-resolution and broader-sky products at multiple frequencies. Key releases are summarized below.

Release	Coverage and angular scale	Catalogue and performance
RACS-low DR1	799 tiles convolved to 25″; $\sim 30\,000$8 to $\sim 30\,000$9; 28 020 deg² outside $\delta_{\rm J2000}\lesssim +49^\circ$0	2 123 638 sources; median noise $\delta_{\rm J2000}\lesssim +49^\circ$1 mJy beam$\delta_{\rm J2000}\lesssim +49^\circ$2; 95% overall completeness at $\delta_{\rm J2000}\lesssim +49^\circ$3 mJy (Hale et al., 2021)
RACS-mid DR2	$\delta_{\rm J2000}\lesssim +49^\circ$4 deg² to $\delta_{\rm J2000}\lesssim +49^\circ$5; median PSF $\delta_{\rm J2000}\lesssim +49^\circ$6	3 105 668 sources; median local rms 182 $\delta_{\rm J2000}\lesssim +49^\circ$7Jy beam$\delta_{\rm J2000}\lesssim +49^\circ$8; 95% completeness at $\delta_{\rm J2000}\lesssim +49^\circ$9 mJy (Duchesne et al., 2023)
RACS-high	1493 images to $\sim 15$0; median PSF $\sim 15$1	2 677 509 sources; median rms $\sim 15$2 $\sim 15$3Jy PSF$\sim 15$4; overall reliability 99.18% (Duchesne et al., 9 Jan 2025)
RACS-low2	whole sky to $\sim 15$5; median PSF $\sim 15$6 in catalogue images	3 922 151 Stokes $\sim 15$7 sources; median rms 195 $\sim 15$8Jy in $\sim 15$9, 163 $+48^\circ$0Jy in $+48^\circ$1; reliability $+48^\circ$2 at $+48^\circ$3 (Duchesne et al., 15 Jun 2026)

The first low-band Stokes $+48^\circ$4 catalogue established the survey’s basic photometric and completeness properties. Simulations with injected point sources gave 50% completeness at $+48^\circ$5 mJy and 95% point-source completeness at $+48^\circ$6 mJy; folding in the intrinsic source-count model yielded an overall 95% completeness near 3 mJy. Negative-image tests indicated a false detection rate $+48^\circ$7, and astrometric offsets relative to SUMSS, NVSS, and AT20G were generally confined to $+48^\circ$8 (Hale et al., 2021).

RACS-mid shifted the balance toward higher angular resolution and all-sky coverage below $+48^\circ$9. Its primary catalogue has variable resolution to maximize sensitivity and sky coverage, while two auxiliary catalogues provide a 25″ fixed-resolution product approximately matched to the low-band footprint and a time-domain catalogue that preserves independently imaged epochs without convolution, mosaicking, or de-duplication (Duchesne et al., 2023). RACS-high extended this strategy to higher frequency with a median rms of $\approx 90\%$0 $\approx 90\%$1Jy PSF$\approx 90\%$2, a brightness-scale accuracy of 10%, and final astrometric uncertainties of $\approx 90\%$3 in right ascension and $\approx 90\%$4 in declination after correction of a declination-dependent systematic trend (Duchesne et al., 9 Jan 2025).

Relative to NVSS and SUMSS, the early survey summaries emphasize that RACS offers roughly three times better angular resolution and about two times lower rms noise over comparable southern-sky regions, while retaining genuinely rapid sky coverage (Koribalski, 2022). That combination is the operational signature of the survey.

5. Spectral, polarization, and pulsar products

Although the initial public emphasis was on Stokes $\approx 90\%$5, RACS was designed from the outset as a full-Stokes survey. Early descriptions explicitly anticipated catalogues with spectral-index and polarization information, and later releases delivered both circular-polarization images and spectro-polarimetric source products (McConnell et al., 2020, Duchesne et al., 2023).

SPICE-RACS is the principal polarimetric derivative. The first data release processed 30 RACS-low fields over $\approx 90\%$6 deg², imaged Stokes $\approx 90\%$7, $\approx 90\%$8, and $\approx 90\%$9 across 744–1032 MHz at 1 MHz spectral resolution and 25″ angular resolution, and produced 5818 reliable rotation measures (RMs) with an areal density of $\sim 37\,000$0 RMs deg$\sim 37\,000$1, mean RM error $\sim 37\,000$2 rad m$\sim 37\,000$3, and median linear polarization fraction $\sim 37\,000$4 (Thomson et al., 2023). In SPICE-RACS DR2, based on the third low-band epoch, the survey covers approximately 87.5% of the celestial sphere, produces $\sim 37\,000$5 cutout cubes around 4 million radio sources and spectra toward 5 million components, reaches $\sim 37\,000$6 $\sim 37\,000$7Jy PSF$\sim 37\,000$8 noise with $\sim 37\,000$9 resolution, and yields 246 509 unique Faraday rotation measures in the conservative $\delta_{\rm J2000}\approx +48^\circ$0 subset, with an areal density of $\delta_{\rm J2000}\approx +48^\circ$1 deg$\delta_{\rm J2000}\approx +48^\circ$2 and median RM uncertainty $\delta_{\rm J2000}\approx +48^\circ$3 rad m$\delta_{\rm J2000}\approx +48^\circ$4 (Thomson et al., 16 May 2026).

The formalism used in SPICE-RACS follows standard RM-synthesis. The complex polarization spectrum $\delta_{\rm J2000}\approx +48^\circ$5 and Faraday dispersion function $\delta_{\rm J2000}\approx +48^\circ$6 satisfy

$\delta_{\rm J2000}\approx +48^\circ$7

and the Faraday depth is

$\delta_{\rm J2000}\approx +48^\circ$8

Within DR2, residual wide-field instrumental polarization is reported to be on the order of $\delta_{\rm J2000}\approx +48^\circ$9, and complexity metrics such as $w$0 and $w$1 are provided to flag Faraday-complex spectra (Thomson et al., 16 May 2026).

RACS has also supported targeted pulsar analysis. A dedicated study detected 661 known pulsars at 888 MHz through astrometric coincidence and estimated a false alarm rate of $w$2. Using archival 400 and 1400 MHz data, it found a mean power-law spectral index of $w$3, while also showing that a single power law is inadequate for all sources; when flux densities between 150 MHz and 3 GHz are included, up to 40% of the sample shows deviations from a simple power-law model. Stokes $w$4 measurements yielded circular-polarization fractions for 9% of the sample with a mean of $w$5, and pseudo-luminosity estimates based on dispersion-measure distances showed no strong evidence for correlation with intrinsic pulsar properties (Anumarlapudi et al., 2023).

The second low-band epoch added a survey-wide Stokes $w$6 candidate catalogue. From 221 measurements above the leakage and detection thresholds, the likely identifications comprised 61 radio stars, 85 pulsars, 43 AGN—many probably residual leakage—and one source not associated with a known astronomical object (Duchesne et al., 15 Jun 2026). This underscores that RACS circular-polarization products are both scientifically useful and calibration-sensitive.

6. Scientific exploitation across source populations

RACS has been used as a morphology survey for powerful radio galaxies. Using RACS-low1, RACS-mid, and RACS-high together over the sky south of $w$7, ASKAP imaging of the G4Jy-3CRE sample classified 173 of 264 sources (66%) as showing jet morphology, including 37 not previously seen to host jets in archival radio maps. With $w$8 beams, FRI/FRII classification was reported as reliable out to $w$9, and the data revealed six optical counterparts that had previously been unidentified or ambiguous (Gawenda et al., 21 May 2026).

RACS has also become a selection engine for rare high-redshift AGN. By cross-matching RACS-low with deep optical and near-infrared surveys over 16 000 deg², one study selected 45 high-$\delta=+41^\circ$00 radio quasar candidates with $\delta=+41^\circ$01 mJy and $\delta=+41^\circ$02, then spectroscopically confirmed 24 new quasars, including 11 at $\delta=+41^\circ$03. After incorporating literature objects with similar flux and magnitude limits, the resulting $\delta=+41^\circ$04 RACS sample contains 33 quasars and is estimated to be $\delta=+41^\circ$05 complete for $\delta=+41^\circ$06 and $\delta=+41^\circ$07 mJy; 22 of the 33 have flat radio spectra $\delta=+41^\circ$08 (Ighina et al., 14 Apr 2025).

At low redshift, RACS has been used for radio-activity censuses in nearby galaxies. A study of very massive early-type galaxies with 887.5 MHz RACS DR1 measurements found that 408 of 587 systems passing the adopted noise cuts are detected above $\delta=+41^\circ$09, and every one of the 40 galaxies with $\delta=+41^\circ$10 is detected. The inferred relation between radio power and $\delta=+41^\circ$11-band luminosity is $\delta=+41^\circ$12, albeit with scatter spanning several orders of magnitude (Brown et al., 2023).

The survey’s full-Stokes capability has opened Galactic stellar parameter space as well. A circular-polarization search in early RACS data identified 33 known stars after excluding artefacts, leakage, and known pulsars; 23 of these had no previous radio detections. The sample spans M-dwarfs, interacting binaries, young stellar objects, and magnetic chemically peculiar stars, demonstrating that RACS can isolate coherent stellar emitters efficiently through a fractional circular-polarization filter (Pritchard et al., 2021).

RACS catalogues have also become testbeds for algorithmic morphology work. A self-organizing-map analysis of 251 259 multi-Gaussian RACS sources assigned broad morphology labels to the complex-source subset and found that sources with Euclidean distance $\delta=+41^\circ$13 to their best-matching neuron, about 79% of the sample, have an estimated $\delta=+41^\circ$14 reliability for their SOM-derived labels (Alam et al., 2024). This suggests that RACS has sufficient scale and homogeneity to function as a training and validation substrate for next-generation radio-source classification.

7. Limitations, astrometric refinement, and long-term role

RACS is a snapshot survey, and its limitations are correspondingly specific. High-band documentation notes modest sensitivity roll-off at beam overlaps, residual circular-polarization leakage of a few percent near mosaic edges, partial resolution loss for sources larger than a few arcminutes, and an astrometric floor of $\delta=+41^\circ$15 set by the absence of phase-referenced calibrators (Duchesne et al., 9 Jan 2025). Mid-band imaging similarly reports residual wide-field $\delta=+41^\circ$16 leakage after correction at the level of $\delta=+41^\circ$17 for the main-survey SBIDs and up to $\delta=+41^\circ$18 for a specific subset (Duchesne et al., 2023). These are not incidental caveats: they define the regime in which catalogue users must treat polarization, source extent, and local systematics carefully.

Astrometry has been a major area of post-release refinement. A dedicated correction procedure for RACS-Low1 and RACS-Low3, based on cross-matching to WISE and modelling separate scan- and beam-dependent offsets, removed the all-sky median offsets and reduced the 68% confidence interval in the residuals to $\delta=+41^\circ$19 or better. Validation against FIRST, VLASS, and the Radio Fundamental Catalogue led to adopted per-source positional uncertainties of $\delta=+41^\circ$20 over most of the survey area, degrading to $\delta=+41^\circ$21 near the Galactic plane (Jaini et al., 19 Mar 2025). A plausible implication is that RACS has evolved from a survey optimized primarily for continuum imaging into a precision reference frame for transient localization, including fast radio bursts.

The survey’s institutional role remains the same as at inception: it supplies calibrated visibilities, images, and source catalogues through CASDA, with no proprietary period emphasized for RACS products, and it underpins ASKAP projects such as EMU, POSSUM, VAST, DINGO, FLASH, and FRB localization workflows (Duchesne et al., 2023, McConnell et al., 2020). Taken together, the release history indicates that RACS is both a science archive and a calibration backbone: a public, repeatedly updated radio-continuum map of most of the southern sky, with sufficient spectral, polarimetric, morphological, and astrometric structure to support both standalone studies and the operation of the wider ASKAP survey system.