GaLactic and Extragalactic All-sky Murchison Widefield Array survey eXtended (GLEAM-X) III: Galactic Plane

Abstract: We present the third data release for the Galactic and Extragalactic All-Sky Murchison Widefield Array eXtended (GLEAM-X) survey, covering = 3800 deg2 of the southern Galactic Plane (GP) with \ang{233} < l < \ang{44} and |b| < \ang{11} across a frequency range of 72 - 231 MHz divided into 20 sub-bands. GLEAM-X observations were taken using the "extended" Phase-II configuration of the Murchison Widefield Array (MWA), which features baselines ranging from approximately 12 m to 5 km. This configuration limits sensitivity to the diffuse structure of the GP, with an angular resolution range of about 45'' to 2'. To achieve lower noise levels while being sensitive to a wide range of spatial scales (45''- \ang{15}), we combined these observations with the previous Galactic and Extragalactic All-Sky Murchison Widefield Array (GLEAM) survey. For the area covered, we provide images spanning the whole frequency range. A wide-band image over 170 - 231 MHz, with RMS noise of = 3 - 6 mJy/beam and source position accuracy within 1 arcseconds, is then used to perform source-finding, which yields 98,207 elements measured across 20 x 7.68 MHz frequency bands. The catalogue is 90% complete at 50 mJy within \ang{233} < l < \ang{324} and at 125 mJy in \ang{290} < l < \ang{44}, while it is 99.3% reliable overall. All the images and the catalogue are available online for download.

• The paper presents a 3,800 deg² low-frequency survey of the southern Galactic Plane, combining GLEAM and GLEAM-X data to recover emission from 45″ to 15° scales at 72–231 MHz.
• The joint Image Domain Gridding deconvolution achieves 3–6 mJy RMS noise and produces a catalogue of 98,207 sources with 99.3–99.4% reliability, validated flux densities, and approximately arcsecond-level astrometric accuracy.
• The paper demonstrates applications for finding faint supernova remnants, studying free-free absorption in H II regions and planetary nebulae, and identifying pulsar candidates, while cautioning that completeness varies strongly with Galactic longitude and latitude.

The third data release of the GaLactic and Extragalactic All-sky Murchison Widefield Array survey eXtended (GLEAM-X) presents a low-frequency radio continuum survey of the southern Galactic Plane (GP), covering approximately 3800 deg$^2$ over $233^\circ < l < 44^\circ$ and $|b| < 11^\circ$ at frequencies of 72–231 MHz (2602.08475). The defining methodological feature of this release is the joint deconvolution of data from two MWA configurations: the Phase I GLEAM survey, which is sensitive to large angular scales ($2'$–$15^\circ$), and the Phase II "extended" configuration used for GLEAM-X, which reaches finer scales ($45''$–$20'$). By combining both datasets through Image Domain Gridding (IDG) within WSClean, the survey achieves sensitivity to spatial scales from $45''$ to $15^\circ$, addressing a well-known deficiency of mid-frequency Galactic Plane surveys such as EMU and SMGPS, whose spectra become unreliable for sources larger than about $2'$ in radius.

Survey design and observations

Observations were conducted in drift-scan mode using seven declination pointings between $233^\circ < l < 44^\circ$0 and $233^\circ < l < 44^\circ$1, each observed for roughly two minutes per band across five 30.72 MHz bands spanning 72–103, 103–134, 139–170, 170–200, and 200–231 MHz, with the 137 MHz OrbComm satellite band excluded. GLEAM observations spanned August 2013 to June 2014 (~28 nights), while GLEAM-X observations covered January 2018 to October 2020 (113 nights). The extended configuration reduces the confusion limit from approximately 2 mJy beam$233^\circ < l < 44^\circ$2 to 0.3 mJy beam$233^\circ < l < 44^\circ$3 at 200 MHz relative to Phase I, enabling the deeper integrations exploited here.

Data reduction

The reduction pipeline builds on the GLEAM-X pipeline of Hurley-Walker et al., with several modifications tailored to the bright, extended sources that dominate the plane. Calibration transfers amplitude and phase gains from nearby bright calibrators, with an in-field alternative using the GLEAM sky model when calibrator solutions are unavailable. An additional flagging stage targets RFI in the lowest band, and sidelobe contamination from bright off-field sources is handled by phase-rotating visibilities onto each contaminant, forming a shallow model image, and subtracting it; Centaurus A receives special treatment with a larger modelling box due to its high surface brightness.

Astrometric calibration follows the established approach of cross-matching compact sources against NVSS, SUMSS, and — to fill the gap where those surveys do not cover the plane — a bright subset of RACS. Position shifts are stored as FITS cubes and applied as direction-dependent ($233^\circ < l < 44^\circ$4-term) corrections during joint imaging. Of all processed observations, only 57% were retained; the remaining 43% were discarded primarily due to calibration failures, residual artefacts, and ionospheric activity, with losses concentrated in the lowest frequency band. This substantial rejection rate is an honest reflection of the difficulty of low-frequency processing along the plane, though the surviving dataset remains balanced between surveys (53% GLEAM-X, 47% GLEAM).

Joint deconvolution via IDG runs on groups of 10–15 observations sharing a declination pointing, applying ionospheric corrections, the Full Embedded Element primary beam model, and $233^\circ < l < 44^\circ$5-term corrections during gridding rather than post hoc. The resulting group images are co-added with swarp into 20 narrow-band (7.68 MHz), five 30.72 MHz, and one wideband (170–231 MHz) mosaic per region, using ZEA projections in two longitude ranges to maintain consistent PSF properties. A position-dependent blur correction normalises flux density scales for residual ionospheric smearing, interpolated over Galactic latitudes $233^\circ < l < 44^\circ$6 where insufficient compact reference sources exist — a region where the correction therefore rests on extrapolation.

Noise properties and catalogue

The wideband source-finding image reaches RMS noise levels of approximately 3–6 mJy, which the authors note is 5–8 times lower than the original GLEAM Galactic Plane data, attributable to the improved resolution of the extended configuration and roughly three times longer total integration time. Pixel distributions confirm Gaussian noise statistics after source subtraction. Source finding on the wideband image with aegean yields a catalogue of 98,207 sources, measured across all 20 narrow bands via priorised fitting. Spectral fitting assigns a simple power law to 89,655 sources and a curved power law to 3,132, with curved-spectrum candidates selected by reduced-$233^\circ < l < 44^\circ$7 comparison and the criterion $233^\circ < l < 44^\circ$8 with $233^\circ < l < 44^\circ$9. The median spectral indices per flux-density bin range from $|b| < 11^\circ$0 to $|b| < 11^\circ$1, and the authors highlight an excess of steep-spectrum sources toward the Galactic centre, consistent with the interpretation by de Gasperin et al. that these may be pulsars missed by time-domain searches owing to scattering broadening.

Astrometric accuracy, assessed against 8,421 isolated, compact, high-S/N reference sources, shows mean offsets of $|b| < 11^\circ$2 mas in RA and $|b| < 11^\circ$3 mas in Dec. Reliability, estimated from negative-source detections restricted to $|b| < 11^\circ$4, is 99.3% and 99.4% in the two regions. Completeness at 200 MHz differs markedly by region: 90% completeness is reached at 50 mJy for $|b| < 11^\circ$5 but requires 125 mJy for $|b| < 11^\circ$6, where large error bars reflect strong longitudinal variation driven by bright structures near Centaurus A versus quieter regions beyond Vela. Flux densities and spectral indices agree with the GLEAM-GP catalogue after Eddington bias correction, validating the flux scale, and the new data deliver substantially smaller spectral-index uncertainties thanks to the higher S/N.

Science applications

The paper outlines several applications enabled by the combination of low-frequency coverage and sensitivity to large angular scales. For supernova remnants, the expected population of roughly 2,000 Galactic SNRs contrasts sharply with only ~300 classified objects; the authors argue the survey's recovery of scales up to $|b| < 11^\circ$7 makes it well suited to detecting faint, old remnants, and demonstrate the quality of broadband spectral fits with the example of Kes 67 (G18.8+0.3). Free-free absorption signatures below 100 MHz can constrain unshocked ejecta and environmental interactions, and radio spectra connect to gamma-ray studies of cosmic-ray acceleration. For H II regions, the distinctive blue colour of thermally absorbed sources in RGB composites enables separation from non-thermal emission, extending the emissivity-mapping methodology of Hindson et al. and Su et al. toward constraining cosmic-ray electrons and the Galactic magnetic field. Planetary nebulae benefit from coverage of their optically thick spectral regime, illustrated with NGC 3918 ($|b| < 11^\circ$8). Finally, continuum imaging offers a complementary route to pulsar detection for sources rendered undetectable in periodicity searches by scattering and dispersion, following the approach demonstrated by Sett et al.; cross-matching steep-spectrum sources with unassociated gamma-ray sources may yield further candidates.

Limitations and open questions

Several caveats bear directly on the use of this release. Reliability and completeness estimates are restricted to $|b| < 11^\circ$9 and are explicitly not expected to hold amid the bright diffuse emission at low latitudes; within $2'$0, bane does not properly isolate background levels, so users are cautioned to inspect flux densities against the images directly. The latitude limit itself reflects the complexity of the joint deconvolution approach, leaving the extragalactic sky to the standard pipeline. The longitude range is bounded by elevation constraints and degradation caused by Cygnus A. Completeness varies strongly with longitude in the inner-Galaxy region, so flux-limited population studies there require position-dependent corrections. Whether the excess of steep-spectrum sources indeed traces a hidden pulsar population remains an inference adopted from prior work rather than a result demonstrated here, and confirming it would require targeted follow-up.

Conclusion

This data release delivers approximately 3800 deg$2'$1 of the southern Galactic Plane at 72–231 MHz with sensitivity to angular scales from $2'$2 to $2'$3, achieved through GPU-accelerated joint deconvolution of GLEAM and GLEAM-X visibilities. The accompanying catalogue of 98,207 sources, with 99.3% reliability, arcsecond-level astrometry, and validated flux calibration, provides the low-frequency, large-scale-sensitive counterpart that mid-frequency Plane surveys lack, and is positioned to advance SNR searches, free-free absorption studies, and continuum-based pulsar identification.