# Source counts and confusion at 72-231 MHz in the MWA GLEAM survey

## TL;DR

This paper presents accurate source counts from the GLEAM survey at 72-231 MHz, revealing discrepancies with models and highlighting confusion noise issues, with implications for future low-frequency radio surveys.

## Contribution

It provides the most accurate low-frequency source counts to date from GLEAM, compares them with models, and discusses confusion noise limitations and improvements in data processing.

## Key findings

- Source counts are more accurate due to large sky coverage and sensitivity to extended emission.
- No flattening of spectral index observed at frequencies above 0.5 Jy.
- Confusion noise dominates thermal noise at frequencies above ~100 MHz.

## Abstract

The GaLactic and Extragalactic All-sky MWA survey (GLEAM) is a radio continuum survey at 72-231 MHz of the whole sky south of declination +30 deg, carried out with the Murchison Widefield Array (MWA). In this paper, we derive source counts from the GLEAM data at 200, 154, 118 and 88 MHz, to a flux density limit of 50, 80, 120 and 290 mJy respectively, correcting for ionospheric smearing, incompleteness and source blending. These counts are more accurate than other counts in the literature at similar frequencies as a result of the large area of sky covered and this survey's sensitivity to extended emission missed by other surveys. At S_154MHz > 0.5 Jy, there is no evidence of flattening in the average spectral index (alpha approx. -0.8 where S proportional to nu^alpha) towards the lower frequencies. We demonstrate that the SKA Design Study (SKADS) model by Wilman et al. (2008) significantly underpredicts the observed 154 MHz GLEAM counts, particularly at the bright end. Using deeper LOFAR counts and the SKADS model, we find that sidelobe confusion dominates the thermal noise and classical confusion at nu >~ 100 MHz due to both the limited CLEANing depth and undeconvolved sources outside the field-of-view. We show that we can approach the theoretical noise limit using a more efficient and automated CLEAN algorithm.