Two procedures to flag radio frequency interference in the UV plane

TL;DR

This paper introduces two algorithms for detecting and flagging radio frequency interference in radio interferometric data, significantly reducing noise and systematics in images, and improving source detection accuracy.

Contribution

The paper presents novel algorithms that utilize UV plane redundancy and ripple removal techniques to effectively identify faint RFI in radio interferometric data.

Findings

20-50% reduction in image noise across datasets

Increased number of detected sources after RFI flagging

Minimal impact on flux densities of sources after flagging

Abstract

We present two algorithms to identify and flag radio frequency interference (RFI) in radio interferometric imaging data. The first algorithm utilizes the redundancy of visibilities inside a UV cell in the visibility plane to identify corrupted data, while varying the detection threshold in accordance with the observed reduction in noise with radial UV distance. In the second algorithm, we propose a scheme to detect faint RFI in the visibility time-channel plane of baselines. The efficacy of identifying RFI in the residual visibilities is reduced by the presence of ripples due to inaccurate subtraction of the strongest sources. This can be due to several reasons including primary beam asymmetries and other direction dependent calibration errors. We eliminated these ripples by clipping the corresponding peaks in the associated Fourier plane. RFI was detected in the ripple-free time-channel plane but was flagged in the original visibilities. Application of these two algorithms to 5 different 150 MHz datasets from the GMRT resulted in a reduction in image noise of 20-50% throughout the field along with a reduction in systematics and a corresponding increase in the number of detected sources. However, on comparing the mean flux densities before and after flagging RFI we find a differential change with the fainter sources ($25\sigma <$ S $< 100$ mJy) showing a change of -6% to +1% relative to the stronger sources (S $>$ 100 mJy). We are unable to explain this effect but it could be related to the CLEAN bias known for interferometers.