RFI Mitigation for the Parkes Galactic All-Sky Survey (GASS)

TL;DR

This paper details RFI mitigation techniques applied to the GASS survey, enhancing the accuracy of Galactic HI emission data by using statistical filtering, prior information, and interpolation methods to remove interference.

Contribution

The paper introduces specific RFI mitigation strategies for the GASS survey, combining median filtering, prior HI distribution data, and statistical comparison to improve data quality.

Findings

Most RFI was cleaned using median filtering and prior information.

Constant RFI spikes were flagged and interpolated from LAB survey.

Remaining RFI was identified through statistical deviations and replaced.

Abstract

The GASS is a survey of Galactic atomic hydrogen (HI) emission in the southern sky observed with the Parkes 64-m Radio Telescope. With a sensitivity of 60 mK for a channel width of 1 km/s the GASS is the most sensitive and most accurate survey of the Galactic HI emission in the southern sky. We discuss RFI mitigation strategies that have been applied during the data reduction. Most of the RFI could be cleaned by using prior information on the HI distribution as well as statistical methods based on median filtering. Narrow line RFI spikes have been flagged during the first steps of the data processing. Most of these lines were found to be constant over long periods of time, such data were replaced by interpolating profiles from the Leiden/Argentine/Bonn (LAB) survey. Remaining RFI was searched for at any position by a statistical comparison of all observations within a distance of 0.1 deg. The median and mean of the line emission was calculated. In cases of significant deviations between both it was checked in addition whether the associated rms fluctuations exceeded the typical scatter by a factor of 3. Such data were replaced by the median, which is found to be least biased by RFI and other artifacts. The median estimator was found to be equivalent to the mean, which was obtained after rejecting outliers.