Singular spectrum analysis of time series data from low frequency radiometers, with an application to SITARA data

TL;DR

This paper explores the use of singular spectrum analysis (SSA) for analyzing low frequency radio telescope data, demonstrating its connection to Fourier methods and applying it to SITARA data to understand calibration challenges.

Contribution

It introduces a novel SSA-based technique for long-term gain change detection and applies it to real SITARA data, revealing temperature fluctuations as a key factor.

Findings

SSA reveals temperature-related variations in SITARA data

Diurnal gain variations limit SSA calibration effectiveness

Proposed SSA method captures long-term gain changes

Abstract

Understanding the temporal characteristics of data from low frequency radio telescopes is of importance in devising suitable calibration strategies. Application of time series analysis techniques to data from radio telescopes can reveal a wealth of information that can aid in calibration. In this paper, we investigate singular spectrum analysis (SSA) as an analysis tool for radio data. We show the intimate connection between SSA and Fourier techniques. We develop the relevant mathematics starting with an idealised periodic dataset and proceeding to include various non-ideal behaviours. We propose a novel technique to obtain long-term gain changes in data, leveraging the periodicity arising from sky drift through the antenna beams. We also simulate several plausible scenarios and apply the techniques to a 30-day time series data collected during June 2021 from SITARA - a short-spacing two element interferometer for global 21-cm detection. Applying the techniques to real data, we find that the first reconstructed component - the trend - has a strong anti-correlation with the local temperature suggesting temperature fluctuations as the most likely origin for the observed variations in the data. We also study the limitations of the calibration in the presence of diurnal gain variations and find that such variations are the likely impediment to calibrating SITARA data with SSA.