[Paper Review] Progress with the LOFAR Imaging Pipeline

G. Heald, J. P. McKean+15 more|UvA-DARE (University of Amsterdam)|Aug 27, 2010

Radio Astronomy Observations and Technology17 references3 citations

TL;DR

This paper presents the development and early performance of the LOFAR Imaging Pipeline, an automated system for processing interferometric data into calibrated Stokes-I images. The pipeline enables end-to-end, user-independent imaging of large-area sky surveys, achieving sub-arcminute resolution with rms noise levels of a few mJy beam⁻¹, as demonstrated on extended sources like M51 and point-source fields during commissioning.

ABSTRACT

One of the science drivers of the new Low Frequency Array (LOFAR) is large-area surveys of the low-frequency radio sky. Realizing this goal requires automated processing of the interferometric data, such that fully calibrated images are produced by the system during survey operations. The LOFAR Imaging Pipeline is the tool intended for this purpose, and is now undergoing significant commissioning work. The pipeline is now functional as an automated processing chain. Here we present several recent LOFAR images that have been produced during the still ongoing commissioning period. These early LOFAR images are representative of some of the science goals of the commissioning team members.

Motivation & Objective

• To develop an automated, end-to-end imaging pipeline for processing LOFAR interferometric data without human intervention.
• To enable rapid production of calibrated, high dynamic range images suitable for large-area low-frequency sky surveys.
• To support imaging of both point sources and extended sources like spiral galaxies, such as M51.
• To integrate advanced calibration techniques, including direction-independent and future direction-dependent calibration, for improved image fidelity.
• To lay the foundation for future polarization imaging and station-based calibration enhancements.

Proposed method

• The pipeline uses the DPPP (Default Pre-Processing Pipeline) for initial flagging of RFI and data compression using median filtering in time and frequency domains.
• It employs the BBS (BlackBoard Selfcal) system to generate a local sky model (LSM) from a global sky model (GSM) for calibration of complex station gains.
• Calibrated data are imaged using either CImager or CASA’s imager, followed by deconvolution via the Clark CLEAN algorithm.
• The pipeline performs iterative major cycles of calibration, flagging, imaging, and LSM updates to refine image quality and dynamic range.
• Future enhancements include source-finding software for more accurate LSM updates and integration of direction-dependent calibration.
• The system is implemented in Python and designed to process full datasets automatically after observation.

Experimental results

Research questions

• RQ1 Can the LOFAR Imaging Pipeline achieve fully automated, end-to-end processing of interferometric data to produce calibrated images?
• RQ2 What is the achievable image quality, in terms of noise and dynamic range, for extended and point-source fields during the commissioning phase?
• RQ3 How effective is the DPPP flagging system in mitigating RFI in the dense low-frequency radio environment of the Netherlands?
• RQ4 To what extent can the pipeline handle fields with bright off-axis sources, such as 3C289 near M51, without significant artifacts?
• RQ5 What improvements are needed to enable polarization imaging and station-based calibration in the future?

Key findings

• The LOFAR Imaging Pipeline successfully produces fully calibrated Stokes-I images with an rms noise level of approximately 3 mJy beam⁻¹ for typical observations.
• For the M51 field, the pipeline achieved a dynamic range of 550 in a widefield image with 2-arcminute resolution, limited by deconvolution errors from a bright off-axis source.
• A high-resolution image of M51 was produced with a synthesized beam size of 59″ × 53″ and an rms noise of 6 mJy beam⁻¹, demonstrating capability for extended source imaging.
• The pipeline has been run end-to-end on known fields without user intervention, confirming its automation potential.
• The current dynamic range limitations are primarily due to deconvolution errors from bright off-axis sources, which will be mitigated by future direction-dependent calibration.
• The pipeline is being extended to support polarization imaging and will benefit from upcoming station calibration, which will improve sensitivity and beam behavior.