An optimized transient detection pipeline for the ASKAP Variables and Slow Transients (VAST) survey

Abstract

ABSTRACT In this paper, we present an optimized version of the detection pipeline for the ASKAP Variables and Slow Transients (VAST) survey, offering significant performance improvement. The key to this optimization is the replacement of the original w-projection algorithm integrated in the Common Astronomy Software Applications package with the w-stacking algorithm implemented in the WSClean software. Our experiments demonstrate that this optimization improves the overall processing efficiency of the pipeline by approximately a factor of 3. Moreover, the residual images generated by the optimized pipeline exhibit lower noise levels and fewer artefact sources, suggesting that our optimized pipeline not only enhances detection accuracy but also improves imaging fidelity. This optimized VAST detection pipeline is integrated into the Data Activated Liu Graph Engine (DALiuGE) execution framework, specifically designed for SKA-scale big data processing. Experimental results show that the performance and scalability advantages of the pipeline using DALiuGE over traditional MPI or BASH techniques increase with the data size. In summary, the optimized transient detection pipeline significantly reduces runtime, increases operational efficiency, and decreases implementation costs, offering a practical optimization solution for other ASKAP imaging pipelines as well.