Square Kilometre Array Science Data Challenge 3a: foreground removal for an EoR experiment

TL;DR

SDC3a demonstrates that current foreground-mitigation methods can recover substantial portions of the faint EoR power spectrum from SKA-Low-like data, even in the presence of realistic Galactic and extragalactic foregrounds and instrumental systematics. The study analyzes 17 participant pipelines using six 15-MHz bands across 106–196 MHz, assessing both the recovered signal and the reliability of error bars, with a focus on cross-pipeline comparability. While several teams achieve near-ground-truth recovery in some bins, error bars are frequently underestimated, highlighting the challenge of robust uncertainty quantification in the presence of residual systematics. The results underscore the value of multi-pipeline comparisons and community-driven benchmarks for diagnosing residual biases and guiding methodological improvements ahead of real SKA observations.

Abstract

We present and analyse the results of the Science data challenge 3a (SDC3a, https://sdc3.skao.int/challenges/foregrounds), an EoR foreground-removal community-wide exercise organised by the Square Kilometre Array Observatory (SKAO). The challenge ran for 8 months, from March to October 2023. Participants were provided with realistic simulations of SKA-Low data between 106 MHz and 196 MHz, including foreground contamination from extragalactic as well as Galactic emission, instrumental and systematic effects. They were asked to deliver cylindrical power spectra of the EoR signal, cleaned from all corruptions, and the corresponding confidence levels. Here we describe the approaches taken by the 17 teams that completed the challenge, and we assess their performance using different metrics. The challenge results provide a positive outlook on the capabilities of current foreground-mitigation approaches to recover the faint EoR signal from SKA-Low observations. The median error committed in the EoR power spectrum recovery is below the true signal for seven teams, although in some cases there are some significant outliers. The smallest residual overall is $4.2_{-4.2}^{+20} \times 10^{-4}\,\rm{K}^2h^{-3}$cMpc$^{3}$ across all considered scales and frequencies. The estimation of confidence levels provided by the teams is overall less accurate, with the true error being typically under-estimated, sometimes very significantly. The most accurate error bars account for $60 \pm 20$\% of the true errors committed. The challenge results provide a means for all teams to understand and improve their performance. This challenge indicates that the comparison between independent pipelines could be a powerful tool to assess residual biases and improve error estimation.

Figures (10)

• Figure 1: 21cmFAST parameters used for the generation of the EoR signal and corresponding mean (red) and standard deviation (blue) of the brightness temperature as a function of frequency.
• Figure 2: Cylindrical power spectrum of the true noiseless EoR, $P'$ (computed with the $H_0=100$ convention)
• Figure 3: Diagonal terms ($k_\parallel=k_\bot$ elements only) of the power spectra or the simulation (dot-dashed line), clean EoR (solid line) and instrumental noise (dashed line) for all the frequency channels (Top: 106--121 MHz, 121-136 MHz, 136-151 MHz) from left to right. Bottom: 151--166 MHz, 166--181 MHz, 181--196 MHz from left to right.
• Figure 4: Illustration of the flagging method in delay space from the HAMSTER team. The top panel shows the amplitude of the visibility data for antenna pair 0 and 1. The bottom panel shows the amplitude of the delay-transformed visibility data for the same pair. Two visible tripes of excess power can be seen around the delay $\eta = 1\,\mu$s. The shaded region shows the delay modes that are excluded from the power spectrum estimation.
• Figure 5: The estimated cylindrical power spectrum of the whole band from the hamster team. The red dashed line shows the region where foregrounds dominate.