[Paper Review] Matching Radio Catalogs with Realistic Geometry: Application to SWIRE and ATLAS

Dongwei Fan, Tamás Budavári+2 more|arXiv (Cornell University)|May 4, 2015

Radio Astronomy Observations and Technology1 references3 citations

TL;DR

This paper presents a Bayesian hypothesis testing method that automates crossmatching of radio catalogs with infrared/optical catalogs by modeling realistic radio source geometry—specifically, core-and-lobe structures of double-lobed radio galaxies. The algorithm improves upon traditional point-source matching by assessing likelihoods of associations, including core-lobes, lobe-lobes, and physically associated components, achieving strong agreement with manual classifications in the SWIRE and ATLAS CDF-S surveys without human intervention.

ABSTRACT

Crossmatching catalogs at different wavelengths is a difficult problem in astronomy, especially when the objects are not point-like. At radio wavelengths an object can have several components corresponding, for example, to a core and lobes. {Considering not all radio detections correspond to visible or infrared sources, matching these catalogs can be challenging.} Traditionally this is done by eye for better quality, which does not scale to the large data volumes expected from the next-generation of radio telescopes. We present a novel automated procedure, using Bayesian hypothesis testing, to achieve reliable associations by explicit modelling of a particular class of radio-source morphology. {The new algorithm not only assesses the likelihood of an association between data at two different wavelengths, but also tries to assess whether different radio sources are physically associated, are double-lobed radio galaxies, or just distinct nearby objects.} Application to the SWIRE and ATLAS CDF-S catalogs shows that this method performs well without human intervention.

Motivation & Objective

• To address the challenge of crossmatching radio catalogs with extended, non-point-like sources such as radio galaxies with cores and lobes.
• To automate a process traditionally done by visual inspection, which is infeasible for upcoming large-scale radio surveys.
• To improve matching reliability by explicitly modeling source morphology rather than relying on simple positional proximity.
• To distinguish between physically associated components (e.g., core and lobes) and chance alignments of separate sources.
• To develop a scalable, objective method for large-volume data from SKA pathfinder telescopes.

Proposed method

• Uses Bayesian hypothesis testing to compute the likelihood of associations between radio sources and point-like counterparts in other wavebands.
• Models radio sources as a core and two symmetrically placed lobes, assuming a linear or near-linear configuration.
• Applies the flat-sky approximation to project local sky coordinates onto tangent planes for efficient computation.
• Employs likelihood integrals over parameter space to evaluate hypotheses: single source, core-lobes, or lobe-lobe pairs.
• Computes Bayes Factors to compare competing association hypotheses and select the most probable one.
• Incorporates astrometric uncertainties via a Gaussian likelihood function for each source position.

Experimental results

Research questions

• RQ1 Can a Bayesian framework reliably associate radio sources with complex morphologies (e.g., core and lobes) to optical/infrared counterparts?
• RQ2 How does modeling realistic radio source geometry improve matching accuracy compared to point-source-only methods?
• RQ3 To what extent can automated matching reproduce results from manual visual inspection in deep surveys?
• RQ4 What types of false positives or misclassifications arise from geometric modeling limitations?
• RQ5 How can automated systems handle ambiguous cases where multiple components are close but not physically linked?

Key findings

• The algorithm successfully identified 29 additional radio triplets not in the reference catalog, most of which were consistent with known radio galaxy structures.
• The method achieved strong agreement with manual classifications from Norris et al. (2006), particularly in identifying core-lobed and lobe-lobed configurations.
• 14 of the 29 extra associations were classified as lobe-lobe pairs, indicating the model’s sensitivity to symmetric, extended radio structures.
• False positives arose when two radio components were embedded within a single extended infrared source, leading to incorrect lobe-lobe associations.
• The algorithm struggled with non-linear or asymmetric morphologies due to the current model’s assumption of a straight-line configuration.
• Visual inspection revealed that extended or low-signal-to-noise features (e.g., faint bridges between lobes) were not captured by the catalog data, limiting algorithm performance.