A New Approach to Multi-wavelength Associations of Astronomical Sources

TL;DR

This paper introduces a Bayesian-based automated method for associating astronomical sources across multiple wavelengths, improving accuracy in crowded fields with high source density and positional uncertainties.

Contribution

The paper presents a novel Bayesian framework combining spatial and SED data for multi-wavelength source association, validated on simulated and real datasets.

Findings

Bayes factor effectively measures match confidence.

Method outperforms simple nearest neighbor associations.

Applicable to future Herschel datasets.

Abstract

One of the biggest problems faced by current and next-generation astronomical surveys is trying to produce large numbers of accurate cross identifications across a range of wavelength regimes with varying data quality and positional uncertainty. Until recently simple spatial "nearest neighbour" associations have been sufficient for most applications. However as advances in instrumentation allow more sensitive images to be made the rapid increase in the source density has meant that source confusion across multiple wavelengths is a serious problem. The field of far-IR and sub-mm astronomy has been particularly hampered by such problems. The poor angular resolution of current sub-mm and far-IR instruments is such that in a lot of cases there are multiple plausible counterparts for each source at other wavelengths. Here we present a new automated method of producing associations between sources at different wavelengths using a combination of spatial and SED information set in the Bayesian framework presented by Budavari & Szalay (2008). Testing of the technique is performed on both simulated catalogues of sources from GaLICS and real data from multi-wavelength observations of the SXDF. It is found that a single figure of merit, the Bayes factor, can be effectively used to describe the confidence in the match. Further applications of this technique to future Herschel datasets are discussed.