| Abstract |
Spatially varying dust extinction is a leading source of photometric non-uniformity in wide-area imaging surveys and can directly limit the precision of Roman cosmology. We will address this vulnerability by developing DUSTER, a field-level inference framework for reconstructing Galactic dust attenuation maps over the Roman HLWAS medium-tier footprint using joint Roman–LSST photometry. Our goal is to deliver dust maps and empirically calibrated attenuation coefficients for the Roman and LSST bands that can be propagated into downstream large-scale structure analyses.
The proposed method uses REDMAGIC-selected red-sequence galaxies as standard crayons. We will first measure the dust-induced fluctuations in the color of red-sequence galaxies. We will then develop a hierarchical Bayesian field-level inference framework that models the dust field as a log-normal random field to jointly infer both the dust map and its uncertainty. We will also measure attenuation coefficients by requiring that dereddened galaxy density maps be uncorrelated with the inferred dust field, targeting 1–3 mmag dereddening accuracy across Roman and LSST bands We will validate the method on simulated data and then apply it to HLWAS Field 1 of the medium tier. Our feasibility study on simulated data already recovers the input dust map with 6 mmag rms accuracy, consistent with Roman cosmology requirements.
This medium proposal will deliver: 1) dust attenuation maps over the Roman HLWAS medium tier footprint available during the award period; 2) empirically calibrated attenuation coefficients for Roman and LSST bands; and 3) a public release of the DUSTER code and associated data products. These products will provide enabling calibration infrastructure for Roman cosmology while also supporting Galactic dust studies. |