Speaker
Description
Many ground- and space-based photometric surveys are quickly approaching sensitivities where correlations between point-like and diffuse emission can lead to significant biases and mis-estimated uncertainties if ignored. Probabilistic cataloging (Portillo et al. 2017, Daylan et al. 2017) is a Bayesian hierarchical modeling framework where covariances due to blending can be explored by sampling from the (transdimensional) model space of catalogs consistent with a given image dataset. In this work, we extend the formalism of probabilistic cataloging to jointly model point-like and diffuse emission through a Fourier component template-based approach, implemented in the code Diffuse Background Cataloger (DBCAT). Using a combination of mock and real Herschel-SPIRE sub-millimeter multiband map data, we demonstrate that point source and diffuse emission can be reliably separated and estimated, including in the confusion-limited regime. This is validated using catalog- and field-based summary statistics of the reconstructed components. Beyond the set of global Fourier basis templates used by DBCAT, additional templates can be included to infer the contributions of unique extended emission components. As an example, we demonstrate that the thermal Sunyaev-Zel'dovich (tSZ) effect can be reliably estimated in cluster fields observed by SPIRE with contamination from cosmic infrared background (CIB) galaxies and cirrus dust, with proper marginalization over a CIB model in which the number of galaxies is unknown a priori.
