Data-driven Cosmology from Three-dimensional Light Cones

TL;DR

This paper introduces a novel data-driven method for analyzing large-scale cosmological survey data that simultaneously constrains structure, source properties, and noise without relying on source detection or redshift estimates.

Contribution

The paper presents a formalism that leverages two-point statistics to analyze multifrequency cosmological data without prior assumptions beyond homogeneity and isotropy.

Findings

Successfully recovers input signals and noise in mock data

Quantifies uncertainties in parameter constraints

Flexible framework applicable to various future surveys

Abstract

We present a data-driven technique to analyze multifrequency images from upcoming cosmological surveys mapping large sky area. Using full information from the data at the two-point level, our method can simultaneously constrain the large-scale structure (LSS), the spectra and redshift distribution of emitting sources, and the noise in the observed data without any prior assumptions beyond the homogeneity and isotropy of cosmological perturbations. In particular, the method does not rely on source detection or photometric or spectroscopic redshift estimates. Here, we present the formalism and demonstrate our technique with a mock observation from nine optical and near-infrared photometric bands. Our method can recover the input signal and noise without bias, and quantify the uncertainty on the constraints. Our technique provides a flexible framework to analyze the LSS observation traced by different types of sources, which has potential for wide application to current or future cosmological datasets such as SPHEREx, Rubin Observatory, Euclid, or the Nancy Grace Roman Space Telescope.