Galaxy Clustering with LSST: Effects of Number Count Bias from Blending

TL;DR

This study investigates how galaxy blending biases number counts and affects clustering measurements in LSST data, finding small but significant impacts on redshift and clustering at certain scales, with minimal effect on cosmological parameters.

Contribution

It provides the first detailed analysis of how blending influences galaxy number counts and clustering statistics in LSST, highlighting scale-dependent biases and their implications for cosmological inference.

Findings

Blending causes small but statistically significant biases in redshift measurements.

Clustering differences due to blending are >3σ below 10 arcminutes scale.

Cosmological parameters are largely unaffected by blending within LSST Y1 and Y5 scales.

Abstract

The Vera C. Rubin Observatory Legacy Survey of Space and Time (LSST) will survey the southern sky to create the largest galaxy catalog to date, and its statistical power demands an improved understanding of systematic effects such as source overlaps, also known as blending. In this work we study how blending introduces a bias in the number counts of galaxies (instead of the flux and colors), and how it propagates into galaxy clustering statistics. We use the $300\,$deg$^2$ DC2 image simulation and its resulting galaxy catalog (LSST Dark Energy Science Collaboration et al. 2021) to carry out this study. We find that, for a LSST Year 1 (Y1)-like cosmological analyses, the number count bias due to blending leads to small but statistically significant differences in mean redshift measurements when comparing an observed sample to an unblended calibration sample. In the two-point correlation function, blending causes differences greater than 3$\sigma$ on scales below approximately $10'$, but large scales are unaffected. We fit $\Omega_{\rm m}$ and linear galaxy bias in a Bayesian cosmological analysis and find that the recovered parameters from this limited area sample, with the LSST Y1 scale cuts, are largely unaffected by blending. Our main results hold when considering photometric redshift and a LSST Year 5 (Y5)-like sample.