Cosmology with imaging galaxy surveys: the impact of evolving galaxy bias and magnification

TL;DR

This paper investigates how evolving galaxy bias and magnification affect cosmological parameter inference in imaging surveys, highlighting potential biases and proposing mitigation strategies for Stage IV surveys like Rubin-LSST.

Contribution

It introduces models of bias and magnification evolution and demonstrates their significant impact on cosmological constraints, emphasizing the need for careful modeling in future surveys.

Findings

Ignoring evolution causes ~1σ biases in Rubin-LSST constraints.

Evolution impacts are comparable to statistical errors for DES and larger for Rubin-LSST.

Cross-correlations with CMB lensing help mitigate these biases.

Abstract

The joint analysis of galaxy clustering and galaxy-shear cross-correlations (galaxy-galaxy lensing) in imaging surveys constitutes one of the main avenues to obtain cosmological information. Analyses from Stage III surveys have assumed constant galaxy bias and weak lensing magnification coefficients within redshift bins. We use simple models of bias and magnification evolution to show that plausible levels of evolution impact cosmological parameter inference significantly for Stage IV surveys and possibly for current datasets as well. The distance weighting factors in galaxy-galaxy lensing cause it to have a different effective mean redshift within a lens bin than galaxy clustering. If the evolution is not included carefully in the theory modeling, this produces an apparent de-correlation of clustering and lensing even on large scales and leads to biased cosmological constraints. The level of bias is sensitive to the redshift and the width of the lens redshift distributions: for magnitude-limited-like samples it is comparable to the statistical precision even for the full DES survey, and exceeds it for Rubin-LSST. We find that ignoring the evolution can lead to approximately $1\sigma$ biases in the cosmological constraints in future Rubin-LSST analyses. We describe how these effects can be mitigated from the modeling and data sides, in particular by using cross-correlations to constrain the evolution parameters. We also demonstrate that the combination of galaxy clustering and CMB lensing does not suffer from this problem, due to the different distance weighting.