The impact of cosmic variance on simulating weak lensing surveys

TL;DR

This paper investigates how cosmic variance in small HST datasets affects galaxy shape and morphology distributions, potentially biasing shear calibration in upcoming weak lensing surveys, and suggests mitigation strategies.

Contribution

It provides the first direct analysis of cosmic variance effects on galaxy shape distributions using the COSMOS survey, highlighting their impact on shear calibration accuracy.

Findings

Cosmic variance causes ~5% variation in RMS ellipticity within redshift bins.

Such variations can lead to ~1% shear calibration uncertainty.

Mitigation requires larger datasets and optimized redshift binning.

Abstract

Upcoming weak lensing surveys will survey large cosmological volumes to measure the growth of cosmological structure with time and thereby constrain dark energy. One major systematic uncertainty in this process is the calibration of the weak lensing shape distortions, or shears. Most upcoming surveys plan to test several aspects of their shear estimation algorithms using sophisticated image simulations that include realistic galaxy populations based on high-resolution data from the Hubble Space Telescope (HST). However, existing datasets from the (HST) cover very small cosmological volumes, so cosmic variance could cause the galaxy populations in them to be atypical. A narrow redshift slice from such surveys could be dominated by a single large overdensity or underdensity. In that case, the morphology-density relation could alter the local galaxy populations and yield an incorrect calibration of shear estimates as a function of redshift. We directly test this scenario using the COSMOS survey, the largest-area (HST) survey to date, and show how the statistical distributions of galaxy shapes and morphological parameters (e.g., S\'{e}rsic $n$) are influenced by redshift-dependent cosmic variance. The typical variation in RMS ellipticity due to environmental effects is 5 per cent (absolute, not relative) for redshift bins of width $\Delta z=0.05$, which could result in uncertain shear calibration at the 1 per cent level. We conclude that the cosmic variance effects are large enough to exceed the systematic error budget of future surveys, but can be mitigated with careful choice of training dataset and sufficiently large redshift binning.