Mask Effects on Cosmological Studies with Weak Lensing Peak Statistics

TL;DR

This study investigates how mask effects influence weak lensing peak statistics, revealing that masks can significantly bias cosmological parameter estimates, and proposes a two-noise-level model to mitigate this bias.

Contribution

The paper introduces a two-noise-level model that effectively accounts for mask effects on peak statistics, reducing bias in cosmological parameter estimation.

Findings

Mask effects increase high peak fractions, biasing cosmological results.

Excluding large masks reduces bias but loses survey area.

A two-noise-level model accurately reproduces mask effects.

Abstract

In this paper, we analyze in detail with numerical simulations how the mask effect can influence the weak lensing peak statistics reconstructed from the shear measurement of background galaxies. It is found that high peak fractions are systematically enhanced due to masks, the larger the masked area, the higher the enhancement. In the case with about $13\%$ of the total masked area, the fraction of peaks with SNR $\nu\ge 3$ is $\sim 11\%$ in comparison with $\sim 7\%$ of the mask-free case in our considered cosmological model. This can induce a large bias on cosmological studies with weak lensing peak statistics. Even for a survey area of $9\hbox{ deg}^2$, the bias in $(\Omega_m, \sigma_8)$ is already close to $3\sigma$. It is noted that most of the affected peaks are close to the masked regions. Therefore excluding peaks in those regions can reduce the bias but at the expense of loosing usable survey areas. Further investigations find that the enhancement of high peaks number can be largely attributed to higher noise led by the fewer number of galaxies usable in the reconstruction. Based on Fan et al. (2010), we develop a model in which we exclude only those large masks with radius larger than $3\arcmin. For the remained part, we treat the areas close to and away from the masked regions separately with different noise levels. It is shown that this two-noise-level model can account for the mask effect on peak statistics very well and the cosmological bias is significantly reduced.