What Lies Beneath: Using p(z) to Reduce Systematic Photometric Redshift Errors

TL;DR

This paper demonstrates that utilizing the full photometric redshift probability distribution p(z) significantly reduces systematic errors in cosmological measurements, and introduces a simple estimator that retains these benefits with minimal modifications.

Contribution

The paper introduces a simple, efficient estimator that leverages p(z) to reduce systematic errors in photometric redshift estimates for cosmology.

Findings

Using p(z) reduces systematic errors in dark energy parameter estimation.

The simple estimator performs nearly as well as using the full p(z).

Systematic error reduction is achieved at no additional survey cost.

Abstract

We use simulations to demonstrate that photometric redshift "errors" can be greatly reduced by using the photometric redshift probability distribution p(z) rather than a one-point estimate such as the most likely redshift. In principle this involves tracking a large array of numbers rather than a single number for each galaxy. We introduce a very simple estimator that requires tracking only a single number for each galaxy, while retaining the systematic-error-reducing properties of using the full p(z) and requiring only very minor modifications to existing photometric redshift codes. We find that using this redshift estimator (or using the full p(z)) can substantially reduce systematics in dark energy parameter estimation from weak lensing, at no cost to the survey.