The Next Generation Virgo Cluster Survey. XV. The photometric redshift estimation for background sources

TL;DR

This paper details the development and validation of photometric redshift estimation methods for background sources in the Next Generation Virgo Cluster Survey, achieving high accuracy with specific band combinations and extending the applicable magnitude range.

Contribution

It introduces a new prior for photometric redshift estimation extending to iAB = 12.5 mag and evaluates the impact of different band combinations on redshift accuracy.

Findings

Photometric redshifts for 15.5 ≤ i ≤ 23 mag have bias < 0.02 and less than 5% outliers.

Using u*griz bands yields more accurate redshifts than u*giz bands, especially in 0.3 ≲ zphot ≲ 0.8.

The method's accuracy is validated through comparison with spectroscopic data and correlation function analysis.

Abstract

The Next Generation Virgo Cluster Survey is an optical imaging survey covering 104 deg^2 centered on the Virgo cluster. Currently, the complete survey area has been observed in the u*giz-bands and one third in the r-band. We present the photometric redshift estimation for the NGVS background sources. After a dedicated data reduction, we perform accurate photometry, with special attention to precise color measurements through point spread function-homogenization. We then estimate the photometric redshifts with the Le Phare and BPZ codes. We add a new prior which extends to iAB = 12.5 mag. When using the u*griz-bands, our photometric redshifts for 15.5 \le i \lesssim 23 mag or zphot \lesssim 1 galaxies have a bias |\Delta z| < 0.02, less than 5% outliers, and a scatter \sigma_{outl.rej.} and an individual error on zphot that increase with magnitude (from 0.02 to 0.05 and from 0.03 to 0.10, respectively). When using the u*giz-bands over the same magnitude and redshift range, the lack of the r-band increases the uncertainties in the 0.3 \lesssim zphot \lesssim 0.8 range (-0.05 < \Delta z < -0.02, \sigma_{outl.rej} ~ 0.06, 10-15% outliers, and zphot.err. ~ 0.15). We also present a joint analysis of the photometric redshift accuracy as a function of redshift and magnitude. We assess the quality of our photometric redshifts by comparison to spectroscopic samples and by verifying that the angular auto- and cross-correlation function w(\theta) of the entire NGVS photometric redshift sample across redshift bins is in agreement with the expectations.