Clustering Redshift Estimation for the Hyper Suprime-Cam Survey

Abstract

Accurate redshift estimates of astronomical objects has been a long-standing goal in astrophysics since it would allow us to determine distances to imaged sources and thus, explore the large-scale structure of the Universe. However, spectroscopic redshifts are typically limited to bright objects and are quite expensive to conduct for a large number of sources. Hence, most modern surveys use photometric redshifts, which rely on spectral energy distribution (SED) templates and do not always provide robust or accurate estimates due to their strong assumptions regarding SEDs. In this work, we attempt to use the clustering redshift technique proposed by Ménard et al. (2013) to estimate the redshift distribution of the galaxy population imaged by the Hyper Suprime-Cam (HSC) survey. This method enables the inference of redshift distributions from measurements of the spatial clustering of arbitrary sources, using a set of reference objects with known redshifts. We test the accuracy of this method on a sample of galaxies with spectroscopic information from the Sloan Digital Sky Survey (SDSS). We show that we can recover the corresponding mean redshifts with an error of order 10%. Compared to using a constant galaxy bias, assuming a linear bias correction reduces the noise in the clustering distribution and improves the accuracy of the mean redshift by about 0.01. Then, we compare clustering and photometric redshifts for subsamples of the HSC data within narrow redshift ranges. Although the clustering technique is able to detect the absence of objects outside the redshift range of the sample, it does not reproduce the exact shape or small scale structure of the photometric distribution. We also map a two dimensional color space to redshift space in order to estimate the redshift probability distributions of individual galaxies. This analysis indicates that the clustering-based redshift inference method provides a good alternative to photometric redshift estimation since it does not assume any prior knowledge of the the spectral energy distribution of the objects.