Photo-z quality cuts and their effect on the measured galaxy clustering

Abstract

Photometric galaxy surveys are an essential tool to further our understanding of the large-scale structure of the universe, its matter and energy content and its evolution. These surveys necessitate the determination of the galaxy redshifts using photometric techniques (photo-z). Oftentimes, it is advantageous to remove from the galaxy sample those for which one suspects that the photo-z estimation might be unreliable. In this paper, we show that applying these photo-z quality cuts blindly can grossly bias the measured galaxy correlations within and across photometric redshift bins. We then extend the work of Ho et al. (2012) and Ross et al. (2011) to develop a simple and effective method to correct for this using the data themselves. Finally, we apply the method to the Mega-Z catalog, containing about a million luminous red galaxies in the redshift range 0.45 < z < 0.65. After splitting the sample into four \Delta z = 0.05 photo-z bins using the BPZ algorithm, we see how our corrections bring the measured galaxy auto- and cross-correlations into agreement with expectations. We then look for the BAO feature in the four bins, with and without applying the photo-z quality cuts, and find a broad agreement between the BAO scales extracted in both cases. Intriguingly, we observe a correlation between galaxy density and photo-z quality even before any photo-z quality cuts are applied. This may be due to uncorrected observational effects that result in correlated gradients across the sky of the galaxy density and the galaxy photo-z precision. Our correction procedure could also help to mitigate some of these systematic effects.