Mitigating systematic errors in angular correlation function measurements from wide field surveys

Abstract

We present an investigation into the effects of survey systematics such as varying depth, point spread function (PSF) size, and extinction on the galaxy selection and correlation in photometric, multi-epoch, wide area surveys. We take the Canada-France-Hawaii Telescope Lensing Survey (CFHTLenS) as an example. Variations in galaxy selection due to systematics are found to cause density fluctuations of up to 10% for some small fraction of the area for most galaxy redshift slices and as much as 50% for some extreme cases of faint high-redshift samples. This results in correlations of galaxies against survey systematics of order $\sim$1% when averaged over the survey area. We present an empirical method for mitigating these systematic correlations from measurements of angular correlation functions using weighted random points. These weighted random catalogs are estimated from the observed galaxy over densities by mapping these to survey parameters. We are able to model and mitigate the effect of systematic correlations allowing for non-linear dependencies of density on systematics. Applied to CFHTLenS we find that the method reduces spurious correlations in the data by a factor two for most galaxy samples and as much as an order of magnitude in others. Such a treatment is particularly important for an unbiased estimation of very small correlation signals, as e.g. from weak gravitational lensing magnification bias. We impose a criterion for using a galaxy sample in a magnification measurement of the majority of the systematic correlations show improvement and are less than 10% of the expected magnification signal when combined in the galaxy cross correlation. After correction the galaxy samples in CFHTLenS satisfy this criterion for $z_{\rm phot}<0.9$ and will be used in a future analysis of magnification.