BAO from angular clustering: optimization and mitigation of theoretical systematics

We study the methodology and potential theoretical systematics of measuring baryon acoustic oscillations (BAO) using the angular correlation functions in tomographic bins. We calibrate and optimize the pipeline for the Dark Energy Survey Year 1 data set using 1800 mocks. We compare the BAO fitting results obtained with three estimators: the Maximum Likelihood Estimator (MLE), Profile Likelihood, and Markov Chain Monte Carlo. The fit results from the MLE arc the least biased and their derived la error bar are closest to the Gaussian distribution value after removing the extreme mocks with non-detected BAO signal. We show that incorrect assumptions in constructing the template, such as mismatches from the cosmology of the mocks or the underlying photo-z errors, can lead to BAO angular shifts. We find that MLE is the method that best traces this systematic biases, allowing to recover the true angular distance values. In a real survey analysis, it may happen that the final data sample properties are slightly different from those of the mock catalogue. We show that the effect on the mock covariance due to the sample differences can be corrected with the help of the Gaussian covariance matrix or more effectively using the eigenmode expansion of the mock covariance. In the eigenmode expansion, the eigenmodes are provided by some proxy covariance matrix. The eigenmode expansion is significantly less susceptible to statistical fluctuations relative to the direct measurements of the covariance matrix because of the number of free parameters is substantially reduced.