Quantifying the accuracy of the Alcock-Paczyński scaling of baryon acoustic oscillation measurements

We investigate - in a generic setting - the regime of applicability of the Alcock-Paczynski (AP) scaling conventionally applied to test different cosmological models, given a fiducial measurement of the baryon acoustic oscillation (BAO) characteristic scale in the galaxy 2-point correlation function. We quantify the error in conventional AP scaling methods, for which our ignorance about the true cosmology is parameterised in terms of two constant AP scaling parameters. We propose a new, and as it turns out, improved version of the constant AP scaling, also consisting of two scaling parameters. The two constant AP scaling methods are almost indistinguishable when the fiducial model used in data reduction and the "true" underlying cosmology are not differing substantially in terms of metric gradients, but are otherwise expected to differ. Our new methods can be applied to existing analyses through a reinterpretation of the results of the conventional AP scaling. This reinterpretation might be important in model universes where curvature gradients above the scale of galaxies are significant. We test our theoretical findings on $\Lambda$CDM mock catalogues. The conventional constant AP scaling methods are surprisingly successful for pairs of large-scale metrics, but eventually break down when toy models allowing for large metric gradients are tested. The new constant AP scaling methods proposed in this paper are efficient for all test models examined. We find systematic errors of ~1% in the recovery of the BAO scale when the true model is distant from the fiducial, which cannot be attributed to any constant AP approximation. The level of systematic uncertainty is robust to the exact fitting method employed. This indicates that caution must be taken with the error budget when extrapolating the BAO acoustic scale measurements obtained in the standard literature.