On the limitations of statistical absorption studies with the Sloan Digital Sky Surveys I-III

We investigate the limitations of statistical absorption measurements with the Sloan Digital Sky Survey (SDSS) optical spectroscopic surveys.We show that changes in the data reduction strategy throughout different data releases have led to a better accuracy at long wavelengths, in particular for sky line subtraction, but a degradation at shortwavelengths with the emergence of systematic spectral features with an amplitude of about 1 per cent.We show that these features originate from inaccuracy in the fitting of modelled F-star spectra used for flux calibration. The best-fitting models for those stars are found to systematically overestimate the strength of metal lines and underestimate that of Lithium. We also identify the existence of artefacts due to masking and interpolation procedures at the wavelengths of the hydrogen Balmer series leading to the existence of artificial Balmer a absorption in all SDSS optical spectra. All these effects occur in the rest frame of the standard stars and therefore present Galactic longitude variations due to the rotation of the Galaxy.We demonstrate that the detection of certain weak absorption lines reported in the literature is solely due to calibration effects. Finally, we discuss new strategies to mitigate these issues.