Atmospheric dispersion effects in weak lensing measurements

Abstract

The wavelength dependence of atmospheric refraction causes elongation of finite-bandwidth images along the elevation vector, which produces spurious signals in weak gravitational lensing shear measurements unless this atmospheric dispersion is calibrated and removed to high precision. Because astrometric solutions and PSF characteristics are typically calibrated from stellar images, differences between the reference stars' spectra and the galaxies' spectra will leave residual errors in both the astrometric positions (dr) and in the second moment (width) of the wavelength-averaged PSF (dv) for galaxies.We estimate the level of dv that will induce spurious weak lensing signals in PSF-corrected galaxy shapes that exceed the statistical errors of the DES and the LSST cosmic-shear experiments. We also estimate the dr signals that will produce unacceptable spurious distortions after stacking of exposures taken at different airmasses and hour angles. We also calculate the errors in the griz bands, and find that dispersion systematics, uncorrected, are up to 6 and 2 times larger in g and r bands,respectively, than the requirements for the DES error budget, but can be safely ignored in i and z bands. For the LSST requirements, the factors are about 30, 10, and 3 in g, r, and i bands,respectively. We find that a simple correction linear in galaxy color is accurate enough to reduce dispersion shear systematics to insignificant levels in the r band for DES and i band for LSST,but still as much as 5 times than the requirements for LSST r-band observations. More complex corrections will likely be able to reduce the systematic cosmic-shear errors below statistical errors for LSST r band. But g-band effects remain large enough that it seems likely that induced systematics will dominate the statistical errors of both surveys, and cosmic-shear measurements should rely on the redder bands.