arXiv1998
The measurement of shape parameters of sources in astronomical images is usually performed by assuming that the underlying noise is uncorrelated. Spatial noise correlation is however present in practice due to various observational effects and can affect source shape parameters. This effect is particularly important for measurements of weak gravitational lensing, for which the sought image distortions are typically of the order of only 1%. We compute the effect of correlated noise on two-dimensional gaussian fits in full generality. The noise properties are naturally quantified by the noise autocorrelation function (ACF), which is easily measured in practice. We compute the resulting bias on the mean, variance and covariance of the source parameters, and the induced correlation between the shapes of neighboring sources. We show that these biases are of second order in the inverse signal-to-noise ratio of the source, and could thus be overlooked if bright stars are used to monitor systematic distortions. Radio interferometric surveys are particularly prone to this effect because of the long-range pixel correlations produced by the Fourier inversion involved in their image construction. As a concrete application, we consider the search for weak lensing by large-scale structure with the FIRST radio survey. We measure the noise ACF for a FIRST coadded field, and compute the resulting ellipticity correlation function induced by the noise. In comparison with the weak-lensing signal expected in CDM models, the noise correlation effect is important on small angular scales, but is negligible for source separations greater than about 1 arcmin. We also discuss how noise correlation can affect weak-lensing studies with optical surveys.