A new algorithm for computing statistics of weak lensing by large-scale structure

We describe an efficient algorithm for calculating the statistics of weak lensing by large-scale structure based on a tiled set of independent particle-mesh N-body simulations that telescope in resolution along the line of sight. This efficiency allows us to predict not only the mean properties of lensing observables such as the power spectrum, skewness and kurtosis of the convergence, but also their sampling errors for finite fields of view, which are themselves crucial for assessing the cosmological significance of observations. We find that the non-Gaussianity of the distribution substantially increases the sampling errors for the skewness and kurtosis in the several to tens of arcminutes regime, whereas those for the power spectrum are only fractionally increased even out to wavenumbers where shot noise from the intrinsic ellipticities of the galaxies will likely dominate the errors.