Problems with Pencils: Lensing Covariance of Supernova Distance Measurements

While luminosity distances from Type Ia supernovae (SNe) provide a powerful probe of cosmological parameters, the accuracy with which these distances can be measured is limited by cosmic magnification due to gravitational lensing by the intervening large-scale structure. Spatial clustering of foreground mass fluctuations leads to correlated errors in distance estimates from SNe. By including the full covariance matrix of supernova distance measurements, we show that a future survey covering more than a few square degrees on the sky, and assuming a total of ~2000 SNe, will be largely unaffected by covariance noise. ``Pencil beam'' surveys with small fields of view, however, will be prone to the lensing covariance, leading to potentially significant degradations in cosmological parameter estimates. For a survey with 30 arcmin mean separation between SNe, lensing covariance leads to a ~45% increase in the expected errors in dark energy parameters compared to fully neglecting lensing, and a ~20% increase compared to including just the lensing variance. Given that the lensing covariance is cosmology dependent and cannot be mapped out sufficiently accurately with direct weak lensing observations, surveys with small mean SN separation must incorporate the effects of lensing covariance, including its dependence on the cosmological parameters.