We study correlated fluctuations of Type~Ia supernova observables due to peculiar velocities of both the observer and the supernova host galaxies, and their impact on cosmological parameter estimation. We demonstrate using the CosmicFlows-3 dataset that at low redshifts the corrections for peculiar velocities in the JLA catalogue have been systematically underestimated. By querying a horizon-size N-body simulation we find that compared to a randomly placed observer, an observer in an environment like our local Universe will see 2-8 times stronger correlations between supernovae in the JLA catalogue. Hence the covariances usually employed assuming a typical observer are unphysical and underestimate the effects of coherent motion of the supernova host galaxies. Contrary to previous studies which asserted that this should have negligible effect on cosmological parameter estimation, we find that when peculiar velocities are treated consistently the JLA data favours significantly smaller values of the dark energy density than in the standard $Lambda$CDM model. A joint fit to simultaneously determine the cosmological parameters and the bulk flow indicates that the latter is around 250 km/s even beyond 200$h^{-1}$ Mpc. The local bulk flow is thus an essential nuisance parameter which must be included in cosmological model fitting when analysing supernova data.
Download