2018MNRAS.481.3033S

The Gaia spacecraft has just released a large set of parallaxes, including 41 novae for which the fractional error is <30 per cent. I have used these to evaluate the accuracy and bias of the many prior methods for getting nova distances. The best of the prior methods is the geometrical parallaxes from Hubble Space Telescope for just four novae, although the real error bars are 3x larger than stated. The canonical method for prior nova distances has been the expansion parallaxes from the nova shells, but this method is found to have real 1σ uncertainty of 0.95 mag in the distance modulus, and the prior quoted error bars are on average 3.6x worse than advertised. The many variations on the `maximum-magnitude-rate-of-decline' (MMRD) relation are all found to be poor, too poor to be useable, and even to be non-applicable for five out of seven samples of nova, so the MMRD should no longer be used. The prior method of using various measures of the extinction from the interstellar medium have been notoriously bad, but now a new version by Ozdonmez and coworkers has improved this to an unbiased method with 1σ uncertainty of 1.14 mag in the distance modulus. For the future, I recommend in order (1) using the Gaia parallax, (2) using the catalogue of Ozdonmez, (3) using Mmax = -7.0 ± 1.4 mag as an empirical method of poor accuracy, and (4) if none of these methods is available, then to not use the nova for purposes where a distance is needed.