Analysing surveys of our Galaxy - II. Determining the potential

We consider the problem of determining the Galaxy's gravitational potential from a star catalogue. We show that orbit-based approaches to this problem suffer from unacceptable numerical noise deriving from the use of only a finite number of orbits. An alternative approach, which requires an ability to determine the model's phase-space density at predetermined positions and velocities, has a level of numerical noise that lies well below the intrinsic uncertainty associated with the finite size of the catalogue analysed. A catalogue of 10 000 stars brighter than V = 17 and distributed over the sky at b > 30 degrees enables us to determine the scaleheight of the disc that contributes to the potential with an uncertainty below 20 pc if the catalogue gives proper motions, line-of-sight velocities and parallaxes with errors typical of the Gaia Catalogue, rising to 36 pc if only proper motions are available. The uncertainty in the disc's scalelength is significantly smaller than 0.25 kpc.