A Gaussian Process Model for the Local Galactic Velocity Field

Abstract

The Milky Way Galaxy has been a source of inspiration and challenge for astronomers who have focused their efforts to understand the motions of stars, the spin of the Galaxy, and the connection beyond to dark matter and our universe. These efforts are being propelled by the Gaia mission, a space observatory dedicated to cataloging more than one billion stars in the Galaxy. Gaia has observed more than 33 million stars with both the positions and velocities in each dimension to give a full six dimensional solution. The next challenge in Galactic astronomy is to analyze and understand how the Galaxy’s motion fits with our knowledge of theory, and the departures within. The size and complexity of the data make it a challenge to analyse the underlying motion. In this thesis we focus on the mean velocity field in the vicinity of the Sun. A number of researchers have constructed views of Galactic motion via binning. While conceptually simple, the connection to theory can be tenuous. Here we use Gaussian process regression, a non-parametric machine learning method that presents an opportunity for modelling the bulk motion of the Galaxy. This method can inform both the current motion and future state of the Galaxy, calculating continuous derivatives of the velocity field. We discuss the use of observational data with Gaussian process regression, modifications to modelling for use with large data, and highlight the observed motions and their connection to our understanding of the Galaxy. Models agree strongly with Galactic motions explored in literature, and provide a measure of the divergence for examining time evolution of regions of the Galaxy.