Deriving the gas kinematics of disc galaxies via a Bayesian approach

This thesis presents a new 3D-Bayesian global fitting tool designed to model the kinematics of line-emitting gas in disc galaxies. The method represents a novel contribution to the field, whereas current approaches are typically non-Bayesian and/or based on independent, circular, ring-by-ring fits, or rely on moment maps (e.g., velocity maps) rather than exploiting the full information contained in the datacube (i.e., two spatial dimensions plus one spectral/velocity dimension). The Bayesian and global nature of this method is intended to improve error estimation, and provide more robust kinematic measurements in cases of low signal-to-noise ratios or limited spatial resolution, conditions often encountered in observations at high redshift. Additionally, the method enables fitting of the vertical structure of the gaseous disc, a component that is frequently neglected, but may introduce biases in the recovered kinematic parameters. In this work, we present the first tests of the method using simplified models of galactic discs. The fitting performance was evaluated under varying inclinations, spatial resolutions, and signal-to-noise ratios. We found that most input parameters were recovered within 20\% of their true values, in line with the accuracy of existing tools used widely in the community. However, the reported uncertainties were found to be overly optimistic, with fewer than 30\% of the recovered parameters falling within $1\sigma$ of their true values. Our preliminary analysis suggests that this overconfidence may arise from issues related to the model normalisation and how observational noise is incorporated into the Bayesian likelihood. Although still in the early stages of development, this new method demonstrates promising potential for future applications to real data. Further refinements in the likelihood formulation and noise modeling will be critical to ensure accurate uncertainty quantification and reliable scientific interpretations.