exoALMA. VIII. Probabilistic Moment Maps and Data Products using Non-parametric Linear Models

Extracting robust inferences on physical quantities from disk kinematics measured from Doppler-shifted molecular line emission is challenging due to the data's size and complexity. In this paper we develop a flexible linear model of the intensity distribution in each frequency channel, accounting for spatial correlations from the point spread function. The analytic form of the model's posterior enables probabilistic data products through sampling. Our method debiases peak intensity, peak velocity, and line width maps, particularly in disk substructures that are only partially resolved. These are needed in order to measure disk mass, turbulence, pressure gradients, and to detect embedded planets. We analyse HD 135344B, MWC 758, and CQ Tau, finding velocity substructures 50--200 ${\rm m s^{-1}}$ greater than with conventional methods. Additionally, we combine our approach with discminer in a case study of J1842. We find that uncertainties in stellar mass and inclination increase by an order of magnitude due to the more realistic noise model. More broadly, our method can be applied to any problem requiring a probabilistic model of an intensity distribution conditioned on a point spread function.