Title: Probabilistic detection of spectral line components

Abstract: Resolved kinematical information, such as from molecular gas in star forming regions, is obtained from spectral line observations. However, these observations often contain multiple line-of-sight components, making estimates harder to obtain and interpret. We present a fully automatic method that determines the number of components along the line of sight, or the spectral multiplicity, through Bayesian model selection. The underlying open-source framework, based on nested sampling and conventional spectral line modeling, is tested using the large area ammonia maps of NGC 1333 in Perseus molecular cloud obtained by the Green Bank Ammonia Survey (GAS). Compared to classic approaches, the presented method constrains velocities and velocity dispersions in a larger area. In addition, we find that the velocity dispersion distribution among multiple components did not change substantially from that of a single fit component analysis of the GAS data. These results showcase the power and relative ease of the fitting and model selection method, which makes it a unique tool to extract maximum information from complex spectral data.