A Bayesian framework for incorporating line data uncertainties into the evaluation of TDLAS traces using spectroscopic fits

Abstract

Tunable diode laser absorption spectroscopy (TDLAS) is a well-established and robust technique for the analysis of gases. The properties of interest are typically inferred by fitting model spectra to experimentally obtained transmission traces. The required model parameters are taken from databases such as HITRAN and are inherently subject to uncertainty. However, the propagation of line data errors through spectroscopic fits is generally not considered in the literature. Not accounting for such model uncertainties can lead to considerable underestimation of the uncertainties in the derived gas properties. In this article, a Bayesian framework is presented that enables the incorporation of uncertainties of model spectra, computed using line data, into the evaluation of TDLAS traces. It is validated using simulated transmission spectra and found to provide reliable estimates for the quantities of interest and their uncertainties. Thus, it provides a practical tool that contributes to the advancement of spectroscopic analysis.